2022-12-22 10:29:13,705 INFO [train.py:966] (3/4) Training started 2022-12-22 10:29:13,705 INFO [train.py:976] (3/4) Device: cuda:3 2022-12-22 10:29:13,707 INFO [train.py:985] (3/4) {'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': '1.23.2', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'efd83642a940dc7db08688cc0791985bed1fafcd', 'k2-git-date': 'Sun Nov 27 19:12:00 2022', 'lhotse-version': '1.4.0.dev+git.21b5b3f.dirty', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'zipformer_streaming', 'icefall-git-sha1': '2f75b00-dirty', 'icefall-git-date': 'Wed Dec 21 22:15:21 2022', 'icefall-path': '/ceph-zw/workspace/zipformer/icefall_streaming', 'k2-path': '/ceph-zw/workspace/share/k2-last/k2/python/k2/__init__.py', 'lhotse-path': '/ceph-jb/yaozengwei/workspace/rnnt/lhotse/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1216192652-5bcf7587b4-n6q9m', 'IP address': '10.177.74.211'}, 'world_size': 4, 'master_port': 12345, 'tensorboard': True, 'num_epochs': 30, 'start_epoch': 1, 'start_batch': 0, 'exp_dir': PosixPath('pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'base_lr': 0.05, 'lr_batches': 5000, 'lr_epochs': 3.5, 'context_size': 2, 'prune_range': 5, 'lm_scale': 0.25, 'am_scale': 0.0, 'simple_loss_scale': 0.5, 'seed': 42, 'print_diagnostics': False, 'inf_check': False, 'save_every_n': 2000, 'keep_last_k': 30, 'average_period': 200, 'use_fp16': True, '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, 'short_chunk_size': 50, 'num_left_chunks': 4, 'full_libri': True, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 750, '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} 2022-12-22 10:29:13,707 INFO [train.py:987] (3/4) About to create model 2022-12-22 10:29:14,178 INFO [zipformer.py:185] (3/4) 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. 2022-12-22 10:29:14,203 INFO [train.py:991] (3/4) Number of model parameters: 70369391 2022-12-22 10:29:18,756 INFO [train.py:1006] (3/4) Using DDP 2022-12-22 10:29:19,179 INFO [asr_datamodule.py:398] (3/4) About to get train-clean-100 cuts 2022-12-22 10:29:19,188 INFO [asr_datamodule.py:405] (3/4) About to get train-clean-360 cuts 2022-12-22 10:29:19,197 INFO [asr_datamodule.py:412] (3/4) About to get train-other-500 cuts 2022-12-22 10:29:19,207 INFO [asr_datamodule.py:224] (3/4) Enable MUSAN 2022-12-22 10:29:19,208 INFO [asr_datamodule.py:225] (3/4) About to get Musan cuts 2022-12-22 10:29:21,333 INFO [asr_datamodule.py:249] (3/4) Enable SpecAugment 2022-12-22 10:29:21,334 INFO [asr_datamodule.py:250] (3/4) Time warp factor: 80 2022-12-22 10:29:21,334 INFO [asr_datamodule.py:260] (3/4) Num frame mask: 10 2022-12-22 10:29:21,334 INFO [asr_datamodule.py:273] (3/4) About to create train dataset 2022-12-22 10:29:21,334 INFO [asr_datamodule.py:300] (3/4) Using DynamicBucketingSampler. 2022-12-22 10:29:25,007 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp1.1 from training. Duration: 22.2954375 2022-12-22 10:29:26,172 INFO [asr_datamodule.py:315] (3/4) About to create train dataloader 2022-12-22 10:29:26,173 INFO [asr_datamodule.py:429] (3/4) About to get dev-clean cuts 2022-12-22 10:29:26,174 INFO [asr_datamodule.py:436] (3/4) About to get dev-other cuts 2022-12-22 10:29:26,176 INFO [asr_datamodule.py:346] (3/4) About to create dev dataset 2022-12-22 10:29:26,429 INFO [asr_datamodule.py:363] (3/4) About to create dev dataloader 2022-12-22 10:29:26,430 INFO [train.py:1206] (3/4) Sanity check -- see if any of the batches in epoch 1 would cause OOM. 2022-12-22 10:29:30,239 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp1.1 from training. Duration: 22.2954375 2022-12-22 10:29:35,306 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp1.1 from training. Duration: 22.2954375 2022-12-22 10:29:39,139 WARNING [train.py:1060] (3/4) Exclude cut with ID 298-126791-0067-24026-0_sp0.9 from training. Duration: 21.438875 2022-12-22 10:29:39,332 WARNING [train.py:1060] (3/4) Exclude cut with ID 5652-39938-0025-23684-0_sp0.9 from training. Duration: 22.2055625 2022-12-22 10:29:46,952 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0 from training. Duration: 24.525 2022-12-22 10:29:48,155 WARNING [train.py:1060] (3/4) Exclude cut with ID 3699-47246-0007-3408-0_sp0.9 from training. Duration: 20.26675 2022-12-22 10:29:48,586 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp0.9 from training. Duration: 27.25 2022-12-22 10:29:51,706 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0 from training. Duration: 21.68 2022-12-22 10:29:52,123 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0 from training. Duration: 21.6300625 2022-12-22 10:29:52,900 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0_sp0.9 from training. Duration: 24.033375 2022-12-22 10:29:54,957 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0 from training. Duration: 22.905 2022-12-22 10:29:55,001 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp1.1 from training. Duration: 23.4318125 2022-12-22 10:30:00,301 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp1.1 from training. Duration: 20.82275 2022-12-22 10:30:00,353 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp0.9 from training. Duration: 25.45 2022-12-22 10:30:02,398 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0 from training. Duration: 25.775 2022-12-22 10:30:03,114 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0_sp0.9 from training. Duration: 22.25 2022-12-22 10:30:04,077 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0 from training. Duration: 26.205 2022-12-22 10:30:05,274 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp0.9 from training. Duration: 30.1555625 2022-12-22 10:30:05,457 WARNING [train.py:1060] (3/4) Exclude cut with ID 1265-135635-0050-6781-0_sp0.9 from training. Duration: 21.8333125 2022-12-22 10:30:05,757 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp1.1 from training. Duration: 20.6545625 2022-12-22 10:30:07,227 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0045-39920-0_sp0.9 from training. Duration: 20.52225 2022-12-22 10:30:07,943 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp0.9 from training. Duration: 29.1166875 2022-12-22 10:30:10,574 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133211-0007-59831-0_sp0.9 from training. Duration: 21.388875 2022-12-22 10:30:11,839 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0 from training. Duration: 22.72 2022-12-22 10:30:11,882 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0_sp0.9 from training. Duration: 22.7444375 2022-12-22 10:30:13,998 WARNING [train.py:1060] (3/4) Exclude cut with ID 4133-6541-0027-40495-0_sp1.1 from training. Duration: 0.9681875 2022-12-22 10:30:14,132 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0_sp0.9 from training. Duration: 22.3166875 2022-12-22 10:30:14,309 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133212-0015-59917-0_sp0.9 from training. Duration: 21.8166875 2022-12-22 10:30:18,478 WARNING [train.py:1060] (3/4) Exclude cut with ID 4957-30119-0041-23990-0_sp0.9 from training. Duration: 20.22775 2022-12-22 10:30:20,024 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp1.1 from training. Duration: 24.67275 2022-12-22 10:30:21,031 WARNING [train.py:1060] (3/4) Exclude cut with ID 3082-165428-0081-50734-0_sp0.9 from training. Duration: 21.8055625 2022-12-22 10:30:22,918 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0_sp0.9 from training. Duration: 22.6666875 2022-12-22 10:30:25,081 WARNING [train.py:1060] (3/4) Exclude cut with ID 2411-132532-0017-82279-0_sp1.1 from training. Duration: 0.9681875 2022-12-22 10:30:26,001 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0 from training. Duration: 22.485 2022-12-22 10:30:27,926 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp1.1 from training. Duration: 23.82275 2022-12-22 10:30:28,406 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0 from training. Duration: 20.77 2022-12-22 10:30:28,687 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0_sp0.9 from training. Duration: 24.088875 2022-12-22 10:30:29,815 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp1.1 from training. Duration: 20.4409375 2022-12-22 10:30:33,047 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0_sp0.9 from training. Duration: 22.511125 2022-12-22 10:30:33,072 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0 from training. Duration: 20.675 2022-12-22 10:30:37,182 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp0.9 from training. Duration: 24.9833125 2022-12-22 10:30:38,254 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0 from training. Duration: 27.14 2022-12-22 10:30:38,829 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0 from training. Duration: 22.44 2022-12-22 10:30:42,113 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0060-62364-0_sp0.9 from training. Duration: 21.361125 2022-12-22 10:30:42,333 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp1.1 from training. Duration: 27.0318125 2022-12-22 10:30:42,679 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp0.9 from training. Duration: 28.638875 2022-12-22 10:30:43,218 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0 from training. Duration: 20.4 2022-12-22 10:30:44,338 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0 from training. Duration: 20.025 2022-12-22 10:30:44,346 WARNING [train.py:1060] (3/4) Exclude cut with ID 2364-131735-0112-64612-0_sp0.9 from training. Duration: 20.488875 2022-12-22 10:30:44,544 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0 from training. Duration: 29.735 2022-12-22 10:30:47,902 WARNING [train.py:1060] (3/4) Exclude cut with ID 7276-92427-0014-12983-0_sp0.9 from training. Duration: 21.3055625 2022-12-22 10:30:47,958 WARNING [train.py:1060] (3/4) Exclude cut with ID 1025-75365-0008-79168-0_sp0.9 from training. Duration: 22.0666875 2022-12-22 10:30:52,398 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0 from training. Duration: 20.26 2022-12-22 10:30:52,901 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0030-9324-0_sp0.9 from training. Duration: 21.3444375 2022-12-22 10:30:55,512 WARNING [train.py:1060] (3/4) Exclude cut with ID 497-129325-0061-62254-0_sp1.1 from training. Duration: 0.97725 2022-12-22 10:30:56,948 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0_sp0.9 from training. Duration: 22.97225 2022-12-22 10:30:58,032 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0047-39922-0_sp0.9 from training. Duration: 21.97775 2022-12-22 10:30:59,125 WARNING [train.py:1060] (3/4) Exclude cut with ID 1112-1043-0006-89194-0_sp0.9 from training. Duration: 21.8333125 2022-12-22 10:30:59,473 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0 from training. Duration: 20.47 2022-12-22 10:31:01,792 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0037-39912-0_sp0.9 from training. Duration: 20.67225 2022-12-22 10:31:02,463 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp0.9 from training. Duration: 25.2444375 2022-12-22 10:31:03,928 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0021-76797-0_sp0.9 from training. Duration: 21.1445 2022-12-22 10:31:06,617 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp0.9 from training. Duration: 33.038875 2022-12-22 10:31:08,620 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64291-0000-16059-0_sp0.9 from training. Duration: 20.0944375 2022-12-22 10:31:09,160 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp1.1 from training. Duration: 20.4 2022-12-22 10:31:09,435 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0 from training. Duration: 20.085 2022-12-22 10:31:09,827 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0_sp0.9 from training. Duration: 23.07775 2022-12-22 10:31:11,517 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp0.9 from training. Duration: 24.9333125 2022-12-22 10:31:13,663 WARNING [train.py:1060] (3/4) Exclude cut with ID 5118-111612-0016-124680-0_sp0.9 from training. Duration: 20.388875 2022-12-22 10:31:13,874 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp1.1 from training. Duration: 20.3590625 2022-12-22 10:31:16,327 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0_sp1.1 from training. Duration: 0.836375 2022-12-22 10:31:18,453 WARNING [train.py:1060] (3/4) Exclude cut with ID 8565-290391-0049-67394-0_sp0.9 from training. Duration: 21.3166875 2022-12-22 10:31:19,538 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0029-104863-0_sp0.9 from training. Duration: 22.1055625 2022-12-22 10:31:19,914 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp1.1 from training. Duration: 21.77725 2022-12-22 10:31:20,609 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp0.9 from training. Duration: 27.8166875 2022-12-22 10:31:22,133 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp1.1 from training. Duration: 22.5090625 2022-12-22 10:31:22,358 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0 from training. Duration: 25.035 2022-12-22 10:31:22,992 WARNING [train.py:1060] (3/4) Exclude cut with ID 774-127930-0014-10412-0_sp1.1 from training. Duration: 0.95 2022-12-22 10:31:23,665 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp0.9 from training. Duration: 0.92225 2022-12-22 10:31:24,581 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0 from training. Duration: 21.97 2022-12-22 10:31:25,249 WARNING [train.py:1060] (3/4) Exclude cut with ID 7492-105653-0055-62765-0_sp0.9 from training. Duration: 21.97225 2022-12-22 10:31:25,286 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp0.9 from training. Duration: 25.3333125 2022-12-22 10:31:26,220 WARNING [train.py:1060] (3/4) Exclude cut with ID 5172-29468-0015-19128-0_sp0.9 from training. Duration: 21.5055625 2022-12-22 10:31:26,566 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp1.1 from training. Duration: 20.72725 2022-12-22 10:31:27,671 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp0.9 from training. Duration: 26.32775 2022-12-22 10:31:28,725 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0 from training. Duration: 20.025 2022-12-22 10:31:28,910 WARNING [train.py:1060] (3/4) Exclude cut with ID 6709-74022-0004-86860-0_sp1.1 from training. Duration: 0.9409375 2022-12-22 10:31:28,920 WARNING [train.py:1060] (3/4) Exclude cut with ID 4757-1811-0023-62229-0_sp0.9 from training. Duration: 21.37775 2022-12-22 10:31:29,623 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0004-25974-0_sp0.9 from training. Duration: 21.17225 2022-12-22 10:31:29,633 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp0.9 from training. Duration: 27.511125 2022-12-22 10:31:31,383 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0 from training. Duration: 22.8 2022-12-22 10:31:31,564 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0 from training. Duration: 22.585 2022-12-22 10:31:32,655 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0001-146967-0_sp0.9 from training. Duration: 22.0166875 2022-12-22 10:31:33,234 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp1.1 from training. Duration: 24.395375 2022-12-22 10:31:33,459 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp0.9 from training. Duration: 27.47775 2022-12-22 10:31:33,617 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp0.9 from training. Duration: 24.8833125 2022-12-22 10:31:33,711 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0 from training. Duration: 23.39 2022-12-22 10:31:33,924 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp0.9 from training. Duration: 28.72225 2022-12-22 10:31:34,852 WARNING [train.py:1060] (3/4) Exclude cut with ID 585-294811-0110-133686-0_sp0.9 from training. Duration: 20.8944375 2022-12-22 10:31:35,388 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0_sp0.9 from training. Duration: 23.8444375 2022-12-22 10:31:36,231 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0 from training. Duration: 25.85 2022-12-22 10:31:36,242 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0 from training. Duration: 21.39 2022-12-22 10:31:36,654 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0 from training. Duration: 27.92 2022-12-22 10:31:37,676 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0039-130165-0_sp0.9 from training. Duration: 20.661125 2022-12-22 10:31:38,530 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0043-15874-0_sp0.9 from training. Duration: 20.07225 2022-12-22 10:31:39,444 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0 from training. Duration: 21.01 2022-12-22 10:31:41,909 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0 from training. Duration: 20.65 2022-12-22 10:31:42,166 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0 from training. Duration: 21.46 2022-12-22 10:31:44,465 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0 from training. Duration: 0.92 2022-12-22 10:31:44,696 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0_sp0.9 from training. Duration: 23.7666875 2022-12-22 10:31:45,944 WARNING [train.py:1060] (3/4) Exclude cut with ID 8544-281189-0060-101339-0_sp0.9 from training. Duration: 20.861125 2022-12-22 10:31:46,305 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0_sp0.9 from training. Duration: 22.711125 2022-12-22 10:31:48,466 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp1.1 from training. Duration: 22.986375 2022-12-22 10:31:49,042 WARNING [train.py:1060] (3/4) Exclude cut with ID 8040-260924-0003-80960-0_sp0.9 from training. Duration: 22.07225 2022-12-22 10:31:49,215 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0045-26330-0_sp0.9 from training. Duration: 20.3055625 2022-12-22 10:31:49,315 WARNING [train.py:1060] (3/4) Exclude cut with ID 6356-271890-0060-94317-0_sp0.9 from training. Duration: 20.72225 2022-12-22 10:31:49,957 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp1.1 from training. Duration: 22.4818125 2022-12-22 10:31:50,832 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp0.9 from training. Duration: 25.0944375 2022-12-22 10:31:50,964 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0 from training. Duration: 21.515 2022-12-22 10:31:51,166 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp0.9 from training. Duration: 27.02225 2022-12-22 10:31:51,344 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0010-62480-0_sp0.9 from training. Duration: 22.22225 2022-12-22 10:31:51,609 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0085-44554-0_sp0.9 from training. Duration: 20.85 2022-12-22 10:31:53,356 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0 from training. Duration: 21.54 2022-12-22 10:31:53,508 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp1.1 from training. Duration: 20.5318125 2022-12-22 10:31:53,815 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0012-134311-0_sp0.9 from training. Duration: 21.9333125 2022-12-22 10:31:55,422 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0025-130151-0_sp0.9 from training. Duration: 21.7944375 2022-12-22 10:31:55,876 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0_sp0.9 from training. Duration: 22.4666875 2022-12-22 10:31:56,128 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0 from training. Duration: 21.635 2022-12-22 10:31:56,925 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0_sp0.9 from training. Duration: 24.038875 2022-12-22 10:31:58,382 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp1.1 from training. Duration: 21.786375 2022-12-22 10:31:58,776 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0 from training. Duration: 20.22 2022-12-22 10:32:03,436 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0 from training. Duration: 25.285 2022-12-22 10:32:06,198 WARNING [train.py:1060] (3/4) Exclude cut with ID 811-130148-0001-63453-0_sp0.9 from training. Duration: 20.861125 2022-12-22 10:32:06,933 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0 from training. Duration: 20.88 2022-12-22 10:32:08,038 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0_sp0.9 from training. Duration: 23.4166875 2022-12-22 10:32:11,256 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0 from training. Duration: 21.24 2022-12-22 10:32:11,272 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0_sp0.9 from training. Duration: 23.9055625 2022-12-22 10:32:12,388 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp0.9 from training. Duration: 25.988875 2022-12-22 10:32:12,683 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0001-134300-0_sp0.9 from training. Duration: 20.67225 2022-12-22 10:32:14,899 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0 from training. Duration: 20.34 2022-12-22 10:32:17,403 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp0.9 from training. Duration: 25.061125 2022-12-22 10:32:18,386 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0 from training. Duration: 0.83 2022-12-22 10:32:19,274 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0 from training. Duration: 24.73 2022-12-22 10:32:19,706 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0 from training. Duration: 23.965 2022-12-22 10:32:19,994 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0030-146996-0_sp0.9 from training. Duration: 22.088875 2022-12-22 10:32:20,552 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0_sp0.9 from training. Duration: 23.6 2022-12-22 10:32:24,626 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0 from training. Duration: 23.795 2022-12-22 10:32:25,193 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp1.1 from training. Duration: 21.5409375 2022-12-22 10:32:25,285 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp0.9 from training. Duration: 24.97775 2022-12-22 10:32:25,669 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0_sp0.9 from training. Duration: 23.3444375 2022-12-22 10:32:27,221 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0_sp0.9 from training. Duration: 23.2 2022-12-22 10:32:27,415 WARNING [train.py:1060] (3/4) Exclude cut with ID 5653-46179-0060-117930-0_sp0.9 from training. Duration: 21.17225 2022-12-22 10:32:28,060 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp0.9 from training. Duration: 24.6555625 2022-12-22 10:32:29,914 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0 from training. Duration: 20.44 2022-12-22 10:32:31,028 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0_sp0.9 from training. Duration: 23.45 2022-12-22 10:32:32,004 WARNING [train.py:1060] (3/4) Exclude cut with ID 6945-60535-0076-12784-0_sp0.9 from training. Duration: 20.52225 2022-12-22 10:32:32,276 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0 from training. Duration: 22.19 2022-12-22 10:32:32,649 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp1.1 from training. Duration: 25.3818125 2022-12-22 10:32:33,235 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp0.9 from training. Duration: 28.0944375 2022-12-22 10:32:33,468 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0_sp0.9 from training. Duration: 22.9444375 2022-12-22 10:32:33,734 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp1.1 from training. Duration: 21.6318125 2022-12-22 10:32:34,339 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0 from training. Duration: 23.695 2022-12-22 10:32:35,899 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0 from training. Duration: 23.955 2022-12-22 10:32:37,123 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp0.9 from training. Duration: 26.438875 2022-12-22 10:32:38,715 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0021-26306-0_sp0.9 from training. Duration: 21.2444375 2022-12-22 10:32:38,739 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp0.9 from training. Duration: 31.02225 2022-12-22 10:32:39,746 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0 from training. Duration: 22.395 2022-12-22 10:32:40,345 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0 from training. Duration: 21.075 2022-12-22 10:32:40,533 WARNING [train.py:1060] (3/4) Exclude cut with ID 6482-98857-0025-147532-0_sp0.9 from training. Duration: 20.0055625 2022-12-22 10:32:40,548 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0037-132304-0_sp0.9 from training. Duration: 22.05 2022-12-22 10:32:40,562 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0 from training. Duration: 26.8349375 2022-12-22 10:32:40,687 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp1.1 from training. Duration: 22.1090625 2022-12-22 10:32:40,955 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp0.9 from training. Duration: 26.6166875 2022-12-22 10:32:41,582 WARNING [train.py:1060] (3/4) Exclude cut with ID 2046-178027-0000-53705-0_sp0.9 from training. Duration: 20.3055625 2022-12-22 10:32:42,394 WARNING [train.py:1060] (3/4) Exclude cut with ID 7205-50138-0008-5373-0_sp0.9 from training. Duration: 20.7 2022-12-22 10:32:44,541 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0 from training. Duration: 22.48 2022-12-22 10:32:45,079 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp0.9 from training. Duration: 29.816625 2022-12-22 10:32:45,905 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp1.1 from training. Duration: 22.7590625 2022-12-22 10:32:46,086 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0 from training. Duration: 22.555 2022-12-22 10:32:46,834 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0005-25975-0_sp0.9 from training. Duration: 21.688875 2022-12-22 10:32:48,625 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0_sp0.9 from training. Duration: 22.6 2022-12-22 10:32:49,905 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0 from training. Duration: 24.32 2022-12-22 10:32:52,018 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-276745-0093-13116-0_sp0.9 from training. Duration: 21.061125 2022-12-22 10:32:53,197 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0024-15855-0_sp0.9 from training. Duration: 20.32225 2022-12-22 10:32:53,643 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp1.1 from training. Duration: 0.7545625 2022-12-22 10:32:54,156 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0_sp0.9 from training. Duration: 23.9333125 2022-12-22 10:32:55,293 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp1.1 from training. Duration: 20.17275 2022-12-22 10:32:55,482 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp1.1 from training. Duration: 20.436375 2022-12-22 10:32:58,972 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0_sp0.9 from training. Duration: 23.1055625 2022-12-22 10:32:59,046 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp1.1 from training. Duration: 23.5 2022-12-22 10:32:59,399 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp0.9 from training. Duration: 26.62775 2022-12-22 10:32:59,837 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0 from training. Duration: 21.105 2022-12-22 10:32:59,995 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0_sp0.9 from training. Duration: 24.411125 2022-12-22 10:33:00,951 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp1.1 from training. Duration: 21.263625 2022-12-22 10:33:02,687 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0 from training. Duration: 20.795 2022-12-22 10:33:03,034 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0 from training. Duration: 24.76 2022-12-22 10:33:03,052 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0_sp0.9 from training. Duration: 22.25 2022-12-22 10:33:03,947 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp1.1 from training. Duration: 20.5045625 2022-12-22 10:33:08,509 INFO [train.py:1234] (3/4) Maximum memory allocated so far is 20098MB 2022-12-22 10:33:11,370 INFO [train.py:1234] (3/4) Maximum memory allocated so far is 21658MB 2022-12-22 10:33:15,384 INFO [train.py:1234] (3/4) Maximum memory allocated so far is 21658MB 2022-12-22 10:33:18,699 INFO [train.py:1234] (3/4) Maximum memory allocated so far is 21658MB 2022-12-22 10:33:23,087 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=4.81 vs. limit=2.0 2022-12-22 10:33:23,720 INFO [train.py:1234] (3/4) Maximum memory allocated so far is 22754MB 2022-12-22 10:33:27,336 INFO [train.py:1234] (3/4) Maximum memory allocated so far is 22754MB 2022-12-22 10:33:38,637 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp1.1 from training. Duration: 22.2954375 2022-12-22 10:33:43,145 INFO [train.py:894] (3/4) Epoch 1, batch 0, loss[loss=7.402, simple_loss=6.703, pruned_loss=6.971, over 18411.00 frames. ], tot_loss[loss=7.402, simple_loss=6.703, pruned_loss=6.971, over 18411.00 frames. ], batch size: 42, lr: 2.50e-02, grad_scale: 2.0 2022-12-22 10:33:43,145 INFO [train.py:919] (3/4) Computing validation loss 2022-12-22 10:33:55,184 INFO [train.py:928] (3/4) Epoch 1, validation: loss=6.913, simple_loss=6.239, pruned_loss=6.726, over 944034.00 frames. 2022-12-22 10:33:55,185 INFO [train.py:929] (3/4) Maximum memory allocated so far is 22754MB 2022-12-22 10:33:56,568 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=3.16 vs. limit=2.0 2022-12-22 10:33:59,110 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5.0, num_to_drop=2, layers_to_drop={1, 2} 2022-12-22 10:34:01,172 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=4.29 vs. limit=2.0 2022-12-22 10:34:08,610 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([6.5641, 6.4293, 6.5850, 6.5819, 6.5906, 6.5847, 6.5753, 6.5752], device='cuda:3'), covar=tensor([0.0062, 0.0036, 0.0159, 0.0074, 0.0061, 0.0072, 0.0034, 0.0071], device='cuda:3'), in_proj_covar=tensor([0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009], device='cuda:3'), out_proj_covar=tensor([9.0315e-06, 9.0522e-06, 8.9825e-06, 8.8194e-06, 9.0207e-06, 8.8961e-06, 8.7820e-06, 8.9276e-06], device='cuda:3') 2022-12-22 10:34:17,401 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=23.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 10:34:31,890 WARNING [train.py:1060] (3/4) Exclude cut with ID 298-126791-0067-24026-0_sp0.9 from training. Duration: 21.438875 2022-12-22 10:34:34,416 WARNING [train.py:1060] (3/4) Exclude cut with ID 5652-39938-0025-23684-0_sp0.9 from training. Duration: 22.2055625 2022-12-22 10:34:39,239 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.6021, 3.5717, 3.5573, 3.5899, 3.5962, 3.6016, 3.5990, 3.5614], device='cuda:3'), covar=tensor([0.0018, 0.0014, 0.0019, 0.0019, 0.0021, 0.0033, 0.0023, 0.0015], device='cuda:3'), in_proj_covar=tensor([0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009], device='cuda:3'), out_proj_covar=tensor([8.9683e-06, 9.0940e-06, 8.9387e-06, 9.1167e-06, 8.9277e-06, 9.0123e-06, 9.0311e-06, 9.1188e-06], device='cuda:3') 2022-12-22 10:34:43,576 INFO [train.py:894] (3/4) Epoch 1, batch 50, loss[loss=1.262, simple_loss=1.114, pruned_loss=1.319, over 18601.00 frames. ], tot_loss[loss=2.161, simple_loss=1.956, pruned_loss=1.972, over 836487.97 frames. ], batch size: 45, lr: 2.75e-02, grad_scale: 1.0 2022-12-22 10:34:46,906 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=55.21 vs. limit=5.0 2022-12-22 10:34:57,145 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=5.08 vs. limit=2.0 2022-12-22 10:35:15,729 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 10:35:19,977 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=5.33 vs. limit=2.0 2022-12-22 10:35:32,558 INFO [train.py:894] (3/4) Epoch 1, batch 100, loss[loss=1.154, simple_loss=0.9916, pruned_loss=1.292, over 18611.00 frames. ], tot_loss[loss=1.634, simple_loss=1.454, pruned_loss=1.622, over 1474344.75 frames. ], batch size: 78, lr: 3.00e-02, grad_scale: 0.0625 2022-12-22 10:35:38,780 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 6.402e+01 1.918e+02 5.720e+02 1.444e+03 1.131e+05, threshold=1.144e+03, percent-clipped=0.0 2022-12-22 10:35:48,734 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=72.91 vs. limit=5.0 2022-12-22 10:35:51,990 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.4407, 4.4407, 4.4407, 4.4408, 4.4405, 4.4404, 4.4407, 4.4404], device='cuda:3'), covar=tensor([0.0006, 0.0006, 0.0006, 0.0011, 0.0007, 0.0011, 0.0006, 0.0009], device='cuda:3'), in_proj_covar=tensor([0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009], device='cuda:3'), out_proj_covar=tensor([9.3439e-06, 9.1948e-06, 9.0958e-06, 9.0341e-06, 9.1198e-06, 9.2090e-06, 8.9818e-06, 9.1989e-06], device='cuda:3') 2022-12-22 10:36:14,142 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=46.19 vs. limit=5.0 2022-12-22 10:36:16,951 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=144.0, num_to_drop=2, layers_to_drop={0, 2} 2022-12-22 10:36:23,423 INFO [train.py:894] (3/4) Epoch 1, batch 150, loss[loss=1.05, simple_loss=0.8906, pruned_loss=1.151, over 18528.00 frames. ], tot_loss[loss=1.394, simple_loss=1.222, pruned_loss=1.443, over 1970842.18 frames. ], batch size: 58, lr: 3.25e-02, grad_scale: 0.0625 2022-12-22 10:36:29,768 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0 from training. Duration: 24.525 2022-12-22 10:36:53,907 WARNING [train.py:1060] (3/4) Exclude cut with ID 3699-47246-0007-3408-0_sp0.9 from training. Duration: 20.26675 2022-12-22 10:36:57,305 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.72 vs. limit=2.0 2022-12-22 10:36:57,949 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([5.8441, 5.8441, 5.8441, 5.8442, 5.8441, 5.8442, 5.8441, 5.8439], device='cuda:3'), covar=tensor([0.0007, 0.0008, 0.0009, 0.0007, 0.0006, 0.0007, 0.0006, 0.0006], device='cuda:3'), in_proj_covar=tensor([0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009], device='cuda:3'), out_proj_covar=tensor([9.3581e-06, 9.1229e-06, 9.1413e-06, 8.9988e-06, 9.1473e-06, 9.1677e-06, 8.9615e-06, 9.1894e-06], device='cuda:3') 2022-12-22 10:37:02,442 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp0.9 from training. Duration: 27.25 2022-12-22 10:37:13,470 INFO [train.py:894] (3/4) Epoch 1, batch 200, loss[loss=1.053, simple_loss=0.8916, pruned_loss=1.085, over 18397.00 frames. ], tot_loss[loss=1.252, simple_loss=1.087, pruned_loss=1.307, over 2357665.35 frames. ], batch size: 53, lr: 3.50e-02, grad_scale: 0.125 2022-12-22 10:37:18,533 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.5068, 3.5460, 3.6160, 3.4278, 3.5471, 3.6186, 3.6102, 3.4088], device='cuda:3'), covar=tensor([0.0085, 0.0104, 0.0158, 0.0105, 0.0106, 0.0155, 0.0122, 0.0310], device='cuda:3'), in_proj_covar=tensor([0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009], device='cuda:3'), out_proj_covar=tensor([8.8480e-06, 8.8790e-06, 8.7543e-06, 8.9791e-06, 8.7632e-06, 8.7860e-06, 8.8949e-06, 8.8967e-06], device='cuda:3') 2022-12-22 10:37:19,084 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 8.110e+01 1.446e+02 1.990e+02 2.489e+02 9.125e+02, threshold=3.981e+02, percent-clipped=0.0 2022-12-22 10:37:50,391 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0 from training. Duration: 21.68 2022-12-22 10:37:57,868 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0 from training. Duration: 21.6300625 2022-12-22 10:38:02,155 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=3.02 vs. limit=2.0 2022-12-22 10:38:03,577 INFO [train.py:894] (3/4) Epoch 1, batch 250, loss[loss=0.9874, simple_loss=0.8304, pruned_loss=0.9819, over 18642.00 frames. ], tot_loss[loss=1.162, simple_loss=1.001, pruned_loss=1.205, over 2659041.46 frames. ], batch size: 77, lr: 3.75e-02, grad_scale: 0.125 2022-12-22 10:38:13,622 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0_sp0.9 from training. Duration: 24.033375 2022-12-22 10:38:23,589 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.50 vs. limit=2.0 2022-12-22 10:38:48,882 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 10:38:51,364 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0 from training. Duration: 22.905 2022-12-22 10:38:52,199 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp1.1 from training. Duration: 23.4318125 2022-12-22 10:38:52,395 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=300.0, num_to_drop=2, layers_to_drop={0, 2} 2022-12-22 10:38:53,076 INFO [train.py:894] (3/4) Epoch 1, batch 300, loss[loss=0.9563, simple_loss=0.7983, pruned_loss=0.9263, over 18548.00 frames. ], tot_loss[loss=1.103, simple_loss=0.9439, pruned_loss=1.13, over 2892385.17 frames. ], batch size: 55, lr: 4.00e-02, grad_scale: 0.25 2022-12-22 10:38:58,484 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.051e+02 1.633e+02 2.114e+02 2.790e+02 8.264e+02, threshold=4.229e+02, percent-clipped=8.0 2022-12-22 10:39:41,577 INFO [train.py:894] (3/4) Epoch 1, batch 350, loss[loss=0.9985, simple_loss=0.8295, pruned_loss=0.9374, over 18582.00 frames. ], tot_loss[loss=1.065, simple_loss=0.9052, pruned_loss=1.072, over 3074973.35 frames. ], batch size: 57, lr: 4.25e-02, grad_scale: 0.25 2022-12-22 10:39:47,446 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=357.0, num_to_drop=2, layers_to_drop={2, 3} 2022-12-22 10:40:09,237 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp1.1 from training. Duration: 20.82275 2022-12-22 10:40:10,152 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp0.9 from training. Duration: 25.45 2022-12-22 10:40:17,558 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=387.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 10:40:31,707 INFO [train.py:894] (3/4) Epoch 1, batch 400, loss[loss=0.929, simple_loss=0.7746, pruned_loss=0.8274, over 18445.00 frames. ], tot_loss[loss=1.031, simple_loss=0.872, pruned_loss=1.015, over 3216405.16 frames. ], batch size: 50, lr: 4.50e-02, grad_scale: 0.5 2022-12-22 10:40:37,773 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 9.684e+01 1.284e+02 1.583e+02 1.935e+02 3.190e+02, threshold=3.166e+02, percent-clipped=0.0 2022-12-22 10:40:46,871 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0 from training. Duration: 25.775 2022-12-22 10:40:50,700 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=3.28 vs. limit=2.0 2022-12-22 10:41:02,136 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0_sp0.9 from training. Duration: 22.25 2022-12-22 10:41:05,146 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.3887, 2.6112, 2.8970, 4.2692, 4.2253, 3.1571, 4.3629, 2.8904], device='cuda:3'), covar=tensor([0.0692, 0.3362, 0.1252, 0.0449, 0.0438, 0.0995, 0.0409, 0.1095], device='cuda:3'), in_proj_covar=tensor([0.0010, 0.0011, 0.0011, 0.0010, 0.0009, 0.0010, 0.0010, 0.0010], device='cuda:3'), out_proj_covar=tensor([9.6447e-06, 1.0461e-05, 9.7346e-06, 9.9233e-06, 9.4562e-06, 9.7909e-06, 9.8677e-06, 9.7924e-06], device='cuda:3') 2022-12-22 10:41:09,531 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=439.0, num_to_drop=2, layers_to_drop={0, 3} 2022-12-22 10:41:18,086 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=448.0, num_to_drop=2, layers_to_drop={1, 2} 2022-12-22 10:41:18,711 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0 from training. Duration: 26.205 2022-12-22 10:41:20,481 INFO [train.py:894] (3/4) Epoch 1, batch 450, loss[loss=0.8121, simple_loss=0.6833, pruned_loss=0.6794, over 18531.00 frames. ], tot_loss[loss=1, simple_loss=0.844, pruned_loss=0.9578, over 3325636.70 frames. ], batch size: 44, lr: 4.75e-02, grad_scale: 0.5 2022-12-22 10:41:29,748 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp0.9 from training. Duration: 30.1555625 2022-12-22 10:41:34,494 WARNING [train.py:1060] (3/4) Exclude cut with ID 1265-135635-0050-6781-0_sp0.9 from training. Duration: 21.8333125 2022-12-22 10:41:41,114 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp1.1 from training. Duration: 20.6545625 2022-12-22 10:41:56,870 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.8945, 0.8595, 1.2059, 0.7992, 0.9918, 0.9428, 0.9026, 0.8787], device='cuda:3'), covar=tensor([0.0702, 0.0967, 0.0765, 0.0872, 0.0623, 0.0804, 0.0587, 0.0782], device='cuda:3'), in_proj_covar=tensor([0.0011, 0.0011, 0.0011, 0.0011, 0.0011, 0.0011, 0.0011, 0.0011], device='cuda:3'), out_proj_covar=tensor([1.0426e-05, 1.0703e-05, 1.0597e-05, 1.0605e-05, 1.0766e-05, 1.0601e-05, 1.0642e-05, 1.0629e-05], device='cuda:3') 2022-12-22 10:42:08,586 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0045-39920-0_sp0.9 from training. Duration: 20.52225 2022-12-22 10:42:10,414 INFO [train.py:894] (3/4) Epoch 1, batch 500, loss[loss=0.8258, simple_loss=0.7053, pruned_loss=0.6417, over 18383.00 frames. ], tot_loss[loss=0.9726, simple_loss=0.8208, pruned_loss=0.9012, over 3412125.35 frames. ], batch size: 46, lr: 4.99e-02, grad_scale: 1.0 2022-12-22 10:42:15,914 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.309e+02 2.085e+02 3.413e+02 4.603e+02 1.381e+03, threshold=6.826e+02, percent-clipped=53.0 2022-12-22 10:42:22,166 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp0.9 from training. Duration: 29.1166875 2022-12-22 10:42:55,994 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=9.86 vs. limit=5.0 2022-12-22 10:42:59,386 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133211-0007-59831-0_sp0.9 from training. Duration: 21.388875 2022-12-22 10:43:00,264 INFO [train.py:894] (3/4) Epoch 1, batch 550, loss[loss=0.8577, simple_loss=0.7379, pruned_loss=0.6364, over 18482.00 frames. ], tot_loss[loss=0.9411, simple_loss=0.7957, pruned_loss=0.8413, over 3479082.22 frames. ], batch size: 64, lr: 4.98e-02, grad_scale: 1.0 2022-12-22 10:43:12,031 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=562.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 10:43:17,543 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=568.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 10:43:23,675 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0 from training. Duration: 22.72 2022-12-22 10:43:24,758 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0_sp0.9 from training. Duration: 22.7444375 2022-12-22 10:43:38,677 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=590.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 10:43:38,799 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9840, 2.6666, 1.7175, 2.3997, 3.3284, 1.7156, 4.5351, 1.6020], device='cuda:3'), covar=tensor([0.8154, 0.7317, 1.3548, 0.7905, 0.4375, 1.0335, 0.1955, 1.2805], device='cuda:3'), in_proj_covar=tensor([0.0024, 0.0021, 0.0029, 0.0022, 0.0021, 0.0026, 0.0020, 0.0027], device='cuda:3'), out_proj_covar=tensor([2.1769e-05, 2.0604e-05, 2.5254e-05, 2.0564e-05, 1.9660e-05, 2.5577e-05, 1.7996e-05, 2.5886e-05], device='cuda:3') 2022-12-22 10:43:44,565 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5215, 1.9433, 1.1815, 1.8661, 2.3853, 1.3159, 3.1562, 1.3142], device='cuda:3'), covar=tensor([0.8961, 0.8489, 1.4839, 0.8845, 0.5727, 1.1229, 0.3176, 1.2839], device='cuda:3'), in_proj_covar=tensor([0.0024, 0.0021, 0.0029, 0.0023, 0.0021, 0.0026, 0.0020, 0.0027], device='cuda:3'), out_proj_covar=tensor([2.1972e-05, 2.0805e-05, 2.5336e-05, 2.1097e-05, 2.0015e-05, 2.5465e-05, 1.8234e-05, 2.5894e-05], device='cuda:3') 2022-12-22 10:43:48,834 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=600.0, num_to_drop=2, layers_to_drop={0, 2} 2022-12-22 10:43:49,536 INFO [train.py:894] (3/4) Epoch 1, batch 600, loss[loss=0.6667, simple_loss=0.5814, pruned_loss=0.4662, over 18404.00 frames. ], tot_loss[loss=0.9047, simple_loss=0.7681, pruned_loss=0.7786, over 3531421.19 frames. ], batch size: 42, lr: 4.98e-02, grad_scale: 1.0 2022-12-22 10:43:52,478 WARNING [train.py:1060] (3/4) Exclude cut with ID 4133-6541-0027-40495-0_sp1.1 from training. Duration: 0.9681875 2022-12-22 10:43:52,885 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=7.29 vs. limit=5.0 2022-12-22 10:43:55,943 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.031e+02 4.953e+02 7.048e+02 1.005e+03 3.491e+03, threshold=1.410e+03, percent-clipped=51.0 2022-12-22 10:43:56,006 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0_sp0.9 from training. Duration: 22.3166875 2022-12-22 10:43:58,827 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133212-0015-59917-0_sp0.9 from training. Duration: 21.8166875 2022-12-22 10:44:12,186 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=623.0, num_to_drop=2, layers_to_drop={2, 3} 2022-12-22 10:44:18,487 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=629.0, num_to_drop=2, layers_to_drop={0, 3} 2022-12-22 10:44:23,183 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=3.38 vs. limit=2.0 2022-12-22 10:44:35,836 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=648.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 10:44:38,611 INFO [train.py:894] (3/4) Epoch 1, batch 650, loss[loss=0.7901, simple_loss=0.6917, pruned_loss=0.5369, over 18554.00 frames. ], tot_loss[loss=0.8634, simple_loss=0.7369, pruned_loss=0.7149, over 3571884.27 frames. ], batch size: 77, lr: 4.98e-02, grad_scale: 1.0 2022-12-22 10:44:38,904 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=651.0, num_to_drop=2, layers_to_drop={0, 2} 2022-12-22 10:44:39,633 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=652.0, num_to_drop=2, layers_to_drop={0, 1} 2022-12-22 10:45:00,817 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.78 vs. limit=2.0 2022-12-22 10:45:05,408 WARNING [train.py:1060] (3/4) Exclude cut with ID 4957-30119-0041-23990-0_sp0.9 from training. Duration: 20.22775 2022-12-22 10:45:22,311 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.8229, 3.5772, 3.8077, 4.6326, 4.6738, 3.3500, 3.5004, 2.0573], device='cuda:3'), covar=tensor([0.1619, 0.1920, 0.1040, 0.0245, 0.0160, 0.1939, 0.0819, 0.3569], device='cuda:3'), in_proj_covar=tensor([0.0020, 0.0022, 0.0020, 0.0018, 0.0017, 0.0022, 0.0020, 0.0028], device='cuda:3'), out_proj_covar=tensor([1.9290e-05, 2.1511e-05, 1.8041e-05, 1.6017e-05, 1.6264e-05, 2.0823e-05, 1.8558e-05, 2.9903e-05], device='cuda:3') 2022-12-22 10:45:26,783 INFO [train.py:894] (3/4) Epoch 1, batch 700, loss[loss=0.6949, simple_loss=0.6109, pruned_loss=0.4598, over 18587.00 frames. ], tot_loss[loss=0.8281, simple_loss=0.7107, pruned_loss=0.66, over 3603338.19 frames. ], batch size: 51, lr: 4.98e-02, grad_scale: 1.0 2022-12-22 10:45:32,844 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.871e+02 5.831e+02 8.351e+02 1.310e+03 8.148e+03, threshold=1.670e+03, percent-clipped=20.0 2022-12-22 10:45:33,731 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp1.1 from training. Duration: 24.67275 2022-12-22 10:45:52,848 WARNING [train.py:1060] (3/4) Exclude cut with ID 3082-165428-0081-50734-0_sp0.9 from training. Duration: 21.8055625 2022-12-22 10:46:03,972 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=739.0, num_to_drop=2, layers_to_drop={1, 2} 2022-12-22 10:46:07,527 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=743.0, num_to_drop=2, layers_to_drop={0, 1} 2022-12-22 10:46:14,944 INFO [train.py:894] (3/4) Epoch 1, batch 750, loss[loss=0.6202, simple_loss=0.5519, pruned_loss=0.3927, over 18442.00 frames. ], tot_loss[loss=0.7935, simple_loss=0.6852, pruned_loss=0.609, over 3627083.37 frames. ], batch size: 48, lr: 4.97e-02, grad_scale: 1.0 2022-12-22 10:46:15,022 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0_sp0.9 from training. Duration: 22.6666875 2022-12-22 10:46:40,834 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.7270, 0.9372, 1.0942, 0.8901, 0.9689, 0.9910, 0.7813, 0.7910], device='cuda:3'), covar=tensor([0.8604, 0.7085, 0.7494, 0.7339, 0.6566, 0.6997, 0.7361, 0.6107], device='cuda:3'), in_proj_covar=tensor([0.0045, 0.0038, 0.0046, 0.0040, 0.0039, 0.0044, 0.0041, 0.0038], device='cuda:3'), out_proj_covar=tensor([3.9902e-05, 3.4898e-05, 4.3261e-05, 3.6780e-05, 3.8726e-05, 4.2326e-05, 3.8226e-05, 3.3871e-05], device='cuda:3') 2022-12-22 10:46:50,899 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=787.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 10:46:56,441 WARNING [train.py:1060] (3/4) Exclude cut with ID 2411-132532-0017-82279-0_sp1.1 from training. Duration: 0.9681875 2022-12-22 10:47:04,406 INFO [train.py:894] (3/4) Epoch 1, batch 800, loss[loss=0.5939, simple_loss=0.5342, pruned_loss=0.362, over 18514.00 frames. ], tot_loss[loss=0.7624, simple_loss=0.6625, pruned_loss=0.5642, over 3645951.87 frames. ], batch size: 44, lr: 4.97e-02, grad_scale: 2.0 2022-12-22 10:47:10,445 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.426e+02 6.087e+02 8.345e+02 1.128e+03 4.122e+03, threshold=1.669e+03, percent-clipped=11.0 2022-12-22 10:47:13,290 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0 from training. Duration: 22.485 2022-12-22 10:47:18,977 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3732, 1.0509, 0.9251, 1.0461, 1.3474, 1.0692, 1.1131, 1.1587], device='cuda:3'), covar=tensor([1.1046, 1.6379, 2.3848, 3.1945, 1.1739, 1.7185, 2.1660, 1.4011], device='cuda:3'), in_proj_covar=tensor([0.0023, 0.0025, 0.0035, 0.0036, 0.0025, 0.0030, 0.0027, 0.0028], device='cuda:3'), out_proj_covar=tensor([2.0058e-05, 2.3050e-05, 2.9793e-05, 3.5293e-05, 2.3404e-05, 2.7534e-05, 2.6375e-05, 2.5973e-05], device='cuda:3') 2022-12-22 10:47:38,753 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp1.1 from training. Duration: 23.82275 2022-12-22 10:47:46,940 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0 from training. Duration: 20.77 2022-12-22 10:47:49,034 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=847.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 10:47:52,634 INFO [train.py:894] (3/4) Epoch 1, batch 850, loss[loss=0.6401, simple_loss=0.5807, pruned_loss=0.3787, over 18691.00 frames. ], tot_loss[loss=0.7338, simple_loss=0.6416, pruned_loss=0.5247, over 3660906.01 frames. ], batch size: 62, lr: 4.96e-02, grad_scale: 2.0 2022-12-22 10:47:52,658 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0_sp0.9 from training. Duration: 24.088875 2022-12-22 10:47:53,970 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=2.49 vs. limit=2.0 2022-12-22 10:48:06,014 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=2.12 vs. limit=2.0 2022-12-22 10:48:14,059 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp1.1 from training. Duration: 20.4409375 2022-12-22 10:48:41,500 INFO [train.py:894] (3/4) Epoch 1, batch 900, loss[loss=0.6309, simple_loss=0.577, pruned_loss=0.3638, over 18694.00 frames. ], tot_loss[loss=0.7079, simple_loss=0.6223, pruned_loss=0.4907, over 3673375.08 frames. ], batch size: 69, lr: 4.96e-02, grad_scale: 2.0 2022-12-22 10:48:47,840 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.705e+02 7.532e+02 9.958e+02 1.427e+03 3.845e+03, threshold=1.992e+03, percent-clipped=19.0 2022-12-22 10:48:49,099 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=908.0, num_to_drop=2, layers_to_drop={0, 3} 2022-12-22 10:48:49,368 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.75 vs. limit=2.0 2022-12-22 10:48:58,783 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=918.0, num_to_drop=2, layers_to_drop={0, 1} 2022-12-22 10:49:02,096 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0_sp0.9 from training. Duration: 22.511125 2022-12-22 10:49:03,699 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0 from training. Duration: 20.675 2022-12-22 10:49:05,640 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=924.0, num_to_drop=2, layers_to_drop={0, 3} 2022-12-22 10:49:25,739 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=946.0, num_to_drop=2, layers_to_drop={0, 1} 2022-12-22 10:49:30,614 INFO [train.py:894] (3/4) Epoch 1, batch 950, loss[loss=0.5809, simple_loss=0.5372, pruned_loss=0.3248, over 18630.00 frames. ], tot_loss[loss=0.6811, simple_loss=0.6034, pruned_loss=0.4567, over 3681947.09 frames. ], batch size: 53, lr: 4.96e-02, grad_scale: 2.0 2022-12-22 10:49:31,766 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=952.0, num_to_drop=2, layers_to_drop={0, 2} 2022-12-22 10:50:08,488 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp0.9 from training. Duration: 24.9833125 2022-12-22 10:50:19,380 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=1000.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 10:50:20,154 INFO [train.py:894] (3/4) Epoch 1, batch 1000, loss[loss=0.6119, simple_loss=0.5457, pruned_loss=0.3683, over 18538.00 frames. ], tot_loss[loss=0.661, simple_loss=0.5888, pruned_loss=0.431, over 3689700.67 frames. ], batch size: 47, lr: 4.95e-02, grad_scale: 2.0 2022-12-22 10:50:26,434 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.198e+02 6.579e+02 9.009e+02 1.268e+03 5.908e+03, threshold=1.802e+03, percent-clipped=7.0 2022-12-22 10:50:28,228 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.5598, 3.0877, 3.1007, 3.5007, 3.7716, 3.0067, 3.1556, 1.6234], device='cuda:3'), covar=tensor([0.0694, 0.0968, 0.1205, 0.0447, 0.0340, 0.1247, 0.1020, 0.4306], device='cuda:3'), in_proj_covar=tensor([0.0023, 0.0027, 0.0028, 0.0023, 0.0024, 0.0028, 0.0028, 0.0041], device='cuda:3'), out_proj_covar=tensor([2.1629e-05, 2.5048e-05, 2.4611e-05, 2.0286e-05, 2.0633e-05, 2.6417e-05, 2.6464e-05, 4.3998e-05], device='cuda:3') 2022-12-22 10:50:29,115 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0 from training. Duration: 27.14 2022-12-22 10:50:35,537 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=2.24 vs. limit=2.0 2022-12-22 10:50:39,634 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0 from training. Duration: 22.44 2022-12-22 10:51:04,285 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1043.0, num_to_drop=2, layers_to_drop={1, 2} 2022-12-22 10:51:12,174 INFO [train.py:894] (3/4) Epoch 1, batch 1050, loss[loss=0.6375, simple_loss=0.5844, pruned_loss=0.3607, over 18536.00 frames. ], tot_loss[loss=0.6398, simple_loss=0.5741, pruned_loss=0.4052, over 3695332.18 frames. ], batch size: 58, lr: 4.95e-02, grad_scale: 2.0 2022-12-22 10:51:26,334 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=5.03 vs. limit=5.0 2022-12-22 10:51:36,162 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0060-62364-0_sp0.9 from training. Duration: 21.361125 2022-12-22 10:51:40,715 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp1.1 from training. Duration: 27.0318125 2022-12-22 10:51:48,446 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp0.9 from training. Duration: 28.638875 2022-12-22 10:51:54,227 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=1091.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 10:51:58,247 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0 from training. Duration: 20.4 2022-12-22 10:52:04,452 INFO [train.py:894] (3/4) Epoch 1, batch 1100, loss[loss=0.598, simple_loss=0.5599, pruned_loss=0.3228, over 18477.00 frames. ], tot_loss[loss=0.6224, simple_loss=0.5621, pruned_loss=0.3841, over 3698207.81 frames. ], batch size: 54, lr: 4.94e-02, grad_scale: 2.0 2022-12-22 10:52:10,278 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.217e+02 6.903e+02 9.658e+02 1.239e+03 4.082e+03, threshold=1.932e+03, percent-clipped=8.0 2022-12-22 10:52:19,810 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0 from training. Duration: 20.025 2022-12-22 10:52:19,826 WARNING [train.py:1060] (3/4) Exclude cut with ID 2364-131735-0112-64612-0_sp0.9 from training. Duration: 20.488875 2022-12-22 10:52:24,911 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0 from training. Duration: 29.735 2022-12-22 10:52:57,056 INFO [train.py:894] (3/4) Epoch 1, batch 1150, loss[loss=0.5019, simple_loss=0.4818, pruned_loss=0.2568, over 18631.00 frames. ], tot_loss[loss=0.6059, simple_loss=0.5512, pruned_loss=0.3646, over 3700987.02 frames. ], batch size: 53, lr: 4.94e-02, grad_scale: 2.0 2022-12-22 10:52:59,387 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1153.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 10:53:20,528 WARNING [train.py:1060] (3/4) Exclude cut with ID 7276-92427-0014-12983-0_sp0.9 from training. Duration: 21.3055625 2022-12-22 10:53:21,405 WARNING [train.py:1060] (3/4) Exclude cut with ID 1025-75365-0008-79168-0_sp0.9 from training. Duration: 22.0666875 2022-12-22 10:53:21,826 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7903, 1.3460, 1.2294, 1.2797, 1.3510, 1.0865, 1.3556, 1.2953], device='cuda:3'), covar=tensor([0.7638, 1.2280, 0.8746, 0.9331, 0.8668, 1.1333, 1.8964, 0.9165], device='cuda:3'), in_proj_covar=tensor([0.0026, 0.0028, 0.0026, 0.0025, 0.0025, 0.0029, 0.0033, 0.0027], device='cuda:3'), out_proj_covar=tensor([2.2804e-05, 2.4831e-05, 2.2686e-05, 2.1398e-05, 2.2287e-05, 2.5622e-05, 3.1019e-05, 2.3931e-05], device='cuda:3') 2022-12-22 10:53:49,548 INFO [train.py:894] (3/4) Epoch 1, batch 1200, loss[loss=0.5077, simple_loss=0.4898, pruned_loss=0.2576, over 18604.00 frames. ], tot_loss[loss=0.5882, simple_loss=0.5389, pruned_loss=0.3458, over 3703503.19 frames. ], batch size: 51, lr: 4.93e-02, grad_scale: 4.0 2022-12-22 10:53:51,687 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1203.0, num_to_drop=2, layers_to_drop={1, 3} 2022-12-22 10:53:55,438 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.627e+02 7.086e+02 8.753e+02 1.222e+03 5.052e+03, threshold=1.751e+03, percent-clipped=6.0 2022-12-22 10:54:03,010 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1214.0, num_to_drop=2, layers_to_drop={2, 3} 2022-12-22 10:54:07,653 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1218.0, num_to_drop=2, layers_to_drop={0, 2} 2022-12-22 10:54:14,058 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1224.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 10:54:34,888 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0 from training. Duration: 20.26 2022-12-22 10:54:36,149 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1246.0, num_to_drop=1, layers_to_drop={2} 2022-12-22 10:54:39,279 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4430, 1.3543, 1.5251, 1.3950, 1.1629, 1.1952, 0.9891, 1.4667], device='cuda:3'), covar=tensor([0.4811, 0.5789, 0.2756, 0.5194, 0.4409, 0.3052, 0.7258, 0.2505], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0075, 0.0066, 0.0108, 0.0072, 0.0064, 0.0079, 0.0062], device='cuda:3'), out_proj_covar=tensor([8.8052e-05, 7.1062e-05, 5.7036e-05, 1.0167e-04, 6.8383e-05, 5.8097e-05, 7.1962e-05, 5.3234e-05], device='cuda:3') 2022-12-22 10:54:40,927 INFO [train.py:894] (3/4) Epoch 1, batch 1250, loss[loss=0.4324, simple_loss=0.4185, pruned_loss=0.2184, over 18622.00 frames. ], tot_loss[loss=0.573, simple_loss=0.5291, pruned_loss=0.3291, over 3705126.41 frames. ], batch size: 45, lr: 4.92e-02, grad_scale: 4.0 2022-12-22 10:54:43,658 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0030-9324-0_sp0.9 from training. Duration: 21.3444375 2022-12-22 10:54:56,282 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=1266.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 10:55:03,198 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=1272.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 10:55:25,104 WARNING [train.py:1060] (3/4) Exclude cut with ID 497-129325-0061-62254-0_sp1.1 from training. Duration: 0.97725 2022-12-22 10:55:27,073 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=1294.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 10:55:33,992 INFO [train.py:894] (3/4) Epoch 1, batch 1300, loss[loss=0.525, simple_loss=0.4899, pruned_loss=0.2836, over 18510.00 frames. ], tot_loss[loss=0.5624, simple_loss=0.522, pruned_loss=0.3176, over 3706983.21 frames. ], batch size: 47, lr: 4.92e-02, grad_scale: 4.0 2022-12-22 10:55:39,954 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.931e+02 6.636e+02 8.494e+02 1.247e+03 3.025e+03, threshold=1.699e+03, percent-clipped=9.0 2022-12-22 10:55:53,660 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0_sp0.9 from training. Duration: 22.97225 2022-12-22 10:56:18,641 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0047-39922-0_sp0.9 from training. Duration: 21.97775 2022-12-22 10:56:27,900 INFO [train.py:894] (3/4) Epoch 1, batch 1350, loss[loss=0.4915, simple_loss=0.4713, pruned_loss=0.2534, over 18465.00 frames. ], tot_loss[loss=0.5501, simple_loss=0.5142, pruned_loss=0.3049, over 3707848.17 frames. ], batch size: 50, lr: 4.91e-02, grad_scale: 4.0 2022-12-22 10:56:27,958 WARNING [train.py:1060] (3/4) Exclude cut with ID 1112-1043-0006-89194-0_sp0.9 from training. Duration: 21.8333125 2022-12-22 10:56:37,071 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0 from training. Duration: 20.47 2022-12-22 10:57:23,822 INFO [train.py:894] (3/4) Epoch 1, batch 1400, loss[loss=0.5022, simple_loss=0.4888, pruned_loss=0.2527, over 18508.00 frames. ], tot_loss[loss=0.5387, simple_loss=0.5069, pruned_loss=0.2937, over 3708027.07 frames. ], batch size: 52, lr: 4.91e-02, grad_scale: 4.0 2022-12-22 10:57:26,041 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0037-39912-0_sp0.9 from training. Duration: 20.67225 2022-12-22 10:57:30,219 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.771e+02 6.721e+02 8.735e+02 1.207e+03 3.329e+03, threshold=1.747e+03, percent-clipped=8.0 2022-12-22 10:57:40,225 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp0.9 from training. Duration: 25.2444375 2022-12-22 10:57:59,614 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0021-76797-0_sp0.9 from training. Duration: 21.1445 2022-12-22 10:58:02,181 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1434.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 10:58:20,573 INFO [train.py:894] (3/4) Epoch 1, batch 1450, loss[loss=0.5703, simple_loss=0.5321, pruned_loss=0.3071, over 18508.00 frames. ], tot_loss[loss=0.5299, simple_loss=0.5017, pruned_loss=0.2848, over 3708836.32 frames. ], batch size: 52, lr: 4.90e-02, grad_scale: 4.0 2022-12-22 10:58:54,374 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp0.9 from training. Duration: 33.038875 2022-12-22 10:59:11,447 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1495.0, num_to_drop=2, layers_to_drop={0, 2} 2022-12-22 10:59:17,561 INFO [train.py:894] (3/4) Epoch 1, batch 1500, loss[loss=0.4475, simple_loss=0.4475, pruned_loss=0.2166, over 18718.00 frames. ], tot_loss[loss=0.5193, simple_loss=0.495, pruned_loss=0.2752, over 3710356.34 frames. ], batch size: 50, lr: 4.89e-02, grad_scale: 4.0 2022-12-22 10:59:20,096 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1503.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 10:59:20,966 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64291-0000-16059-0_sp0.9 from training. Duration: 20.0944375 2022-12-22 10:59:24,965 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.791e+02 8.013e+02 9.741e+02 1.259e+03 2.345e+03, threshold=1.948e+03, percent-clipped=8.0 2022-12-22 10:59:27,305 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1509.0, num_to_drop=2, layers_to_drop={1, 2} 2022-12-22 10:59:32,753 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp1.1 from training. Duration: 20.4 2022-12-22 10:59:38,382 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0 from training. Duration: 20.085 2022-12-22 10:59:48,206 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0_sp0.9 from training. Duration: 23.07775 2022-12-22 11:00:15,529 INFO [train.py:894] (3/4) Epoch 1, batch 1550, loss[loss=0.5622, simple_loss=0.5398, pruned_loss=0.2904, over 18654.00 frames. ], tot_loss[loss=0.5088, simple_loss=0.4889, pruned_loss=0.2657, over 3711939.78 frames. ], batch size: 69, lr: 4.89e-02, grad_scale: 4.0 2022-12-22 11:00:15,688 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=1551.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 11:00:23,702 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp0.9 from training. Duration: 24.9333125 2022-12-22 11:00:32,865 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1566.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 11:00:57,521 WARNING [train.py:1060] (3/4) Exclude cut with ID 5118-111612-0016-124680-0_sp0.9 from training. Duration: 20.388875 2022-12-22 11:01:01,032 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp1.1 from training. Duration: 20.3590625 2022-12-22 11:01:13,471 INFO [train.py:894] (3/4) Epoch 1, batch 1600, loss[loss=0.4737, simple_loss=0.4515, pruned_loss=0.2474, over 18692.00 frames. ], tot_loss[loss=0.5007, simple_loss=0.4835, pruned_loss=0.2591, over 3711885.66 frames. ], batch size: 46, lr: 4.88e-02, grad_scale: 8.0 2022-12-22 11:01:14,835 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0065, 2.2484, 2.3050, 2.2878, 1.9133, 2.0410, 1.2924, 2.7844], device='cuda:3'), covar=tensor([0.5964, 0.4832, 0.1901, 0.5897, 0.4546, 0.2844, 0.6650, 0.1237], device='cuda:3'), in_proj_covar=tensor([0.0114, 0.0088, 0.0080, 0.0128, 0.0090, 0.0080, 0.0105, 0.0068], device='cuda:3'), out_proj_covar=tensor([1.0601e-04, 8.2952e-05, 6.9844e-05, 1.1883e-04, 8.3304e-05, 7.2882e-05, 9.6473e-05, 5.9311e-05], device='cuda:3') 2022-12-22 11:01:20,614 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.215e+02 7.636e+02 9.611e+02 1.412e+03 2.370e+03, threshold=1.922e+03, percent-clipped=7.0 2022-12-22 11:01:25,051 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1611.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 11:01:33,888 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2595, 1.5525, 1.1821, 1.3304, 2.4980, 1.6673, 1.5916, 1.9325], device='cuda:3'), covar=tensor([0.5030, 0.6561, 0.8857, 1.5761, 0.4664, 0.6106, 1.0307, 0.5875], device='cuda:3'), in_proj_covar=tensor([0.0036, 0.0034, 0.0045, 0.0064, 0.0038, 0.0041, 0.0040, 0.0042], device='cuda:3'), out_proj_covar=tensor([3.5096e-05, 3.3223e-05, 3.9270e-05, 5.9615e-05, 3.5318e-05, 3.7706e-05, 3.7643e-05, 4.0786e-05], device='cuda:3') 2022-12-22 11:01:43,986 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1627.0, num_to_drop=2, layers_to_drop={1, 3} 2022-12-22 11:01:54,334 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0_sp1.1 from training. Duration: 0.836375 2022-12-22 11:02:10,610 INFO [train.py:894] (3/4) Epoch 1, batch 1650, loss[loss=0.6032, simple_loss=0.5469, pruned_loss=0.3346, over 18617.00 frames. ], tot_loss[loss=0.4974, simple_loss=0.4805, pruned_loss=0.2567, over 3711900.24 frames. ], batch size: 178, lr: 4.87e-02, grad_scale: 8.0 2022-12-22 11:02:17,412 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.15 vs. limit=5.0 2022-12-22 11:02:25,763 WARNING [train.py:1060] (3/4) Exclude cut with ID 8565-290391-0049-67394-0_sp0.9 from training. Duration: 21.3166875 2022-12-22 11:02:28,831 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.1711, 1.3798, 1.2874, 1.1687, 1.0151, 1.1868, 0.9632, 1.8182], device='cuda:3'), covar=tensor([0.4269, 0.1897, 0.3769, 0.2611, 0.4260, 0.3055, 0.3170, 0.1747], device='cuda:3'), in_proj_covar=tensor([0.0062, 0.0051, 0.0089, 0.0056, 0.0071, 0.0074, 0.0056, 0.0058], device='cuda:3'), out_proj_covar=tensor([6.3306e-05, 5.4278e-05, 8.5846e-05, 5.7407e-05, 7.1502e-05, 7.4497e-05, 5.8104e-05, 5.7339e-05], device='cuda:3') 2022-12-22 11:02:35,909 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1672.0, num_to_drop=2, layers_to_drop={0, 3} 2022-12-22 11:02:50,747 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0029-104863-0_sp0.9 from training. Duration: 22.1055625 2022-12-22 11:02:58,819 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp1.1 from training. Duration: 21.77725 2022-12-22 11:03:09,739 INFO [train.py:894] (3/4) Epoch 1, batch 1700, loss[loss=0.4777, simple_loss=0.4659, pruned_loss=0.2426, over 18584.00 frames. ], tot_loss[loss=0.4978, simple_loss=0.4806, pruned_loss=0.2569, over 3711626.50 frames. ], batch size: 49, lr: 4.86e-02, grad_scale: 8.0 2022-12-22 11:03:13,225 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp0.9 from training. Duration: 27.8166875 2022-12-22 11:03:17,161 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.788e+02 7.530e+02 9.640e+02 1.219e+03 2.819e+03, threshold=1.928e+03, percent-clipped=4.0 2022-12-22 11:03:33,948 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp1.1 from training. Duration: 22.5090625 2022-12-22 11:03:38,336 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0 from training. Duration: 25.035 2022-12-22 11:03:53,494 WARNING [train.py:1060] (3/4) Exclude cut with ID 774-127930-0014-10412-0_sp1.1 from training. Duration: 0.95 2022-12-22 11:04:07,236 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2022-12-22 11:04:08,720 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp0.9 from training. Duration: 0.92225 2022-12-22 11:04:09,842 INFO [train.py:894] (3/4) Epoch 1, batch 1750, loss[loss=0.4913, simple_loss=0.4778, pruned_loss=0.2508, over 18559.00 frames. ], tot_loss[loss=0.4966, simple_loss=0.48, pruned_loss=0.2558, over 3712273.59 frames. ], batch size: 49, lr: 4.86e-02, grad_scale: 8.0 2022-12-22 11:04:25,627 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5477, 3.3006, 4.1001, 1.5263, 4.3261, 2.1524, 1.2086, 2.1583], device='cuda:3'), covar=tensor([0.1219, 0.0786, 0.0941, 0.2398, 0.0445, 0.1578, 0.3025, 0.1952], device='cuda:3'), in_proj_covar=tensor([0.0082, 0.0076, 0.0092, 0.0082, 0.0061, 0.0063, 0.0102, 0.0076], device='cuda:3'), out_proj_covar=tensor([9.0218e-05, 7.8773e-05, 1.0322e-04, 8.2776e-05, 6.7196e-05, 6.8543e-05, 9.7526e-05, 8.1406e-05], device='cuda:3') 2022-12-22 11:04:29,872 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0 from training. Duration: 21.97 2022-12-22 11:04:39,945 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3389, 1.3351, 1.3108, 1.2708, 1.3188, 1.4477, 1.0695, 2.2963], device='cuda:3'), covar=tensor([0.3871, 0.2042, 0.3857, 0.2745, 0.4023, 0.3147, 0.3028, 0.1533], device='cuda:3'), in_proj_covar=tensor([0.0068, 0.0058, 0.0098, 0.0063, 0.0080, 0.0085, 0.0062, 0.0068], device='cuda:3'), out_proj_covar=tensor([7.1239e-05, 6.3314e-05, 9.7829e-05, 6.4935e-05, 8.1957e-05, 8.7123e-05, 6.6088e-05, 6.7911e-05], device='cuda:3') 2022-12-22 11:04:44,008 WARNING [train.py:1060] (3/4) Exclude cut with ID 7492-105653-0055-62765-0_sp0.9 from training. Duration: 21.97225 2022-12-22 11:04:45,316 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp0.9 from training. Duration: 25.3333125 2022-12-22 11:04:52,182 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.1286, 1.1915, 0.9025, 1.1860, 1.3430, 1.0616, 1.7709, 1.2194], device='cuda:3'), covar=tensor([0.5086, 0.6237, 0.7862, 0.7654, 0.6484, 0.4929, 0.2453, 0.4126], device='cuda:3'), in_proj_covar=tensor([0.0046, 0.0049, 0.0050, 0.0069, 0.0048, 0.0045, 0.0031, 0.0041], device='cuda:3'), out_proj_covar=tensor([4.1424e-05, 4.6540e-05, 4.5127e-05, 6.4045e-05, 4.5323e-05, 3.9993e-05, 2.3189e-05, 3.7049e-05], device='cuda:3') 2022-12-22 11:04:52,977 WARNING [train.py:1060] (3/4) Exclude cut with ID 5172-29468-0015-19128-0_sp0.9 from training. Duration: 21.5055625 2022-12-22 11:04:56,686 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1790.0, num_to_drop=1, layers_to_drop={3} 2022-12-22 11:05:00,936 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp1.1 from training. Duration: 20.72725 2022-12-22 11:05:08,950 INFO [train.py:894] (3/4) Epoch 1, batch 1800, loss[loss=0.4248, simple_loss=0.41, pruned_loss=0.219, over 18319.00 frames. ], tot_loss[loss=0.4962, simple_loss=0.4788, pruned_loss=0.256, over 3712203.22 frames. ], batch size: 40, lr: 4.85e-02, grad_scale: 8.0 2022-12-22 11:05:16,282 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 5.367e+02 7.495e+02 9.120e+02 1.278e+03 3.607e+03, threshold=1.824e+03, percent-clipped=3.0 2022-12-22 11:05:18,611 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1809.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 11:05:27,570 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp0.9 from training. Duration: 26.32775 2022-12-22 11:05:51,140 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0 from training. Duration: 20.025 2022-12-22 11:05:51,385 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.1653, 1.3479, 1.4845, 1.3885, 1.2750, 1.4662, 1.1203, 2.0895], device='cuda:3'), covar=tensor([0.3471, 0.2002, 0.3557, 0.2551, 0.3691, 0.2983, 0.3287, 0.1692], device='cuda:3'), in_proj_covar=tensor([0.0068, 0.0060, 0.0100, 0.0065, 0.0080, 0.0088, 0.0063, 0.0069], device='cuda:3'), out_proj_covar=tensor([7.3004e-05, 6.5728e-05, 1.0059e-04, 6.8346e-05, 8.4403e-05, 9.0572e-05, 6.8184e-05, 7.0729e-05], device='cuda:3') 2022-12-22 11:05:53,713 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1838.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 11:05:55,624 WARNING [train.py:1060] (3/4) Exclude cut with ID 6709-74022-0004-86860-0_sp1.1 from training. Duration: 0.9409375 2022-12-22 11:05:55,630 WARNING [train.py:1060] (3/4) Exclude cut with ID 4757-1811-0023-62229-0_sp0.9 from training. Duration: 21.37775 2022-12-22 11:06:08,807 INFO [train.py:894] (3/4) Epoch 1, batch 1850, loss[loss=0.4899, simple_loss=0.4623, pruned_loss=0.2589, over 18408.00 frames. ], tot_loss[loss=0.5001, simple_loss=0.4808, pruned_loss=0.259, over 3712132.05 frames. ], batch size: 42, lr: 4.84e-02, grad_scale: 8.0 2022-12-22 11:06:12,298 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0004-25974-0_sp0.9 from training. Duration: 21.17225 2022-12-22 11:06:12,310 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp0.9 from training. Duration: 27.511125 2022-12-22 11:06:15,908 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=1857.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 11:06:16,054 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1857.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 11:06:32,253 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1870.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 11:06:38,694 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0 from training. Duration: 22.8 2022-12-22 11:06:44,470 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0 from training. Duration: 22.585 2022-12-22 11:07:08,463 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1899.0, num_to_drop=2, layers_to_drop={0, 1} 2022-12-22 11:07:09,403 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0001-146967-0_sp0.9 from training. Duration: 22.0166875 2022-12-22 11:07:10,551 INFO [train.py:894] (3/4) Epoch 1, batch 1900, loss[loss=0.5615, simple_loss=0.5215, pruned_loss=0.3013, over 18724.00 frames. ], tot_loss[loss=0.5005, simple_loss=0.4804, pruned_loss=0.2598, over 3712541.44 frames. ], batch size: 52, lr: 4.83e-02, grad_scale: 8.0 2022-12-22 11:07:17,482 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.376e+02 8.337e+02 1.120e+03 1.422e+03 4.063e+03, threshold=2.239e+03, percent-clipped=10.0 2022-12-22 11:07:22,969 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp1.1 from training. Duration: 24.395375 2022-12-22 11:07:27,658 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp0.9 from training. Duration: 27.47775 2022-12-22 11:07:30,960 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp0.9 from training. Duration: 24.8833125 2022-12-22 11:07:31,270 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1918.0, num_to_drop=2, layers_to_drop={0, 1} 2022-12-22 11:07:32,208 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0 from training. Duration: 23.39 2022-12-22 11:07:36,581 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1922.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 11:07:37,512 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp0.9 from training. Duration: 28.72225 2022-12-22 11:07:43,586 WARNING [train.py:1060] (3/4) Exclude cut with ID 585-294811-0110-133686-0_sp0.9 from training. Duration: 20.8944375 2022-12-22 11:07:47,845 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1931.0, num_to_drop=2, layers_to_drop={1, 2} 2022-12-22 11:07:55,794 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0_sp0.9 from training. Duration: 23.8444375 2022-12-22 11:08:11,715 INFO [train.py:894] (3/4) Epoch 1, batch 1950, loss[loss=0.4481, simple_loss=0.444, pruned_loss=0.2256, over 18499.00 frames. ], tot_loss[loss=0.4979, simple_loss=0.4794, pruned_loss=0.2577, over 3713381.35 frames. ], batch size: 44, lr: 4.83e-02, grad_scale: 8.0 2022-12-22 11:08:14,327 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0 from training. Duration: 25.85 2022-12-22 11:08:14,340 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0 from training. Duration: 21.39 2022-12-22 11:08:23,956 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0 from training. Duration: 27.92 2022-12-22 11:08:30,704 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1967.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 11:08:34,447 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.44 vs. limit=5.0 2022-12-22 11:08:46,745 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0039-130165-0_sp0.9 from training. Duration: 20.661125 2022-12-22 11:09:05,216 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0043-15874-0_sp0.9 from training. Duration: 20.07225 2022-12-22 11:09:11,457 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0 from training. Duration: 21.01 2022-12-22 11:09:16,248 INFO [train.py:894] (3/4) Epoch 1, batch 2000, loss[loss=0.5004, simple_loss=0.4798, pruned_loss=0.2605, over 18702.00 frames. ], tot_loss[loss=0.4959, simple_loss=0.4781, pruned_loss=0.2564, over 3713732.88 frames. ], batch size: 65, lr: 4.82e-02, grad_scale: 8.0 2022-12-22 11:09:24,809 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.889e+02 7.270e+02 9.125e+02 1.253e+03 3.242e+03, threshold=1.825e+03, percent-clipped=1.0 2022-12-22 11:10:17,843 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0 from training. Duration: 20.65 2022-12-22 11:10:24,439 INFO [train.py:894] (3/4) Epoch 1, batch 2050, loss[loss=0.4251, simple_loss=0.4363, pruned_loss=0.207, over 18711.00 frames. ], tot_loss[loss=0.4892, simple_loss=0.4745, pruned_loss=0.2516, over 3715421.79 frames. ], batch size: 50, lr: 4.81e-02, grad_scale: 8.0 2022-12-22 11:10:24,447 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0 from training. Duration: 21.46 2022-12-22 11:11:06,299 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0 from training. Duration: 0.92 2022-12-22 11:11:11,600 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0_sp0.9 from training. Duration: 23.7666875 2022-12-22 11:11:17,666 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2090.0, num_to_drop=2, layers_to_drop={1, 2} 2022-12-22 11:11:32,147 INFO [train.py:894] (3/4) Epoch 1, batch 2100, loss[loss=0.4112, simple_loss=0.4277, pruned_loss=0.1974, over 18561.00 frames. ], tot_loss[loss=0.4794, simple_loss=0.4687, pruned_loss=0.2448, over 3714638.38 frames. ], batch size: 49, lr: 4.80e-02, grad_scale: 16.0 2022-12-22 11:11:39,985 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.126e+02 6.724e+02 7.895e+02 9.819e+02 2.004e+03, threshold=1.579e+03, percent-clipped=3.0 2022-12-22 11:11:44,967 WARNING [train.py:1060] (3/4) Exclude cut with ID 8544-281189-0060-101339-0_sp0.9 from training. Duration: 20.861125 2022-12-22 11:11:55,711 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0_sp0.9 from training. Duration: 22.711125 2022-12-22 11:12:20,904 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=2138.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 11:12:30,737 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp1.1 from training. Duration: 22.986375 2022-12-22 11:12:39,161 INFO [train.py:894] (3/4) Epoch 1, batch 2150, loss[loss=0.4529, simple_loss=0.4634, pruned_loss=0.2212, over 18580.00 frames. ], tot_loss[loss=0.4748, simple_loss=0.4659, pruned_loss=0.2416, over 3714382.89 frames. ], batch size: 56, lr: 4.79e-02, grad_scale: 16.0 2022-12-22 11:12:44,444 WARNING [train.py:1060] (3/4) Exclude cut with ID 8040-260924-0003-80960-0_sp0.9 from training. Duration: 22.07225 2022-12-22 11:12:50,020 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0045-26330-0_sp0.9 from training. Duration: 20.3055625 2022-12-22 11:12:52,521 WARNING [train.py:1060] (3/4) Exclude cut with ID 6356-271890-0060-94317-0_sp0.9 from training. Duration: 20.72225 2022-12-22 11:13:08,516 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp1.1 from training. Duration: 22.4818125 2022-12-22 11:13:23,031 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.5837, 1.1844, 1.6306, 1.0412, 1.6820, 1.4960, 1.0550, 1.1997], device='cuda:3'), covar=tensor([0.3059, 0.1922, 0.0801, 0.2455, 0.1343, 0.0867, 0.1550, 0.1417], device='cuda:3'), in_proj_covar=tensor([0.0082, 0.0078, 0.0057, 0.0090, 0.0066, 0.0059, 0.0071, 0.0065], device='cuda:3'), out_proj_covar=tensor([8.6125e-05, 7.2187e-05, 4.7587e-05, 8.4187e-05, 6.7572e-05, 5.0447e-05, 6.5374e-05, 5.8592e-05], device='cuda:3') 2022-12-22 11:13:29,152 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp0.9 from training. Duration: 25.0944375 2022-12-22 11:13:32,208 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0 from training. Duration: 21.515 2022-12-22 11:13:37,444 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp0.9 from training. Duration: 27.02225 2022-12-22 11:13:37,574 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2194.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 11:13:42,810 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0010-62480-0_sp0.9 from training. Duration: 22.22225 2022-12-22 11:13:46,887 INFO [train.py:894] (3/4) Epoch 1, batch 2200, loss[loss=0.4812, simple_loss=0.4611, pruned_loss=0.2507, over 18715.00 frames. ], tot_loss[loss=0.4707, simple_loss=0.4636, pruned_loss=0.2387, over 3713890.57 frames. ], batch size: 50, lr: 4.78e-02, grad_scale: 16.0 2022-12-22 11:13:50,193 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0085-44554-0_sp0.9 from training. Duration: 20.85 2022-12-22 11:13:55,273 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.692e+02 7.033e+02 8.615e+02 1.094e+03 2.363e+03, threshold=1.723e+03, percent-clipped=9.0 2022-12-22 11:14:03,738 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2213.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 11:14:15,441 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2222.0, num_to_drop=2, layers_to_drop={1, 2} 2022-12-22 11:14:20,259 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0 from training. Duration: 21.54 2022-12-22 11:14:20,467 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2226.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 11:14:23,105 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7389, 1.9413, 2.1704, 1.3995, 1.1712, 1.7376, 1.4159, 1.4301], device='cuda:3'), covar=tensor([0.2800, 0.1830, 0.1774, 0.1638, 0.2278, 0.2357, 0.2189, 0.1640], device='cuda:3'), in_proj_covar=tensor([0.0072, 0.0070, 0.0080, 0.0069, 0.0069, 0.0093, 0.0091, 0.0067], device='cuda:3'), out_proj_covar=tensor([8.8343e-05, 8.3888e-05, 9.5059e-05, 8.6232e-05, 8.3964e-05, 1.0781e-04, 1.0267e-04, 8.4530e-05], device='cuda:3') 2022-12-22 11:14:24,014 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp1.1 from training. Duration: 20.5318125 2022-12-22 11:14:26,868 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.87 vs. limit=2.0 2022-12-22 11:14:33,774 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0012-134311-0_sp0.9 from training. Duration: 21.9333125 2022-12-22 11:14:53,776 INFO [train.py:894] (3/4) Epoch 1, batch 2250, loss[loss=0.467, simple_loss=0.4793, pruned_loss=0.2274, over 18689.00 frames. ], tot_loss[loss=0.4647, simple_loss=0.4603, pruned_loss=0.2344, over 3714319.88 frames. ], batch size: 60, lr: 4.77e-02, grad_scale: 16.0 2022-12-22 11:15:14,987 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5650, 1.5055, 1.3819, 2.5195, 1.7908, 1.7323, 2.4333, 1.4989], device='cuda:3'), covar=tensor([0.1737, 0.3303, 0.2939, 0.1689, 0.1859, 0.2305, 0.2658, 0.2563], device='cuda:3'), in_proj_covar=tensor([0.0038, 0.0045, 0.0039, 0.0031, 0.0033, 0.0041, 0.0039, 0.0039], device='cuda:3'), out_proj_covar=tensor([3.4125e-05, 3.7828e-05, 3.3636e-05, 2.6587e-05, 2.7105e-05, 3.4642e-05, 3.3630e-05, 3.3680e-05], device='cuda:3') 2022-12-22 11:15:15,757 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0025-130151-0_sp0.9 from training. Duration: 21.7944375 2022-12-22 11:15:15,936 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2267.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 11:15:19,875 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=2270.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 11:15:28,273 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0_sp0.9 from training. Duration: 22.4666875 2022-12-22 11:15:34,133 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0 from training. Duration: 21.635 2022-12-22 11:15:39,980 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0_sp0.9 from training. Duration: 24.038875 2022-12-22 11:16:01,175 INFO [train.py:894] (3/4) Epoch 1, batch 2300, loss[loss=0.3747, simple_loss=0.407, pruned_loss=0.1712, over 18453.00 frames. ], tot_loss[loss=0.4612, simple_loss=0.4591, pruned_loss=0.2315, over 3713848.60 frames. ], batch size: 50, lr: 4.77e-02, grad_scale: 16.0 2022-12-22 11:16:08,972 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.591e+02 7.252e+02 9.061e+02 1.161e+03 2.853e+03, threshold=1.812e+03, percent-clipped=4.0 2022-12-22 11:16:17,717 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp1.1 from training. Duration: 21.786375 2022-12-22 11:16:20,580 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=2315.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 11:16:27,686 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0 from training. Duration: 20.22 2022-12-22 11:16:54,400 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2340.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 11:17:08,641 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2022-12-22 11:17:09,290 INFO [train.py:894] (3/4) Epoch 1, batch 2350, loss[loss=0.4325, simple_loss=0.4519, pruned_loss=0.2066, over 18514.00 frames. ], tot_loss[loss=0.4591, simple_loss=0.4583, pruned_loss=0.2299, over 3714668.07 frames. ], batch size: 55, lr: 4.76e-02, grad_scale: 16.0 2022-12-22 11:18:16,136 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0 from training. Duration: 25.285 2022-12-22 11:18:17,759 INFO [train.py:894] (3/4) Epoch 1, batch 2400, loss[loss=0.4638, simple_loss=0.4646, pruned_loss=0.2315, over 18669.00 frames. ], tot_loss[loss=0.455, simple_loss=0.4554, pruned_loss=0.2272, over 3715109.00 frames. ], batch size: 69, lr: 4.75e-02, grad_scale: 16.0 2022-12-22 11:18:18,183 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2401.0, num_to_drop=2, layers_to_drop={0, 1} 2022-12-22 11:18:25,228 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 5.700e+02 7.707e+02 9.733e+02 1.236e+03 3.171e+03, threshold=1.947e+03, percent-clipped=7.0 2022-12-22 11:19:15,015 WARNING [train.py:1060] (3/4) Exclude cut with ID 811-130148-0001-63453-0_sp0.9 from training. Duration: 20.861125 2022-12-22 11:19:25,848 INFO [train.py:894] (3/4) Epoch 1, batch 2450, loss[loss=0.4258, simple_loss=0.4275, pruned_loss=0.212, over 18576.00 frames. ], tot_loss[loss=0.452, simple_loss=0.4539, pruned_loss=0.225, over 3715207.37 frames. ], batch size: 41, lr: 4.74e-02, grad_scale: 16.0 2022-12-22 11:19:33,871 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0 from training. Duration: 20.88 2022-12-22 11:20:04,173 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0_sp0.9 from training. Duration: 23.4166875 2022-12-22 11:20:24,690 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2494.0, num_to_drop=1, layers_to_drop={2} 2022-12-22 11:20:34,035 INFO [train.py:894] (3/4) Epoch 1, batch 2500, loss[loss=0.4441, simple_loss=0.4448, pruned_loss=0.2216, over 18449.00 frames. ], tot_loss[loss=0.4473, simple_loss=0.4516, pruned_loss=0.2214, over 3715304.84 frames. ], batch size: 50, lr: 4.73e-02, grad_scale: 16.0 2022-12-22 11:20:41,986 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.819e+02 7.053e+02 8.815e+02 1.099e+03 3.368e+03, threshold=1.763e+03, percent-clipped=3.0 2022-12-22 11:20:50,393 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2513.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 11:21:08,589 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2526.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 11:21:12,464 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0 from training. Duration: 21.24 2022-12-22 11:21:12,482 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0_sp0.9 from training. Duration: 23.9055625 2022-12-22 11:21:31,254 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=2542.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 11:21:43,233 INFO [train.py:894] (3/4) Epoch 1, batch 2550, loss[loss=0.4579, simple_loss=0.4618, pruned_loss=0.227, over 18527.00 frames. ], tot_loss[loss=0.4443, simple_loss=0.4498, pruned_loss=0.2194, over 3715557.62 frames. ], batch size: 52, lr: 4.72e-02, grad_scale: 16.0 2022-12-22 11:21:43,259 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp0.9 from training. Duration: 25.988875 2022-12-22 11:21:51,034 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0001-134300-0_sp0.9 from training. Duration: 20.67225 2022-12-22 11:21:57,779 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=2561.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 11:22:16,100 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=2574.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 11:22:35,072 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0 from training. Duration: 20.34 2022-12-22 11:22:52,225 INFO [train.py:894] (3/4) Epoch 1, batch 2600, loss[loss=0.3995, simple_loss=0.4294, pruned_loss=0.1848, over 18725.00 frames. ], tot_loss[loss=0.4425, simple_loss=0.4485, pruned_loss=0.2183, over 3715531.61 frames. ], batch size: 52, lr: 4.71e-02, grad_scale: 16.0 2022-12-22 11:23:00,099 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.621e+02 7.074e+02 9.179e+02 1.186e+03 2.558e+03, threshold=1.836e+03, percent-clipped=4.0 2022-12-22 11:23:06,836 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0382, 2.3594, 1.7409, 2.0777, 2.1302, 1.8339, 1.3227, 2.9339], device='cuda:3'), covar=tensor([0.2585, 0.1626, 0.2981, 0.2049, 0.3116, 0.2760, 0.2814, 0.1471], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0071, 0.0113, 0.0075, 0.0087, 0.0099, 0.0073, 0.0079], device='cuda:3'), out_proj_covar=tensor([9.4759e-05, 9.0223e-05, 1.2886e-04, 9.5373e-05, 1.0920e-04, 1.1390e-04, 9.4336e-05, 9.5317e-05], device='cuda:3') 2022-12-22 11:23:34,203 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4978, 1.9504, 2.4432, 1.4278, 1.9362, 2.8402, 1.9547, 1.8387], device='cuda:3'), covar=tensor([0.2165, 0.1468, 0.0800, 0.2076, 0.1381, 0.0485, 0.1320, 0.1016], device='cuda:3'), in_proj_covar=tensor([0.0083, 0.0078, 0.0060, 0.0091, 0.0070, 0.0056, 0.0077, 0.0066], device='cuda:3'), out_proj_covar=tensor([8.5906e-05, 7.4316e-05, 5.7314e-05, 8.6253e-05, 7.3493e-05, 5.1912e-05, 7.4479e-05, 6.2569e-05], device='cuda:3') 2022-12-22 11:23:43,989 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp0.9 from training. Duration: 25.061125 2022-12-22 11:23:52,049 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0 from training. Duration: 0.83 2022-12-22 11:24:01,862 INFO [train.py:894] (3/4) Epoch 1, batch 2650, loss[loss=0.498, simple_loss=0.4817, pruned_loss=0.2571, over 18661.00 frames. ], tot_loss[loss=0.438, simple_loss=0.4457, pruned_loss=0.2151, over 3715397.61 frames. ], batch size: 179, lr: 4.70e-02, grad_scale: 8.0 2022-12-22 11:24:15,297 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0 from training. Duration: 24.73 2022-12-22 11:24:27,701 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0 from training. Duration: 23.965 2022-12-22 11:24:28,134 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.3047, 2.7614, 2.1789, 1.4119, 1.9603, 2.9193, 2.0064, 1.5516], device='cuda:3'), covar=tensor([0.0617, 0.1308, 0.1655, 0.3044, 0.1375, 0.0782, 0.1833, 0.3804], device='cuda:3'), in_proj_covar=tensor([0.0058, 0.0080, 0.0080, 0.0097, 0.0071, 0.0063, 0.0083, 0.0109], device='cuda:3'), out_proj_covar=tensor([5.7887e-05, 7.9539e-05, 7.6495e-05, 9.2093e-05, 7.1429e-05, 6.0536e-05, 7.9576e-05, 1.0301e-04], device='cuda:3') 2022-12-22 11:24:35,952 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0030-146996-0_sp0.9 from training. Duration: 22.088875 2022-12-22 11:24:41,784 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.2861, 1.5270, 0.7362, 1.6295, 1.1887, 0.9804, 2.1313, 1.0367], device='cuda:3'), covar=tensor([0.2308, 0.2407, 0.4398, 0.2758, 0.4673, 0.2567, 0.0731, 0.2384], device='cuda:3'), in_proj_covar=tensor([0.0051, 0.0050, 0.0059, 0.0070, 0.0071, 0.0052, 0.0030, 0.0046], device='cuda:3'), out_proj_covar=tensor([4.7667e-05, 4.7169e-05, 5.3045e-05, 6.3283e-05, 6.6727e-05, 4.6420e-05, 2.3758e-05, 4.1406e-05], device='cuda:3') 2022-12-22 11:24:50,447 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2685.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 11:24:52,983 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0_sp0.9 from training. Duration: 23.6 2022-12-22 11:25:05,297 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2696.0, num_to_drop=1, layers_to_drop={3} 2022-12-22 11:25:12,221 INFO [train.py:894] (3/4) Epoch 1, batch 2700, loss[loss=0.4337, simple_loss=0.4525, pruned_loss=0.2074, over 18553.00 frames. ], tot_loss[loss=0.4329, simple_loss=0.4425, pruned_loss=0.2117, over 3716030.18 frames. ], batch size: 57, lr: 4.69e-02, grad_scale: 8.0 2022-12-22 11:25:22,295 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.615e+02 7.385e+02 8.951e+02 1.108e+03 2.259e+03, threshold=1.790e+03, percent-clipped=4.0 2022-12-22 11:25:56,778 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=2.05 vs. limit=2.0 2022-12-22 11:26:17,167 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2746.0, num_to_drop=2, layers_to_drop={1, 2} 2022-12-22 11:26:23,450 INFO [train.py:894] (3/4) Epoch 1, batch 2750, loss[loss=0.4524, simple_loss=0.4608, pruned_loss=0.222, over 18583.00 frames. ], tot_loss[loss=0.4312, simple_loss=0.4412, pruned_loss=0.2106, over 3716001.70 frames. ], batch size: 51, lr: 4.68e-02, grad_scale: 8.0 2022-12-22 11:26:23,496 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0 from training. Duration: 23.795 2022-12-22 11:26:38,545 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp1.1 from training. Duration: 21.5409375 2022-12-22 11:26:41,235 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp0.9 from training. Duration: 24.97775 2022-12-22 11:26:51,898 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0_sp0.9 from training. Duration: 23.3444375 2022-12-22 11:27:01,057 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7253, 3.2520, 3.7679, 2.0041, 3.6432, 2.3213, 0.7620, 2.5871], device='cuda:3'), covar=tensor([0.1391, 0.0839, 0.1355, 0.2515, 0.1033, 0.1896, 0.5536, 0.2267], device='cuda:3'), in_proj_covar=tensor([0.0087, 0.0062, 0.0108, 0.0081, 0.0068, 0.0073, 0.0122, 0.0086], device='cuda:3'), out_proj_covar=tensor([1.0441e-04, 7.2727e-05, 1.3712e-04, 8.8464e-05, 8.6700e-05, 8.7503e-05, 1.2035e-04, 1.0244e-04], device='cuda:3') 2022-12-22 11:27:16,568 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0_sp0.9 from training. Duration: 23.2 2022-12-22 11:27:22,992 WARNING [train.py:1060] (3/4) Exclude cut with ID 5653-46179-0060-117930-0_sp0.9 from training. Duration: 21.17225 2022-12-22 11:27:33,909 INFO [train.py:894] (3/4) Epoch 1, batch 2800, loss[loss=0.4438, simple_loss=0.4527, pruned_loss=0.2175, over 18591.00 frames. ], tot_loss[loss=0.4278, simple_loss=0.4388, pruned_loss=0.2083, over 3715309.33 frames. ], batch size: 56, lr: 4.67e-02, grad_scale: 8.0 2022-12-22 11:27:39,315 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp0.9 from training. Duration: 24.6555625 2022-12-22 11:27:43,871 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.105e+02 7.234e+02 8.983e+02 1.149e+03 3.907e+03, threshold=1.797e+03, percent-clipped=2.0 2022-12-22 11:27:58,497 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.3057, 3.6857, 3.4672, 3.7211, 3.7788, 3.7685, 4.3283, 1.1990], device='cuda:3'), covar=tensor([0.0333, 0.0487, 0.0744, 0.0374, 0.1103, 0.0446, 0.0319, 0.4477], device='cuda:3'), in_proj_covar=tensor([0.0076, 0.0086, 0.0090, 0.0064, 0.0116, 0.0077, 0.0090, 0.0143], device='cuda:3'), out_proj_covar=tensor([7.9179e-05, 9.1048e-05, 9.5879e-05, 6.5227e-05, 1.0850e-04, 8.3487e-05, 9.6051e-05, 1.2569e-04], device='cuda:3') 2022-12-22 11:28:05,887 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=2.28 vs. limit=2.0 2022-12-22 11:28:07,083 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2824.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 11:28:08,360 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8932, 1.9124, 1.4595, 1.0584, 2.0244, 2.1970, 1.3986, 1.3122], device='cuda:3'), covar=tensor([0.1949, 0.1730, 0.3562, 0.3103, 0.1113, 0.1137, 0.2349, 0.3578], device='cuda:3'), in_proj_covar=tensor([0.0119, 0.0108, 0.0156, 0.0131, 0.0092, 0.0105, 0.0125, 0.0130], device='cuda:3'), out_proj_covar=tensor([1.2744e-04, 1.1884e-04, 1.5695e-04, 1.3367e-04, 9.9330e-05, 1.1310e-04, 1.2826e-04, 1.3243e-04], device='cuda:3') 2022-12-22 11:28:11,886 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2132, 1.4655, 1.4140, 2.5366, 2.5558, 2.8976, 2.4200, 1.9087], device='cuda:3'), covar=tensor([0.2170, 0.2788, 0.3058, 0.2048, 0.1475, 0.1069, 0.1381, 0.2635], device='cuda:3'), in_proj_covar=tensor([0.0082, 0.0073, 0.0092, 0.0080, 0.0078, 0.0077, 0.0072, 0.0081], device='cuda:3'), out_proj_covar=tensor([1.1116e-04, 9.5792e-05, 1.1807e-04, 1.1211e-04, 1.0092e-04, 1.1645e-04, 9.8443e-05, 1.0124e-04], device='cuda:3') 2022-12-22 11:28:28,559 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0 from training. Duration: 20.44 2022-12-22 11:28:42,831 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0_sp0.9 from training. Duration: 23.45 2022-12-22 11:28:44,155 INFO [train.py:894] (3/4) Epoch 1, batch 2850, loss[loss=0.3956, simple_loss=0.3959, pruned_loss=0.1976, over 18558.00 frames. ], tot_loss[loss=0.4265, simple_loss=0.4385, pruned_loss=0.2073, over 3715388.37 frames. ], batch size: 41, lr: 4.66e-02, grad_scale: 8.0 2022-12-22 11:29:09,139 WARNING [train.py:1060] (3/4) Exclude cut with ID 6945-60535-0076-12784-0_sp0.9 from training. Duration: 20.52225 2022-12-22 11:29:15,878 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0 from training. Duration: 22.19 2022-12-22 11:29:25,877 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp1.1 from training. Duration: 25.3818125 2022-12-22 11:29:32,172 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2885.0, num_to_drop=2, layers_to_drop={1, 2} 2022-12-22 11:29:41,151 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp0.9 from training. Duration: 28.0944375 2022-12-22 11:29:41,553 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0741, 2.6951, 2.5267, 2.0491, 2.4899, 2.7431, 1.5072, 3.3072], device='cuda:3'), covar=tensor([0.3370, 0.1987, 0.2954, 0.4977, 0.2560, 0.1952, 0.3673, 0.1079], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0113, 0.0125, 0.0146, 0.0124, 0.0112, 0.0132, 0.0104], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:3') 2022-12-22 11:29:48,282 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0_sp0.9 from training. Duration: 22.9444375 2022-12-22 11:29:53,523 INFO [train.py:894] (3/4) Epoch 1, batch 2900, loss[loss=0.3861, simple_loss=0.414, pruned_loss=0.1791, over 18458.00 frames. ], tot_loss[loss=0.4236, simple_loss=0.4364, pruned_loss=0.2054, over 3714913.27 frames. ], batch size: 50, lr: 4.65e-02, grad_scale: 8.0 2022-12-22 11:29:53,537 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp1.1 from training. Duration: 21.6318125 2022-12-22 11:30:04,353 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.831e+02 7.832e+02 9.330e+02 1.210e+03 2.996e+03, threshold=1.866e+03, percent-clipped=5.0 2022-12-22 11:30:09,861 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0 from training. Duration: 23.695 2022-12-22 11:30:33,861 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0 from training. Duration: 23.955 2022-12-22 11:30:45,595 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2937.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 11:31:04,254 INFO [train.py:894] (3/4) Epoch 1, batch 2950, loss[loss=0.3843, simple_loss=0.4189, pruned_loss=0.1749, over 18537.00 frames. ], tot_loss[loss=0.421, simple_loss=0.4352, pruned_loss=0.2034, over 3715302.03 frames. ], batch size: 55, lr: 4.64e-02, grad_scale: 8.0 2022-12-22 11:31:05,531 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp0.9 from training. Duration: 26.438875 2022-12-22 11:31:46,378 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0021-26306-0_sp0.9 from training. Duration: 21.2444375 2022-12-22 11:31:46,401 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp0.9 from training. Duration: 31.02225 2022-12-22 11:31:57,576 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0 from training. Duration: 22.395 2022-12-22 11:32:06,991 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2996.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 11:32:09,683 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2998.0, num_to_drop=2, layers_to_drop={2, 3} 2022-12-22 11:32:10,547 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0 from training. Duration: 21.075 2022-12-22 11:32:14,381 INFO [train.py:894] (3/4) Epoch 1, batch 3000, loss[loss=0.3996, simple_loss=0.4084, pruned_loss=0.1954, over 18439.00 frames. ], tot_loss[loss=0.4169, simple_loss=0.4323, pruned_loss=0.2008, over 3714056.64 frames. ], batch size: 48, lr: 4.63e-02, grad_scale: 8.0 2022-12-22 11:32:14,381 INFO [train.py:919] (3/4) Computing validation loss 2022-12-22 11:32:25,520 INFO [train.py:928] (3/4) Epoch 1, validation: loss=0.3402, simple_loss=0.4207, pruned_loss=0.1299, over 944034.00 frames. 2022-12-22 11:32:25,521 INFO [train.py:929] (3/4) Maximum memory allocated so far is 22754MB 2022-12-22 11:32:28,961 WARNING [train.py:1060] (3/4) Exclude cut with ID 6482-98857-0025-147532-0_sp0.9 from training. Duration: 20.0055625 2022-12-22 11:32:28,973 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0037-132304-0_sp0.9 from training. Duration: 22.05 2022-12-22 11:32:30,383 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0 from training. Duration: 26.8349375 2022-12-22 11:32:33,104 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp1.1 from training. Duration: 22.1090625 2022-12-22 11:32:35,857 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.958e+02 7.073e+02 9.090e+02 1.113e+03 2.521e+03, threshold=1.818e+03, percent-clipped=3.0 2022-12-22 11:32:40,436 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp0.9 from training. Duration: 26.6166875 2022-12-22 11:32:44,782 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.6749, 3.8479, 3.8634, 4.0397, 4.2450, 4.0250, 4.5095, 1.5256], device='cuda:3'), covar=tensor([0.0373, 0.0580, 0.0679, 0.0394, 0.1054, 0.0500, 0.0323, 0.4161], device='cuda:3'), in_proj_covar=tensor([0.0084, 0.0094, 0.0097, 0.0070, 0.0124, 0.0085, 0.0095, 0.0152], device='cuda:3'), out_proj_covar=tensor([8.7522e-05, 9.9715e-05, 1.0499e-04, 7.2501e-05, 1.1884e-04, 9.3750e-05, 1.0170e-04, 1.3306e-04], device='cuda:3') 2022-12-22 11:32:51,769 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3019.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 11:32:56,328 WARNING [train.py:1060] (3/4) Exclude cut with ID 2046-178027-0000-53705-0_sp0.9 from training. Duration: 20.3055625 2022-12-22 11:33:10,471 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2022-12-22 11:33:18,490 WARNING [train.py:1060] (3/4) Exclude cut with ID 7205-50138-0008-5373-0_sp0.9 from training. Duration: 20.7 2022-12-22 11:33:23,989 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3041.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 11:33:28,123 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=3044.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 11:33:38,546 INFO [train.py:894] (3/4) Epoch 1, batch 3050, loss[loss=0.3877, simple_loss=0.4212, pruned_loss=0.1771, over 18517.00 frames. ], tot_loss[loss=0.4143, simple_loss=0.4304, pruned_loss=0.1991, over 3714319.37 frames. ], batch size: 52, lr: 4.62e-02, grad_scale: 8.0 2022-12-22 11:33:59,962 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0 from training. Duration: 22.48 2022-12-22 11:34:14,216 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp0.9 from training. Duration: 29.816625 2022-12-22 11:34:20,657 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3080.0, num_to_drop=2, layers_to_drop={0, 1} 2022-12-22 11:34:32,420 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp1.1 from training. Duration: 22.7590625 2022-12-22 11:34:37,955 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0 from training. Duration: 22.555 2022-12-22 11:34:48,852 INFO [train.py:894] (3/4) Epoch 1, batch 3100, loss[loss=0.4268, simple_loss=0.4336, pruned_loss=0.21, over 18557.00 frames. ], tot_loss[loss=0.4102, simple_loss=0.4282, pruned_loss=0.1961, over 3714566.69 frames. ], batch size: 49, lr: 4.61e-02, grad_scale: 8.0 2022-12-22 11:34:58,945 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.605e+02 6.687e+02 8.494e+02 1.046e+03 2.071e+03, threshold=1.699e+03, percent-clipped=1.0 2022-12-22 11:34:58,969 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0005-25975-0_sp0.9 from training. Duration: 21.688875 2022-12-22 11:35:30,836 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2022-12-22 11:35:31,294 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0_sp0.9 from training. Duration: 22.6 2022-12-22 11:35:58,875 INFO [train.py:894] (3/4) Epoch 1, batch 3150, loss[loss=0.3993, simple_loss=0.4243, pruned_loss=0.1871, over 18585.00 frames. ], tot_loss[loss=0.4086, simple_loss=0.4271, pruned_loss=0.1951, over 3715109.96 frames. ], batch size: 51, lr: 4.60e-02, grad_scale: 8.0 2022-12-22 11:36:03,797 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0 from training. Duration: 24.32 2022-12-22 11:36:36,949 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9115, 0.9507, 2.0862, 1.2417, 1.5880, 1.9203, 2.0344, 1.8137], device='cuda:3'), covar=tensor([0.0863, 0.2382, 0.0547, 0.2227, 0.1186, 0.0991, 0.0741, 0.1085], device='cuda:3'), in_proj_covar=tensor([0.0065, 0.0090, 0.0059, 0.0094, 0.0080, 0.0063, 0.0060, 0.0081], device='cuda:3'), out_proj_covar=tensor([6.1714e-05, 8.8416e-05, 5.7333e-05, 8.9059e-05, 7.8382e-05, 6.0935e-05, 5.6109e-05, 8.0646e-05], device='cuda:3') 2022-12-22 11:36:37,182 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2022-12-22 11:36:38,115 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6905, 1.5508, 2.0321, 1.3575, 1.6090, 1.7603, 0.8719, 1.4083], device='cuda:3'), covar=tensor([0.2520, 0.2101, 0.2168, 0.1285, 0.2092, 0.1789, 0.2908, 0.1319], device='cuda:3'), in_proj_covar=tensor([0.0080, 0.0083, 0.0101, 0.0072, 0.0077, 0.0100, 0.0113, 0.0074], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:3') 2022-12-22 11:36:40,652 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3180.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 11:36:58,540 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-276745-0093-13116-0_sp0.9 from training. Duration: 21.061125 2022-12-22 11:37:09,802 INFO [train.py:894] (3/4) Epoch 1, batch 3200, loss[loss=0.3842, simple_loss=0.409, pruned_loss=0.1798, over 18369.00 frames. ], tot_loss[loss=0.4097, simple_loss=0.4278, pruned_loss=0.1958, over 3715008.92 frames. ], batch size: 51, lr: 4.59e-02, grad_scale: 8.0 2022-12-22 11:37:12,499 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0024-15855-0_sp0.9 from training. Duration: 20.32225 2022-12-22 11:37:19,192 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 5.026e+02 7.855e+02 1.051e+03 1.351e+03 2.916e+03, threshold=2.101e+03, percent-clipped=7.0 2022-12-22 11:37:22,069 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp1.1 from training. Duration: 0.7545625 2022-12-22 11:37:36,503 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0_sp0.9 from training. Duration: 23.9333125 2022-12-22 11:38:06,614 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp1.1 from training. Duration: 20.17275 2022-12-22 11:38:11,071 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp1.1 from training. Duration: 20.436375 2022-12-22 11:38:18,737 INFO [train.py:894] (3/4) Epoch 1, batch 3250, loss[loss=0.4377, simple_loss=0.4615, pruned_loss=0.2069, over 18564.00 frames. ], tot_loss[loss=0.4101, simple_loss=0.428, pruned_loss=0.196, over 3714392.92 frames. ], batch size: 57, lr: 4.58e-02, grad_scale: 8.0 2022-12-22 11:39:10,543 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4109, 1.9216, 1.8684, 1.3387, 1.4545, 2.5448, 1.3504, 1.4093], device='cuda:3'), covar=tensor([0.1618, 0.2052, 0.2145, 0.3889, 0.2071, 0.0822, 0.3050, 0.3967], device='cuda:3'), in_proj_covar=tensor([0.0077, 0.0106, 0.0120, 0.0139, 0.0098, 0.0083, 0.0117, 0.0136], device='cuda:3'), out_proj_covar=tensor([9.0189e-05, 1.1214e-04, 1.2252e-04, 1.3790e-04, 1.0673e-04, 8.6775e-05, 1.1836e-04, 1.3398e-04], device='cuda:3') 2022-12-22 11:39:10,546 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3288.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 11:39:17,222 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3293.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 11:39:28,602 INFO [train.py:894] (3/4) Epoch 1, batch 3300, loss[loss=0.3626, simple_loss=0.4115, pruned_loss=0.1569, over 18578.00 frames. ], tot_loss[loss=0.4061, simple_loss=0.4253, pruned_loss=0.1934, over 3714121.05 frames. ], batch size: 78, lr: 4.57e-02, grad_scale: 8.0 2022-12-22 11:39:28,656 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0_sp0.9 from training. Duration: 23.1055625 2022-12-22 11:39:31,178 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp1.1 from training. Duration: 23.5 2022-12-22 11:39:37,652 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.762e+02 7.922e+02 9.750e+02 1.198e+03 1.925e+03, threshold=1.950e+03, percent-clipped=0.0 2022-12-22 11:39:40,588 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp0.9 from training. Duration: 26.62775 2022-12-22 11:39:52,404 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0 from training. Duration: 21.105 2022-12-22 11:39:56,660 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0_sp0.9 from training. Duration: 24.411125 2022-12-22 11:40:01,767 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2022-12-22 11:40:21,445 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp1.1 from training. Duration: 21.263625 2022-12-22 11:40:25,231 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3341.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 11:40:37,061 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3349.0, num_to_drop=2, layers_to_drop={0, 1} 2022-12-22 11:40:39,255 INFO [train.py:894] (3/4) Epoch 1, batch 3350, loss[loss=0.386, simple_loss=0.4219, pruned_loss=0.1751, over 18579.00 frames. ], tot_loss[loss=0.4029, simple_loss=0.4231, pruned_loss=0.1913, over 3714160.87 frames. ], batch size: 57, lr: 4.56e-02, grad_scale: 8.0 2022-12-22 11:40:51,658 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0 from training. Duration: 20.795 2022-12-22 11:41:02,171 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0 from training. Duration: 24.76 2022-12-22 11:41:03,267 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0_sp0.9 from training. Duration: 22.25 2022-12-22 11:41:13,720 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3375.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 11:41:24,773 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp1.1 from training. Duration: 20.5045625 2022-12-22 11:41:34,529 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=3389.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 11:41:51,778 INFO [train.py:894] (3/4) Epoch 1, batch 3400, loss[loss=0.4151, simple_loss=0.4357, pruned_loss=0.1972, over 18635.00 frames. ], tot_loss[loss=0.4029, simple_loss=0.4234, pruned_loss=0.1911, over 3714273.01 frames. ], batch size: 98, lr: 4.55e-02, grad_scale: 8.0 2022-12-22 11:42:01,145 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.101e+02 7.385e+02 8.854e+02 1.159e+03 2.009e+03, threshold=1.771e+03, percent-clipped=1.0 2022-12-22 11:42:24,658 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2022-12-22 11:42:58,273 INFO [train.py:894] (3/4) Epoch 1, batch 3450, loss[loss=0.3786, simple_loss=0.4195, pruned_loss=0.1689, over 18735.00 frames. ], tot_loss[loss=0.4012, simple_loss=0.4222, pruned_loss=0.1901, over 3715536.33 frames. ], batch size: 54, lr: 4.54e-02, grad_scale: 4.0 2022-12-22 11:43:37,984 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3480.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 11:44:07,898 INFO [train.py:894] (3/4) Epoch 1, batch 3500, loss[loss=0.4516, simple_loss=0.4483, pruned_loss=0.2274, over 18602.00 frames. ], tot_loss[loss=0.4017, simple_loss=0.423, pruned_loss=0.1902, over 3715226.33 frames. ], batch size: 180, lr: 4.53e-02, grad_scale: 4.0 2022-12-22 11:44:28,513 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp1.1 from training. Duration: 22.2954375 2022-12-22 11:44:37,380 INFO [train.py:894] (3/4) Epoch 2, batch 0, loss[loss=0.444, simple_loss=0.4563, pruned_loss=0.2158, over 18607.00 frames. ], tot_loss[loss=0.444, simple_loss=0.4563, pruned_loss=0.2158, over 18607.00 frames. ], batch size: 45, lr: 4.44e-02, grad_scale: 8.0 2022-12-22 11:44:37,380 INFO [train.py:919] (3/4) Computing validation loss 2022-12-22 11:44:49,009 INFO [train.py:928] (3/4) Epoch 2, validation: loss=0.3008, simple_loss=0.3844, pruned_loss=0.1086, over 944034.00 frames. 2022-12-22 11:44:49,010 INFO [train.py:929] (3/4) Maximum memory allocated so far is 22754MB 2022-12-22 11:44:51,671 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.103e+02 7.617e+02 9.580e+02 1.359e+03 2.916e+03, threshold=1.916e+03, percent-clipped=9.0 2022-12-22 11:45:15,239 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2022-12-22 11:45:20,035 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=3528.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 11:45:26,969 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7834, 1.7292, 1.2452, 2.4482, 2.3612, 3.0797, 2.0357, 1.5422], device='cuda:3'), covar=tensor([0.1484, 0.2399, 0.2913, 0.2153, 0.1792, 0.0675, 0.1825, 0.3100], device='cuda:3'), in_proj_covar=tensor([0.0088, 0.0081, 0.0098, 0.0092, 0.0094, 0.0080, 0.0086, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:3') 2022-12-22 11:45:36,179 WARNING [train.py:1060] (3/4) Exclude cut with ID 298-126791-0067-24026-0_sp0.9 from training. Duration: 21.438875 2022-12-22 11:45:40,170 WARNING [train.py:1060] (3/4) Exclude cut with ID 5652-39938-0025-23684-0_sp0.9 from training. Duration: 22.2055625 2022-12-22 11:45:44,499 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3546.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 11:45:59,794 INFO [train.py:894] (3/4) Epoch 2, batch 50, loss[loss=0.3033, simple_loss=0.3587, pruned_loss=0.124, over 18561.00 frames. ], tot_loss[loss=0.3519, simple_loss=0.3977, pruned_loss=0.1531, over 838190.58 frames. ], batch size: 49, lr: 4.43e-02, grad_scale: 8.0 2022-12-22 11:46:42,303 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([5.8819, 4.9371, 4.9235, 5.4135, 5.4571, 4.9408, 5.3703, 1.4262], device='cuda:3'), covar=tensor([0.0265, 0.0374, 0.0522, 0.0245, 0.0736, 0.0407, 0.0297, 0.4484], device='cuda:3'), in_proj_covar=tensor([0.0103, 0.0103, 0.0101, 0.0075, 0.0140, 0.0096, 0.0102, 0.0163], device='cuda:3'), out_proj_covar=tensor([1.1352e-04, 1.1198e-04, 1.1586e-04, 8.4553e-05, 1.3836e-04, 1.0886e-04, 1.1314e-04, 1.4563e-04], device='cuda:3') 2022-12-22 11:46:51,215 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=2.04 vs. limit=2.0 2022-12-22 11:46:52,043 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3593.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 11:46:56,662 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3136, 1.6563, 1.4164, 1.0324, 1.2051, 1.8297, 1.0578, 1.3093], device='cuda:3'), covar=tensor([0.1056, 0.1625, 0.2386, 0.3333, 0.1931, 0.1088, 0.3109, 0.2905], device='cuda:3'), in_proj_covar=tensor([0.0078, 0.0112, 0.0129, 0.0144, 0.0101, 0.0088, 0.0127, 0.0138], device='cuda:3'), out_proj_covar=tensor([9.8556e-05, 1.1960e-04, 1.3467e-04, 1.4531e-04, 1.1309e-04, 9.4860e-05, 1.3182e-04, 1.3800e-04], device='cuda:3') 2022-12-22 11:47:11,766 INFO [train.py:894] (3/4) Epoch 2, batch 100, loss[loss=0.3258, simple_loss=0.3861, pruned_loss=0.1327, over 18649.00 frames. ], tot_loss[loss=0.3453, simple_loss=0.3926, pruned_loss=0.149, over 1475542.97 frames. ], batch size: 69, lr: 4.42e-02, grad_scale: 8.0 2022-12-22 11:47:12,803 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3607.0, num_to_drop=2, layers_to_drop={1, 3} 2022-12-22 11:47:15,573 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.976e+02 8.232e+02 1.042e+03 1.328e+03 2.766e+03, threshold=2.084e+03, percent-clipped=5.0 2022-12-22 11:48:01,636 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=3641.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 11:48:05,751 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3644.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 11:48:24,882 INFO [train.py:894] (3/4) Epoch 2, batch 150, loss[loss=0.2848, simple_loss=0.3505, pruned_loss=0.1096, over 18646.00 frames. ], tot_loss[loss=0.3378, simple_loss=0.3878, pruned_loss=0.1439, over 1971725.04 frames. ], batch size: 62, lr: 4.40e-02, grad_scale: 8.0 2022-12-22 11:48:33,317 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0 from training. Duration: 24.525 2022-12-22 11:48:52,296 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3675.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 11:48:53,720 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3676.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 11:49:07,520 WARNING [train.py:1060] (3/4) Exclude cut with ID 3699-47246-0007-3408-0_sp0.9 from training. Duration: 20.26675 2022-12-22 11:49:20,127 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp0.9 from training. Duration: 27.25 2022-12-22 11:49:36,037 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2022-12-22 11:49:38,189 INFO [train.py:894] (3/4) Epoch 2, batch 200, loss[loss=0.2858, simple_loss=0.3513, pruned_loss=0.1101, over 18531.00 frames. ], tot_loss[loss=0.3317, simple_loss=0.3835, pruned_loss=0.14, over 2357422.37 frames. ], batch size: 47, lr: 4.39e-02, grad_scale: 8.0 2022-12-22 11:49:41,128 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.513e+02 7.302e+02 9.000e+02 1.037e+03 2.044e+03, threshold=1.800e+03, percent-clipped=0.0 2022-12-22 11:49:43,214 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2022-12-22 11:49:57,784 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([5.8317, 4.9189, 4.8998, 5.3716, 5.0895, 4.8806, 5.2473, 1.3182], device='cuda:3'), covar=tensor([0.0366, 0.0389, 0.0472, 0.0233, 0.1289, 0.0430, 0.0487, 0.4441], device='cuda:3'), in_proj_covar=tensor([0.0105, 0.0105, 0.0101, 0.0074, 0.0139, 0.0099, 0.0105, 0.0161], device='cuda:3'), out_proj_covar=tensor([1.1711e-04, 1.1562e-04, 1.1699e-04, 8.3171e-05, 1.3873e-04, 1.1228e-04, 1.1596e-04, 1.4480e-04], device='cuda:3') 2022-12-22 11:50:02,788 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=3723.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 11:50:23,516 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3737.0, num_to_drop=2, layers_to_drop={0, 2} 2022-12-22 11:50:34,639 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0 from training. Duration: 21.68 2022-12-22 11:50:43,901 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8752, 1.8597, 1.9001, 1.7894, 2.1008, 2.2876, 1.4114, 2.4978], device='cuda:3'), covar=tensor([0.1455, 0.1508, 0.1637, 0.2195, 0.1196, 0.0921, 0.3047, 0.0614], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0130, 0.0145, 0.0158, 0.0139, 0.0131, 0.0161, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001], device='cuda:3') 2022-12-22 11:50:46,804 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0 from training. Duration: 21.6300625 2022-12-22 11:50:52,892 INFO [train.py:894] (3/4) Epoch 2, batch 250, loss[loss=0.3305, simple_loss=0.3882, pruned_loss=0.1364, over 18452.00 frames. ], tot_loss[loss=0.3293, simple_loss=0.3811, pruned_loss=0.1388, over 2657920.43 frames. ], batch size: 54, lr: 4.38e-02, grad_scale: 8.0 2022-12-22 11:51:11,831 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0_sp0.9 from training. Duration: 24.033375 2022-12-22 11:52:06,375 INFO [train.py:894] (3/4) Epoch 2, batch 300, loss[loss=0.367, simple_loss=0.4086, pruned_loss=0.1627, over 18637.00 frames. ], tot_loss[loss=0.3325, simple_loss=0.3834, pruned_loss=0.1408, over 2892050.14 frames. ], batch size: 179, lr: 4.37e-02, grad_scale: 8.0 2022-12-22 11:52:06,443 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0 from training. Duration: 22.905 2022-12-22 11:52:07,897 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp1.1 from training. Duration: 23.4318125 2022-12-22 11:52:09,151 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.194e+02 6.939e+02 8.430e+02 1.162e+03 2.242e+03, threshold=1.686e+03, percent-clipped=6.0 2022-12-22 11:53:19,701 INFO [train.py:894] (3/4) Epoch 2, batch 350, loss[loss=0.3945, simple_loss=0.4258, pruned_loss=0.1816, over 18394.00 frames. ], tot_loss[loss=0.3331, simple_loss=0.3836, pruned_loss=0.1413, over 3074307.17 frames. ], batch size: 53, lr: 4.36e-02, grad_scale: 8.0 2022-12-22 11:53:29,551 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8756, 0.6833, 2.0062, 1.0640, 1.5607, 1.8056, 1.9468, 1.8185], device='cuda:3'), covar=tensor([0.0566, 0.1869, 0.0503, 0.1652, 0.0981, 0.0799, 0.0611, 0.0834], device='cuda:3'), in_proj_covar=tensor([0.0061, 0.0097, 0.0068, 0.0104, 0.0087, 0.0063, 0.0062, 0.0086], device='cuda:3'), out_proj_covar=tensor([6.4122e-05, 9.3426e-05, 7.0507e-05, 9.9358e-05, 8.8304e-05, 6.8118e-05, 6.4494e-05, 8.5872e-05], device='cuda:3') 2022-12-22 11:54:05,299 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp1.1 from training. Duration: 20.82275 2022-12-22 11:54:06,797 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp0.9 from training. Duration: 25.45 2022-12-22 11:54:29,253 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3902.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 11:54:35,946 INFO [train.py:894] (3/4) Epoch 2, batch 400, loss[loss=0.3254, simple_loss=0.3829, pruned_loss=0.1339, over 18372.00 frames. ], tot_loss[loss=0.3349, simple_loss=0.3853, pruned_loss=0.1422, over 3215161.38 frames. ], batch size: 51, lr: 4.35e-02, grad_scale: 8.0 2022-12-22 11:54:38,882 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.591e+02 7.650e+02 9.752e+02 1.196e+03 2.094e+03, threshold=1.950e+03, percent-clipped=3.0 2022-12-22 11:55:03,902 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8202, 1.2946, 0.8954, 1.9517, 1.3284, 2.8482, 1.8398, 1.3108], device='cuda:3'), covar=tensor([0.1502, 0.2074, 0.2523, 0.1495, 0.1736, 0.0589, 0.1447, 0.2342], device='cuda:3'), in_proj_covar=tensor([0.0089, 0.0081, 0.0099, 0.0091, 0.0094, 0.0079, 0.0086, 0.0088], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:3') 2022-12-22 11:55:05,446 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3926.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 11:55:06,539 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0 from training. Duration: 25.775 2022-12-22 11:55:27,424 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.0170, 2.2747, 2.6630, 1.5370, 1.6977, 3.0091, 1.7429, 1.7836], device='cuda:3'), covar=tensor([0.1710, 0.2198, 0.2814, 0.4047, 0.2119, 0.0793, 0.2596, 0.3971], device='cuda:3'), in_proj_covar=tensor([0.0088, 0.0123, 0.0142, 0.0152, 0.0112, 0.0100, 0.0136, 0.0146], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:3') 2022-12-22 11:55:28,411 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0_sp0.9 from training. Duration: 22.25 2022-12-22 11:55:31,590 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3944.0, num_to_drop=1, layers_to_drop={2} 2022-12-22 11:55:50,780 INFO [train.py:894] (3/4) Epoch 2, batch 450, loss[loss=0.2974, simple_loss=0.3455, pruned_loss=0.1247, over 18467.00 frames. ], tot_loss[loss=0.3346, simple_loss=0.3855, pruned_loss=0.1419, over 3326103.72 frames. ], batch size: 43, lr: 4.34e-02, grad_scale: 8.0 2022-12-22 11:55:55,131 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0 from training. Duration: 26.205 2022-12-22 11:56:04,130 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.80 vs. limit=2.0 2022-12-22 11:56:11,102 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp0.9 from training. Duration: 30.1555625 2022-12-22 11:56:16,999 WARNING [train.py:1060] (3/4) Exclude cut with ID 1265-135635-0050-6781-0_sp0.9 from training. Duration: 21.8333125 2022-12-22 11:56:25,368 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp1.1 from training. Duration: 20.6545625 2022-12-22 11:56:36,303 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3987.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 11:56:43,464 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=3992.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 11:57:01,379 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8666, 1.9645, 1.0078, 1.7883, 1.7486, 1.5439, 2.7709, 1.8242], device='cuda:3'), covar=tensor([0.0866, 0.0959, 0.1715, 0.1113, 0.1807, 0.0827, 0.0200, 0.0837], device='cuda:3'), in_proj_covar=tensor([0.0072, 0.0066, 0.0080, 0.0085, 0.0102, 0.0071, 0.0040, 0.0070], device='cuda:3'), out_proj_covar=tensor([6.5773e-05, 6.3107e-05, 7.2989e-05, 7.4707e-05, 9.5981e-05, 6.1485e-05, 3.4337e-05, 6.3633e-05], device='cuda:3') 2022-12-22 11:57:09,980 INFO [train.py:894] (3/4) Epoch 2, batch 500, loss[loss=0.2883, simple_loss=0.3491, pruned_loss=0.1138, over 18517.00 frames. ], tot_loss[loss=0.3377, simple_loss=0.3877, pruned_loss=0.1439, over 3412086.18 frames. ], batch size: 47, lr: 4.33e-02, grad_scale: 8.0 2022-12-22 11:57:12,287 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0045-39920-0_sp0.9 from training. Duration: 20.52225 2022-12-22 11:57:13,499 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.346e+02 6.819e+02 8.199e+02 9.839e+02 2.413e+03, threshold=1.640e+03, percent-clipped=2.0 2022-12-22 11:57:14,635 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.67 vs. limit=5.0 2022-12-22 11:57:33,810 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp0.9 from training. Duration: 29.1166875 2022-12-22 11:57:48,762 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4032.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 11:58:28,033 INFO [train.py:894] (3/4) Epoch 2, batch 550, loss[loss=0.3379, simple_loss=0.3801, pruned_loss=0.1478, over 18379.00 frames. ], tot_loss[loss=0.3391, simple_loss=0.3884, pruned_loss=0.1449, over 3478738.41 frames. ], batch size: 46, lr: 4.32e-02, grad_scale: 8.0 2022-12-22 11:58:34,167 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133211-0007-59831-0_sp0.9 from training. Duration: 21.388875 2022-12-22 11:59:09,617 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0 from training. Duration: 22.72 2022-12-22 11:59:11,227 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0_sp0.9 from training. Duration: 22.7444375 2022-12-22 11:59:44,276 INFO [train.py:894] (3/4) Epoch 2, batch 600, loss[loss=0.3189, simple_loss=0.378, pruned_loss=0.1299, over 18594.00 frames. ], tot_loss[loss=0.3372, simple_loss=0.3868, pruned_loss=0.1438, over 3530122.09 frames. ], batch size: 51, lr: 4.31e-02, grad_scale: 8.0 2022-12-22 11:59:47,290 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.183e+02 7.677e+02 9.029e+02 1.105e+03 3.557e+03, threshold=1.806e+03, percent-clipped=7.0 2022-12-22 11:59:56,490 WARNING [train.py:1060] (3/4) Exclude cut with ID 4133-6541-0027-40495-0_sp1.1 from training. Duration: 0.9681875 2022-12-22 12:00:01,057 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0_sp0.9 from training. Duration: 22.3166875 2022-12-22 12:00:05,454 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133212-0015-59917-0_sp0.9 from training. Duration: 21.8166875 2022-12-22 12:00:35,710 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=4141.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:00:59,845 INFO [train.py:894] (3/4) Epoch 2, batch 650, loss[loss=0.4074, simple_loss=0.4357, pruned_loss=0.1895, over 18693.00 frames. ], tot_loss[loss=0.3388, simple_loss=0.3878, pruned_loss=0.1449, over 3570088.61 frames. ], batch size: 62, lr: 4.30e-02, grad_scale: 8.0 2022-12-22 12:01:03,275 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2155, 1.9558, 1.4739, 2.6302, 1.8179, 1.4893, 2.4796, 2.0430], device='cuda:3'), covar=tensor([0.0583, 0.0692, 0.1610, 0.0569, 0.0574, 0.0768, 0.0489, 0.0588], device='cuda:3'), in_proj_covar=tensor([0.0037, 0.0038, 0.0048, 0.0036, 0.0034, 0.0039, 0.0032, 0.0038], device='cuda:3'), out_proj_covar=tensor([3.3124e-05, 2.8850e-05, 4.2075e-05, 3.1613e-05, 2.8773e-05, 3.1819e-05, 2.6192e-05, 3.1655e-05], device='cuda:3') 2022-12-22 12:01:47,925 WARNING [train.py:1060] (3/4) Exclude cut with ID 4957-30119-0041-23990-0_sp0.9 from training. Duration: 20.22775 2022-12-22 12:02:10,269 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4202.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 12:02:10,375 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=4202.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:02:17,454 INFO [train.py:894] (3/4) Epoch 2, batch 700, loss[loss=0.3202, simple_loss=0.3755, pruned_loss=0.1324, over 18697.00 frames. ], tot_loss[loss=0.3385, simple_loss=0.3876, pruned_loss=0.1447, over 3602965.09 frames. ], batch size: 50, lr: 4.29e-02, grad_scale: 8.0 2022-12-22 12:02:20,234 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.497e+02 7.136e+02 8.417e+02 1.055e+03 2.568e+03, threshold=1.683e+03, percent-clipped=3.0 2022-12-22 12:02:21,251 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4794, 1.8635, 2.4887, 2.0967, 2.1603, 4.2040, 2.1507, 3.0065], device='cuda:3'), covar=tensor([0.5626, 0.3245, 0.1850, 0.3206, 0.2604, 0.0218, 0.2666, 0.1324], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0122, 0.0122, 0.0121, 0.0131, 0.0074, 0.0124, 0.0109], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-22 12:02:34,055 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp1.1 from training. Duration: 24.67275 2022-12-22 12:02:58,922 WARNING [train.py:1060] (3/4) Exclude cut with ID 3082-165428-0081-50734-0_sp0.9 from training. Duration: 21.8055625 2022-12-22 12:03:16,328 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2022-12-22 12:03:24,562 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=4250.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 12:03:34,916 INFO [train.py:894] (3/4) Epoch 2, batch 750, loss[loss=0.3171, simple_loss=0.3783, pruned_loss=0.128, over 18631.00 frames. ], tot_loss[loss=0.3359, simple_loss=0.386, pruned_loss=0.1429, over 3627195.19 frames. ], batch size: 53, lr: 4.28e-02, grad_scale: 8.0 2022-12-22 12:03:39,633 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0_sp0.9 from training. Duration: 22.6666875 2022-12-22 12:04:03,676 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8710, 2.2253, 1.2655, 2.3466, 2.8560, 1.3147, 2.0490, 0.8704], device='cuda:3'), covar=tensor([0.1360, 0.1238, 0.1245, 0.0814, 0.0702, 0.1193, 0.1169, 0.1564], device='cuda:3'), in_proj_covar=tensor([0.0120, 0.0093, 0.0098, 0.0091, 0.0098, 0.0098, 0.0087, 0.0100], device='cuda:3'), out_proj_covar=tensor([1.1867e-04, 9.6182e-05, 9.5857e-05, 9.1261e-05, 9.7257e-05, 9.5707e-05, 9.0361e-05, 9.5828e-05], device='cuda:3') 2022-12-22 12:04:12,161 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4282.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:04:22,691 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=4288.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:04:43,628 WARNING [train.py:1060] (3/4) Exclude cut with ID 2411-132532-0017-82279-0_sp1.1 from training. Duration: 0.9681875 2022-12-22 12:04:51,497 INFO [train.py:894] (3/4) Epoch 2, batch 800, loss[loss=0.3233, simple_loss=0.3931, pruned_loss=0.1267, over 18462.00 frames. ], tot_loss[loss=0.336, simple_loss=0.3865, pruned_loss=0.1428, over 3646629.90 frames. ], batch size: 54, lr: 4.27e-02, grad_scale: 8.0 2022-12-22 12:04:54,410 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.666e+02 6.822e+02 8.702e+02 1.097e+03 2.358e+03, threshold=1.740e+03, percent-clipped=4.0 2022-12-22 12:05:09,487 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0 from training. Duration: 22.485 2022-12-22 12:05:14,464 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4193, 1.4653, 1.0580, 1.6035, 1.3503, 1.1276, 1.6297, 1.5069], device='cuda:3'), covar=tensor([0.0818, 0.0976, 0.1403, 0.0868, 0.1657, 0.0841, 0.0395, 0.0760], device='cuda:3'), in_proj_covar=tensor([0.0076, 0.0073, 0.0083, 0.0096, 0.0106, 0.0076, 0.0044, 0.0075], device='cuda:3'), out_proj_covar=tensor([7.1245e-05, 7.2676e-05, 7.7973e-05, 8.6921e-05, 1.0207e-04, 6.7826e-05, 4.0473e-05, 6.9757e-05], device='cuda:3') 2022-12-22 12:05:28,819 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4332.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 12:05:49,998 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp1.1 from training. Duration: 23.82275 2022-12-22 12:05:55,901 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=4349.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:06:03,202 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0 from training. Duration: 20.77 2022-12-22 12:06:07,420 INFO [train.py:894] (3/4) Epoch 2, batch 850, loss[loss=0.3393, simple_loss=0.3877, pruned_loss=0.1455, over 18581.00 frames. ], tot_loss[loss=0.3359, simple_loss=0.3867, pruned_loss=0.1426, over 3661983.81 frames. ], batch size: 51, lr: 4.26e-02, grad_scale: 8.0 2022-12-22 12:06:10,562 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0_sp0.9 from training. Duration: 24.088875 2022-12-22 12:06:28,728 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([5.3944, 4.7842, 4.6959, 4.7404, 4.7672, 4.7402, 5.1524, 1.2258], device='cuda:3'), covar=tensor([0.0554, 0.0401, 0.0524, 0.0311, 0.1547, 0.0501, 0.0549, 0.4867], device='cuda:3'), in_proj_covar=tensor([0.0116, 0.0106, 0.0104, 0.0075, 0.0150, 0.0105, 0.0100, 0.0168], device='cuda:3'), out_proj_covar=tensor([1.3061e-04, 1.1742e-04, 1.2185e-04, 8.7074e-05, 1.5046e-04, 1.1798e-04, 1.0993e-04, 1.5291e-04], device='cuda:3') 2022-12-22 12:06:34,974 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2022-12-22 12:06:41,371 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp1.1 from training. Duration: 20.4409375 2022-12-22 12:06:41,514 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=4380.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 12:07:23,241 INFO [train.py:894] (3/4) Epoch 2, batch 900, loss[loss=0.3257, simple_loss=0.3784, pruned_loss=0.1365, over 18624.00 frames. ], tot_loss[loss=0.3332, simple_loss=0.3843, pruned_loss=0.141, over 3673089.95 frames. ], batch size: 53, lr: 4.25e-02, grad_scale: 8.0 2022-12-22 12:07:26,209 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.602e+02 7.350e+02 9.194e+02 1.218e+03 2.503e+03, threshold=1.839e+03, percent-clipped=6.0 2022-12-22 12:07:34,228 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5106, 0.8698, 1.0019, 1.6373, 1.3541, 1.2056, 1.5420, 1.2040], device='cuda:3'), covar=tensor([0.0690, 0.0991, 0.2049, 0.0683, 0.0627, 0.0833, 0.0601, 0.0808], device='cuda:3'), in_proj_covar=tensor([0.0044, 0.0048, 0.0063, 0.0044, 0.0042, 0.0048, 0.0040, 0.0046], device='cuda:3'), out_proj_covar=tensor([4.0412e-05, 3.7291e-05, 5.5551e-05, 4.0691e-05, 3.5538e-05, 4.0341e-05, 3.4200e-05, 3.9405e-05], device='cuda:3') 2022-12-22 12:08:00,133 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0_sp0.9 from training. Duration: 22.511125 2022-12-22 12:08:00,159 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0 from training. Duration: 20.675 2022-12-22 12:08:39,521 INFO [train.py:894] (3/4) Epoch 2, batch 950, loss[loss=0.3266, simple_loss=0.3916, pruned_loss=0.1307, over 18507.00 frames. ], tot_loss[loss=0.3339, simple_loss=0.385, pruned_loss=0.1414, over 3681983.51 frames. ], batch size: 52, lr: 4.24e-02, grad_scale: 8.0 2022-12-22 12:09:40,297 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp0.9 from training. Duration: 24.9833125 2022-12-22 12:09:41,868 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4497.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:09:56,180 INFO [train.py:894] (3/4) Epoch 2, batch 1000, loss[loss=0.342, simple_loss=0.3968, pruned_loss=0.1436, over 18496.00 frames. ], tot_loss[loss=0.3304, simple_loss=0.3827, pruned_loss=0.1391, over 3689258.08 frames. ], batch size: 52, lr: 4.23e-02, grad_scale: 8.0 2022-12-22 12:09:59,100 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.331e+02 6.849e+02 8.394e+02 1.021e+03 1.875e+03, threshold=1.679e+03, percent-clipped=1.0 2022-12-22 12:10:11,286 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0 from training. Duration: 27.14 2022-12-22 12:10:27,475 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0 from training. Duration: 22.44 2022-12-22 12:10:29,381 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4383, 1.5911, 1.1775, 1.6881, 2.2528, 1.1769, 1.3666, 0.8570], device='cuda:3'), covar=tensor([0.1478, 0.1328, 0.1238, 0.0895, 0.0706, 0.1185, 0.1262, 0.1576], device='cuda:3'), in_proj_covar=tensor([0.0127, 0.0100, 0.0104, 0.0094, 0.0109, 0.0104, 0.0093, 0.0107], device='cuda:3'), out_proj_covar=tensor([1.2597e-04, 1.0518e-04, 1.0254e-04, 9.6257e-05, 1.1034e-04, 1.0376e-04, 9.8853e-05, 1.0379e-04], device='cuda:3') 2022-12-22 12:11:12,766 INFO [train.py:894] (3/4) Epoch 2, batch 1050, loss[loss=0.2992, simple_loss=0.3412, pruned_loss=0.1286, over 18533.00 frames. ], tot_loss[loss=0.3288, simple_loss=0.3813, pruned_loss=0.1381, over 3694427.51 frames. ], batch size: 44, lr: 4.22e-02, grad_scale: 8.0 2022-12-22 12:11:45,480 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0060-62364-0_sp0.9 from training. Duration: 21.361125 2022-12-22 12:11:50,919 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4582.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:11:52,085 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp1.1 from training. Duration: 27.0318125 2022-12-22 12:12:00,678 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp0.9 from training. Duration: 28.638875 2022-12-22 12:12:15,323 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0 from training. Duration: 20.4 2022-12-22 12:12:22,044 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8110, 2.1015, 1.1558, 1.7353, 1.6510, 1.5920, 1.7838, 2.2060], device='cuda:3'), covar=tensor([0.1443, 0.0762, 0.1151, 0.1210, 0.0932, 0.0696, 0.1245, 0.0483], device='cuda:3'), in_proj_covar=tensor([0.0092, 0.0056, 0.0068, 0.0080, 0.0061, 0.0066, 0.0071, 0.0057], device='cuda:3'), out_proj_covar=tensor([9.2495e-05, 5.6071e-05, 6.2092e-05, 7.5089e-05, 6.4176e-05, 6.1505e-05, 7.3421e-05, 5.7800e-05], device='cuda:3') 2022-12-22 12:12:30,561 INFO [train.py:894] (3/4) Epoch 2, batch 1100, loss[loss=0.3091, simple_loss=0.3694, pruned_loss=0.1245, over 18586.00 frames. ], tot_loss[loss=0.3275, simple_loss=0.3803, pruned_loss=0.1373, over 3698455.30 frames. ], batch size: 51, lr: 4.20e-02, grad_scale: 8.0 2022-12-22 12:12:33,503 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.704e+02 6.945e+02 8.446e+02 1.102e+03 2.369e+03, threshold=1.689e+03, percent-clipped=6.0 2022-12-22 12:12:47,122 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0 from training. Duration: 20.025 2022-12-22 12:12:47,131 WARNING [train.py:1060] (3/4) Exclude cut with ID 2364-131735-0112-64612-0_sp0.9 from training. Duration: 20.488875 2022-12-22 12:12:53,629 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0 from training. Duration: 29.735 2022-12-22 12:13:06,190 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=4630.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:13:27,774 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4644.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:13:47,614 INFO [train.py:894] (3/4) Epoch 2, batch 1150, loss[loss=0.2876, simple_loss=0.3397, pruned_loss=0.1177, over 18606.00 frames. ], tot_loss[loss=0.3255, simple_loss=0.3783, pruned_loss=0.1363, over 3701629.26 frames. ], batch size: 45, lr: 4.19e-02, grad_scale: 8.0 2022-12-22 12:14:19,141 WARNING [train.py:1060] (3/4) Exclude cut with ID 7276-92427-0014-12983-0_sp0.9 from training. Duration: 21.3055625 2022-12-22 12:14:19,193 WARNING [train.py:1060] (3/4) Exclude cut with ID 1025-75365-0008-79168-0_sp0.9 from training. Duration: 22.0666875 2022-12-22 12:15:04,744 INFO [train.py:894] (3/4) Epoch 2, batch 1200, loss[loss=0.328, simple_loss=0.3641, pruned_loss=0.146, over 18603.00 frames. ], tot_loss[loss=0.3231, simple_loss=0.3761, pruned_loss=0.135, over 3704324.01 frames. ], batch size: 45, lr: 4.18e-02, grad_scale: 8.0 2022-12-22 12:15:07,492 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.237e+02 6.724e+02 8.402e+02 1.025e+03 3.359e+03, threshold=1.680e+03, percent-clipped=3.0 2022-12-22 12:15:26,332 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.85 vs. limit=2.0 2022-12-22 12:16:10,643 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0 from training. Duration: 20.26 2022-12-22 12:16:22,818 INFO [train.py:894] (3/4) Epoch 2, batch 1250, loss[loss=0.3174, simple_loss=0.3786, pruned_loss=0.1281, over 18545.00 frames. ], tot_loss[loss=0.3225, simple_loss=0.3754, pruned_loss=0.1348, over 3705478.67 frames. ], batch size: 58, lr: 4.17e-02, grad_scale: 8.0 2022-12-22 12:16:24,279 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0030-9324-0_sp0.9 from training. Duration: 21.3444375 2022-12-22 12:17:14,784 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2022-12-22 12:17:22,516 WARNING [train.py:1060] (3/4) Exclude cut with ID 497-129325-0061-62254-0_sp1.1 from training. Duration: 0.97725 2022-12-22 12:17:24,351 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4797.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:17:35,552 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2022-12-22 12:17:39,049 INFO [train.py:894] (3/4) Epoch 2, batch 1300, loss[loss=0.3535, simple_loss=0.403, pruned_loss=0.152, over 18689.00 frames. ], tot_loss[loss=0.3239, simple_loss=0.3762, pruned_loss=0.1358, over 3707713.72 frames. ], batch size: 69, lr: 4.16e-02, grad_scale: 8.0 2022-12-22 12:17:42,186 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.728e+02 7.714e+02 8.750e+02 1.105e+03 2.805e+03, threshold=1.750e+03, percent-clipped=4.0 2022-12-22 12:18:03,982 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0_sp0.9 from training. Duration: 22.97225 2022-12-22 12:18:04,134 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4803, 3.5295, 3.4375, 1.7389, 3.4390, 2.1885, 0.9615, 2.0728], device='cuda:3'), covar=tensor([0.1976, 0.0620, 0.1542, 0.3018, 0.1000, 0.1936, 0.5529, 0.2526], device='cuda:3'), in_proj_covar=tensor([0.0096, 0.0063, 0.0127, 0.0091, 0.0073, 0.0083, 0.0129, 0.0098], device='cuda:3'), out_proj_covar=tensor([1.2762e-04, 8.3644e-05, 1.6806e-04, 1.1060e-04, 1.0250e-04, 1.1053e-04, 1.4187e-04, 1.2769e-04], device='cuda:3') 2022-12-22 12:18:12,486 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3529, 2.6483, 2.5396, 1.5419, 2.7178, 1.7383, 0.9051, 1.6487], device='cuda:3'), covar=tensor([0.1701, 0.0823, 0.1923, 0.2727, 0.1009, 0.1835, 0.4738, 0.2448], device='cuda:3'), in_proj_covar=tensor([0.0095, 0.0062, 0.0126, 0.0090, 0.0072, 0.0083, 0.0128, 0.0097], device='cuda:3'), out_proj_covar=tensor([1.2663e-04, 8.3606e-05, 1.6719e-04, 1.0939e-04, 1.0195e-04, 1.0980e-04, 1.4098e-04, 1.2658e-04], device='cuda:3') 2022-12-22 12:18:35,158 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0047-39922-0_sp0.9 from training. Duration: 21.97775 2022-12-22 12:18:38,283 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=4845.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:18:49,865 WARNING [train.py:1060] (3/4) Exclude cut with ID 1112-1043-0006-89194-0_sp0.9 from training. Duration: 21.8333125 2022-12-22 12:18:55,618 INFO [train.py:894] (3/4) Epoch 2, batch 1350, loss[loss=0.3047, simple_loss=0.3548, pruned_loss=0.1273, over 18660.00 frames. ], tot_loss[loss=0.3222, simple_loss=0.3749, pruned_loss=0.1347, over 3708785.49 frames. ], batch size: 46, lr: 4.15e-02, grad_scale: 8.0 2022-12-22 12:19:00,308 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0 from training. Duration: 20.47 2022-12-22 12:19:45,276 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.92 vs. limit=2.0 2022-12-22 12:20:08,125 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0037-39912-0_sp0.9 from training. Duration: 20.67225 2022-12-22 12:20:12,449 INFO [train.py:894] (3/4) Epoch 2, batch 1400, loss[loss=0.3017, simple_loss=0.3425, pruned_loss=0.1304, over 18565.00 frames. ], tot_loss[loss=0.3207, simple_loss=0.3739, pruned_loss=0.1337, over 3710085.31 frames. ], batch size: 44, lr: 4.14e-02, grad_scale: 8.0 2022-12-22 12:20:16,046 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.445e+02 6.272e+02 7.556e+02 9.448e+02 1.744e+03, threshold=1.511e+03, percent-clipped=0.0 2022-12-22 12:20:26,259 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp0.9 from training. Duration: 25.2444375 2022-12-22 12:20:49,156 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2022-12-22 12:20:52,485 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0021-76797-0_sp0.9 from training. Duration: 21.1445 2022-12-22 12:21:08,968 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4944.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:21:28,044 INFO [train.py:894] (3/4) Epoch 2, batch 1450, loss[loss=0.267, simple_loss=0.3184, pruned_loss=0.1078, over 18562.00 frames. ], tot_loss[loss=0.3202, simple_loss=0.3734, pruned_loss=0.1335, over 3710695.88 frames. ], batch size: 41, lr: 4.13e-02, grad_scale: 8.0 2022-12-22 12:22:06,349 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp0.9 from training. Duration: 33.038875 2022-12-22 12:22:21,206 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=4992.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:22:44,248 INFO [train.py:894] (3/4) Epoch 2, batch 1500, loss[loss=0.2899, simple_loss=0.3331, pruned_loss=0.1234, over 18405.00 frames. ], tot_loss[loss=0.3186, simple_loss=0.3719, pruned_loss=0.1327, over 3710961.25 frames. ], batch size: 42, lr: 4.12e-02, grad_scale: 8.0 2022-12-22 12:22:44,292 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64291-0000-16059-0_sp0.9 from training. Duration: 20.0944375 2022-12-22 12:22:47,216 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.674e+02 6.252e+02 7.708e+02 1.003e+03 2.600e+03, threshold=1.542e+03, percent-clipped=7.0 2022-12-22 12:22:57,570 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp1.1 from training. Duration: 20.4 2022-12-22 12:23:00,002 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([5.9038, 5.2243, 5.1864, 5.4362, 5.5299, 5.2121, 5.6899, 1.4164], device='cuda:3'), covar=tensor([0.0335, 0.0272, 0.0330, 0.0194, 0.0919, 0.0377, 0.0271, 0.3794], device='cuda:3'), in_proj_covar=tensor([0.0124, 0.0108, 0.0102, 0.0077, 0.0150, 0.0111, 0.0100, 0.0166], device='cuda:3'), out_proj_covar=tensor([1.3949e-04, 1.1799e-04, 1.1988e-04, 9.0219e-05, 1.5249e-04, 1.2309e-04, 1.0966e-04, 1.5459e-04], device='cuda:3') 2022-12-22 12:23:04,749 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0 from training. Duration: 20.085 2022-12-22 12:23:17,186 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0_sp0.9 from training. Duration: 23.07775 2022-12-22 12:23:59,411 INFO [train.py:894] (3/4) Epoch 2, batch 1550, loss[loss=0.3452, simple_loss=0.3984, pruned_loss=0.146, over 18667.00 frames. ], tot_loss[loss=0.3188, simple_loss=0.3726, pruned_loss=0.1326, over 3711604.61 frames. ], batch size: 78, lr: 4.11e-02, grad_scale: 8.0 2022-12-22 12:24:02,447 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp0.9 from training. Duration: 24.9333125 2022-12-22 12:24:24,246 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6420, 2.3772, 1.6768, 2.2519, 1.7574, 2.0777, 2.0403, 2.8188], device='cuda:3'), covar=tensor([0.1202, 0.0817, 0.0953, 0.1311, 0.1068, 0.0571, 0.1445, 0.0437], device='cuda:3'), in_proj_covar=tensor([0.0102, 0.0064, 0.0074, 0.0091, 0.0068, 0.0071, 0.0082, 0.0064], device='cuda:3'), out_proj_covar=tensor([1.0406e-04, 6.7899e-05, 6.8763e-05, 8.7848e-05, 7.5487e-05, 6.8721e-05, 8.7280e-05, 6.8027e-05], device='cuda:3') 2022-12-22 12:24:34,438 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4700, 1.3347, 0.9366, 1.3041, 1.2619, 1.2926, 1.9481, 1.5121], device='cuda:3'), covar=tensor([0.0879, 0.1071, 0.1561, 0.1177, 0.1807, 0.0793, 0.0307, 0.0867], device='cuda:3'), in_proj_covar=tensor([0.0092, 0.0087, 0.0101, 0.0119, 0.0124, 0.0089, 0.0050, 0.0089], device='cuda:3'), out_proj_covar=tensor([9.1762e-05, 9.0275e-05, 9.8016e-05, 1.1451e-04, 1.2138e-04, 8.5273e-05, 5.1565e-05, 8.7337e-05], device='cuda:3') 2022-12-22 12:24:45,949 WARNING [train.py:1060] (3/4) Exclude cut with ID 5118-111612-0016-124680-0_sp0.9 from training. Duration: 20.388875 2022-12-22 12:24:53,298 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp1.1 from training. Duration: 20.3590625 2022-12-22 12:25:15,875 INFO [train.py:894] (3/4) Epoch 2, batch 1600, loss[loss=0.4077, simple_loss=0.4305, pruned_loss=0.1925, over 18591.00 frames. ], tot_loss[loss=0.3193, simple_loss=0.3729, pruned_loss=0.1328, over 3711298.94 frames. ], batch size: 56, lr: 4.10e-02, grad_scale: 8.0 2022-12-22 12:25:20,225 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.493e+02 7.158e+02 8.150e+02 1.051e+03 2.578e+03, threshold=1.630e+03, percent-clipped=2.0 2022-12-22 12:26:02,096 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0_sp1.1 from training. Duration: 0.836375 2022-12-22 12:26:31,558 INFO [train.py:894] (3/4) Epoch 2, batch 1650, loss[loss=0.3159, simple_loss=0.3538, pruned_loss=0.1391, over 18559.00 frames. ], tot_loss[loss=0.3217, simple_loss=0.3733, pruned_loss=0.135, over 3711472.70 frames. ], batch size: 41, lr: 4.09e-02, grad_scale: 8.0 2022-12-22 12:26:48,289 WARNING [train.py:1060] (3/4) Exclude cut with ID 8565-290391-0049-67394-0_sp0.9 from training. Duration: 21.3166875 2022-12-22 12:27:09,884 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.91 vs. limit=5.0 2022-12-22 12:27:19,610 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0029-104863-0_sp0.9 from training. Duration: 22.1055625 2022-12-22 12:27:28,603 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp1.1 from training. Duration: 21.77725 2022-12-22 12:27:49,270 INFO [train.py:894] (3/4) Epoch 2, batch 1700, loss[loss=0.4075, simple_loss=0.4341, pruned_loss=0.1904, over 18607.00 frames. ], tot_loss[loss=0.3278, simple_loss=0.3765, pruned_loss=0.1395, over 3711273.42 frames. ], batch size: 69, lr: 4.08e-02, grad_scale: 8.0 2022-12-22 12:27:49,357 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp0.9 from training. Duration: 27.8166875 2022-12-22 12:27:53,990 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.221e+02 7.752e+02 9.535e+02 1.253e+03 3.396e+03, threshold=1.907e+03, percent-clipped=8.0 2022-12-22 12:28:16,144 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp1.1 from training. Duration: 22.5090625 2022-12-22 12:28:21,938 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0 from training. Duration: 25.035 2022-12-22 12:28:41,506 WARNING [train.py:1060] (3/4) Exclude cut with ID 774-127930-0014-10412-0_sp1.1 from training. Duration: 0.95 2022-12-22 12:28:59,824 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp0.9 from training. Duration: 0.92225 2022-12-22 12:29:05,315 INFO [train.py:894] (3/4) Epoch 2, batch 1750, loss[loss=0.3988, simple_loss=0.4292, pruned_loss=0.1842, over 18653.00 frames. ], tot_loss[loss=0.3317, simple_loss=0.3776, pruned_loss=0.1428, over 3712288.52 frames. ], batch size: 98, lr: 4.07e-02, grad_scale: 8.0 2022-12-22 12:29:23,475 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4876, 1.8063, 1.8791, 1.7330, 1.9243, 3.4246, 1.4910, 2.2114], device='cuda:3'), covar=tensor([0.4476, 0.2757, 0.1849, 0.3219, 0.2235, 0.0292, 0.2428, 0.1610], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0125, 0.0129, 0.0126, 0.0135, 0.0080, 0.0121, 0.0116], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-22 12:29:26,169 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0 from training. Duration: 21.97 2022-12-22 12:29:45,524 WARNING [train.py:1060] (3/4) Exclude cut with ID 7492-105653-0055-62765-0_sp0.9 from training. Duration: 21.97225 2022-12-22 12:29:46,907 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp0.9 from training. Duration: 25.3333125 2022-12-22 12:29:59,189 WARNING [train.py:1060] (3/4) Exclude cut with ID 5172-29468-0015-19128-0_sp0.9 from training. Duration: 21.5055625 2022-12-22 12:30:09,167 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp1.1 from training. Duration: 20.72725 2022-12-22 12:30:20,454 INFO [train.py:894] (3/4) Epoch 2, batch 1800, loss[loss=0.3909, simple_loss=0.4238, pruned_loss=0.179, over 18681.00 frames. ], tot_loss[loss=0.3388, simple_loss=0.3819, pruned_loss=0.1479, over 3714046.84 frames. ], batch size: 69, lr: 4.06e-02, grad_scale: 8.0 2022-12-22 12:30:25,429 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.783e+02 7.590e+02 9.192e+02 1.099e+03 2.002e+03, threshold=1.838e+03, percent-clipped=3.0 2022-12-22 12:30:41,032 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp0.9 from training. Duration: 26.32775 2022-12-22 12:30:47,431 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7809, 0.5714, 1.6918, 1.1893, 1.6669, 1.8949, 2.0201, 1.4806], device='cuda:3'), covar=tensor([0.0873, 0.1525, 0.0644, 0.1240, 0.0716, 0.0555, 0.0505, 0.0749], device='cuda:3'), in_proj_covar=tensor([0.0061, 0.0103, 0.0083, 0.0120, 0.0095, 0.0065, 0.0066, 0.0097], device='cuda:3'), out_proj_covar=tensor([7.0609e-05, 1.0383e-04, 9.0516e-05, 1.2190e-04, 9.9136e-05, 7.5652e-05, 7.3986e-05, 9.8153e-05], device='cuda:3') 2022-12-22 12:31:15,460 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0 from training. Duration: 20.025 2022-12-22 12:31:21,325 WARNING [train.py:1060] (3/4) Exclude cut with ID 6709-74022-0004-86860-0_sp1.1 from training. Duration: 0.9409375 2022-12-22 12:31:21,335 WARNING [train.py:1060] (3/4) Exclude cut with ID 4757-1811-0023-62229-0_sp0.9 from training. Duration: 21.37775 2022-12-22 12:31:36,691 INFO [train.py:894] (3/4) Epoch 2, batch 1850, loss[loss=0.3531, simple_loss=0.3794, pruned_loss=0.1634, over 18455.00 frames. ], tot_loss[loss=0.3447, simple_loss=0.385, pruned_loss=0.1522, over 3713708.36 frames. ], batch size: 50, lr: 4.05e-02, grad_scale: 8.0 2022-12-22 12:31:40,269 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0004-25974-0_sp0.9 from training. Duration: 21.17225 2022-12-22 12:31:40,279 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp0.9 from training. Duration: 27.511125 2022-12-22 12:32:10,912 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3987, 1.7184, 1.6856, 1.1802, 1.8750, 1.8705, 1.3640, 1.7704], device='cuda:3'), covar=tensor([0.0854, 0.0907, 0.1556, 0.1999, 0.1118, 0.0988, 0.1605, 0.1433], device='cuda:3'), in_proj_covar=tensor([0.0098, 0.0129, 0.0163, 0.0165, 0.0141, 0.0123, 0.0136, 0.0160], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-22 12:32:12,016 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0 from training. Duration: 22.8 2022-12-22 12:32:18,030 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0 from training. Duration: 22.585 2022-12-22 12:32:23,646 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.80 vs. limit=2.0 2022-12-22 12:32:48,611 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0001-146967-0_sp0.9 from training. Duration: 22.0166875 2022-12-22 12:32:54,492 INFO [train.py:894] (3/4) Epoch 2, batch 1900, loss[loss=0.2764, simple_loss=0.3186, pruned_loss=0.1172, over 18487.00 frames. ], tot_loss[loss=0.3477, simple_loss=0.386, pruned_loss=0.1547, over 3714068.34 frames. ], batch size: 43, lr: 4.04e-02, grad_scale: 8.0 2022-12-22 12:32:59,281 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 5.057e+02 7.619e+02 8.990e+02 1.158e+03 2.314e+03, threshold=1.798e+03, percent-clipped=4.0 2022-12-22 12:33:04,365 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp1.1 from training. Duration: 24.395375 2022-12-22 12:33:10,226 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp0.9 from training. Duration: 27.47775 2022-12-22 12:33:15,978 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp0.9 from training. Duration: 24.8833125 2022-12-22 12:33:19,063 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0 from training. Duration: 23.39 2022-12-22 12:33:24,879 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp0.9 from training. Duration: 28.72225 2022-12-22 12:33:35,391 WARNING [train.py:1060] (3/4) Exclude cut with ID 585-294811-0110-133686-0_sp0.9 from training. Duration: 20.8944375 2022-12-22 12:33:50,036 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0_sp0.9 from training. Duration: 23.8444375 2022-12-22 12:34:11,304 INFO [train.py:894] (3/4) Epoch 2, batch 1950, loss[loss=0.3594, simple_loss=0.3893, pruned_loss=0.1648, over 18457.00 frames. ], tot_loss[loss=0.3489, simple_loss=0.3861, pruned_loss=0.1559, over 3713672.01 frames. ], batch size: 50, lr: 4.03e-02, grad_scale: 8.0 2022-12-22 12:34:17,344 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0 from training. Duration: 25.85 2022-12-22 12:34:17,354 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0 from training. Duration: 21.39 2022-12-22 12:34:27,522 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0 from training. Duration: 27.92 2022-12-22 12:34:56,295 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0039-130165-0_sp0.9 from training. Duration: 20.661125 2022-12-22 12:35:19,846 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0043-15874-0_sp0.9 from training. Duration: 20.07225 2022-12-22 12:35:27,423 INFO [train.py:894] (3/4) Epoch 2, batch 2000, loss[loss=0.321, simple_loss=0.3742, pruned_loss=0.1339, over 18521.00 frames. ], tot_loss[loss=0.354, simple_loss=0.3894, pruned_loss=0.1593, over 3714643.75 frames. ], batch size: 77, lr: 4.02e-02, grad_scale: 8.0 2022-12-22 12:35:28,915 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0 from training. Duration: 21.01 2022-12-22 12:35:31,583 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.702e+02 7.346e+02 8.906e+02 1.153e+03 2.441e+03, threshold=1.781e+03, percent-clipped=4.0 2022-12-22 12:36:38,591 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0 from training. Duration: 20.65 2022-12-22 12:36:42,812 INFO [train.py:894] (3/4) Epoch 2, batch 2050, loss[loss=0.3846, simple_loss=0.41, pruned_loss=0.1796, over 18670.00 frames. ], tot_loss[loss=0.3541, simple_loss=0.3891, pruned_loss=0.1595, over 3714811.19 frames. ], batch size: 62, lr: 4.01e-02, grad_scale: 8.0 2022-12-22 12:36:45,670 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0 from training. Duration: 21.46 2022-12-22 12:36:59,478 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8448, 0.5560, 1.6007, 1.2016, 1.4190, 1.6726, 1.8324, 1.3380], device='cuda:3'), covar=tensor([0.1013, 0.1657, 0.0735, 0.1370, 0.0920, 0.0757, 0.0590, 0.1059], device='cuda:3'), in_proj_covar=tensor([0.0062, 0.0107, 0.0085, 0.0123, 0.0098, 0.0067, 0.0065, 0.0099], device='cuda:3'), out_proj_covar=tensor([7.1119e-05, 1.0852e-04, 9.2729e-05, 1.2565e-04, 1.0217e-04, 7.8395e-05, 7.3786e-05, 1.0179e-04], device='cuda:3') 2022-12-22 12:37:33,515 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0 from training. Duration: 0.92 2022-12-22 12:37:41,197 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0_sp0.9 from training. Duration: 23.7666875 2022-12-22 12:38:00,488 INFO [train.py:894] (3/4) Epoch 2, batch 2100, loss[loss=0.3278, simple_loss=0.3634, pruned_loss=0.1461, over 18664.00 frames. ], tot_loss[loss=0.3528, simple_loss=0.3882, pruned_loss=0.1587, over 3715633.37 frames. ], batch size: 41, lr: 4.00e-02, grad_scale: 8.0 2022-12-22 12:38:04,908 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.064e+02 6.944e+02 8.737e+02 1.108e+03 7.286e+03, threshold=1.747e+03, percent-clipped=11.0 2022-12-22 12:38:17,065 WARNING [train.py:1060] (3/4) Exclude cut with ID 8544-281189-0060-101339-0_sp0.9 from training. Duration: 20.861125 2022-12-22 12:38:22,510 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.1828, 1.4885, 1.4494, 0.1917, 1.0507, 1.3547, 1.0499, 1.3408], device='cuda:3'), covar=tensor([0.1316, 0.0707, 0.0763, 0.1471, 0.0811, 0.0534, 0.1183, 0.0837], device='cuda:3'), in_proj_covar=tensor([0.0085, 0.0087, 0.0064, 0.0102, 0.0079, 0.0055, 0.0097, 0.0071], device='cuda:3'), out_proj_covar=tensor([9.1280e-05, 9.7359e-05, 8.0372e-05, 1.1249e-04, 9.2168e-05, 6.5024e-05, 1.1039e-04, 8.1225e-05], device='cuda:3') 2022-12-22 12:38:27,876 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0_sp0.9 from training. Duration: 22.711125 2022-12-22 12:39:06,134 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5650.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 12:39:08,835 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp1.1 from training. Duration: 22.986375 2022-12-22 12:39:15,889 INFO [train.py:894] (3/4) Epoch 2, batch 2150, loss[loss=0.3538, simple_loss=0.3904, pruned_loss=0.1586, over 18669.00 frames. ], tot_loss[loss=0.3527, simple_loss=0.3879, pruned_loss=0.1588, over 3715651.82 frames. ], batch size: 62, lr: 3.99e-02, grad_scale: 8.0 2022-12-22 12:39:19,436 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6923, 1.5300, 1.3896, 1.5044, 1.5219, 1.3743, 1.5816, 2.0754], device='cuda:3'), covar=tensor([0.2533, 0.2370, 0.3047, 0.2559, 0.3282, 0.2714, 0.2464, 0.2045], device='cuda:3'), in_proj_covar=tensor([0.0083, 0.0097, 0.0130, 0.0095, 0.0101, 0.0100, 0.0085, 0.0092], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 12:39:24,885 WARNING [train.py:1060] (3/4) Exclude cut with ID 8040-260924-0003-80960-0_sp0.9 from training. Duration: 22.07225 2022-12-22 12:39:25,383 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.2238, 1.4412, 1.4581, 0.1492, 1.3325, 1.5238, 1.5624, 1.2573], device='cuda:3'), covar=tensor([0.1169, 0.0717, 0.0735, 0.1236, 0.0578, 0.0369, 0.0640, 0.0686], device='cuda:3'), in_proj_covar=tensor([0.0083, 0.0085, 0.0065, 0.0100, 0.0077, 0.0055, 0.0095, 0.0071], device='cuda:3'), out_proj_covar=tensor([8.8821e-05, 9.5483e-05, 8.0829e-05, 1.1071e-04, 8.9427e-05, 6.4630e-05, 1.0763e-04, 8.0661e-05], device='cuda:3') 2022-12-22 12:39:29,388 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0045-26330-0_sp0.9 from training. Duration: 20.3055625 2022-12-22 12:39:32,980 WARNING [train.py:1060] (3/4) Exclude cut with ID 6356-271890-0060-94317-0_sp0.9 from training. Duration: 20.72225 2022-12-22 12:39:52,256 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp1.1 from training. Duration: 22.4818125 2022-12-22 12:40:17,809 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp0.9 from training. Duration: 25.0944375 2022-12-22 12:40:21,788 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0 from training. Duration: 21.515 2022-12-22 12:40:28,897 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp0.9 from training. Duration: 27.02225 2022-12-22 12:40:31,691 INFO [train.py:894] (3/4) Epoch 2, batch 2200, loss[loss=0.3393, simple_loss=0.3694, pruned_loss=0.1546, over 18678.00 frames. ], tot_loss[loss=0.3505, simple_loss=0.3867, pruned_loss=0.1571, over 3715872.29 frames. ], batch size: 46, lr: 3.98e-02, grad_scale: 8.0 2022-12-22 12:40:34,689 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0010-62480-0_sp0.9 from training. Duration: 22.22225 2022-12-22 12:40:36,069 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.664e+02 7.690e+02 8.741e+02 1.076e+03 1.668e+03, threshold=1.748e+03, percent-clipped=0.0 2022-12-22 12:40:38,051 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5711.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 12:40:41,880 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0085-44554-0_sp0.9 from training. Duration: 20.85 2022-12-22 12:41:15,290 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0 from training. Duration: 21.54 2022-12-22 12:41:20,991 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp1.1 from training. Duration: 20.5318125 2022-12-22 12:41:29,877 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0012-134311-0_sp0.9 from training. Duration: 21.9333125 2022-12-22 12:41:44,797 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5755.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:41:47,255 INFO [train.py:894] (3/4) Epoch 2, batch 2250, loss[loss=0.4433, simple_loss=0.4494, pruned_loss=0.2186, over 18675.00 frames. ], tot_loss[loss=0.3514, simple_loss=0.3876, pruned_loss=0.1576, over 3715324.76 frames. ], batch size: 98, lr: 3.97e-02, grad_scale: 8.0 2022-12-22 12:42:07,891 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 2022-12-22 12:42:12,088 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1452, 1.6416, 2.2776, 1.9766, 2.3689, 2.7182, 2.7572, 2.1585], device='cuda:3'), covar=tensor([0.1696, 0.1611, 0.0575, 0.1236, 0.0980, 0.0626, 0.0581, 0.0872], device='cuda:3'), in_proj_covar=tensor([0.0060, 0.0104, 0.0087, 0.0118, 0.0099, 0.0067, 0.0064, 0.0097], device='cuda:3'), out_proj_covar=tensor([6.9403e-05, 1.0598e-04, 9.4358e-05, 1.2118e-04, 1.0458e-04, 7.9088e-05, 7.3438e-05, 1.0083e-04], device='cuda:3') 2022-12-22 12:42:17,571 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0025-130151-0_sp0.9 from training. Duration: 21.7944375 2022-12-22 12:42:26,042 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2022-12-22 12:42:31,840 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0_sp0.9 from training. Duration: 22.4666875 2022-12-22 12:42:37,525 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0 from training. Duration: 21.635 2022-12-22 12:42:43,739 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0_sp0.9 from training. Duration: 24.038875 2022-12-22 12:42:49,949 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3230, 0.6236, 1.1867, 1.3388, 1.2693, 1.2856, 0.9733, 0.9796], device='cuda:3'), covar=tensor([0.0711, 0.1169, 0.1820, 0.0881, 0.0586, 0.0624, 0.0875, 0.0773], device='cuda:3'), in_proj_covar=tensor([0.0070, 0.0085, 0.0101, 0.0082, 0.0068, 0.0074, 0.0071, 0.0073], device='cuda:3'), out_proj_covar=tensor([7.1998e-05, 7.4078e-05, 9.4139e-05, 8.5059e-05, 6.5037e-05, 7.1322e-05, 6.8890e-05, 6.8431e-05], device='cuda:3') 2022-12-22 12:43:04,011 INFO [train.py:894] (3/4) Epoch 2, batch 2300, loss[loss=0.3588, simple_loss=0.4021, pruned_loss=0.1578, over 18682.00 frames. ], tot_loss[loss=0.3519, simple_loss=0.3879, pruned_loss=0.158, over 3715717.34 frames. ], batch size: 60, lr: 3.96e-02, grad_scale: 8.0 2022-12-22 12:43:08,409 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 5.429e+02 8.168e+02 9.351e+02 1.113e+03 2.351e+03, threshold=1.870e+03, percent-clipped=5.0 2022-12-22 12:43:17,809 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5816.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:43:26,157 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp1.1 from training. Duration: 21.786375 2022-12-22 12:43:39,431 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0 from training. Duration: 20.22 2022-12-22 12:43:53,170 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.5720, 3.9287, 3.9005, 4.0838, 3.9851, 4.1061, 4.6075, 1.1631], device='cuda:3'), covar=tensor([0.0589, 0.0477, 0.0600, 0.0345, 0.1547, 0.0597, 0.0392, 0.5058], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0122, 0.0112, 0.0090, 0.0170, 0.0127, 0.0115, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002], device='cuda:3') 2022-12-22 12:44:01,255 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2022-12-22 12:44:19,728 INFO [train.py:894] (3/4) Epoch 2, batch 2350, loss[loss=0.2787, simple_loss=0.3316, pruned_loss=0.113, over 18390.00 frames. ], tot_loss[loss=0.3499, simple_loss=0.3861, pruned_loss=0.1568, over 3714930.14 frames. ], batch size: 46, lr: 3.95e-02, grad_scale: 8.0 2022-12-22 12:44:21,617 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5858.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:45:35,757 INFO [train.py:894] (3/4) Epoch 2, batch 2400, loss[loss=0.3415, simple_loss=0.3809, pruned_loss=0.1511, over 18598.00 frames. ], tot_loss[loss=0.3485, simple_loss=0.3846, pruned_loss=0.1562, over 3715406.05 frames. ], batch size: 56, lr: 3.94e-02, grad_scale: 8.0 2022-12-22 12:45:40,241 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.433e+02 7.534e+02 9.546e+02 1.123e+03 1.593e+03, threshold=1.909e+03, percent-clipped=0.0 2022-12-22 12:45:40,321 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0 from training. Duration: 25.285 2022-12-22 12:45:53,622 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5919.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:46:44,677 WARNING [train.py:1060] (3/4) Exclude cut with ID 811-130148-0001-63453-0_sp0.9 from training. Duration: 20.861125 2022-12-22 12:46:46,761 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6489, 1.7179, 0.8781, 1.6707, 2.0598, 1.4993, 3.4429, 1.8179], device='cuda:3'), covar=tensor([0.0992, 0.1394, 0.1708, 0.1651, 0.1618, 0.0866, 0.0135, 0.0964], device='cuda:3'), in_proj_covar=tensor([0.0107, 0.0108, 0.0123, 0.0153, 0.0148, 0.0104, 0.0063, 0.0106], device='cuda:3'), out_proj_covar=tensor([1.1003e-04, 1.1723e-04, 1.2466e-04, 1.4896e-04, 1.4686e-04, 1.0334e-04, 6.9707e-05, 1.0762e-04], device='cuda:3') 2022-12-22 12:46:52,386 INFO [train.py:894] (3/4) Epoch 2, batch 2450, loss[loss=0.3319, simple_loss=0.3596, pruned_loss=0.1521, over 18417.00 frames. ], tot_loss[loss=0.3491, simple_loss=0.3852, pruned_loss=0.1565, over 3715114.19 frames. ], batch size: 42, lr: 3.93e-02, grad_scale: 8.0 2022-12-22 12:47:08,230 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0 from training. Duration: 20.88 2022-12-22 12:47:28,086 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9118, 1.0053, 1.2759, 1.8141, 1.4513, 1.3677, 1.6759, 1.1262], device='cuda:3'), covar=tensor([0.0544, 0.1210, 0.1667, 0.0925, 0.0514, 0.0604, 0.0755, 0.0718], device='cuda:3'), in_proj_covar=tensor([0.0072, 0.0091, 0.0105, 0.0088, 0.0071, 0.0078, 0.0073, 0.0076], device='cuda:3'), out_proj_covar=tensor([7.4764e-05, 7.9452e-05, 9.8969e-05, 9.2034e-05, 6.7923e-05, 7.5336e-05, 7.2099e-05, 7.1793e-05], device='cuda:3') 2022-12-22 12:47:41,329 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0_sp0.9 from training. Duration: 23.4166875 2022-12-22 12:47:43,121 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5990.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:48:09,257 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6006.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 12:48:10,616 INFO [train.py:894] (3/4) Epoch 2, batch 2500, loss[loss=0.3219, simple_loss=0.3725, pruned_loss=0.1357, over 18703.00 frames. ], tot_loss[loss=0.3492, simple_loss=0.3849, pruned_loss=0.1568, over 3715458.29 frames. ], batch size: 60, lr: 3.92e-02, grad_scale: 8.0 2022-12-22 12:48:14,727 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.130e+02 7.982e+02 9.439e+02 1.211e+03 2.213e+03, threshold=1.888e+03, percent-clipped=1.0 2022-12-22 12:48:59,392 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0 from training. Duration: 21.24 2022-12-22 12:48:59,407 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0_sp0.9 from training. Duration: 23.9055625 2022-12-22 12:49:05,947 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6043.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:49:17,314 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6051.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:49:26,048 INFO [train.py:894] (3/4) Epoch 2, batch 2550, loss[loss=0.3702, simple_loss=0.4127, pruned_loss=0.1639, over 18645.00 frames. ], tot_loss[loss=0.3491, simple_loss=0.385, pruned_loss=0.1566, over 3715396.50 frames. ], batch size: 78, lr: 3.91e-02, grad_scale: 8.0 2022-12-22 12:49:32,408 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp0.9 from training. Duration: 25.988875 2022-12-22 12:49:40,849 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0001-134300-0_sp0.9 from training. Duration: 20.67225 2022-12-22 12:49:41,278 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.1764, 1.3479, 1.0424, 1.3931, 1.7046, 1.2326, 1.1070, 0.9917], device='cuda:3'), covar=tensor([0.1890, 0.1705, 0.1586, 0.1293, 0.0997, 0.1434, 0.1508, 0.1750], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0127, 0.0128, 0.0115, 0.0145, 0.0125, 0.0117, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001], device='cuda:3') 2022-12-22 12:50:29,592 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0 from training. Duration: 20.34 2022-12-22 12:50:38,798 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6104.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:50:43,048 INFO [train.py:894] (3/4) Epoch 2, batch 2600, loss[loss=0.2998, simple_loss=0.3433, pruned_loss=0.1281, over 18538.00 frames. ], tot_loss[loss=0.3473, simple_loss=0.3834, pruned_loss=0.1556, over 3715121.55 frames. ], batch size: 47, lr: 3.90e-02, grad_scale: 8.0 2022-12-22 12:50:47,637 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.975e+02 7.215e+02 8.875e+02 1.232e+03 1.892e+03, threshold=1.775e+03, percent-clipped=1.0 2022-12-22 12:50:49,979 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6111.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:51:26,210 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6135.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:51:42,721 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp0.9 from training. Duration: 25.061125 2022-12-22 12:51:53,975 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0 from training. Duration: 0.83 2022-12-22 12:52:01,066 INFO [train.py:894] (3/4) Epoch 2, batch 2650, loss[loss=0.3551, simple_loss=0.39, pruned_loss=0.1601, over 18459.00 frames. ], tot_loss[loss=0.3481, simple_loss=0.3844, pruned_loss=0.1559, over 3715175.94 frames. ], batch size: 50, lr: 3.89e-02, grad_scale: 8.0 2022-12-22 12:52:16,611 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.3042, 1.6631, 1.7412, 0.1664, 1.0645, 1.4639, 1.2266, 1.1955], device='cuda:3'), covar=tensor([0.1250, 0.0507, 0.0621, 0.1524, 0.0805, 0.0427, 0.0709, 0.0713], device='cuda:3'), in_proj_covar=tensor([0.0083, 0.0086, 0.0062, 0.0099, 0.0080, 0.0054, 0.0096, 0.0069], device='cuda:3'), out_proj_covar=tensor([8.9892e-05, 9.6743e-05, 7.5497e-05, 1.0961e-04, 9.3041e-05, 6.4235e-05, 1.1109e-04, 7.8870e-05], device='cuda:3') 2022-12-22 12:52:18,855 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0 from training. Duration: 24.73 2022-12-22 12:52:20,465 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6170.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:52:32,344 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0 from training. Duration: 23.965 2022-12-22 12:52:39,656 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0030-146996-0_sp0.9 from training. Duration: 22.088875 2022-12-22 12:52:57,008 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0_sp0.9 from training. Duration: 23.6 2022-12-22 12:53:00,910 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6196.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:53:11,425 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6203.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:53:17,467 INFO [train.py:894] (3/4) Epoch 2, batch 2700, loss[loss=0.3576, simple_loss=0.3964, pruned_loss=0.1594, over 18654.00 frames. ], tot_loss[loss=0.349, simple_loss=0.3848, pruned_loss=0.1566, over 3714339.16 frames. ], batch size: 65, lr: 3.88e-02, grad_scale: 8.0 2022-12-22 12:53:22,119 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.865e+02 7.324e+02 9.004e+02 1.200e+03 2.759e+03, threshold=1.801e+03, percent-clipped=6.0 2022-12-22 12:53:23,848 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6211.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:53:28,112 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6214.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:53:53,593 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6231.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:54:23,387 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4004, 1.5294, 1.1102, 1.8045, 2.0339, 1.2847, 1.3427, 1.0665], device='cuda:3'), covar=tensor([0.1554, 0.1461, 0.1540, 0.1039, 0.0973, 0.1291, 0.1449, 0.1590], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0126, 0.0129, 0.0113, 0.0148, 0.0126, 0.0121, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001], device='cuda:3') 2022-12-22 12:54:33,803 INFO [train.py:894] (3/4) Epoch 2, batch 2750, loss[loss=0.2802, simple_loss=0.3275, pruned_loss=0.1165, over 18576.00 frames. ], tot_loss[loss=0.3474, simple_loss=0.3835, pruned_loss=0.1556, over 3714128.53 frames. ], batch size: 41, lr: 3.87e-02, grad_scale: 8.0 2022-12-22 12:54:39,487 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0 from training. Duration: 23.795 2022-12-22 12:54:44,650 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6264.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:54:47,513 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6266.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:54:54,452 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp1.1 from training. Duration: 21.5409375 2022-12-22 12:54:56,075 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp0.9 from training. Duration: 24.97775 2022-12-22 12:54:56,448 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6272.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:55:06,619 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0_sp0.9 from training. Duration: 23.3444375 2022-12-22 12:55:32,618 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0_sp0.9 from training. Duration: 23.2 2022-12-22 12:55:40,597 WARNING [train.py:1060] (3/4) Exclude cut with ID 5653-46179-0060-117930-0_sp0.9 from training. Duration: 21.17225 2022-12-22 12:55:48,363 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6306.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 12:55:49,452 INFO [train.py:894] (3/4) Epoch 2, batch 2800, loss[loss=0.3718, simple_loss=0.4126, pruned_loss=0.1654, over 18636.00 frames. ], tot_loss[loss=0.3479, simple_loss=0.3843, pruned_loss=0.1558, over 3712987.08 frames. ], batch size: 77, lr: 3.86e-02, grad_scale: 8.0 2022-12-22 12:55:53,788 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.276e+02 7.690e+02 9.238e+02 1.166e+03 2.440e+03, threshold=1.848e+03, percent-clipped=6.0 2022-12-22 12:55:59,689 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp0.9 from training. Duration: 24.6555625 2022-12-22 12:56:18,855 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6327.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:56:48,185 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6346.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:56:57,067 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0 from training. Duration: 20.44 2022-12-22 12:57:00,660 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6354.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 12:57:04,357 INFO [train.py:894] (3/4) Epoch 2, batch 2850, loss[loss=0.3576, simple_loss=0.3981, pruned_loss=0.1586, over 18709.00 frames. ], tot_loss[loss=0.3476, simple_loss=0.3838, pruned_loss=0.1557, over 3712565.23 frames. ], batch size: 52, lr: 3.85e-02, grad_scale: 8.0 2022-12-22 12:57:11,449 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0_sp0.9 from training. Duration: 23.45 2022-12-22 12:57:43,973 WARNING [train.py:1060] (3/4) Exclude cut with ID 6945-60535-0076-12784-0_sp0.9 from training. Duration: 20.52225 2022-12-22 12:57:50,397 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0 from training. Duration: 22.19 2022-12-22 12:58:00,022 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp1.1 from training. Duration: 25.3818125 2022-12-22 12:58:09,006 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6399.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:58:16,841 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp0.9 from training. Duration: 28.0944375 2022-12-22 12:58:21,113 INFO [train.py:894] (3/4) Epoch 2, batch 2900, loss[loss=0.339, simple_loss=0.381, pruned_loss=0.1484, over 18504.00 frames. ], tot_loss[loss=0.3458, simple_loss=0.3823, pruned_loss=0.1546, over 3713750.80 frames. ], batch size: 58, lr: 3.85e-02, grad_scale: 8.0 2022-12-22 12:58:24,103 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0_sp0.9 from training. Duration: 22.9444375 2022-12-22 12:58:25,477 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.875e+02 7.430e+02 9.420e+02 1.139e+03 3.139e+03, threshold=1.884e+03, percent-clipped=2.0 2022-12-22 12:58:27,241 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6411.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:58:33,030 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp1.1 from training. Duration: 21.6318125 2022-12-22 12:58:37,777 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4496, 2.5473, 2.8353, 1.5401, 2.7389, 1.9580, 0.9638, 1.7887], device='cuda:3'), covar=tensor([0.1442, 0.0946, 0.1545, 0.3190, 0.1014, 0.1446, 0.4758, 0.2195], device='cuda:3'), in_proj_covar=tensor([0.0104, 0.0074, 0.0133, 0.0103, 0.0085, 0.0088, 0.0136, 0.0104], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001], device='cuda:3') 2022-12-22 12:58:48,055 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0 from training. Duration: 23.695 2022-12-22 12:58:51,979 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0448, 1.4209, 1.2528, 2.5310, 1.9974, 3.8698, 1.9094, 1.3224], device='cuda:3'), covar=tensor([0.1380, 0.2190, 0.2081, 0.1390, 0.1906, 0.0404, 0.1721, 0.2574], device='cuda:3'), in_proj_covar=tensor([0.0091, 0.0089, 0.0099, 0.0092, 0.0108, 0.0077, 0.0097, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 12:59:16,827 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0 from training. Duration: 23.955 2022-12-22 12:59:37,906 INFO [train.py:894] (3/4) Epoch 2, batch 2950, loss[loss=0.3162, simple_loss=0.3795, pruned_loss=0.1265, over 18712.00 frames. ], tot_loss[loss=0.3446, simple_loss=0.3818, pruned_loss=0.1537, over 3713439.49 frames. ], batch size: 52, lr: 3.84e-02, grad_scale: 8.0 2022-12-22 12:59:40,989 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6459.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:59:45,923 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6462.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 12:59:51,256 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp0.9 from training. Duration: 26.438875 2022-12-22 12:59:53,014 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6467.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:00:15,133 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6296, 1.1212, 1.9797, 1.8937, 1.5110, 1.8077, 0.9217, 1.5302], device='cuda:3'), covar=tensor([0.2391, 0.2365, 0.1940, 0.0898, 0.2106, 0.1892, 0.3037, 0.1524], device='cuda:3'), in_proj_covar=tensor([0.0096, 0.0102, 0.0114, 0.0074, 0.0092, 0.0107, 0.0130, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 13:00:22,183 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.8316, 3.3388, 3.5327, 2.1381, 3.4726, 2.3591, 1.2950, 2.5129], device='cuda:3'), covar=tensor([0.1538, 0.0692, 0.1405, 0.2928, 0.0996, 0.1451, 0.4844, 0.2099], device='cuda:3'), in_proj_covar=tensor([0.0104, 0.0072, 0.0131, 0.0103, 0.0084, 0.0086, 0.0133, 0.0103], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001], device='cuda:3') 2022-12-22 13:00:30,435 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6491.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:00:36,343 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0021-26306-0_sp0.9 from training. Duration: 21.2444375 2022-12-22 13:00:37,701 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp0.9 from training. Duration: 31.02225 2022-12-22 13:00:48,194 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0 from training. Duration: 22.395 2022-12-22 13:00:54,416 INFO [train.py:894] (3/4) Epoch 2, batch 3000, loss[loss=0.3385, simple_loss=0.3802, pruned_loss=0.1483, over 18651.00 frames. ], tot_loss[loss=0.3435, simple_loss=0.3812, pruned_loss=0.1529, over 3713133.02 frames. ], batch size: 69, lr: 3.83e-02, grad_scale: 8.0 2022-12-22 13:00:54,416 INFO [train.py:919] (3/4) Computing validation loss 2022-12-22 13:01:05,452 INFO [train.py:928] (3/4) Epoch 2, validation: loss=0.2458, simple_loss=0.336, pruned_loss=0.07777, over 944034.00 frames. 2022-12-22 13:01:05,453 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24511MB 2022-12-22 13:01:10,819 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.073e+02 7.196e+02 8.571e+02 1.049e+03 2.506e+03, threshold=1.714e+03, percent-clipped=2.0 2022-12-22 13:01:13,920 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0 from training. Duration: 21.075 2022-12-22 13:01:17,118 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6514.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:01:18,340 WARNING [train.py:1060] (3/4) Exclude cut with ID 6482-98857-0025-147532-0_sp0.9 from training. Duration: 20.0055625 2022-12-22 13:01:18,352 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0037-132304-0_sp0.9 from training. Duration: 22.05 2022-12-22 13:01:18,365 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0 from training. Duration: 26.8349375 2022-12-22 13:01:23,587 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp1.1 from training. Duration: 22.1090625 2022-12-22 13:01:30,748 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp0.9 from training. Duration: 26.6166875 2022-12-22 13:01:31,185 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6523.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 13:01:35,694 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6526.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:01:38,810 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6528.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:01:50,282 WARNING [train.py:1060] (3/4) Exclude cut with ID 2046-178027-0000-53705-0_sp0.9 from training. Duration: 20.3055625 2022-12-22 13:02:11,626 WARNING [train.py:1060] (3/4) Exclude cut with ID 7205-50138-0008-5373-0_sp0.9 from training. Duration: 20.7 2022-12-22 13:02:22,263 INFO [train.py:894] (3/4) Epoch 2, batch 3050, loss[loss=0.3659, simple_loss=0.3878, pruned_loss=0.172, over 18425.00 frames. ], tot_loss[loss=0.3448, simple_loss=0.3823, pruned_loss=0.1536, over 3713532.70 frames. ], batch size: 48, lr: 3.82e-02, grad_scale: 8.0 2022-12-22 13:02:25,597 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6559.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:02:30,749 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6562.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:02:38,373 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6567.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:02:48,524 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.3873, 1.4400, 1.5064, 0.2810, 1.2969, 1.6671, 1.0951, 1.1058], device='cuda:3'), covar=tensor([0.1177, 0.0464, 0.0773, 0.1242, 0.0750, 0.0441, 0.0864, 0.0724], device='cuda:3'), in_proj_covar=tensor([0.0086, 0.0088, 0.0063, 0.0102, 0.0084, 0.0056, 0.0099, 0.0070], device='cuda:3'), out_proj_covar=tensor([9.4869e-05, 1.0039e-04, 7.7983e-05, 1.1369e-04, 9.7473e-05, 6.7536e-05, 1.1439e-04, 8.1685e-05], device='cuda:3') 2022-12-22 13:02:55,484 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0 from training. Duration: 22.48 2022-12-22 13:03:10,287 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp0.9 from training. Duration: 29.816625 2022-12-22 13:03:29,498 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp1.1 from training. Duration: 22.7590625 2022-12-22 13:03:35,698 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0 from training. Duration: 22.555 2022-12-22 13:03:40,724 INFO [train.py:894] (3/4) Epoch 2, batch 3100, loss[loss=0.3092, simple_loss=0.3493, pruned_loss=0.1345, over 18700.00 frames. ], tot_loss[loss=0.3426, simple_loss=0.3807, pruned_loss=0.1523, over 3713473.71 frames. ], batch size: 46, lr: 3.81e-02, grad_scale: 8.0 2022-12-22 13:03:44,970 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.125e+02 7.487e+02 9.188e+02 1.082e+03 2.139e+03, threshold=1.838e+03, percent-clipped=7.0 2022-12-22 13:03:56,560 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0005-25975-0_sp0.9 from training. Duration: 21.688875 2022-12-22 13:04:03,029 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6622.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:04:13,366 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6629.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:04:16,538 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.4295, 4.4090, 4.4876, 2.4611, 4.2998, 2.8536, 1.4151, 3.2794], device='cuda:3'), covar=tensor([0.1356, 0.0608, 0.1404, 0.2974, 0.0840, 0.1450, 0.5289, 0.1922], device='cuda:3'), in_proj_covar=tensor([0.0108, 0.0074, 0.0137, 0.0104, 0.0086, 0.0089, 0.0139, 0.0105], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001], device='cuda:3') 2022-12-22 13:04:30,218 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0_sp0.9 from training. Duration: 22.6 2022-12-22 13:04:30,470 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.1137, 3.6034, 3.5279, 3.7040, 3.6963, 3.6985, 4.0813, 2.1269], device='cuda:3'), covar=tensor([0.0556, 0.0490, 0.0682, 0.0423, 0.1281, 0.0585, 0.0524, 0.3387], device='cuda:3'), in_proj_covar=tensor([0.0157, 0.0130, 0.0119, 0.0094, 0.0177, 0.0127, 0.0125, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002], device='cuda:3') 2022-12-22 13:04:32,781 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 2022-12-22 13:04:39,996 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6646.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:04:42,892 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9899, 2.1796, 1.9289, 1.3032, 2.1207, 2.3471, 1.3878, 2.6594], device='cuda:3'), covar=tensor([0.1497, 0.0987, 0.1956, 0.2552, 0.1216, 0.1261, 0.2849, 0.0698], device='cuda:3'), in_proj_covar=tensor([0.0187, 0.0138, 0.0187, 0.0183, 0.0173, 0.0180, 0.0182, 0.0160], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 13:04:52,460 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2022-12-22 13:04:55,899 INFO [train.py:894] (3/4) Epoch 2, batch 3150, loss[loss=0.3883, simple_loss=0.4189, pruned_loss=0.1788, over 18664.00 frames. ], tot_loss[loss=0.343, simple_loss=0.3816, pruned_loss=0.1522, over 3713573.80 frames. ], batch size: 60, lr: 3.80e-02, grad_scale: 8.0 2022-12-22 13:05:05,913 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0 from training. Duration: 24.32 2022-12-22 13:05:45,540 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6690.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:05:52,598 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6694.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:06:00,268 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6699.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:06:06,638 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-276745-0093-13116-0_sp0.9 from training. Duration: 21.061125 2022-12-22 13:06:07,423 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 2022-12-22 13:06:11,980 INFO [train.py:894] (3/4) Epoch 2, batch 3200, loss[loss=0.4315, simple_loss=0.4363, pruned_loss=0.2133, over 18627.00 frames. ], tot_loss[loss=0.3427, simple_loss=0.381, pruned_loss=0.1522, over 3713477.87 frames. ], batch size: 97, lr: 3.79e-02, grad_scale: 8.0 2022-12-22 13:06:16,409 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.683e+02 7.806e+02 9.589e+02 1.263e+03 2.859e+03, threshold=1.918e+03, percent-clipped=8.0 2022-12-22 13:06:21,347 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0024-15855-0_sp0.9 from training. Duration: 20.32225 2022-12-22 13:06:33,578 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp1.1 from training. Duration: 0.7545625 2022-12-22 13:06:49,415 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0_sp0.9 from training. Duration: 23.9333125 2022-12-22 13:07:10,066 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3147, 1.9670, 1.8115, 2.3712, 2.0165, 1.8189, 1.9021, 2.6743], device='cuda:3'), covar=tensor([0.1182, 0.1067, 0.0854, 0.1227, 0.1008, 0.0571, 0.1368, 0.0341], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0103, 0.0109, 0.0153, 0.0106, 0.0108, 0.0124, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:3') 2022-12-22 13:07:12,637 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6747.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:07:23,002 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp1.1 from training. Duration: 20.17275 2022-12-22 13:07:26,537 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.97 vs. limit=2.0 2022-12-22 13:07:27,097 INFO [train.py:894] (3/4) Epoch 2, batch 3250, loss[loss=0.3256, simple_loss=0.3722, pruned_loss=0.1395, over 18678.00 frames. ], tot_loss[loss=0.3422, simple_loss=0.3807, pruned_loss=0.1518, over 3713648.16 frames. ], batch size: 60, lr: 3.78e-02, grad_scale: 8.0 2022-12-22 13:07:27,192 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp1.1 from training. Duration: 20.436375 2022-12-22 13:08:19,674 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6791.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:08:44,174 INFO [train.py:894] (3/4) Epoch 2, batch 3300, loss[loss=0.3738, simple_loss=0.4144, pruned_loss=0.1667, over 18484.00 frames. ], tot_loss[loss=0.3403, simple_loss=0.3796, pruned_loss=0.1505, over 3713032.60 frames. ], batch size: 52, lr: 3.77e-02, grad_scale: 8.0 2022-12-22 13:08:48,836 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.425e+02 7.176e+02 8.285e+02 1.070e+03 1.922e+03, threshold=1.657e+03, percent-clipped=1.0 2022-12-22 13:08:51,674 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0_sp0.9 from training. Duration: 23.1055625 2022-12-22 13:08:53,247 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp1.1 from training. Duration: 23.5 2022-12-22 13:09:01,116 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6818.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 13:09:03,818 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp0.9 from training. Duration: 26.62775 2022-12-22 13:09:08,386 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6823.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:09:11,806 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2022-12-22 13:09:12,875 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6826.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:09:18,423 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0 from training. Duration: 21.105 2022-12-22 13:09:23,075 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0_sp0.9 from training. Duration: 24.411125 2022-12-22 13:09:32,607 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6839.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:09:50,859 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp1.1 from training. Duration: 21.263625 2022-12-22 13:09:57,713 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7813, 0.9356, 1.3444, 1.5846, 1.4340, 1.3362, 1.5524, 0.9877], device='cuda:3'), covar=tensor([0.0674, 0.1487, 0.2097, 0.1089, 0.0582, 0.0688, 0.0974, 0.0804], device='cuda:3'), in_proj_covar=tensor([0.0085, 0.0111, 0.0122, 0.0104, 0.0084, 0.0092, 0.0094, 0.0092], device='cuda:3'), out_proj_covar=tensor([9.2960e-05, 1.0099e-04, 1.1768e-04, 1.1274e-04, 8.4440e-05, 9.0911e-05, 9.5956e-05, 8.8994e-05], device='cuda:3') 2022-12-22 13:10:00,173 INFO [train.py:894] (3/4) Epoch 2, batch 3350, loss[loss=0.3363, simple_loss=0.385, pruned_loss=0.1438, over 18652.00 frames. ], tot_loss[loss=0.34, simple_loss=0.3793, pruned_loss=0.1504, over 3713310.55 frames. ], batch size: 78, lr: 3.76e-02, grad_scale: 8.0 2022-12-22 13:10:03,404 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6859.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:10:15,247 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6867.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:10:22,496 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0 from training. Duration: 20.795 2022-12-22 13:10:25,742 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6874.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:10:32,061 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6878.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 13:10:33,208 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0 from training. Duration: 24.76 2022-12-22 13:10:33,229 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0_sp0.9 from training. Duration: 22.25 2022-12-22 13:11:00,069 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp1.1 from training. Duration: 20.5045625 2022-12-22 13:11:17,339 INFO [train.py:894] (3/4) Epoch 2, batch 3400, loss[loss=0.324, simple_loss=0.3702, pruned_loss=0.1389, over 18547.00 frames. ], tot_loss[loss=0.3398, simple_loss=0.3791, pruned_loss=0.1503, over 3713814.44 frames. ], batch size: 58, lr: 3.75e-02, grad_scale: 8.0 2022-12-22 13:11:17,515 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6907.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:11:20,741 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6909.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:11:22,031 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 5.317e+02 7.342e+02 9.217e+02 1.220e+03 2.617e+03, threshold=1.843e+03, percent-clipped=11.0 2022-12-22 13:11:29,551 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6915.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:11:39,475 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6922.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:12:04,254 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6939.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 13:12:11,580 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1957, 2.3629, 1.2275, 2.7068, 1.8838, 4.1692, 2.2026, 1.8864], device='cuda:3'), covar=tensor([0.1223, 0.1658, 0.2049, 0.1326, 0.1878, 0.0287, 0.1476, 0.2056], device='cuda:3'), in_proj_covar=tensor([0.0090, 0.0090, 0.0098, 0.0091, 0.0109, 0.0078, 0.0098, 0.0091], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 13:12:32,078 INFO [train.py:894] (3/4) Epoch 2, batch 3450, loss[loss=0.293, simple_loss=0.3462, pruned_loss=0.1199, over 18675.00 frames. ], tot_loss[loss=0.3395, simple_loss=0.3784, pruned_loss=0.1503, over 3714371.66 frames. ], batch size: 48, lr: 3.75e-02, grad_scale: 8.0 2022-12-22 13:12:51,022 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6970.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:12:51,269 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6970.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:13:09,413 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.42 vs. limit=5.0 2022-12-22 13:13:13,505 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6985.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:13:46,395 INFO [train.py:894] (3/4) Epoch 2, batch 3500, loss[loss=0.4025, simple_loss=0.4221, pruned_loss=0.1915, over 18599.00 frames. ], tot_loss[loss=0.3383, simple_loss=0.3772, pruned_loss=0.1497, over 3713932.96 frames. ], batch size: 176, lr: 3.74e-02, grad_scale: 8.0 2022-12-22 13:13:51,973 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.008e+02 6.904e+02 8.316e+02 1.072e+03 2.058e+03, threshold=1.663e+03, percent-clipped=2.0 2022-12-22 13:14:08,195 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp1.1 from training. Duration: 22.2954375 2022-12-22 13:14:17,623 INFO [train.py:894] (3/4) Epoch 3, batch 0, loss[loss=0.3428, simple_loss=0.3969, pruned_loss=0.1444, over 18532.00 frames. ], tot_loss[loss=0.3428, simple_loss=0.3969, pruned_loss=0.1444, over 18532.00 frames. ], batch size: 77, lr: 3.55e-02, grad_scale: 8.0 2022-12-22 13:14:17,624 INFO [train.py:919] (3/4) Computing validation loss 2022-12-22 13:14:28,602 INFO [train.py:928] (3/4) Epoch 3, validation: loss=0.2448, simple_loss=0.3366, pruned_loss=0.07644, over 944034.00 frames. 2022-12-22 13:14:28,604 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24598MB 2022-12-22 13:14:46,742 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=5.15 vs. limit=5.0 2022-12-22 13:15:21,065 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7048.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 13:15:23,529 WARNING [train.py:1060] (3/4) Exclude cut with ID 298-126791-0067-24026-0_sp0.9 from training. Duration: 21.438875 2022-12-22 13:15:27,947 WARNING [train.py:1060] (3/4) Exclude cut with ID 5652-39938-0025-23684-0_sp0.9 from training. Duration: 22.2055625 2022-12-22 13:15:36,102 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.2431, 0.9517, 0.9929, 1.0186, 1.2607, 0.9052, 1.0459, 1.4102], device='cuda:3'), covar=tensor([0.1475, 0.1748, 0.2129, 0.1638, 0.1983, 0.1744, 0.1466, 0.1201], device='cuda:3'), in_proj_covar=tensor([0.0083, 0.0101, 0.0127, 0.0096, 0.0100, 0.0098, 0.0087, 0.0095], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 13:15:43,246 INFO [train.py:894] (3/4) Epoch 3, batch 50, loss[loss=0.2959, simple_loss=0.3612, pruned_loss=0.1153, over 18513.00 frames. ], tot_loss[loss=0.305, simple_loss=0.3627, pruned_loss=0.1237, over 838401.77 frames. ], batch size: 58, lr: 3.54e-02, grad_scale: 16.0 2022-12-22 13:16:55,261 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7109.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 13:16:58,255 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.649e+02 5.613e+02 6.780e+02 8.484e+02 1.673e+03, threshold=1.356e+03, percent-clipped=1.0 2022-12-22 13:17:01,179 INFO [train.py:894] (3/4) Epoch 3, batch 100, loss[loss=0.2564, simple_loss=0.3359, pruned_loss=0.0885, over 18706.00 frames. ], tot_loss[loss=0.293, simple_loss=0.3535, pruned_loss=0.1162, over 1475475.77 frames. ], batch size: 52, lr: 3.53e-02, grad_scale: 8.0 2022-12-22 13:17:10,093 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7118.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 13:17:17,538 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7123.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:17:20,615 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7653, 2.2906, 1.6511, 0.8301, 2.0109, 2.2510, 1.6166, 1.8931], device='cuda:3'), covar=tensor([0.0651, 0.0792, 0.2162, 0.2826, 0.1532, 0.1267, 0.1559, 0.1632], device='cuda:3'), in_proj_covar=tensor([0.0108, 0.0145, 0.0186, 0.0177, 0.0162, 0.0150, 0.0147, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 13:18:08,099 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.1520, 0.7985, 0.9906, 0.0052, 0.8562, 0.8794, 0.7411, 0.8449], device='cuda:3'), covar=tensor([0.1058, 0.0664, 0.0644, 0.1516, 0.0671, 0.0458, 0.0896, 0.0827], device='cuda:3'), in_proj_covar=tensor([0.0089, 0.0095, 0.0067, 0.0105, 0.0085, 0.0062, 0.0103, 0.0077], device='cuda:3'), out_proj_covar=tensor([9.8819e-05, 1.0873e-04, 8.3322e-05, 1.1761e-04, 9.8530e-05, 7.3817e-05, 1.1876e-04, 8.9657e-05], device='cuda:3') 2022-12-22 13:18:16,841 INFO [train.py:894] (3/4) Epoch 3, batch 150, loss[loss=0.2846, simple_loss=0.3372, pruned_loss=0.116, over 18692.00 frames. ], tot_loss[loss=0.2866, simple_loss=0.3491, pruned_loss=0.1121, over 1970716.14 frames. ], batch size: 50, lr: 3.52e-02, grad_scale: 8.0 2022-12-22 13:18:17,256 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7163.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:18:22,196 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=7166.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:18:26,212 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0 from training. Duration: 24.525 2022-12-22 13:18:30,299 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=7171.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:19:00,640 WARNING [train.py:1060] (3/4) Exclude cut with ID 3699-47246-0007-3408-0_sp0.9 from training. Duration: 20.26675 2022-12-22 13:19:15,770 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp0.9 from training. Duration: 27.25 2022-12-22 13:19:17,422 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3236, 2.2122, 1.6967, 1.1781, 2.7621, 2.7067, 1.2018, 1.4938], device='cuda:3'), covar=tensor([0.0819, 0.0805, 0.1590, 0.1733, 0.0232, 0.0480, 0.1637, 0.1892], device='cuda:3'), in_proj_covar=tensor([0.0119, 0.0112, 0.0141, 0.0128, 0.0081, 0.0113, 0.0134, 0.0145], device='cuda:3'), out_proj_covar=tensor([1.4618e-04, 1.3982e-04, 1.6595e-04, 1.5176e-04, 9.9260e-05, 1.3692e-04, 1.5996e-04, 1.6738e-04], device='cuda:3') 2022-12-22 13:19:28,829 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.213e+02 5.590e+02 6.850e+02 8.940e+02 2.169e+03, threshold=1.370e+03, percent-clipped=7.0 2022-12-22 13:19:32,462 INFO [train.py:894] (3/4) Epoch 3, batch 200, loss[loss=0.2529, simple_loss=0.3293, pruned_loss=0.08822, over 18697.00 frames. ], tot_loss[loss=0.2849, simple_loss=0.3472, pruned_loss=0.1113, over 2356598.84 frames. ], batch size: 62, lr: 3.52e-02, grad_scale: 8.0 2022-12-22 13:19:50,980 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7224.0, num_to_drop=1, layers_to_drop={3} 2022-12-22 13:19:56,645 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7894, 2.1208, 1.7471, 2.4473, 2.2216, 1.7860, 2.0734, 3.2092], device='cuda:3'), covar=tensor([0.0956, 0.1135, 0.0944, 0.1400, 0.1020, 0.0731, 0.1367, 0.0336], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0108, 0.0112, 0.0160, 0.0109, 0.0111, 0.0129, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:3') 2022-12-22 13:20:04,788 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7234.0, num_to_drop=1, layers_to_drop={2} 2022-12-22 13:20:11,518 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2022-12-22 13:20:29,691 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0 from training. Duration: 21.68 2022-12-22 13:20:42,016 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0 from training. Duration: 21.6300625 2022-12-22 13:20:47,784 INFO [train.py:894] (3/4) Epoch 3, batch 250, loss[loss=0.2801, simple_loss=0.3493, pruned_loss=0.1055, over 18509.00 frames. ], tot_loss[loss=0.2827, simple_loss=0.3458, pruned_loss=0.1099, over 2658008.68 frames. ], batch size: 58, lr: 3.51e-02, grad_scale: 8.0 2022-12-22 13:20:51,294 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7265.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:21:06,554 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0_sp0.9 from training. Duration: 24.033375 2022-12-22 13:21:06,892 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7275.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:21:17,564 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3778, 0.4499, 1.3028, 0.9408, 1.6189, 1.5445, 1.5268, 1.1427], device='cuda:3'), covar=tensor([0.0836, 0.1173, 0.1063, 0.0901, 0.0496, 0.0441, 0.0445, 0.0760], device='cuda:3'), in_proj_covar=tensor([0.0060, 0.0101, 0.0096, 0.0115, 0.0096, 0.0065, 0.0065, 0.0096], device='cuda:3'), out_proj_covar=tensor([7.0709e-05, 1.0539e-04, 1.0564e-04, 1.1941e-04, 1.0490e-04, 7.4127e-05, 7.4893e-05, 1.0149e-04], device='cuda:3') 2022-12-22 13:21:21,813 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7285.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:21:27,774 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5074, 1.0890, 2.0344, 2.2859, 1.7636, 1.9191, 0.8629, 1.6445], device='cuda:3'), covar=tensor([0.2359, 0.2291, 0.1716, 0.0685, 0.1587, 0.1716, 0.2978, 0.1329], device='cuda:3'), in_proj_covar=tensor([0.0096, 0.0104, 0.0111, 0.0076, 0.0092, 0.0111, 0.0132, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 13:22:00,481 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.728e+02 5.754e+02 6.972e+02 8.878e+02 1.547e+03, threshold=1.394e+03, percent-clipped=2.0 2022-12-22 13:22:03,713 INFO [train.py:894] (3/4) Epoch 3, batch 300, loss[loss=0.296, simple_loss=0.3649, pruned_loss=0.1135, over 18584.00 frames. ], tot_loss[loss=0.2817, simple_loss=0.345, pruned_loss=0.1091, over 2890758.36 frames. ], batch size: 56, lr: 3.50e-02, grad_scale: 8.0 2022-12-22 13:22:06,454 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0 from training. Duration: 22.905 2022-12-22 13:22:07,946 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp1.1 from training. Duration: 23.4318125 2022-12-22 13:22:28,530 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1662, 1.5052, 3.2112, 3.0244, 2.7334, 2.1100, 0.9018, 1.9551], device='cuda:3'), covar=tensor([0.2175, 0.2183, 0.1285, 0.0644, 0.1309, 0.1855, 0.2896, 0.1406], device='cuda:3'), in_proj_covar=tensor([0.0093, 0.0101, 0.0108, 0.0074, 0.0090, 0.0109, 0.0127, 0.0087], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 13:22:34,159 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=7333.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:22:39,171 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7336.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:23:19,970 INFO [train.py:894] (3/4) Epoch 3, batch 350, loss[loss=0.2643, simple_loss=0.3412, pruned_loss=0.09376, over 18627.00 frames. ], tot_loss[loss=0.2829, simple_loss=0.3459, pruned_loss=0.11, over 3072753.40 frames. ], batch size: 53, lr: 3.49e-02, grad_scale: 8.0 2022-12-22 13:23:30,751 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2022-12-22 13:24:06,656 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp1.1 from training. Duration: 20.82275 2022-12-22 13:24:08,113 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp0.9 from training. Duration: 25.45 2022-12-22 13:24:14,761 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2022-12-22 13:24:22,095 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7404.0, num_to_drop=1, layers_to_drop={2} 2022-12-22 13:24:27,603 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7407.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:24:33,299 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.000e+02 6.102e+02 7.979e+02 1.069e+03 3.939e+03, threshold=1.596e+03, percent-clipped=10.0 2022-12-22 13:24:33,617 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([5.5864, 4.8523, 4.8054, 5.0415, 5.1369, 5.0047, 5.4566, 1.5439], device='cuda:3'), covar=tensor([0.0642, 0.0398, 0.0476, 0.0363, 0.1189, 0.0521, 0.0280, 0.4567], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0132, 0.0116, 0.0097, 0.0177, 0.0132, 0.0133, 0.0186], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002], device='cuda:3') 2022-12-22 13:24:36,869 INFO [train.py:894] (3/4) Epoch 3, batch 400, loss[loss=0.3078, simple_loss=0.3736, pruned_loss=0.121, over 18549.00 frames. ], tot_loss[loss=0.2867, simple_loss=0.3484, pruned_loss=0.1125, over 3214463.09 frames. ], batch size: 55, lr: 3.48e-02, grad_scale: 8.0 2022-12-22 13:25:04,453 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4128, 2.0214, 1.9210, 2.6791, 2.2452, 3.4296, 2.1223, 2.0927], device='cuda:3'), covar=tensor([0.1059, 0.1735, 0.1533, 0.1028, 0.1430, 0.0769, 0.1267, 0.1591], device='cuda:3'), in_proj_covar=tensor([0.0089, 0.0089, 0.0095, 0.0091, 0.0107, 0.0077, 0.0095, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 13:25:05,674 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0 from training. Duration: 25.775 2022-12-22 13:25:26,296 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0_sp0.9 from training. Duration: 22.25 2022-12-22 13:25:52,873 INFO [train.py:894] (3/4) Epoch 3, batch 450, loss[loss=0.3086, simple_loss=0.3597, pruned_loss=0.1288, over 18433.00 frames. ], tot_loss[loss=0.2885, simple_loss=0.3497, pruned_loss=0.1136, over 3325221.63 frames. ], batch size: 48, lr: 3.48e-02, grad_scale: 8.0 2022-12-22 13:25:55,919 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0 from training. Duration: 26.205 2022-12-22 13:26:01,204 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7468.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:26:13,440 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp0.9 from training. Duration: 30.1555625 2022-12-22 13:26:17,601 WARNING [train.py:1060] (3/4) Exclude cut with ID 1265-135635-0050-6781-0_sp0.9 from training. Duration: 21.8333125 2022-12-22 13:26:24,861 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp1.1 from training. Duration: 20.6545625 2022-12-22 13:26:29,036 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.76 vs. limit=2.0 2022-12-22 13:27:04,840 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.631e+02 6.334e+02 7.684e+02 9.294e+02 1.756e+03, threshold=1.537e+03, percent-clipped=3.0 2022-12-22 13:27:07,850 INFO [train.py:894] (3/4) Epoch 3, batch 500, loss[loss=0.2484, simple_loss=0.314, pruned_loss=0.0914, over 18594.00 frames. ], tot_loss[loss=0.2898, simple_loss=0.3506, pruned_loss=0.1144, over 3411383.37 frames. ], batch size: 45, lr: 3.47e-02, grad_scale: 8.0 2022-12-22 13:27:09,250 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0045-39920-0_sp0.9 from training. Duration: 20.52225 2022-12-22 13:27:17,092 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7519.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 13:27:28,582 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp0.9 from training. Duration: 29.1166875 2022-12-22 13:27:38,702 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7534.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 13:28:23,482 INFO [train.py:894] (3/4) Epoch 3, batch 550, loss[loss=0.3007, simple_loss=0.3637, pruned_loss=0.1188, over 18525.00 frames. ], tot_loss[loss=0.2924, simple_loss=0.3531, pruned_loss=0.1159, over 3479665.35 frames. ], batch size: 55, lr: 3.46e-02, grad_scale: 8.0 2022-12-22 13:28:26,647 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7565.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:28:27,697 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133211-0007-59831-0_sp0.9 from training. Duration: 21.388875 2022-12-22 13:28:50,644 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=7582.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 13:28:58,896 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0 from training. Duration: 22.72 2022-12-22 13:29:00,268 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0_sp0.9 from training. Duration: 22.7444375 2022-12-22 13:29:26,196 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2022-12-22 13:29:34,404 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.376e+02 6.274e+02 7.834e+02 9.587e+02 2.337e+03, threshold=1.567e+03, percent-clipped=2.0 2022-12-22 13:29:37,664 INFO [train.py:894] (3/4) Epoch 3, batch 600, loss[loss=0.3456, simple_loss=0.3917, pruned_loss=0.1498, over 18688.00 frames. ], tot_loss[loss=0.2921, simple_loss=0.3525, pruned_loss=0.1159, over 3531625.85 frames. ], batch size: 175, lr: 3.45e-02, grad_scale: 8.0 2022-12-22 13:29:37,829 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=7613.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:29:44,932 WARNING [train.py:1060] (3/4) Exclude cut with ID 4133-6541-0027-40495-0_sp1.1 from training. Duration: 0.9681875 2022-12-22 13:29:48,029 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0_sp0.9 from training. Duration: 22.3166875 2022-12-22 13:29:54,168 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133212-0015-59917-0_sp0.9 from training. Duration: 21.8166875 2022-12-22 13:29:54,470 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7624.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:30:04,469 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7631.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:30:44,133 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 2022-12-22 13:30:53,721 INFO [train.py:894] (3/4) Epoch 3, batch 650, loss[loss=0.2885, simple_loss=0.3585, pruned_loss=0.1092, over 18463.00 frames. ], tot_loss[loss=0.2944, simple_loss=0.3544, pruned_loss=0.1172, over 3571375.88 frames. ], batch size: 50, lr: 3.44e-02, grad_scale: 8.0 2022-12-22 13:31:09,567 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1036, 1.1048, 1.4071, 1.8896, 1.6312, 1.4824, 1.9409, 1.1009], device='cuda:3'), covar=tensor([0.0623, 0.1570, 0.1786, 0.0998, 0.0608, 0.0618, 0.0939, 0.0794], device='cuda:3'), in_proj_covar=tensor([0.0095, 0.0127, 0.0131, 0.0119, 0.0097, 0.0100, 0.0108, 0.0102], device='cuda:3'), out_proj_covar=tensor([1.0782e-04, 1.1782e-04, 1.2912e-04, 1.3038e-04, 9.7861e-05, 9.8564e-05, 1.1292e-04, 9.9612e-05], device='cuda:3') 2022-12-22 13:31:27,229 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7685.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:31:33,191 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9662, 2.1130, 0.9295, 2.1026, 2.1930, 1.7772, 3.2047, 2.2104], device='cuda:3'), covar=tensor([0.0808, 0.1274, 0.2202, 0.1639, 0.1623, 0.0778, 0.0266, 0.0917], device='cuda:3'), in_proj_covar=tensor([0.0119, 0.0124, 0.0142, 0.0179, 0.0162, 0.0117, 0.0085, 0.0124], device='cuda:3'), out_proj_covar=tensor([1.2375e-04, 1.3444e-04, 1.4652e-04, 1.7582e-04, 1.6580e-04, 1.1825e-04, 9.8317e-05, 1.2862e-04], device='cuda:3') 2022-12-22 13:31:37,110 WARNING [train.py:1060] (3/4) Exclude cut with ID 4957-30119-0041-23990-0_sp0.9 from training. Duration: 20.22775 2022-12-22 13:31:56,666 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7704.0, num_to_drop=1, layers_to_drop={2} 2022-12-22 13:32:06,837 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.551e+02 6.142e+02 7.581e+02 9.638e+02 3.494e+03, threshold=1.516e+03, percent-clipped=4.0 2022-12-22 13:32:09,746 INFO [train.py:894] (3/4) Epoch 3, batch 700, loss[loss=0.2912, simple_loss=0.3646, pruned_loss=0.1089, over 18441.00 frames. ], tot_loss[loss=0.2938, simple_loss=0.3543, pruned_loss=0.1166, over 3603072.11 frames. ], batch size: 64, lr: 3.44e-02, grad_scale: 8.0 2022-12-22 13:32:23,259 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp1.1 from training. Duration: 24.67275 2022-12-22 13:32:29,636 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5252, 1.1266, 1.2484, 1.1657, 1.4585, 1.1990, 1.4318, 1.6826], device='cuda:3'), covar=tensor([0.1985, 0.2324, 0.2474, 0.2098, 0.2585, 0.1987, 0.1826, 0.1789], device='cuda:3'), in_proj_covar=tensor([0.0083, 0.0101, 0.0127, 0.0096, 0.0102, 0.0100, 0.0088, 0.0095], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 13:32:51,343 WARNING [train.py:1060] (3/4) Exclude cut with ID 3082-165428-0081-50734-0_sp0.9 from training. Duration: 21.8055625 2022-12-22 13:32:54,668 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7743.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:33:08,109 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=7752.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 13:33:24,435 INFO [train.py:894] (3/4) Epoch 3, batch 750, loss[loss=0.2755, simple_loss=0.3477, pruned_loss=0.1016, over 18544.00 frames. ], tot_loss[loss=0.2952, simple_loss=0.3554, pruned_loss=0.1175, over 3626783.94 frames. ], batch size: 55, lr: 3.43e-02, grad_scale: 8.0 2022-12-22 13:33:24,593 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7763.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:33:28,802 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0_sp0.9 from training. Duration: 22.6666875 2022-12-22 13:33:59,389 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4100, 1.6172, 1.6193, 1.2966, 1.8844, 1.7431, 1.1983, 2.0038], device='cuda:3'), covar=tensor([0.1554, 0.1246, 0.1807, 0.2282, 0.1001, 0.1367, 0.2796, 0.0858], device='cuda:3'), in_proj_covar=tensor([0.0188, 0.0144, 0.0186, 0.0185, 0.0179, 0.0186, 0.0180, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 13:34:06,016 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.0207, 3.4891, 3.3730, 3.5601, 3.5609, 3.6352, 4.1121, 1.3940], device='cuda:3'), covar=tensor([0.0685, 0.0609, 0.0649, 0.0393, 0.1507, 0.0728, 0.0538, 0.4393], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0136, 0.0118, 0.0096, 0.0181, 0.0134, 0.0136, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002], device='cuda:3') 2022-12-22 13:34:23,858 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7802.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:34:27,351 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7804.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:34:29,915 WARNING [train.py:1060] (3/4) Exclude cut with ID 2411-132532-0017-82279-0_sp1.1 from training. Duration: 0.9681875 2022-12-22 13:34:37,380 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.051e+02 6.250e+02 7.704e+02 9.761e+02 2.933e+03, threshold=1.541e+03, percent-clipped=4.0 2022-12-22 13:34:40,176 INFO [train.py:894] (3/4) Epoch 3, batch 800, loss[loss=0.247, simple_loss=0.3076, pruned_loss=0.09326, over 18428.00 frames. ], tot_loss[loss=0.2937, simple_loss=0.3543, pruned_loss=0.1166, over 3646314.43 frames. ], batch size: 42, lr: 3.42e-02, grad_scale: 8.0 2022-12-22 13:34:49,490 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7819.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 13:34:54,597 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0 from training. Duration: 22.485 2022-12-22 13:35:32,159 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp1.1 from training. Duration: 23.82275 2022-12-22 13:35:48,011 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0 from training. Duration: 20.77 2022-12-22 13:35:55,074 INFO [train.py:894] (3/4) Epoch 3, batch 850, loss[loss=0.3176, simple_loss=0.3703, pruned_loss=0.1324, over 18722.00 frames. ], tot_loss[loss=0.2923, simple_loss=0.3531, pruned_loss=0.1158, over 3660509.81 frames. ], batch size: 52, lr: 3.41e-02, grad_scale: 8.0 2022-12-22 13:35:55,102 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0_sp0.9 from training. Duration: 24.088875 2022-12-22 13:35:55,446 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7863.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:36:00,806 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=7867.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:36:25,331 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp1.1 from training. Duration: 20.4409375 2022-12-22 13:37:07,074 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.016e+02 5.777e+02 7.379e+02 9.226e+02 3.443e+03, threshold=1.476e+03, percent-clipped=5.0 2022-12-22 13:37:09,975 INFO [train.py:894] (3/4) Epoch 3, batch 900, loss[loss=0.324, simple_loss=0.3815, pruned_loss=0.1332, over 18533.00 frames. ], tot_loss[loss=0.293, simple_loss=0.3536, pruned_loss=0.1162, over 3672502.58 frames. ], batch size: 58, lr: 3.41e-02, grad_scale: 8.0 2022-12-22 13:37:36,404 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7931.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:37:41,862 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0_sp0.9 from training. Duration: 22.511125 2022-12-22 13:37:41,884 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0 from training. Duration: 20.675 2022-12-22 13:38:06,747 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.5678, 3.8662, 3.8530, 4.2854, 4.0316, 3.9822, 4.6031, 1.2383], device='cuda:3'), covar=tensor([0.0594, 0.0575, 0.0542, 0.0283, 0.1302, 0.0613, 0.0403, 0.4613], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0135, 0.0116, 0.0094, 0.0182, 0.0135, 0.0135, 0.0185], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002], device='cuda:3') 2022-12-22 13:38:24,402 INFO [train.py:894] (3/4) Epoch 3, batch 950, loss[loss=0.2895, simple_loss=0.3598, pruned_loss=0.1096, over 18721.00 frames. ], tot_loss[loss=0.2924, simple_loss=0.3535, pruned_loss=0.1156, over 3681917.46 frames. ], batch size: 62, lr: 3.40e-02, grad_scale: 8.0 2022-12-22 13:38:26,344 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.9966, 0.4162, 0.8374, 0.5273, 0.9947, 1.0926, 1.0588, 0.7756], device='cuda:3'), covar=tensor([0.0716, 0.0934, 0.1177, 0.0994, 0.0622, 0.0546, 0.0467, 0.0777], device='cuda:3'), in_proj_covar=tensor([0.0063, 0.0099, 0.0101, 0.0121, 0.0099, 0.0066, 0.0064, 0.0098], device='cuda:3'), out_proj_covar=tensor([7.3293e-05, 1.0437e-04, 1.1058e-04, 1.2636e-04, 1.0929e-04, 7.7168e-05, 7.3345e-05, 1.0420e-04], device='cuda:3') 2022-12-22 13:38:43,654 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.1389, 1.0513, 0.3573, 1.1400, 1.0837, 2.2411, 1.1774, 1.1119], device='cuda:3'), covar=tensor([0.1473, 0.2478, 0.2175, 0.1421, 0.2215, 0.0528, 0.1839, 0.2356], device='cuda:3'), in_proj_covar=tensor([0.0089, 0.0090, 0.0094, 0.0088, 0.0106, 0.0078, 0.0097, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 13:38:48,052 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=7979.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:38:49,747 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7980.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:39:25,642 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp0.9 from training. Duration: 24.9833125 2022-12-22 13:39:40,171 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.748e+02 5.844e+02 7.550e+02 9.469e+02 1.849e+03, threshold=1.510e+03, percent-clipped=2.0 2022-12-22 13:39:43,201 INFO [train.py:894] (3/4) Epoch 3, batch 1000, loss[loss=0.2899, simple_loss=0.3404, pruned_loss=0.1197, over 18621.00 frames. ], tot_loss[loss=0.2911, simple_loss=0.3525, pruned_loss=0.1149, over 3688256.13 frames. ], batch size: 41, lr: 3.39e-02, grad_scale: 8.0 2022-12-22 13:39:56,296 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0 from training. Duration: 27.14 2022-12-22 13:40:10,690 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0 from training. Duration: 22.44 2022-12-22 13:40:57,482 INFO [train.py:894] (3/4) Epoch 3, batch 1050, loss[loss=0.2496, simple_loss=0.3217, pruned_loss=0.08879, over 18712.00 frames. ], tot_loss[loss=0.29, simple_loss=0.3523, pruned_loss=0.1139, over 3695036.66 frames. ], batch size: 50, lr: 3.38e-02, grad_scale: 8.0 2022-12-22 13:40:57,893 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8063.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:41:31,056 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0060-62364-0_sp0.9 from training. Duration: 21.361125 2022-12-22 13:41:39,291 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp1.1 from training. Duration: 27.0318125 2022-12-22 13:41:47,121 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp0.9 from training. Duration: 28.638875 2022-12-22 13:41:52,776 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=8099.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:41:53,357 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2022-12-22 13:42:04,467 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0 from training. Duration: 20.4 2022-12-22 13:42:10,268 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.547e+02 6.153e+02 7.411e+02 9.093e+02 1.546e+03, threshold=1.482e+03, percent-clipped=1.0 2022-12-22 13:42:10,464 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=8111.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:42:13,245 INFO [train.py:894] (3/4) Epoch 3, batch 1100, loss[loss=0.3331, simple_loss=0.3852, pruned_loss=0.1405, over 18720.00 frames. ], tot_loss[loss=0.2897, simple_loss=0.3522, pruned_loss=0.1136, over 3699387.32 frames. ], batch size: 54, lr: 3.37e-02, grad_scale: 8.0 2022-12-22 13:42:34,099 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0 from training. Duration: 20.025 2022-12-22 13:42:35,646 WARNING [train.py:1060] (3/4) Exclude cut with ID 2364-131735-0112-64612-0_sp0.9 from training. Duration: 20.488875 2022-12-22 13:42:40,993 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0 from training. Duration: 29.735 2022-12-22 13:42:47,924 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6329, 1.1232, 2.1601, 2.2734, 1.8363, 1.8295, 0.9366, 1.5124], device='cuda:3'), covar=tensor([0.2663, 0.2697, 0.1959, 0.0803, 0.1940, 0.1937, 0.3413, 0.1799], device='cuda:3'), in_proj_covar=tensor([0.0096, 0.0107, 0.0115, 0.0078, 0.0095, 0.0110, 0.0135, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 13:42:48,197 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0178, 1.1131, 1.6001, 1.8085, 1.7624, 1.5351, 1.6865, 1.0476], device='cuda:3'), covar=tensor([0.0705, 0.1646, 0.1460, 0.1214, 0.0561, 0.0565, 0.1081, 0.0764], device='cuda:3'), in_proj_covar=tensor([0.0106, 0.0138, 0.0142, 0.0132, 0.0107, 0.0110, 0.0119, 0.0112], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:3') 2022-12-22 13:43:22,492 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=8158.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:43:29,557 INFO [train.py:894] (3/4) Epoch 3, batch 1150, loss[loss=0.2759, simple_loss=0.355, pruned_loss=0.09836, over 18637.00 frames. ], tot_loss[loss=0.2891, simple_loss=0.3516, pruned_loss=0.1133, over 3702912.16 frames. ], batch size: 53, lr: 3.37e-02, grad_scale: 8.0 2022-12-22 13:44:02,513 WARNING [train.py:1060] (3/4) Exclude cut with ID 7276-92427-0014-12983-0_sp0.9 from training. Duration: 21.3055625 2022-12-22 13:44:03,866 WARNING [train.py:1060] (3/4) Exclude cut with ID 1025-75365-0008-79168-0_sp0.9 from training. Duration: 22.0666875 2022-12-22 13:44:22,945 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3610, 1.3882, 1.3474, 1.2531, 1.2094, 2.1612, 1.0165, 1.6692], device='cuda:3'), covar=tensor([0.4215, 0.2584, 0.2340, 0.3034, 0.2019, 0.0496, 0.2101, 0.1412], device='cuda:3'), in_proj_covar=tensor([0.0153, 0.0130, 0.0142, 0.0129, 0.0136, 0.0089, 0.0119, 0.0121], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], device='cuda:3') 2022-12-22 13:44:39,102 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2022-12-22 13:44:42,384 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.950e+02 6.195e+02 7.771e+02 9.937e+02 1.486e+03, threshold=1.554e+03, percent-clipped=1.0 2022-12-22 13:44:45,409 INFO [train.py:894] (3/4) Epoch 3, batch 1200, loss[loss=0.2916, simple_loss=0.3601, pruned_loss=0.1116, over 18468.00 frames. ], tot_loss[loss=0.2902, simple_loss=0.3528, pruned_loss=0.1138, over 3706827.79 frames. ], batch size: 64, lr: 3.36e-02, grad_scale: 8.0 2022-12-22 13:45:12,964 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0360, 3.3349, 2.5844, 1.5634, 3.0426, 3.1796, 1.8650, 3.6242], device='cuda:3'), covar=tensor([0.1776, 0.0781, 0.2141, 0.2841, 0.1152, 0.1259, 0.2361, 0.0697], device='cuda:3'), in_proj_covar=tensor([0.0185, 0.0145, 0.0184, 0.0182, 0.0175, 0.0185, 0.0181, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 13:45:18,774 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.2680, 3.7098, 3.4945, 3.9778, 3.8295, 3.8908, 4.3809, 1.1703], device='cuda:3'), covar=tensor([0.0660, 0.0572, 0.0670, 0.0368, 0.1454, 0.0770, 0.0503, 0.4548], device='cuda:3'), in_proj_covar=tensor([0.0164, 0.0137, 0.0120, 0.0099, 0.0183, 0.0135, 0.0136, 0.0188], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002], device='cuda:3') 2022-12-22 13:45:50,168 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0 from training. Duration: 20.26 2022-12-22 13:45:58,966 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3685, 1.5107, 1.5137, 1.5265, 1.8336, 2.5814, 1.2526, 1.8310], device='cuda:3'), covar=tensor([0.4358, 0.2523, 0.2134, 0.2546, 0.1617, 0.0419, 0.2145, 0.1431], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0133, 0.0142, 0.0131, 0.0139, 0.0090, 0.0121, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], device='cuda:3') 2022-12-22 13:46:00,043 INFO [train.py:894] (3/4) Epoch 3, batch 1250, loss[loss=0.3236, simple_loss=0.3795, pruned_loss=0.1339, over 18511.00 frames. ], tot_loss[loss=0.2908, simple_loss=0.3535, pruned_loss=0.1141, over 3707498.69 frames. ], batch size: 52, lr: 3.35e-02, grad_scale: 8.0 2022-12-22 13:46:03,447 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8265.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:46:04,551 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0030-9324-0_sp0.9 from training. Duration: 21.3444375 2022-12-22 13:46:26,003 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8280.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:46:36,203 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.72 vs. limit=5.0 2022-12-22 13:47:00,053 WARNING [train.py:1060] (3/4) Exclude cut with ID 497-129325-0061-62254-0_sp1.1 from training. Duration: 0.97725 2022-12-22 13:47:11,732 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.563e+02 6.682e+02 7.893e+02 1.062e+03 4.114e+03, threshold=1.579e+03, percent-clipped=7.0 2022-12-22 13:47:14,622 INFO [train.py:894] (3/4) Epoch 3, batch 1300, loss[loss=0.2934, simple_loss=0.3623, pruned_loss=0.1123, over 18579.00 frames. ], tot_loss[loss=0.2891, simple_loss=0.3519, pruned_loss=0.1131, over 3707375.83 frames. ], batch size: 57, lr: 3.34e-02, grad_scale: 8.0 2022-12-22 13:47:35,048 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8326.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:47:37,583 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=8328.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:47:43,516 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0_sp0.9 from training. Duration: 22.97225 2022-12-22 13:48:10,376 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.54 vs. limit=5.0 2022-12-22 13:48:15,431 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0047-39922-0_sp0.9 from training. Duration: 21.97775 2022-12-22 13:48:20,299 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3451, 2.2006, 1.4252, 0.7321, 1.7535, 1.7993, 1.3416, 1.8212], device='cuda:3'), covar=tensor([0.0917, 0.0649, 0.1927, 0.2586, 0.1550, 0.1743, 0.1848, 0.1327], device='cuda:3'), in_proj_covar=tensor([0.0107, 0.0140, 0.0176, 0.0175, 0.0158, 0.0147, 0.0150, 0.0160], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 13:48:22,150 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 2022-12-22 13:48:27,772 WARNING [train.py:1060] (3/4) Exclude cut with ID 1112-1043-0006-89194-0_sp0.9 from training. Duration: 21.8333125 2022-12-22 13:48:30,480 INFO [train.py:894] (3/4) Epoch 3, batch 1350, loss[loss=0.2822, simple_loss=0.337, pruned_loss=0.1137, over 18385.00 frames. ], tot_loss[loss=0.2874, simple_loss=0.3501, pruned_loss=0.1123, over 3707543.79 frames. ], batch size: 46, lr: 3.34e-02, grad_scale: 8.0 2022-12-22 13:48:37,790 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0 from training. Duration: 20.47 2022-12-22 13:49:26,111 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8399.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:49:42,247 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0037-39912-0_sp0.9 from training. Duration: 20.67225 2022-12-22 13:49:43,613 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.311e+02 6.013e+02 7.179e+02 8.761e+02 1.780e+03, threshold=1.436e+03, percent-clipped=1.0 2022-12-22 13:49:46,612 INFO [train.py:894] (3/4) Epoch 3, batch 1400, loss[loss=0.2611, simple_loss=0.3381, pruned_loss=0.09207, over 18601.00 frames. ], tot_loss[loss=0.2869, simple_loss=0.3502, pruned_loss=0.1118, over 3709951.07 frames. ], batch size: 57, lr: 3.33e-02, grad_scale: 8.0 2022-12-22 13:50:00,323 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp0.9 from training. Duration: 25.2444375 2022-12-22 13:50:24,672 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0021-76797-0_sp0.9 from training. Duration: 21.1445 2022-12-22 13:50:32,270 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.2607, 1.3847, 1.0594, 0.5250, 1.8398, 1.6074, 1.0914, 0.9712], device='cuda:3'), covar=tensor([0.0747, 0.0760, 0.1067, 0.1313, 0.0278, 0.0575, 0.1104, 0.2033], device='cuda:3'), in_proj_covar=tensor([0.0111, 0.0103, 0.0128, 0.0118, 0.0075, 0.0105, 0.0128, 0.0137], device='cuda:3'), out_proj_covar=tensor([1.3741e-04, 1.2982e-04, 1.5359e-04, 1.4227e-04, 9.3399e-05, 1.2937e-04, 1.5355e-04, 1.6096e-04], device='cuda:3') 2022-12-22 13:50:37,685 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=8447.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:50:54,331 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8458.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:50:54,481 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5538, 1.6294, 1.0998, 1.8107, 1.5006, 1.2351, 1.3980, 1.6586], device='cuda:3'), covar=tensor([0.1680, 0.1494, 0.1424, 0.1379, 0.1564, 0.1057, 0.1753, 0.0631], device='cuda:3'), in_proj_covar=tensor([0.0166, 0.0134, 0.0129, 0.0188, 0.0130, 0.0126, 0.0149, 0.0107], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2022-12-22 13:51:01,639 INFO [train.py:894] (3/4) Epoch 3, batch 1450, loss[loss=0.3198, simple_loss=0.3606, pruned_loss=0.1395, over 18437.00 frames. ], tot_loss[loss=0.2888, simple_loss=0.3518, pruned_loss=0.1129, over 3711239.76 frames. ], batch size: 48, lr: 3.32e-02, grad_scale: 8.0 2022-12-22 13:51:29,534 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.5430, 2.1158, 2.2793, 0.5746, 1.7405, 3.3527, 1.5062, 2.0688], device='cuda:3'), covar=tensor([0.1463, 0.0645, 0.1520, 0.1248, 0.1087, 0.0162, 0.1013, 0.0911], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0098, 0.0071, 0.0105, 0.0089, 0.0062, 0.0103, 0.0077], device='cuda:3'), out_proj_covar=tensor([1.0516e-04, 1.1130e-04, 8.6029e-05, 1.1755e-04, 1.0170e-04, 7.3452e-05, 1.1888e-04, 8.9515e-05], device='cuda:3') 2022-12-22 13:51:37,937 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2022-12-22 13:51:38,560 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp0.9 from training. Duration: 33.038875 2022-12-22 13:52:05,964 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=8506.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:52:13,219 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.559e+02 6.368e+02 8.032e+02 1.005e+03 2.835e+03, threshold=1.606e+03, percent-clipped=10.0 2022-12-22 13:52:15,315 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64291-0000-16059-0_sp0.9 from training. Duration: 20.0944375 2022-12-22 13:52:16,677 INFO [train.py:894] (3/4) Epoch 3, batch 1500, loss[loss=0.32, simple_loss=0.3709, pruned_loss=0.1346, over 18721.00 frames. ], tot_loss[loss=0.2888, simple_loss=0.3515, pruned_loss=0.1131, over 3712163.58 frames. ], batch size: 52, lr: 3.32e-02, grad_scale: 8.0 2022-12-22 13:52:30,709 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp1.1 from training. Duration: 20.4 2022-12-22 13:52:39,414 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0 from training. Duration: 20.085 2022-12-22 13:52:49,629 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0_sp0.9 from training. Duration: 23.07775 2022-12-22 13:53:12,447 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3627, 2.0431, 1.2282, 2.2232, 2.4973, 1.3299, 1.8340, 1.1232], device='cuda:3'), covar=tensor([0.1798, 0.1246, 0.1421, 0.0740, 0.0962, 0.1336, 0.1324, 0.1631], device='cuda:3'), in_proj_covar=tensor([0.0166, 0.0138, 0.0138, 0.0126, 0.0169, 0.0137, 0.0138, 0.0140], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2022-12-22 13:53:30,919 INFO [train.py:894] (3/4) Epoch 3, batch 1550, loss[loss=0.2781, simple_loss=0.3459, pruned_loss=0.1051, over 18567.00 frames. ], tot_loss[loss=0.2878, simple_loss=0.3505, pruned_loss=0.1126, over 3712657.69 frames. ], batch size: 57, lr: 3.31e-02, grad_scale: 8.0 2022-12-22 13:53:37,289 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp0.9 from training. Duration: 24.9333125 2022-12-22 13:54:06,171 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2022-12-22 13:54:20,301 WARNING [train.py:1060] (3/4) Exclude cut with ID 5118-111612-0016-124680-0_sp0.9 from training. Duration: 20.388875 2022-12-22 13:54:26,227 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp1.1 from training. Duration: 20.3590625 2022-12-22 13:54:43,756 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.316e+02 6.096e+02 7.223e+02 9.116e+02 1.940e+03, threshold=1.445e+03, percent-clipped=4.0 2022-12-22 13:54:47,240 INFO [train.py:894] (3/4) Epoch 3, batch 1600, loss[loss=0.2135, simple_loss=0.2894, pruned_loss=0.06879, over 18586.00 frames. ], tot_loss[loss=0.2863, simple_loss=0.349, pruned_loss=0.1118, over 3712469.33 frames. ], batch size: 45, lr: 3.30e-02, grad_scale: 8.0 2022-12-22 13:54:59,165 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=8621.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:55:33,367 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0_sp1.1 from training. Duration: 0.836375 2022-12-22 13:56:01,309 INFO [train.py:894] (3/4) Epoch 3, batch 1650, loss[loss=0.3095, simple_loss=0.3661, pruned_loss=0.1265, over 18467.00 frames. ], tot_loss[loss=0.2877, simple_loss=0.3495, pruned_loss=0.1129, over 3712480.87 frames. ], batch size: 50, lr: 3.29e-02, grad_scale: 8.0 2022-12-22 13:56:19,160 WARNING [train.py:1060] (3/4) Exclude cut with ID 8565-290391-0049-67394-0_sp0.9 from training. Duration: 21.3166875 2022-12-22 13:56:45,753 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4524, 2.0451, 1.6822, 0.7193, 1.8658, 1.8429, 1.3040, 1.9215], device='cuda:3'), covar=tensor([0.0626, 0.0713, 0.1704, 0.2426, 0.1373, 0.1548, 0.1937, 0.1198], device='cuda:3'), in_proj_covar=tensor([0.0112, 0.0146, 0.0186, 0.0180, 0.0168, 0.0150, 0.0155, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 13:56:49,671 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0029-104863-0_sp0.9 from training. Duration: 22.1055625 2022-12-22 13:56:58,499 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp1.1 from training. Duration: 21.77725 2022-12-22 13:57:12,831 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.680e+02 6.678e+02 8.781e+02 1.140e+03 4.926e+03, threshold=1.756e+03, percent-clipped=10.0 2022-12-22 13:57:15,871 INFO [train.py:894] (3/4) Epoch 3, batch 1700, loss[loss=0.3215, simple_loss=0.3788, pruned_loss=0.1321, over 18637.00 frames. ], tot_loss[loss=0.2947, simple_loss=0.3532, pruned_loss=0.1181, over 3712795.07 frames. ], batch size: 53, lr: 3.29e-02, grad_scale: 8.0 2022-12-22 13:57:18,526 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.76 vs. limit=2.0 2022-12-22 13:57:18,764 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp0.9 from training. Duration: 27.8166875 2022-12-22 13:57:43,458 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp1.1 from training. Duration: 22.5090625 2022-12-22 13:57:49,866 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0 from training. Duration: 25.035 2022-12-22 13:57:54,232 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0206, 2.2777, 1.4193, 1.1116, 2.8301, 2.6248, 1.4674, 1.3583], device='cuda:3'), covar=tensor([0.0836, 0.0617, 0.1385, 0.1460, 0.0191, 0.0483, 0.1259, 0.1848], device='cuda:3'), in_proj_covar=tensor([0.0117, 0.0107, 0.0132, 0.0124, 0.0078, 0.0111, 0.0131, 0.0146], device='cuda:3'), out_proj_covar=tensor([1.4364e-04, 1.3415e-04, 1.5826e-04, 1.4975e-04, 9.7589e-05, 1.3611e-04, 1.5777e-04, 1.7161e-04], device='cuda:3') 2022-12-22 13:58:08,302 WARNING [train.py:1060] (3/4) Exclude cut with ID 774-127930-0014-10412-0_sp1.1 from training. Duration: 0.95 2022-12-22 13:58:26,853 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp0.9 from training. Duration: 0.92225 2022-12-22 13:58:27,125 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8760.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 13:58:31,072 INFO [train.py:894] (3/4) Epoch 3, batch 1750, loss[loss=0.3413, simple_loss=0.3895, pruned_loss=0.1465, over 18519.00 frames. ], tot_loss[loss=0.3015, simple_loss=0.3571, pruned_loss=0.123, over 3713586.16 frames. ], batch size: 77, lr: 3.28e-02, grad_scale: 8.0 2022-12-22 13:58:55,292 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0 from training. Duration: 21.97 2022-12-22 13:59:12,780 WARNING [train.py:1060] (3/4) Exclude cut with ID 7492-105653-0055-62765-0_sp0.9 from training. Duration: 21.97225 2022-12-22 13:59:14,178 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp0.9 from training. Duration: 25.3333125 2022-12-22 13:59:22,892 WARNING [train.py:1060] (3/4) Exclude cut with ID 5172-29468-0015-19128-0_sp0.9 from training. Duration: 21.5055625 2022-12-22 13:59:34,766 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp1.1 from training. Duration: 20.72725 2022-12-22 13:59:43,096 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.523e+02 7.062e+02 8.790e+02 1.163e+03 3.166e+03, threshold=1.758e+03, percent-clipped=4.0 2022-12-22 13:59:46,003 INFO [train.py:894] (3/4) Epoch 3, batch 1800, loss[loss=0.3285, simple_loss=0.3747, pruned_loss=0.1411, over 18728.00 frames. ], tot_loss[loss=0.3078, simple_loss=0.3602, pruned_loss=0.1276, over 3714338.27 frames. ], batch size: 52, lr: 3.27e-02, grad_scale: 8.0 2022-12-22 13:59:59,149 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8821.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:00:03,546 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp0.9 from training. Duration: 26.32775 2022-12-22 14:00:36,950 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0 from training. Duration: 20.025 2022-12-22 14:00:43,604 WARNING [train.py:1060] (3/4) Exclude cut with ID 6709-74022-0004-86860-0_sp1.1 from training. Duration: 0.9409375 2022-12-22 14:00:43,615 WARNING [train.py:1060] (3/4) Exclude cut with ID 4757-1811-0023-62229-0_sp0.9 from training. Duration: 21.37775 2022-12-22 14:01:02,105 INFO [train.py:894] (3/4) Epoch 3, batch 1850, loss[loss=0.3457, simple_loss=0.3887, pruned_loss=0.1513, over 18593.00 frames. ], tot_loss[loss=0.3119, simple_loss=0.3623, pruned_loss=0.1308, over 3714513.17 frames. ], batch size: 51, lr: 3.27e-02, grad_scale: 8.0 2022-12-22 14:01:04,289 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0004-25974-0_sp0.9 from training. Duration: 21.17225 2022-12-22 14:01:04,305 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp0.9 from training. Duration: 27.511125 2022-12-22 14:01:35,353 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0 from training. Duration: 22.8 2022-12-22 14:01:39,523 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0 from training. Duration: 22.585 2022-12-22 14:02:09,077 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0001-146967-0_sp0.9 from training. Duration: 22.0166875 2022-12-22 14:02:15,219 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.126e+02 6.920e+02 8.514e+02 1.073e+03 2.311e+03, threshold=1.703e+03, percent-clipped=5.0 2022-12-22 14:02:18,066 INFO [train.py:894] (3/4) Epoch 3, batch 1900, loss[loss=0.3134, simple_loss=0.3728, pruned_loss=0.127, over 18508.00 frames. ], tot_loss[loss=0.3157, simple_loss=0.3641, pruned_loss=0.1337, over 3713282.03 frames. ], batch size: 58, lr: 3.26e-02, grad_scale: 8.0 2022-12-22 14:02:25,468 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp1.1 from training. Duration: 24.395375 2022-12-22 14:02:30,030 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8921.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:02:32,800 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp0.9 from training. Duration: 27.47775 2022-12-22 14:02:37,729 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp0.9 from training. Duration: 24.8833125 2022-12-22 14:02:39,400 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8927.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:02:40,284 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0 from training. Duration: 23.39 2022-12-22 14:02:40,602 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4131, 1.2573, 0.7420, 1.4250, 1.1820, 2.8288, 1.3281, 1.2904], device='cuda:3'), covar=tensor([0.1338, 0.2295, 0.2122, 0.1409, 0.2217, 0.0456, 0.1756, 0.2121], device='cuda:3'), in_proj_covar=tensor([0.0091, 0.0091, 0.0095, 0.0091, 0.0112, 0.0080, 0.0098, 0.0091], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-22 14:02:45,106 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp0.9 from training. Duration: 28.72225 2022-12-22 14:02:55,068 WARNING [train.py:1060] (3/4) Exclude cut with ID 585-294811-0110-133686-0_sp0.9 from training. Duration: 20.8944375 2022-12-22 14:02:55,506 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.9611, 2.3397, 1.6242, 2.8141, 2.0483, 2.0814, 2.2840, 3.7816], device='cuda:3'), covar=tensor([0.1103, 0.1371, 0.1011, 0.1526, 0.1635, 0.0667, 0.1598, 0.0298], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0136, 0.0134, 0.0199, 0.0135, 0.0129, 0.0153, 0.0112], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2022-12-22 14:03:09,382 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0_sp0.9 from training. Duration: 23.8444375 2022-12-22 14:03:32,684 INFO [train.py:894] (3/4) Epoch 3, batch 1950, loss[loss=0.3129, simple_loss=0.3657, pruned_loss=0.13, over 18723.00 frames. ], tot_loss[loss=0.3172, simple_loss=0.3648, pruned_loss=0.1348, over 3714470.53 frames. ], batch size: 52, lr: 3.25e-02, grad_scale: 8.0 2022-12-22 14:03:34,155 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0 from training. Duration: 25.85 2022-12-22 14:03:34,164 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0 from training. Duration: 21.39 2022-12-22 14:03:41,378 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=8969.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:03:47,645 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0 from training. Duration: 27.92 2022-12-22 14:04:04,713 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2022-12-22 14:04:07,528 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8986.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:04:10,627 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8988.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:04:16,388 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0039-130165-0_sp0.9 from training. Duration: 20.661125 2022-12-22 14:04:40,577 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0043-15874-0_sp0.9 from training. Duration: 20.07225 2022-12-22 14:04:44,895 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.540e+02 6.702e+02 8.505e+02 1.065e+03 2.063e+03, threshold=1.701e+03, percent-clipped=2.0 2022-12-22 14:04:47,920 INFO [train.py:894] (3/4) Epoch 3, batch 2000, loss[loss=0.3364, simple_loss=0.3849, pruned_loss=0.144, over 18660.00 frames. ], tot_loss[loss=0.3176, simple_loss=0.3649, pruned_loss=0.1352, over 3713550.97 frames. ], batch size: 60, lr: 3.24e-02, grad_scale: 8.0 2022-12-22 14:04:47,996 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0 from training. Duration: 21.01 2022-12-22 14:04:48,281 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9013.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:05:38,971 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9047.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:05:57,228 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0 from training. Duration: 20.65 2022-12-22 14:06:03,149 INFO [train.py:894] (3/4) Epoch 3, batch 2050, loss[loss=0.2824, simple_loss=0.3425, pruned_loss=0.1111, over 18418.00 frames. ], tot_loss[loss=0.3209, simple_loss=0.3667, pruned_loss=0.1376, over 3713992.39 frames. ], batch size: 48, lr: 3.24e-02, grad_scale: 8.0 2022-12-22 14:06:03,179 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0 from training. Duration: 21.46 2022-12-22 14:06:03,620 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7124, 1.4818, 1.0958, 0.3313, 1.4818, 1.5133, 1.1102, 1.3170], device='cuda:3'), covar=tensor([0.0717, 0.0577, 0.1168, 0.1734, 0.1010, 0.1475, 0.1566, 0.0909], device='cuda:3'), in_proj_covar=tensor([0.0116, 0.0149, 0.0189, 0.0182, 0.0169, 0.0155, 0.0160, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 14:06:20,853 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9074.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:06:47,926 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0 from training. Duration: 0.92 2022-12-22 14:06:50,155 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7758, 2.3373, 2.1634, 0.9805, 2.1513, 2.1389, 1.4395, 1.7065], device='cuda:3'), covar=tensor([0.0634, 0.0645, 0.1578, 0.2317, 0.1341, 0.1212, 0.1640, 0.1531], device='cuda:3'), in_proj_covar=tensor([0.0117, 0.0150, 0.0191, 0.0183, 0.0171, 0.0156, 0.0164, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 14:06:55,389 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0_sp0.9 from training. Duration: 23.7666875 2022-12-22 14:07:16,204 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.185e+02 7.079e+02 8.812e+02 1.071e+03 1.633e+03, threshold=1.762e+03, percent-clipped=0.0 2022-12-22 14:07:19,199 INFO [train.py:894] (3/4) Epoch 3, batch 2100, loss[loss=0.2816, simple_loss=0.3237, pruned_loss=0.1197, over 18392.00 frames. ], tot_loss[loss=0.3218, simple_loss=0.3672, pruned_loss=0.1382, over 3714611.72 frames. ], batch size: 42, lr: 3.23e-02, grad_scale: 16.0 2022-12-22 14:07:24,387 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9116.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:07:31,849 WARNING [train.py:1060] (3/4) Exclude cut with ID 8544-281189-0060-101339-0_sp0.9 from training. Duration: 20.861125 2022-12-22 14:07:42,337 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0_sp0.9 from training. Duration: 22.711125 2022-12-22 14:08:21,787 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp1.1 from training. Duration: 22.986375 2022-12-22 14:08:30,037 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2022-12-22 14:08:35,389 INFO [train.py:894] (3/4) Epoch 3, batch 2150, loss[loss=0.3279, simple_loss=0.377, pruned_loss=0.1394, over 18449.00 frames. ], tot_loss[loss=0.3218, simple_loss=0.367, pruned_loss=0.1383, over 3714198.44 frames. ], batch size: 64, lr: 3.22e-02, grad_scale: 8.0 2022-12-22 14:08:39,775 WARNING [train.py:1060] (3/4) Exclude cut with ID 8040-260924-0003-80960-0_sp0.9 from training. Duration: 22.07225 2022-12-22 14:08:43,573 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0045-26330-0_sp0.9 from training. Duration: 20.3055625 2022-12-22 14:08:45,348 WARNING [train.py:1060] (3/4) Exclude cut with ID 6356-271890-0060-94317-0_sp0.9 from training. Duration: 20.72225 2022-12-22 14:09:03,833 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp1.1 from training. Duration: 22.4818125 2022-12-22 14:09:30,338 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp0.9 from training. Duration: 25.0944375 2022-12-22 14:09:34,928 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0 from training. Duration: 21.515 2022-12-22 14:09:40,683 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp0.9 from training. Duration: 27.02225 2022-12-22 14:09:46,763 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0010-62480-0_sp0.9 from training. Duration: 22.22225 2022-12-22 14:09:48,382 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.146e+02 7.205e+02 8.913e+02 1.057e+03 2.694e+03, threshold=1.783e+03, percent-clipped=4.0 2022-12-22 14:09:49,871 INFO [train.py:894] (3/4) Epoch 3, batch 2200, loss[loss=0.3254, simple_loss=0.378, pruned_loss=0.1364, over 18530.00 frames. ], tot_loss[loss=0.3224, simple_loss=0.3674, pruned_loss=0.1387, over 3714389.35 frames. ], batch size: 58, lr: 3.22e-02, grad_scale: 8.0 2022-12-22 14:09:52,943 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0085-44554-0_sp0.9 from training. Duration: 20.85 2022-12-22 14:09:54,700 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3193, 1.3428, 1.7829, 1.4752, 1.9600, 2.7864, 2.5015, 2.1538], device='cuda:3'), covar=tensor([0.0870, 0.0950, 0.0864, 0.0864, 0.0614, 0.0397, 0.0423, 0.0743], device='cuda:3'), in_proj_covar=tensor([0.0065, 0.0100, 0.0106, 0.0120, 0.0097, 0.0068, 0.0063, 0.0101], device='cuda:3'), out_proj_covar=tensor([7.3494e-05, 1.0481e-04, 1.1530e-04, 1.2524e-04, 1.0622e-04, 7.4875e-05, 7.1462e-05, 1.0663e-04], device='cuda:3') 2022-12-22 14:10:23,526 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5149, 0.9267, 1.0559, 1.0613, 1.5905, 1.0475, 1.3756, 1.6821], device='cuda:3'), covar=tensor([0.1896, 0.2393, 0.2396, 0.2003, 0.2465, 0.1814, 0.1662, 0.1648], device='cuda:3'), in_proj_covar=tensor([0.0087, 0.0104, 0.0132, 0.0098, 0.0106, 0.0099, 0.0091, 0.0102], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 14:10:29,118 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0 from training. Duration: 21.54 2022-12-22 14:10:33,657 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp1.1 from training. Duration: 20.5318125 2022-12-22 14:10:43,912 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0012-134311-0_sp0.9 from training. Duration: 21.9333125 2022-12-22 14:11:03,641 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8472, 2.0763, 2.2417, 1.4447, 2.2850, 2.2302, 1.4984, 2.6154], device='cuda:3'), covar=tensor([0.1488, 0.1131, 0.1634, 0.2412, 0.1053, 0.1397, 0.2546, 0.0705], device='cuda:3'), in_proj_covar=tensor([0.0190, 0.0150, 0.0187, 0.0184, 0.0178, 0.0196, 0.0186, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 14:11:04,628 INFO [train.py:894] (3/4) Epoch 3, batch 2250, loss[loss=0.3291, simple_loss=0.373, pruned_loss=0.1427, over 18700.00 frames. ], tot_loss[loss=0.32, simple_loss=0.3653, pruned_loss=0.1374, over 3713375.32 frames. ], batch size: 98, lr: 3.21e-02, grad_scale: 8.0 2022-12-22 14:11:32,500 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0025-130151-0_sp0.9 from training. Duration: 21.7944375 2022-12-22 14:11:35,404 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9283.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:11:46,357 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0_sp0.9 from training. Duration: 22.4666875 2022-12-22 14:11:52,190 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0 from training. Duration: 21.635 2022-12-22 14:11:58,065 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0_sp0.9 from training. Duration: 24.038875 2022-12-22 14:12:18,741 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.113e+02 7.064e+02 8.441e+02 1.095e+03 1.855e+03, threshold=1.688e+03, percent-clipped=2.0 2022-12-22 14:12:20,129 INFO [train.py:894] (3/4) Epoch 3, batch 2300, loss[loss=0.2899, simple_loss=0.3464, pruned_loss=0.1166, over 18414.00 frames. ], tot_loss[loss=0.3187, simple_loss=0.3644, pruned_loss=0.1365, over 3713270.28 frames. ], batch size: 48, lr: 3.20e-02, grad_scale: 8.0 2022-12-22 14:12:41,353 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp1.1 from training. Duration: 21.786375 2022-12-22 14:12:52,379 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0 from training. Duration: 20.22 2022-12-22 14:13:06,085 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9342.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:13:37,570 INFO [train.py:894] (3/4) Epoch 3, batch 2350, loss[loss=0.2401, simple_loss=0.3036, pruned_loss=0.08825, over 18714.00 frames. ], tot_loss[loss=0.3189, simple_loss=0.3648, pruned_loss=0.1365, over 3712784.92 frames. ], batch size: 46, lr: 3.20e-02, grad_scale: 8.0 2022-12-22 14:13:48,347 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9369.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:13:55,749 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6431, 1.6518, 0.9000, 1.5184, 1.7822, 1.6286, 2.6356, 1.9446], device='cuda:3'), covar=tensor([0.1015, 0.1338, 0.2508, 0.1746, 0.1800, 0.0848, 0.0501, 0.0959], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0141, 0.0170, 0.0209, 0.0184, 0.0135, 0.0118, 0.0143], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001], device='cuda:3') 2022-12-22 14:14:48,722 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2022-12-22 14:14:50,984 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0 from training. Duration: 25.285 2022-12-22 14:14:53,183 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.774e+02 6.974e+02 8.877e+02 1.045e+03 2.873e+03, threshold=1.775e+03, percent-clipped=4.0 2022-12-22 14:14:54,866 INFO [train.py:894] (3/4) Epoch 3, batch 2400, loss[loss=0.2886, simple_loss=0.3296, pruned_loss=0.1238, over 18437.00 frames. ], tot_loss[loss=0.3186, simple_loss=0.3644, pruned_loss=0.1364, over 3712185.32 frames. ], batch size: 42, lr: 3.19e-02, grad_scale: 8.0 2022-12-22 14:14:56,768 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9414.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:15:00,147 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9416.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:15:01,365 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4808, 1.2877, 0.7835, 1.6335, 1.3174, 2.9695, 1.2392, 1.1959], device='cuda:3'), covar=tensor([0.1449, 0.2407, 0.2171, 0.1391, 0.2239, 0.0443, 0.2048, 0.2332], device='cuda:3'), in_proj_covar=tensor([0.0092, 0.0091, 0.0093, 0.0091, 0.0112, 0.0078, 0.0101, 0.0091], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-22 14:15:57,211 WARNING [train.py:1060] (3/4) Exclude cut with ID 811-130148-0001-63453-0_sp0.9 from training. Duration: 20.861125 2022-12-22 14:16:09,909 INFO [train.py:894] (3/4) Epoch 3, batch 2450, loss[loss=0.3555, simple_loss=0.394, pruned_loss=0.1585, over 18678.00 frames. ], tot_loss[loss=0.3179, simple_loss=0.3638, pruned_loss=0.1361, over 3712543.31 frames. ], batch size: 62, lr: 3.18e-02, grad_scale: 8.0 2022-12-22 14:16:11,467 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=9464.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:16:15,265 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0737, 1.7751, 1.2964, 2.3352, 1.6622, 4.3099, 1.9388, 1.6420], device='cuda:3'), covar=tensor([0.1152, 0.2045, 0.1840, 0.1356, 0.1984, 0.0224, 0.1527, 0.2110], device='cuda:3'), in_proj_covar=tensor([0.0090, 0.0091, 0.0094, 0.0091, 0.0112, 0.0076, 0.0099, 0.0091], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-22 14:16:20,917 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0 from training. Duration: 20.88 2022-12-22 14:16:28,087 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9475.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:16:52,493 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0_sp0.9 from training. Duration: 23.4166875 2022-12-22 14:16:59,756 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2022-12-22 14:17:24,908 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.333e+02 7.255e+02 8.783e+02 1.029e+03 1.927e+03, threshold=1.757e+03, percent-clipped=1.0 2022-12-22 14:17:25,418 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6963, 0.7501, 1.5459, 1.2717, 1.3527, 1.3681, 1.7061, 1.3480], device='cuda:3'), covar=tensor([0.0985, 0.0974, 0.0989, 0.0872, 0.0750, 0.0633, 0.0461, 0.0860], device='cuda:3'), in_proj_covar=tensor([0.0067, 0.0102, 0.0112, 0.0123, 0.0100, 0.0072, 0.0068, 0.0107], device='cuda:3'), out_proj_covar=tensor([7.5607e-05, 1.0575e-04, 1.2083e-04, 1.2799e-04, 1.0930e-04, 7.9336e-05, 7.5615e-05, 1.1305e-04], device='cuda:3') 2022-12-22 14:17:26,938 INFO [train.py:894] (3/4) Epoch 3, batch 2500, loss[loss=0.2902, simple_loss=0.3312, pruned_loss=0.1246, over 18683.00 frames. ], tot_loss[loss=0.318, simple_loss=0.3638, pruned_loss=0.1361, over 3713462.62 frames. ], batch size: 46, lr: 3.18e-02, grad_scale: 8.0 2022-12-22 14:17:27,384 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7816, 1.9404, 1.1335, 0.8874, 2.3175, 2.0065, 1.2741, 1.1370], device='cuda:3'), covar=tensor([0.0651, 0.0528, 0.1291, 0.1461, 0.0253, 0.0601, 0.1270, 0.2093], device='cuda:3'), in_proj_covar=tensor([0.0118, 0.0110, 0.0136, 0.0125, 0.0077, 0.0114, 0.0137, 0.0149], device='cuda:3'), out_proj_covar=tensor([1.4683e-04, 1.3866e-04, 1.6414e-04, 1.5225e-04, 9.5712e-05, 1.3982e-04, 1.6699e-04, 1.7625e-04], device='cuda:3') 2022-12-22 14:18:07,759 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0 from training. Duration: 21.24 2022-12-22 14:18:09,195 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0_sp0.9 from training. Duration: 23.9055625 2022-12-22 14:18:15,798 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2022-12-22 14:18:42,849 INFO [train.py:894] (3/4) Epoch 3, batch 2550, loss[loss=0.2978, simple_loss=0.3618, pruned_loss=0.1169, over 18649.00 frames. ], tot_loss[loss=0.3161, simple_loss=0.3623, pruned_loss=0.1349, over 3713336.06 frames. ], batch size: 62, lr: 3.17e-02, grad_scale: 8.0 2022-12-22 14:18:42,868 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp0.9 from training. Duration: 25.988875 2022-12-22 14:18:51,355 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0001-134300-0_sp0.9 from training. Duration: 20.67225 2022-12-22 14:19:12,534 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9583.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:19:15,059 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9584.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:19:38,077 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0 from training. Duration: 20.34 2022-12-22 14:19:58,220 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.647e+02 7.260e+02 8.987e+02 1.119e+03 2.066e+03, threshold=1.797e+03, percent-clipped=2.0 2022-12-22 14:19:59,699 INFO [train.py:894] (3/4) Epoch 3, batch 2600, loss[loss=0.3521, simple_loss=0.3942, pruned_loss=0.155, over 18498.00 frames. ], tot_loss[loss=0.3154, simple_loss=0.3617, pruned_loss=0.1345, over 3712963.38 frames. ], batch size: 69, lr: 3.16e-02, grad_scale: 8.0 2022-12-22 14:20:20,904 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7637, 2.3627, 1.8866, 1.0990, 1.9629, 2.0909, 1.5106, 1.6635], device='cuda:3'), covar=tensor([0.0753, 0.0617, 0.1757, 0.2433, 0.1668, 0.1390, 0.1741, 0.1472], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0151, 0.0196, 0.0187, 0.0178, 0.0158, 0.0168, 0.0173], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 14:20:26,417 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=9631.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:20:43,645 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9642.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:20:48,849 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9645.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:20:51,252 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp0.9 from training. Duration: 25.061125 2022-12-22 14:21:02,773 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0 from training. Duration: 0.83 2022-12-22 14:21:07,141 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9658.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:21:10,708 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2022-12-22 14:21:14,220 INFO [train.py:894] (3/4) Epoch 3, batch 2650, loss[loss=0.2652, simple_loss=0.31, pruned_loss=0.1102, over 18486.00 frames. ], tot_loss[loss=0.317, simple_loss=0.3632, pruned_loss=0.1354, over 3713014.99 frames. ], batch size: 43, lr: 3.16e-02, grad_scale: 8.0 2022-12-22 14:21:23,371 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9669.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:21:27,992 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2022-12-22 14:21:28,863 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0 from training. Duration: 24.73 2022-12-22 14:21:43,017 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0 from training. Duration: 23.965 2022-12-22 14:21:52,804 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0030-146996-0_sp0.9 from training. Duration: 22.088875 2022-12-22 14:21:55,804 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=9690.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:22:09,365 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0_sp0.9 from training. Duration: 23.6 2022-12-22 14:22:28,387 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.461e+02 7.252e+02 8.983e+02 1.112e+03 2.476e+03, threshold=1.797e+03, percent-clipped=3.0 2022-12-22 14:22:29,673 INFO [train.py:894] (3/4) Epoch 3, batch 2700, loss[loss=0.2877, simple_loss=0.3254, pruned_loss=0.125, over 18589.00 frames. ], tot_loss[loss=0.3165, simple_loss=0.3625, pruned_loss=0.1353, over 3713089.45 frames. ], batch size: 45, lr: 3.15e-02, grad_scale: 8.0 2022-12-22 14:22:35,626 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=9717.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:22:38,877 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9719.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:23:16,447 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9743.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:23:45,413 INFO [train.py:894] (3/4) Epoch 3, batch 2750, loss[loss=0.2498, simple_loss=0.3143, pruned_loss=0.09264, over 18385.00 frames. ], tot_loss[loss=0.3173, simple_loss=0.3632, pruned_loss=0.1357, over 3712886.83 frames. ], batch size: 46, lr: 3.15e-02, grad_scale: 8.0 2022-12-22 14:23:49,551 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0 from training. Duration: 23.795 2022-12-22 14:23:56,056 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9770.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:24:06,601 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp1.1 from training. Duration: 21.5409375 2022-12-22 14:24:09,284 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp0.9 from training. Duration: 24.97775 2022-12-22 14:24:19,240 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0_sp0.9 from training. Duration: 23.3444375 2022-12-22 14:24:24,365 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1702, 1.6884, 1.6413, 2.8190, 1.9431, 4.2951, 1.9187, 1.6676], device='cuda:3'), covar=tensor([0.0976, 0.1878, 0.1435, 0.0847, 0.1602, 0.0208, 0.1364, 0.1812], device='cuda:3'), in_proj_covar=tensor([0.0092, 0.0093, 0.0094, 0.0090, 0.0112, 0.0079, 0.0100, 0.0092], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], device='cuda:3') 2022-12-22 14:24:33,667 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.9594, 3.5550, 3.7974, 2.4510, 3.3777, 2.5481, 1.9892, 2.8170], device='cuda:3'), covar=tensor([0.1839, 0.0822, 0.1173, 0.2543, 0.1009, 0.1353, 0.3504, 0.1800], device='cuda:3'), in_proj_covar=tensor([0.0117, 0.0087, 0.0144, 0.0111, 0.0098, 0.0095, 0.0141, 0.0109], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-22 14:24:45,311 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0_sp0.9 from training. Duration: 23.2 2022-12-22 14:24:47,107 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9804.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:24:51,273 WARNING [train.py:1060] (3/4) Exclude cut with ID 5653-46179-0060-117930-0_sp0.9 from training. Duration: 21.17225 2022-12-22 14:24:57,994 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.440e+02 7.867e+02 9.020e+02 1.115e+03 2.564e+03, threshold=1.804e+03, percent-clipped=3.0 2022-12-22 14:24:59,519 INFO [train.py:894] (3/4) Epoch 3, batch 2800, loss[loss=0.2996, simple_loss=0.3541, pruned_loss=0.1225, over 18528.00 frames. ], tot_loss[loss=0.316, simple_loss=0.3621, pruned_loss=0.1349, over 3712812.80 frames. ], batch size: 55, lr: 3.14e-02, grad_scale: 8.0 2022-12-22 14:25:11,489 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp0.9 from training. Duration: 24.6555625 2022-12-22 14:25:17,616 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3531, 1.4504, 1.4834, 1.7013, 1.7857, 3.2393, 1.3142, 2.1780], device='cuda:3'), covar=tensor([0.4219, 0.2560, 0.2263, 0.2425, 0.1638, 0.0247, 0.1919, 0.1177], device='cuda:3'), in_proj_covar=tensor([0.0157, 0.0133, 0.0148, 0.0130, 0.0136, 0.0093, 0.0118, 0.0120], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], device='cuda:3') 2022-12-22 14:26:06,606 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0 from training. Duration: 20.44 2022-12-22 14:26:15,232 INFO [train.py:894] (3/4) Epoch 3, batch 2850, loss[loss=0.2868, simple_loss=0.3303, pruned_loss=0.1217, over 18522.00 frames. ], tot_loss[loss=0.3163, simple_loss=0.3622, pruned_loss=0.1352, over 3713223.74 frames. ], batch size: 44, lr: 3.13e-02, grad_scale: 8.0 2022-12-22 14:26:21,837 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0_sp0.9 from training. Duration: 23.45 2022-12-22 14:26:50,287 WARNING [train.py:1060] (3/4) Exclude cut with ID 6945-60535-0076-12784-0_sp0.9 from training. Duration: 20.52225 2022-12-22 14:26:52,154 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.5420, 3.7625, 3.8518, 4.2505, 3.9314, 4.0393, 4.6721, 1.2237], device='cuda:3'), covar=tensor([0.0693, 0.0644, 0.0607, 0.0293, 0.1718, 0.0850, 0.0486, 0.4409], device='cuda:3'), in_proj_covar=tensor([0.0184, 0.0146, 0.0135, 0.0105, 0.0200, 0.0153, 0.0151, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-22 14:26:57,805 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0 from training. Duration: 22.19 2022-12-22 14:26:58,464 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.88 vs. limit=5.0 2022-12-22 14:27:08,360 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp1.1 from training. Duration: 25.3818125 2022-12-22 14:27:25,601 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp0.9 from training. Duration: 28.0944375 2022-12-22 14:27:29,796 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.837e+02 6.577e+02 8.557e+02 1.075e+03 2.821e+03, threshold=1.711e+03, percent-clipped=5.0 2022-12-22 14:27:31,272 INFO [train.py:894] (3/4) Epoch 3, batch 2900, loss[loss=0.3225, simple_loss=0.3642, pruned_loss=0.1404, over 18535.00 frames. ], tot_loss[loss=0.3152, simple_loss=0.3617, pruned_loss=0.1343, over 3713457.69 frames. ], batch size: 55, lr: 3.13e-02, grad_scale: 8.0 2022-12-22 14:27:33,444 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0_sp0.9 from training. Duration: 22.9444375 2022-12-22 14:27:41,105 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp1.1 from training. Duration: 21.6318125 2022-12-22 14:27:59,979 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0 from training. Duration: 23.695 2022-12-22 14:28:13,381 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9940.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:28:24,946 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0 from training. Duration: 23.955 2022-12-22 14:28:47,822 INFO [train.py:894] (3/4) Epoch 3, batch 2950, loss[loss=0.3359, simple_loss=0.3775, pruned_loss=0.1472, over 18613.00 frames. ], tot_loss[loss=0.3143, simple_loss=0.3612, pruned_loss=0.1337, over 3713076.60 frames. ], batch size: 51, lr: 3.12e-02, grad_scale: 8.0 2022-12-22 14:28:58,860 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp0.9 from training. Duration: 26.438875 2022-12-22 14:29:40,519 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0021-26306-0_sp0.9 from training. Duration: 21.2444375 2022-12-22 14:29:40,552 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp0.9 from training. Duration: 31.02225 2022-12-22 14:29:54,282 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0 from training. Duration: 22.395 2022-12-22 14:30:01,180 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9746, 1.9090, 1.1107, 2.6286, 1.7982, 4.3770, 2.0345, 2.1141], device='cuda:3'), covar=tensor([0.1132, 0.1869, 0.1796, 0.1156, 0.1735, 0.0239, 0.1353, 0.1734], device='cuda:3'), in_proj_covar=tensor([0.0090, 0.0092, 0.0094, 0.0091, 0.0111, 0.0078, 0.0098, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], device='cuda:3') 2022-12-22 14:30:05,698 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.928e+02 7.043e+02 9.169e+02 1.169e+03 2.051e+03, threshold=1.834e+03, percent-clipped=1.0 2022-12-22 14:30:07,882 INFO [train.py:894] (3/4) Epoch 3, batch 3000, loss[loss=0.3481, simple_loss=0.3766, pruned_loss=0.1599, over 18390.00 frames. ], tot_loss[loss=0.313, simple_loss=0.3601, pruned_loss=0.133, over 3712594.27 frames. ], batch size: 46, lr: 3.11e-02, grad_scale: 8.0 2022-12-22 14:30:07,882 INFO [train.py:919] (3/4) Computing validation loss 2022-12-22 14:30:18,924 INFO [train.py:928] (3/4) Epoch 3, validation: loss=0.2259, simple_loss=0.3201, pruned_loss=0.06583, over 944034.00 frames. 2022-12-22 14:30:18,925 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24676MB 2022-12-22 14:30:20,720 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10014.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:30:23,454 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0 from training. Duration: 21.075 2022-12-22 14:30:29,604 WARNING [train.py:1060] (3/4) Exclude cut with ID 6482-98857-0025-147532-0_sp0.9 from training. Duration: 20.0055625 2022-12-22 14:30:29,614 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0037-132304-0_sp0.9 from training. Duration: 22.05 2022-12-22 14:30:29,628 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0 from training. Duration: 26.8349375 2022-12-22 14:30:32,504 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp1.1 from training. Duration: 22.1090625 2022-12-22 14:30:38,432 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp0.9 from training. Duration: 26.6166875 2022-12-22 14:30:58,390 WARNING [train.py:1060] (3/4) Exclude cut with ID 2046-178027-0000-53705-0_sp0.9 from training. Duration: 20.3055625 2022-12-22 14:30:58,685 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10039.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:31:19,907 WARNING [train.py:1060] (3/4) Exclude cut with ID 7205-50138-0008-5373-0_sp0.9 from training. Duration: 20.7 2022-12-22 14:31:33,693 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6100, 1.0232, 1.9469, 2.6189, 1.8818, 2.2529, 0.5845, 1.7141], device='cuda:3'), covar=tensor([0.2280, 0.2498, 0.1812, 0.0691, 0.1868, 0.1410, 0.3204, 0.1484], device='cuda:3'), in_proj_covar=tensor([0.0100, 0.0112, 0.0122, 0.0085, 0.0101, 0.0117, 0.0138, 0.0096], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 14:31:33,730 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10062.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 14:31:34,709 INFO [train.py:894] (3/4) Epoch 3, batch 3050, loss[loss=0.3093, simple_loss=0.3651, pruned_loss=0.1267, over 18661.00 frames. ], tot_loss[loss=0.3122, simple_loss=0.3598, pruned_loss=0.1323, over 3712336.53 frames. ], batch size: 69, lr: 3.11e-02, grad_scale: 8.0 2022-12-22 14:31:44,984 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10070.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:32:02,287 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0 from training. Duration: 22.48 2022-12-22 14:32:18,983 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp0.9 from training. Duration: 29.816625 2022-12-22 14:32:30,179 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10099.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:32:30,463 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.9087, 2.5256, 1.9817, 0.9996, 1.9901, 2.2977, 1.8136, 1.9913], device='cuda:3'), covar=tensor([0.0793, 0.0640, 0.1781, 0.2518, 0.1562, 0.1121, 0.1395, 0.1219], device='cuda:3'), in_proj_covar=tensor([0.0120, 0.0156, 0.0195, 0.0188, 0.0177, 0.0159, 0.0171, 0.0170], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 14:32:31,828 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10100.0, num_to_drop=1, layers_to_drop={3} 2022-12-22 14:32:39,113 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp1.1 from training. Duration: 22.7590625 2022-12-22 14:32:44,749 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0 from training. Duration: 22.555 2022-12-22 14:32:48,880 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.989e+02 6.948e+02 8.486e+02 9.943e+02 2.176e+03, threshold=1.697e+03, percent-clipped=1.0 2022-12-22 14:32:50,369 INFO [train.py:894] (3/4) Epoch 3, batch 3100, loss[loss=0.3438, simple_loss=0.3832, pruned_loss=0.1522, over 18623.00 frames. ], tot_loss[loss=0.3112, simple_loss=0.3589, pruned_loss=0.1318, over 3711885.67 frames. ], batch size: 69, lr: 3.10e-02, grad_scale: 8.0 2022-12-22 14:32:58,179 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=10118.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:33:02,320 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0005-25975-0_sp0.9 from training. Duration: 21.688875 2022-12-22 14:33:05,410 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10123.0, num_to_drop=1, layers_to_drop={3} 2022-12-22 14:33:37,068 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0_sp0.9 from training. Duration: 22.6 2022-12-22 14:34:04,925 INFO [train.py:894] (3/4) Epoch 3, batch 3150, loss[loss=0.2915, simple_loss=0.3511, pruned_loss=0.1159, over 18599.00 frames. ], tot_loss[loss=0.3139, simple_loss=0.3612, pruned_loss=0.1333, over 3712739.96 frames. ], batch size: 56, lr: 3.10e-02, grad_scale: 4.0 2022-12-22 14:34:11,716 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0 from training. Duration: 24.32 2022-12-22 14:35:11,105 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-276745-0093-13116-0_sp0.9 from training. Duration: 21.061125 2022-12-22 14:35:21,375 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.845e+02 7.390e+02 8.631e+02 1.171e+03 2.300e+03, threshold=1.726e+03, percent-clipped=5.0 2022-12-22 14:35:21,394 INFO [train.py:894] (3/4) Epoch 3, batch 3200, loss[loss=0.3238, simple_loss=0.3766, pruned_loss=0.1355, over 18566.00 frames. ], tot_loss[loss=0.3127, simple_loss=0.3605, pruned_loss=0.1325, over 3713273.01 frames. ], batch size: 69, lr: 3.09e-02, grad_scale: 8.0 2022-12-22 14:35:23,136 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0024-15855-0_sp0.9 from training. Duration: 20.32225 2022-12-22 14:35:31,804 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([6.0193, 5.2256, 5.2863, 5.4120, 5.3944, 5.2743, 5.8097, 1.5708], device='cuda:3'), covar=tensor([0.0548, 0.0456, 0.0441, 0.0273, 0.1526, 0.0879, 0.0385, 0.4578], device='cuda:3'), in_proj_covar=tensor([0.0192, 0.0152, 0.0138, 0.0111, 0.0208, 0.0155, 0.0153, 0.0197], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-22 14:35:34,744 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp1.1 from training. Duration: 0.7545625 2022-12-22 14:35:54,154 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0_sp0.9 from training. Duration: 23.9333125 2022-12-22 14:36:00,829 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2022-12-22 14:36:04,556 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10240.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:36:15,979 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.74 vs. limit=5.0 2022-12-22 14:36:25,410 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp1.1 from training. Duration: 20.17275 2022-12-22 14:36:32,462 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp1.1 from training. Duration: 20.436375 2022-12-22 14:36:38,068 INFO [train.py:894] (3/4) Epoch 3, batch 3250, loss[loss=0.3484, simple_loss=0.3819, pruned_loss=0.1574, over 18583.00 frames. ], tot_loss[loss=0.3129, simple_loss=0.3603, pruned_loss=0.1327, over 3713036.71 frames. ], batch size: 51, lr: 3.08e-02, grad_scale: 8.0 2022-12-22 14:37:17,016 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=10288.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:37:31,260 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7192, 3.8419, 3.8645, 1.8737, 3.6690, 2.9480, 0.8121, 2.5381], device='cuda:3'), covar=tensor([0.1906, 0.0821, 0.1641, 0.3837, 0.1003, 0.1247, 0.6920, 0.2052], device='cuda:3'), in_proj_covar=tensor([0.0122, 0.0091, 0.0151, 0.0117, 0.0101, 0.0097, 0.0147, 0.0111], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 14:37:53,207 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.532e+02 6.935e+02 8.611e+02 1.028e+03 2.623e+03, threshold=1.722e+03, percent-clipped=3.0 2022-12-22 14:37:53,224 INFO [train.py:894] (3/4) Epoch 3, batch 3300, loss[loss=0.2759, simple_loss=0.3523, pruned_loss=0.09977, over 18548.00 frames. ], tot_loss[loss=0.3109, simple_loss=0.3594, pruned_loss=0.1312, over 3713285.71 frames. ], batch size: 55, lr: 3.08e-02, grad_scale: 8.0 2022-12-22 14:37:54,876 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0_sp0.9 from training. Duration: 23.1055625 2022-12-22 14:37:55,169 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10314.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:37:56,543 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp1.1 from training. Duration: 23.5 2022-12-22 14:38:08,720 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp0.9 from training. Duration: 26.62775 2022-12-22 14:38:22,343 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0 from training. Duration: 21.105 2022-12-22 14:38:26,859 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0_sp0.9 from training. Duration: 24.411125 2022-12-22 14:38:33,382 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2022-12-22 14:38:44,414 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7382, 1.3339, 1.4248, 0.2934, 1.4132, 1.5256, 1.1668, 1.5872], device='cuda:3'), covar=tensor([0.0798, 0.0619, 0.1081, 0.1850, 0.0971, 0.1532, 0.1625, 0.0774], device='cuda:3'), in_proj_covar=tensor([0.0121, 0.0154, 0.0188, 0.0186, 0.0175, 0.0156, 0.0168, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 14:38:54,577 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp1.1 from training. Duration: 21.263625 2022-12-22 14:39:07,893 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=10362.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:39:09,165 INFO [train.py:894] (3/4) Epoch 3, batch 3350, loss[loss=0.3253, simple_loss=0.3707, pruned_loss=0.14, over 18601.00 frames. ], tot_loss[loss=0.3116, simple_loss=0.3598, pruned_loss=0.1317, over 3714141.90 frames. ], batch size: 51, lr: 3.07e-02, grad_scale: 8.0 2022-12-22 14:39:25,668 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0 from training. Duration: 20.795 2022-12-22 14:39:37,119 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0 from training. Duration: 24.76 2022-12-22 14:39:38,458 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0_sp0.9 from training. Duration: 22.25 2022-12-22 14:39:43,440 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.3126, 2.3525, 1.7922, 2.7721, 2.1258, 2.1703, 2.1673, 3.7614], device='cuda:3'), covar=tensor([0.1012, 0.1587, 0.1036, 0.1928, 0.1763, 0.0712, 0.1991, 0.0331], device='cuda:3'), in_proj_covar=tensor([0.0192, 0.0164, 0.0153, 0.0234, 0.0160, 0.0151, 0.0174, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2022-12-22 14:39:59,645 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10395.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 14:40:02,557 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp1.1 from training. Duration: 20.5045625 2022-12-22 14:40:05,651 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10399.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:40:21,231 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10409.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:40:27,021 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.373e+02 7.368e+02 9.134e+02 1.174e+03 1.740e+03, threshold=1.827e+03, percent-clipped=1.0 2022-12-22 14:40:27,037 INFO [train.py:894] (3/4) Epoch 3, batch 3400, loss[loss=0.3224, simple_loss=0.3705, pruned_loss=0.1372, over 18592.00 frames. ], tot_loss[loss=0.3102, simple_loss=0.3588, pruned_loss=0.1308, over 3713840.61 frames. ], batch size: 177, lr: 3.07e-02, grad_scale: 8.0 2022-12-22 14:40:34,853 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10418.0, num_to_drop=1, layers_to_drop={2} 2022-12-22 14:41:16,452 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=10447.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:41:23,799 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0787, 1.8489, 1.3198, 2.1146, 1.7403, 1.5753, 1.6923, 2.3467], device='cuda:3'), covar=tensor([0.1343, 0.1609, 0.1167, 0.1725, 0.1602, 0.0737, 0.1622, 0.0458], device='cuda:3'), in_proj_covar=tensor([0.0194, 0.0166, 0.0156, 0.0237, 0.0162, 0.0153, 0.0176, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2022-12-22 14:41:26,385 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10454.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:41:34,505 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([5.8095, 5.0056, 4.9450, 5.3454, 5.1176, 5.1709, 5.7453, 1.3743], device='cuda:3'), covar=tensor([0.0420, 0.0449, 0.0449, 0.0238, 0.1083, 0.0557, 0.0242, 0.3944], device='cuda:3'), in_proj_covar=tensor([0.0198, 0.0156, 0.0141, 0.0116, 0.0212, 0.0164, 0.0165, 0.0204], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-22 14:41:38,779 INFO [train.py:894] (3/4) Epoch 3, batch 3450, loss[loss=0.2437, simple_loss=0.2973, pruned_loss=0.09503, over 18500.00 frames. ], tot_loss[loss=0.3094, simple_loss=0.3582, pruned_loss=0.1303, over 3713311.41 frames. ], batch size: 43, lr: 3.06e-02, grad_scale: 8.0 2022-12-22 14:41:49,353 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10470.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:41:50,741 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5513, 1.6257, 1.6527, 1.4803, 1.7217, 2.5597, 1.6854, 2.1466], device='cuda:3'), covar=tensor([0.3027, 0.1931, 0.1448, 0.2128, 0.1286, 0.0306, 0.1441, 0.0875], device='cuda:3'), in_proj_covar=tensor([0.0156, 0.0133, 0.0145, 0.0132, 0.0136, 0.0098, 0.0121, 0.0118], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], device='cuda:3') 2022-12-22 14:42:07,180 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4966, 1.7413, 2.1426, 2.2495, 2.1043, 1.6574, 2.0297, 1.1818], device='cuda:3'), covar=tensor([0.0791, 0.1675, 0.1276, 0.1286, 0.0616, 0.0566, 0.1357, 0.0772], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0182, 0.0175, 0.0173, 0.0147, 0.0144, 0.0160, 0.0147], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2022-12-22 14:42:51,938 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.608e+02 7.309e+02 8.677e+02 1.147e+03 1.668e+03, threshold=1.735e+03, percent-clipped=0.0 2022-12-22 14:42:51,954 INFO [train.py:894] (3/4) Epoch 3, batch 3500, loss[loss=0.4074, simple_loss=0.415, pruned_loss=0.1999, over 18641.00 frames. ], tot_loss[loss=0.3114, simple_loss=0.3593, pruned_loss=0.1318, over 3714346.59 frames. ], batch size: 181, lr: 3.05e-02, grad_scale: 8.0 2022-12-22 14:42:55,824 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10515.0, num_to_drop=1, layers_to_drop={3} 2022-12-22 14:43:14,891 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp1.1 from training. Duration: 22.2954375 2022-12-22 14:43:26,381 INFO [train.py:894] (3/4) Epoch 4, batch 0, loss[loss=0.2983, simple_loss=0.3575, pruned_loss=0.1196, over 18707.00 frames. ], tot_loss[loss=0.2983, simple_loss=0.3575, pruned_loss=0.1196, over 18707.00 frames. ], batch size: 65, lr: 2.85e-02, grad_scale: 8.0 2022-12-22 14:43:26,381 INFO [train.py:919] (3/4) Computing validation loss 2022-12-22 14:43:37,537 INFO [train.py:928] (3/4) Epoch 4, validation: loss=0.2246, simple_loss=0.3191, pruned_loss=0.06499, over 944034.00 frames. 2022-12-22 14:43:37,538 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24676MB 2022-12-22 14:43:55,880 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3750, 0.5575, 1.1641, 0.7521, 1.4615, 1.5492, 1.3461, 0.8671], device='cuda:3'), covar=tensor([0.0754, 0.0921, 0.1041, 0.0784, 0.0545, 0.0496, 0.0405, 0.0909], device='cuda:3'), in_proj_covar=tensor([0.0071, 0.0104, 0.0114, 0.0123, 0.0100, 0.0075, 0.0071, 0.0107], device='cuda:3'), out_proj_covar=tensor([7.9273e-05, 1.0676e-04, 1.2195e-04, 1.2604e-04, 1.0884e-04, 8.1702e-05, 7.7991e-05, 1.1151e-04], device='cuda:3') 2022-12-22 14:44:28,688 WARNING [train.py:1060] (3/4) Exclude cut with ID 298-126791-0067-24026-0_sp0.9 from training. Duration: 21.438875 2022-12-22 14:44:34,622 WARNING [train.py:1060] (3/4) Exclude cut with ID 5652-39938-0025-23684-0_sp0.9 from training. Duration: 22.2055625 2022-12-22 14:44:53,492 INFO [train.py:894] (3/4) Epoch 4, batch 50, loss[loss=0.2557, simple_loss=0.33, pruned_loss=0.09067, over 18593.00 frames. ], tot_loss[loss=0.2757, simple_loss=0.3427, pruned_loss=0.1044, over 839110.97 frames. ], batch size: 51, lr: 2.85e-02, grad_scale: 8.0 2022-12-22 14:45:59,177 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.242e+02 5.273e+02 6.714e+02 8.694e+02 2.081e+03, threshold=1.343e+03, percent-clipped=1.0 2022-12-22 14:46:10,523 INFO [train.py:894] (3/4) Epoch 4, batch 100, loss[loss=0.2116, simple_loss=0.2883, pruned_loss=0.06751, over 18480.00 frames. ], tot_loss[loss=0.2689, simple_loss=0.3361, pruned_loss=0.1009, over 1475477.53 frames. ], batch size: 43, lr: 2.84e-02, grad_scale: 8.0 2022-12-22 14:47:02,195 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0981, 1.6417, 1.5430, 2.3071, 1.9194, 4.3164, 1.4798, 1.7068], device='cuda:3'), covar=tensor([0.1475, 0.2918, 0.1985, 0.1458, 0.2276, 0.0227, 0.2474, 0.2797], device='cuda:3'), in_proj_covar=tensor([0.0089, 0.0094, 0.0094, 0.0092, 0.0111, 0.0081, 0.0102, 0.0091], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], device='cuda:3') 2022-12-22 14:47:25,416 INFO [train.py:894] (3/4) Epoch 4, batch 150, loss[loss=0.219, simple_loss=0.2985, pruned_loss=0.0698, over 18723.00 frames. ], tot_loss[loss=0.2663, simple_loss=0.3344, pruned_loss=0.09909, over 1971769.66 frames. ], batch size: 52, lr: 2.84e-02, grad_scale: 8.0 2022-12-22 14:47:35,134 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0 from training. Duration: 24.525 2022-12-22 14:47:50,485 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2022-12-22 14:47:54,642 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6419, 0.6901, 1.3730, 1.2441, 1.5819, 1.7404, 1.7395, 1.0929], device='cuda:3'), covar=tensor([0.0643, 0.0767, 0.0867, 0.0605, 0.0433, 0.0397, 0.0277, 0.0702], device='cuda:3'), in_proj_covar=tensor([0.0067, 0.0100, 0.0113, 0.0120, 0.0098, 0.0072, 0.0068, 0.0104], device='cuda:3'), out_proj_covar=tensor([7.4747e-05, 1.0272e-04, 1.2017e-04, 1.2392e-04, 1.0737e-04, 7.8662e-05, 7.5308e-05, 1.0856e-04], device='cuda:3') 2022-12-22 14:48:01,943 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10695.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:48:08,923 WARNING [train.py:1060] (3/4) Exclude cut with ID 3699-47246-0007-3408-0_sp0.9 from training. Duration: 20.26675 2022-12-22 14:48:22,544 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp0.9 from training. Duration: 27.25 2022-12-22 14:48:30,169 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.370e+02 5.528e+02 6.608e+02 7.818e+02 1.640e+03, threshold=1.322e+03, percent-clipped=2.0 2022-12-22 14:48:38,209 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10718.0, num_to_drop=1, layers_to_drop={2} 2022-12-22 14:48:40,693 INFO [train.py:894] (3/4) Epoch 4, batch 200, loss[loss=0.2795, simple_loss=0.3455, pruned_loss=0.1067, over 18388.00 frames. ], tot_loss[loss=0.2636, simple_loss=0.3323, pruned_loss=0.09747, over 2357299.80 frames. ], batch size: 53, lr: 2.83e-02, grad_scale: 8.0 2022-12-22 14:48:52,533 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5754, 3.2342, 3.2503, 1.4622, 3.0924, 2.3128, 1.2130, 2.1074], device='cuda:3'), covar=tensor([0.1804, 0.0776, 0.1538, 0.3846, 0.1052, 0.1452, 0.4565, 0.2281], device='cuda:3'), in_proj_covar=tensor([0.0124, 0.0091, 0.0147, 0.0117, 0.0100, 0.0098, 0.0146, 0.0114], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 14:49:14,307 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=10743.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:49:16,081 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10744.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:49:36,810 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2022-12-22 14:49:42,080 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0 from training. Duration: 21.68 2022-12-22 14:49:47,960 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10765.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:49:49,373 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=10766.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 14:49:52,836 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0 from training. Duration: 21.6300625 2022-12-22 14:49:55,754 INFO [train.py:894] (3/4) Epoch 4, batch 250, loss[loss=0.2164, simple_loss=0.2916, pruned_loss=0.07062, over 18383.00 frames. ], tot_loss[loss=0.2613, simple_loss=0.3299, pruned_loss=0.09639, over 2656217.66 frames. ], batch size: 46, lr: 2.83e-02, grad_scale: 8.0 2022-12-22 14:50:01,973 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10774.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:50:15,074 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0_sp0.9 from training. Duration: 24.033375 2022-12-22 14:50:50,226 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10805.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:50:57,730 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10810.0, num_to_drop=1, layers_to_drop={3} 2022-12-22 14:51:02,534 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.039e+02 5.468e+02 6.455e+02 8.141e+02 3.038e+03, threshold=1.291e+03, percent-clipped=5.0 2022-12-22 14:51:10,871 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0 from training. Duration: 22.905 2022-12-22 14:51:12,325 INFO [train.py:894] (3/4) Epoch 4, batch 300, loss[loss=0.2317, simple_loss=0.2971, pruned_loss=0.08311, over 18516.00 frames. ], tot_loss[loss=0.2605, simple_loss=0.3298, pruned_loss=0.09558, over 2891645.19 frames. ], batch size: 44, lr: 2.82e-02, grad_scale: 8.0 2022-12-22 14:51:12,405 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp1.1 from training. Duration: 23.4318125 2022-12-22 14:51:33,938 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10835.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:52:26,559 INFO [train.py:894] (3/4) Epoch 4, batch 350, loss[loss=0.3252, simple_loss=0.3734, pruned_loss=0.1385, over 18675.00 frames. ], tot_loss[loss=0.2624, simple_loss=0.3309, pruned_loss=0.09699, over 3073742.62 frames. ], batch size: 177, lr: 2.82e-02, grad_scale: 8.0 2022-12-22 14:52:48,913 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10885.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 14:52:57,597 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.10 vs. limit=5.0 2022-12-22 14:53:06,404 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp1.1 from training. Duration: 20.82275 2022-12-22 14:53:08,048 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp0.9 from training. Duration: 25.45 2022-12-22 14:53:33,184 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.068e+02 5.924e+02 7.243e+02 8.972e+02 2.148e+03, threshold=1.449e+03, percent-clipped=9.0 2022-12-22 14:53:43,091 INFO [train.py:894] (3/4) Epoch 4, batch 400, loss[loss=0.2796, simple_loss=0.3479, pruned_loss=0.1056, over 18590.00 frames. ], tot_loss[loss=0.2656, simple_loss=0.3332, pruned_loss=0.09898, over 3215994.39 frames. ], batch size: 57, lr: 2.81e-02, grad_scale: 8.0 2022-12-22 14:53:48,058 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6282, 2.1957, 1.4156, 2.5917, 2.6389, 1.3291, 2.5736, 1.2916], device='cuda:3'), covar=tensor([0.1712, 0.1587, 0.1387, 0.0780, 0.1474, 0.1252, 0.1209, 0.1537], device='cuda:3'), in_proj_covar=tensor([0.0184, 0.0161, 0.0156, 0.0144, 0.0193, 0.0147, 0.0157, 0.0152], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 14:54:06,243 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0 from training. Duration: 25.775 2022-12-22 14:54:22,661 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10946.0, num_to_drop=1, layers_to_drop={2} 2022-12-22 14:54:27,085 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0_sp0.9 from training. Duration: 22.25 2022-12-22 14:54:27,495 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0020, 2.3545, 2.3491, 1.6387, 2.2217, 2.1911, 1.3892, 2.5040], device='cuda:3'), covar=tensor([0.1281, 0.1049, 0.1518, 0.2152, 0.1146, 0.1314, 0.2552, 0.0687], device='cuda:3'), in_proj_covar=tensor([0.0188, 0.0156, 0.0188, 0.0183, 0.0182, 0.0185, 0.0180, 0.0157], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 14:54:38,382 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7137, 0.4746, 1.3564, 1.4100, 1.5741, 1.4340, 1.2235, 1.0367], device='cuda:3'), covar=tensor([0.0951, 0.1592, 0.1732, 0.1178, 0.0853, 0.0741, 0.1222, 0.0952], device='cuda:3'), in_proj_covar=tensor([0.0156, 0.0187, 0.0175, 0.0177, 0.0152, 0.0148, 0.0167, 0.0147], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2022-12-22 14:54:54,945 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0 from training. Duration: 26.205 2022-12-22 14:54:57,936 INFO [train.py:894] (3/4) Epoch 4, batch 450, loss[loss=0.2526, simple_loss=0.3047, pruned_loss=0.1002, over 18621.00 frames. ], tot_loss[loss=0.2687, simple_loss=0.3359, pruned_loss=0.1007, over 3325860.29 frames. ], batch size: 45, lr: 2.81e-02, grad_scale: 8.0 2022-12-22 14:55:12,669 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp0.9 from training. Duration: 30.1555625 2022-12-22 14:55:18,402 WARNING [train.py:1060] (3/4) Exclude cut with ID 1265-135635-0050-6781-0_sp0.9 from training. Duration: 21.8333125 2022-12-22 14:55:26,119 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp1.1 from training. Duration: 20.6545625 2022-12-22 14:56:03,706 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.947e+02 6.290e+02 7.836e+02 1.078e+03 2.986e+03, threshold=1.567e+03, percent-clipped=7.0 2022-12-22 14:56:08,084 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0045-39920-0_sp0.9 from training. Duration: 20.52225 2022-12-22 14:56:13,920 INFO [train.py:894] (3/4) Epoch 4, batch 500, loss[loss=0.262, simple_loss=0.334, pruned_loss=0.09498, over 18704.00 frames. ], tot_loss[loss=0.2702, simple_loss=0.3369, pruned_loss=0.1018, over 3411688.99 frames. ], batch size: 50, lr: 2.80e-02, grad_scale: 8.0 2022-12-22 14:56:29,940 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp0.9 from training. Duration: 29.1166875 2022-12-22 14:57:16,170 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8742, 2.0422, 2.1148, 2.0797, 2.3065, 3.3437, 2.2342, 2.7058], device='cuda:3'), covar=tensor([0.3117, 0.1781, 0.1508, 0.1803, 0.1121, 0.0213, 0.1315, 0.0854], device='cuda:3'), in_proj_covar=tensor([0.0153, 0.0129, 0.0141, 0.0131, 0.0129, 0.0093, 0.0118, 0.0115], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], device='cuda:3') 2022-12-22 14:57:21,973 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11065.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:57:28,743 INFO [train.py:894] (3/4) Epoch 4, batch 550, loss[loss=0.2533, simple_loss=0.3262, pruned_loss=0.09022, over 18590.00 frames. ], tot_loss[loss=0.2711, simple_loss=0.3375, pruned_loss=0.1023, over 3478823.58 frames. ], batch size: 51, lr: 2.80e-02, grad_scale: 8.0 2022-12-22 14:57:32,926 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133211-0007-59831-0_sp0.9 from training. Duration: 21.388875 2022-12-22 14:57:34,682 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5911, 1.7019, 2.0151, 2.1704, 2.5058, 4.8217, 2.5036, 3.1575], device='cuda:3'), covar=tensor([0.3835, 0.2406, 0.1951, 0.2195, 0.1433, 0.0094, 0.1529, 0.1101], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0130, 0.0142, 0.0132, 0.0131, 0.0094, 0.0119, 0.0116], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], device='cuda:3') 2022-12-22 14:58:08,839 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0 from training. Duration: 22.72 2022-12-22 14:58:10,277 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0_sp0.9 from training. Duration: 22.7444375 2022-12-22 14:58:13,851 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11100.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:58:28,869 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11110.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 14:58:33,134 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.935e+02 5.294e+02 6.541e+02 8.578e+02 1.518e+03, threshold=1.308e+03, percent-clipped=0.0 2022-12-22 14:58:33,331 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=11113.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:58:43,498 INFO [train.py:894] (3/4) Epoch 4, batch 600, loss[loss=0.2887, simple_loss=0.355, pruned_loss=0.1113, over 18651.00 frames. ], tot_loss[loss=0.2717, simple_loss=0.338, pruned_loss=0.1027, over 3531287.75 frames. ], batch size: 69, lr: 2.79e-02, grad_scale: 8.0 2022-12-22 14:58:52,570 WARNING [train.py:1060] (3/4) Exclude cut with ID 4133-6541-0027-40495-0_sp1.1 from training. Duration: 0.9681875 2022-12-22 14:58:55,485 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0_sp0.9 from training. Duration: 22.3166875 2022-12-22 14:58:58,530 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11130.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:59:01,155 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133212-0015-59917-0_sp0.9 from training. Duration: 21.8166875 2022-12-22 14:59:42,976 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=11158.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 14:59:55,676 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 2022-12-22 15:00:00,322 INFO [train.py:894] (3/4) Epoch 4, batch 650, loss[loss=0.2402, simple_loss=0.3056, pruned_loss=0.08737, over 18425.00 frames. ], tot_loss[loss=0.2696, simple_loss=0.3368, pruned_loss=0.1012, over 3571750.07 frames. ], batch size: 42, lr: 2.78e-02, grad_scale: 8.0 2022-12-22 15:00:06,662 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11174.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:00:17,069 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2022-12-22 15:00:44,723 WARNING [train.py:1060] (3/4) Exclude cut with ID 4957-30119-0041-23990-0_sp0.9 from training. Duration: 20.22775 2022-12-22 15:01:06,116 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.895e+02 6.331e+02 7.829e+02 9.597e+02 3.856e+03, threshold=1.566e+03, percent-clipped=9.0 2022-12-22 15:01:16,831 INFO [train.py:894] (3/4) Epoch 4, batch 700, loss[loss=0.2344, simple_loss=0.3056, pruned_loss=0.0816, over 18521.00 frames. ], tot_loss[loss=0.269, simple_loss=0.3363, pruned_loss=0.1008, over 3603107.43 frames. ], batch size: 44, lr: 2.78e-02, grad_scale: 8.0 2022-12-22 15:01:27,148 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp1.1 from training. Duration: 24.67275 2022-12-22 15:01:41,168 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11235.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:01:50,561 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11241.0, num_to_drop=1, layers_to_drop={2} 2022-12-22 15:01:59,302 WARNING [train.py:1060] (3/4) Exclude cut with ID 3082-165428-0081-50734-0_sp0.9 from training. Duration: 21.8055625 2022-12-22 15:02:33,317 INFO [train.py:894] (3/4) Epoch 4, batch 750, loss[loss=0.2867, simple_loss=0.3516, pruned_loss=0.1109, over 18684.00 frames. ], tot_loss[loss=0.2682, simple_loss=0.335, pruned_loss=0.1007, over 3627792.47 frames. ], batch size: 69, lr: 2.77e-02, grad_scale: 8.0 2022-12-22 15:02:36,266 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0_sp0.9 from training. Duration: 22.6666875 2022-12-22 15:03:38,711 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.263e+02 5.618e+02 6.637e+02 8.413e+02 1.667e+03, threshold=1.327e+03, percent-clipped=1.0 2022-12-22 15:03:40,066 WARNING [train.py:1060] (3/4) Exclude cut with ID 2411-132532-0017-82279-0_sp1.1 from training. Duration: 0.9681875 2022-12-22 15:03:48,454 INFO [train.py:894] (3/4) Epoch 4, batch 800, loss[loss=0.3076, simple_loss=0.3712, pruned_loss=0.122, over 18649.00 frames. ], tot_loss[loss=0.2683, simple_loss=0.3352, pruned_loss=0.1007, over 3647103.90 frames. ], batch size: 60, lr: 2.77e-02, grad_scale: 8.0 2022-12-22 15:03:52,518 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.95 vs. limit=5.0 2022-12-22 15:04:07,149 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0 from training. Duration: 22.485 2022-12-22 15:04:31,977 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2022-12-22 15:04:46,161 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp1.1 from training. Duration: 23.82275 2022-12-22 15:04:59,210 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0 from training. Duration: 20.77 2022-12-22 15:05:05,214 INFO [train.py:894] (3/4) Epoch 4, batch 850, loss[loss=0.2683, simple_loss=0.3379, pruned_loss=0.09937, over 18379.00 frames. ], tot_loss[loss=0.269, simple_loss=0.3359, pruned_loss=0.101, over 3661260.68 frames. ], batch size: 51, lr: 2.76e-02, grad_scale: 8.0 2022-12-22 15:05:06,700 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0_sp0.9 from training. Duration: 24.088875 2022-12-22 15:05:25,513 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.2166, 1.0006, 0.5012, 1.2584, 1.0710, 2.1575, 1.0078, 0.9983], device='cuda:3'), covar=tensor([0.1204, 0.2021, 0.1772, 0.1124, 0.1800, 0.0579, 0.1581, 0.1887], device='cuda:3'), in_proj_covar=tensor([0.0088, 0.0093, 0.0093, 0.0089, 0.0110, 0.0080, 0.0099, 0.0091], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], device='cuda:3') 2022-12-22 15:05:36,697 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp1.1 from training. Duration: 20.4409375 2022-12-22 15:05:42,034 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4926, 1.9377, 1.6195, 1.8376, 1.9222, 3.7510, 1.9518, 2.5897], device='cuda:3'), covar=tensor([0.4046, 0.1989, 0.1995, 0.2122, 0.1509, 0.0162, 0.1566, 0.1036], device='cuda:3'), in_proj_covar=tensor([0.0155, 0.0131, 0.0143, 0.0131, 0.0129, 0.0095, 0.0118, 0.0115], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], device='cuda:3') 2022-12-22 15:05:52,206 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11400.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:06:10,636 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.650e+02 5.684e+02 6.986e+02 8.807e+02 1.585e+03, threshold=1.397e+03, percent-clipped=2.0 2022-12-22 15:06:21,138 INFO [train.py:894] (3/4) Epoch 4, batch 900, loss[loss=0.3138, simple_loss=0.3731, pruned_loss=0.1272, over 18562.00 frames. ], tot_loss[loss=0.2678, simple_loss=0.335, pruned_loss=0.1003, over 3672514.35 frames. ], batch size: 55, lr: 2.76e-02, grad_scale: 8.0 2022-12-22 15:06:37,102 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11430.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:06:54,329 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0_sp0.9 from training. Duration: 22.511125 2022-12-22 15:06:54,349 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0 from training. Duration: 20.675 2022-12-22 15:07:05,476 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=11448.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:07:37,406 INFO [train.py:894] (3/4) Epoch 4, batch 950, loss[loss=0.2783, simple_loss=0.3548, pruned_loss=0.1009, over 18601.00 frames. ], tot_loss[loss=0.2671, simple_loss=0.3343, pruned_loss=0.09995, over 3680696.35 frames. ], batch size: 56, lr: 2.75e-02, grad_scale: 8.0 2022-12-22 15:07:42,553 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2022-12-22 15:07:50,430 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=11478.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:08:22,431 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4738, 3.9490, 4.0476, 2.0166, 3.8369, 2.8367, 0.8521, 2.7758], device='cuda:3'), covar=tensor([0.1868, 0.0610, 0.1035, 0.2944, 0.0893, 0.1218, 0.5633, 0.1679], device='cuda:3'), in_proj_covar=tensor([0.0120, 0.0090, 0.0145, 0.0111, 0.0099, 0.0097, 0.0144, 0.0109], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 15:08:34,214 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp0.9 from training. Duration: 24.9833125 2022-12-22 15:08:40,170 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2616, 1.5407, 2.0093, 1.3985, 2.0869, 2.8144, 2.4679, 1.9354], device='cuda:3'), covar=tensor([0.0876, 0.0935, 0.0778, 0.0861, 0.0577, 0.0508, 0.0562, 0.0821], device='cuda:3'), in_proj_covar=tensor([0.0068, 0.0103, 0.0115, 0.0118, 0.0095, 0.0077, 0.0072, 0.0106], device='cuda:3'), out_proj_covar=tensor([7.5699e-05, 1.0464e-04, 1.2172e-04, 1.2126e-04, 1.0239e-04, 8.2301e-05, 7.7305e-05, 1.0900e-04], device='cuda:3') 2022-12-22 15:08:42,488 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.740e+02 6.104e+02 6.909e+02 8.530e+02 1.884e+03, threshold=1.382e+03, percent-clipped=3.0 2022-12-22 15:08:50,194 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4959, 0.8964, 1.4327, 2.6351, 1.7228, 1.6401, 0.6083, 1.4872], device='cuda:3'), covar=tensor([0.2029, 0.2189, 0.1776, 0.0572, 0.1476, 0.1755, 0.2942, 0.1531], device='cuda:3'), in_proj_covar=tensor([0.0103, 0.0113, 0.0123, 0.0083, 0.0101, 0.0117, 0.0137, 0.0100], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002], device='cuda:3') 2022-12-22 15:08:52,610 INFO [train.py:894] (3/4) Epoch 4, batch 1000, loss[loss=0.2814, simple_loss=0.3511, pruned_loss=0.1059, over 18447.00 frames. ], tot_loss[loss=0.2678, simple_loss=0.3349, pruned_loss=0.1003, over 3687941.01 frames. ], batch size: 50, lr: 2.75e-02, grad_scale: 8.0 2022-12-22 15:09:05,550 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0 from training. Duration: 27.14 2022-12-22 15:09:08,546 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11530.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:09:11,611 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11532.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:09:20,287 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0 from training. Duration: 22.44 2022-12-22 15:09:25,476 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11541.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 15:09:44,947 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5100, 1.4435, 1.2482, 1.9133, 1.3836, 3.2752, 1.3793, 1.3742], device='cuda:3'), covar=tensor([0.1336, 0.2101, 0.1667, 0.1253, 0.1936, 0.0323, 0.1682, 0.2075], device='cuda:3'), in_proj_covar=tensor([0.0090, 0.0095, 0.0093, 0.0091, 0.0111, 0.0081, 0.0101, 0.0092], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 15:10:07,536 INFO [train.py:894] (3/4) Epoch 4, batch 1050, loss[loss=0.3504, simple_loss=0.4036, pruned_loss=0.1485, over 18533.00 frames. ], tot_loss[loss=0.2669, simple_loss=0.3342, pruned_loss=0.09978, over 3693923.36 frames. ], batch size: 77, lr: 2.74e-02, grad_scale: 8.0 2022-12-22 15:10:33,342 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8132, 1.8280, 0.9406, 1.8000, 2.0129, 1.7817, 2.8161, 1.9072], device='cuda:3'), covar=tensor([0.0859, 0.1351, 0.2487, 0.1670, 0.1596, 0.0771, 0.0536, 0.0986], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0153, 0.0184, 0.0225, 0.0193, 0.0150, 0.0141, 0.0157], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-22 15:10:37,241 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=11589.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 15:10:40,285 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0060-62364-0_sp0.9 from training. Duration: 21.361125 2022-12-22 15:10:43,353 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11593.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:10:45,666 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp1.1 from training. Duration: 27.0318125 2022-12-22 15:10:57,864 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp0.9 from training. Duration: 28.638875 2022-12-22 15:11:12,459 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.335e+02 6.019e+02 7.505e+02 8.612e+02 1.443e+03, threshold=1.501e+03, percent-clipped=1.0 2022-12-22 15:11:13,932 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0 from training. Duration: 20.4 2022-12-22 15:11:22,524 INFO [train.py:894] (3/4) Epoch 4, batch 1100, loss[loss=0.274, simple_loss=0.3468, pruned_loss=0.1006, over 18589.00 frames. ], tot_loss[loss=0.2672, simple_loss=0.3346, pruned_loss=0.09989, over 3698483.21 frames. ], batch size: 56, lr: 2.74e-02, grad_scale: 8.0 2022-12-22 15:11:46,722 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0 from training. Duration: 20.025 2022-12-22 15:11:46,735 WARNING [train.py:1060] (3/4) Exclude cut with ID 2364-131735-0112-64612-0_sp0.9 from training. Duration: 20.488875 2022-12-22 15:11:47,101 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11635.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:11:53,193 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0 from training. Duration: 29.735 2022-12-22 15:12:14,431 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.8544, 2.5612, 1.8816, 1.1153, 1.7779, 2.1173, 1.5194, 2.0956], device='cuda:3'), covar=tensor([0.0611, 0.0471, 0.1313, 0.1659, 0.1466, 0.1105, 0.1409, 0.0824], device='cuda:3'), in_proj_covar=tensor([0.0127, 0.0160, 0.0194, 0.0187, 0.0179, 0.0160, 0.0175, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 15:12:38,691 INFO [train.py:894] (3/4) Epoch 4, batch 1150, loss[loss=0.2662, simple_loss=0.3407, pruned_loss=0.09583, over 18716.00 frames. ], tot_loss[loss=0.2667, simple_loss=0.3343, pruned_loss=0.09956, over 3701773.94 frames. ], batch size: 60, lr: 2.73e-02, grad_scale: 8.0 2022-12-22 15:13:15,090 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4014, 2.0389, 1.4018, 2.3907, 1.8213, 1.7735, 1.9562, 2.7480], device='cuda:3'), covar=tensor([0.1233, 0.1670, 0.1192, 0.1910, 0.1740, 0.0727, 0.1691, 0.0382], device='cuda:3'), in_proj_covar=tensor([0.0201, 0.0179, 0.0163, 0.0253, 0.0175, 0.0161, 0.0189, 0.0134], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2022-12-22 15:13:16,030 WARNING [train.py:1060] (3/4) Exclude cut with ID 7276-92427-0014-12983-0_sp0.9 from training. Duration: 21.3055625 2022-12-22 15:13:17,594 WARNING [train.py:1060] (3/4) Exclude cut with ID 1025-75365-0008-79168-0_sp0.9 from training. Duration: 22.0666875 2022-12-22 15:13:19,386 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11696.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:13:23,716 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11699.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:13:43,179 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.231e+02 5.643e+02 6.547e+02 8.776e+02 2.408e+03, threshold=1.309e+03, percent-clipped=2.0 2022-12-22 15:13:53,672 INFO [train.py:894] (3/4) Epoch 4, batch 1200, loss[loss=0.2314, simple_loss=0.2927, pruned_loss=0.08504, over 18492.00 frames. ], tot_loss[loss=0.2661, simple_loss=0.334, pruned_loss=0.09909, over 3704472.85 frames. ], batch size: 43, lr: 2.73e-02, grad_scale: 8.0 2022-12-22 15:14:52,864 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2022-12-22 15:14:55,564 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11760.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:15:08,702 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0 from training. Duration: 20.26 2022-12-22 15:15:10,222 INFO [train.py:894] (3/4) Epoch 4, batch 1250, loss[loss=0.2458, simple_loss=0.306, pruned_loss=0.09283, over 18442.00 frames. ], tot_loss[loss=0.2648, simple_loss=0.333, pruned_loss=0.09831, over 3705852.77 frames. ], batch size: 42, lr: 2.72e-02, grad_scale: 8.0 2022-12-22 15:15:21,023 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0030-9324-0_sp0.9 from training. Duration: 21.3444375 2022-12-22 15:16:16,649 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.456e+02 5.375e+02 6.633e+02 8.647e+02 2.282e+03, threshold=1.327e+03, percent-clipped=7.0 2022-12-22 15:16:17,984 WARNING [train.py:1060] (3/4) Exclude cut with ID 497-129325-0061-62254-0_sp1.1 from training. Duration: 0.97725 2022-12-22 15:16:21,520 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11816.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:16:27,459 INFO [train.py:894] (3/4) Epoch 4, batch 1300, loss[loss=0.2837, simple_loss=0.3532, pruned_loss=0.1071, over 18664.00 frames. ], tot_loss[loss=0.2642, simple_loss=0.3325, pruned_loss=0.09793, over 3707143.53 frames. ], batch size: 98, lr: 2.72e-02, grad_scale: 8.0 2022-12-22 15:16:43,005 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11830.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:16:58,146 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0_sp0.9 from training. Duration: 22.97225 2022-12-22 15:17:07,603 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11846.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:17:30,588 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0047-39922-0_sp0.9 from training. Duration: 21.97775 2022-12-22 15:17:43,824 INFO [train.py:894] (3/4) Epoch 4, batch 1350, loss[loss=0.3098, simple_loss=0.3666, pruned_loss=0.1265, over 18525.00 frames. ], tot_loss[loss=0.2656, simple_loss=0.3339, pruned_loss=0.09863, over 3708768.43 frames. ], batch size: 58, lr: 2.71e-02, grad_scale: 8.0 2022-12-22 15:17:43,865 WARNING [train.py:1060] (3/4) Exclude cut with ID 1112-1043-0006-89194-0_sp0.9 from training. Duration: 21.8333125 2022-12-22 15:17:55,043 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0 from training. Duration: 20.47 2022-12-22 15:17:55,405 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11877.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:17:56,631 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=11878.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:18:11,676 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11888.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:18:39,481 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11907.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:18:48,468 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.033e+02 5.388e+02 6.547e+02 8.025e+02 2.001e+03, threshold=1.309e+03, percent-clipped=2.0 2022-12-22 15:18:50,891 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=2.10 vs. limit=2.0 2022-12-22 15:18:59,689 INFO [train.py:894] (3/4) Epoch 4, batch 1400, loss[loss=0.2492, simple_loss=0.3303, pruned_loss=0.08403, over 18651.00 frames. ], tot_loss[loss=0.264, simple_loss=0.3331, pruned_loss=0.09748, over 3710827.93 frames. ], batch size: 60, lr: 2.71e-02, grad_scale: 8.0 2022-12-22 15:19:00,700 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11920.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:19:01,917 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0037-39912-0_sp0.9 from training. Duration: 20.67225 2022-12-22 15:19:21,495 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp0.9 from training. Duration: 25.2444375 2022-12-22 15:19:42,347 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11948.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:19:43,527 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0021-76797-0_sp0.9 from training. Duration: 21.1445 2022-12-22 15:20:06,582 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7286, 1.1359, 1.4660, 1.5624, 1.6127, 1.7766, 2.0342, 1.3024], device='cuda:3'), covar=tensor([0.0702, 0.0585, 0.0755, 0.0483, 0.0443, 0.0473, 0.0304, 0.0576], device='cuda:3'), in_proj_covar=tensor([0.0071, 0.0103, 0.0122, 0.0118, 0.0098, 0.0079, 0.0072, 0.0109], device='cuda:3'), out_proj_covar=tensor([7.7782e-05, 1.0397e-04, 1.2814e-04, 1.2011e-04, 1.0575e-04, 8.2357e-05, 7.7088e-05, 1.1233e-04], device='cuda:3') 2022-12-22 15:20:15,658 INFO [train.py:894] (3/4) Epoch 4, batch 1450, loss[loss=0.2208, simple_loss=0.2944, pruned_loss=0.07359, over 18575.00 frames. ], tot_loss[loss=0.2638, simple_loss=0.3329, pruned_loss=0.09738, over 3712810.95 frames. ], batch size: 49, lr: 2.70e-02, grad_scale: 8.0 2022-12-22 15:20:32,829 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11981.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:20:37,202 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11984.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:20:47,004 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11991.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:20:58,181 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp0.9 from training. Duration: 33.038875 2022-12-22 15:21:17,910 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12009.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:21:24,163 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.421e+02 5.689e+02 8.122e+02 1.002e+03 1.607e+03, threshold=1.624e+03, percent-clipped=6.0 2022-12-22 15:21:35,435 INFO [train.py:894] (3/4) Epoch 4, batch 1500, loss[loss=0.2657, simple_loss=0.3492, pruned_loss=0.09111, over 18628.00 frames. ], tot_loss[loss=0.2625, simple_loss=0.3321, pruned_loss=0.09648, over 3713037.74 frames. ], batch size: 53, lr: 2.70e-02, grad_scale: 8.0 2022-12-22 15:21:36,011 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.6214, 2.5093, 1.9197, 2.8634, 2.1873, 2.4616, 2.6516, 3.9945], device='cuda:3'), covar=tensor([0.0886, 0.1680, 0.1039, 0.1961, 0.1820, 0.0638, 0.1663, 0.0260], device='cuda:3'), in_proj_covar=tensor([0.0205, 0.0188, 0.0170, 0.0260, 0.0182, 0.0167, 0.0195, 0.0140], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2022-12-22 15:21:41,157 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64291-0000-16059-0_sp0.9 from training. Duration: 20.0944375 2022-12-22 15:21:56,391 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp1.1 from training. Duration: 20.4 2022-12-22 15:22:03,746 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0 from training. Duration: 20.085 2022-12-22 15:22:04,162 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6028, 0.9830, 1.4081, 1.4802, 1.6365, 1.8247, 1.8906, 1.1927], device='cuda:3'), covar=tensor([0.0634, 0.0559, 0.0784, 0.0487, 0.0368, 0.0412, 0.0348, 0.0602], device='cuda:3'), in_proj_covar=tensor([0.0069, 0.0100, 0.0117, 0.0116, 0.0097, 0.0077, 0.0071, 0.0108], device='cuda:3'), out_proj_covar=tensor([7.5447e-05, 1.0139e-04, 1.2295e-04, 1.1869e-04, 1.0453e-04, 7.9505e-05, 7.6491e-05, 1.1095e-04], device='cuda:3') 2022-12-22 15:22:12,704 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12045.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:22:13,796 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0_sp0.9 from training. Duration: 23.07775 2022-12-22 15:22:27,312 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12055.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:22:50,193 INFO [train.py:894] (3/4) Epoch 4, batch 1550, loss[loss=0.2928, simple_loss=0.3599, pruned_loss=0.1128, over 18587.00 frames. ], tot_loss[loss=0.2622, simple_loss=0.3318, pruned_loss=0.09634, over 3713829.98 frames. ], batch size: 57, lr: 2.70e-02, grad_scale: 8.0 2022-12-22 15:23:00,330 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12076.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:23:01,334 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp0.9 from training. Duration: 24.9333125 2022-12-22 15:23:44,821 WARNING [train.py:1060] (3/4) Exclude cut with ID 5118-111612-0016-124680-0_sp0.9 from training. Duration: 20.388875 2022-12-22 15:23:52,592 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp1.1 from training. Duration: 20.3590625 2022-12-22 15:23:54,473 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4008, 0.8473, 0.7994, 1.0498, 1.6666, 0.7687, 1.0998, 1.2687], device='cuda:3'), covar=tensor([0.2163, 0.2857, 0.2795, 0.2289, 0.2362, 0.1969, 0.1880, 0.2177], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0108, 0.0134, 0.0105, 0.0111, 0.0098, 0.0099, 0.0103], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 15:23:55,524 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.456e+02 6.070e+02 7.487e+02 9.437e+02 1.926e+03, threshold=1.497e+03, percent-clipped=1.0 2022-12-22 15:24:06,569 INFO [train.py:894] (3/4) Epoch 4, batch 1600, loss[loss=0.2733, simple_loss=0.3297, pruned_loss=0.1084, over 18384.00 frames. ], tot_loss[loss=0.2624, simple_loss=0.3312, pruned_loss=0.09675, over 3713305.14 frames. ], batch size: 46, lr: 2.69e-02, grad_scale: 8.0 2022-12-22 15:24:32,409 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12137.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:25:00,644 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0_sp1.1 from training. Duration: 0.836375 2022-12-22 15:25:22,157 INFO [train.py:894] (3/4) Epoch 4, batch 1650, loss[loss=0.299, simple_loss=0.3439, pruned_loss=0.127, over 18707.00 frames. ], tot_loss[loss=0.2679, simple_loss=0.3348, pruned_loss=0.1005, over 3713175.17 frames. ], batch size: 50, lr: 2.69e-02, grad_scale: 16.0 2022-12-22 15:25:25,260 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12172.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:25:41,395 WARNING [train.py:1060] (3/4) Exclude cut with ID 8565-290391-0049-67394-0_sp0.9 from training. Duration: 21.3166875 2022-12-22 15:25:49,278 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12188.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:26:07,309 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.03 vs. limit=5.0 2022-12-22 15:26:09,647 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12202.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:26:13,013 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0029-104863-0_sp0.9 from training. Duration: 22.1055625 2022-12-22 15:26:23,740 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp1.1 from training. Duration: 21.77725 2022-12-22 15:26:27,788 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.397e+02 7.185e+02 8.672e+02 1.039e+03 2.165e+03, threshold=1.734e+03, percent-clipped=6.0 2022-12-22 15:26:37,347 INFO [train.py:894] (3/4) Epoch 4, batch 1700, loss[loss=0.2784, simple_loss=0.3263, pruned_loss=0.1152, over 18420.00 frames. ], tot_loss[loss=0.2723, simple_loss=0.3372, pruned_loss=0.1037, over 3713741.33 frames. ], batch size: 42, lr: 2.68e-02, grad_scale: 8.0 2022-12-22 15:26:44,756 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp0.9 from training. Duration: 27.8166875 2022-12-22 15:27:01,076 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12236.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:27:09,604 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp1.1 from training. Duration: 22.5090625 2022-12-22 15:27:15,112 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2022-12-22 15:27:17,036 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0 from training. Duration: 25.035 2022-12-22 15:27:34,412 WARNING [train.py:1060] (3/4) Exclude cut with ID 774-127930-0014-10412-0_sp1.1 from training. Duration: 0.95 2022-12-22 15:27:54,443 INFO [train.py:894] (3/4) Epoch 4, batch 1750, loss[loss=0.3179, simple_loss=0.3618, pruned_loss=0.137, over 18660.00 frames. ], tot_loss[loss=0.277, simple_loss=0.3399, pruned_loss=0.107, over 3713776.73 frames. ], batch size: 69, lr: 2.68e-02, grad_scale: 4.0 2022-12-22 15:27:54,517 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp0.9 from training. Duration: 0.92225 2022-12-22 15:28:03,256 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12276.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:28:20,579 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0 from training. Duration: 21.97 2022-12-22 15:28:25,169 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12291.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:28:40,105 WARNING [train.py:1060] (3/4) Exclude cut with ID 7492-105653-0055-62765-0_sp0.9 from training. Duration: 21.97225 2022-12-22 15:28:41,564 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp0.9 from training. Duration: 25.3333125 2022-12-22 15:28:44,645 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12304.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:28:54,013 WARNING [train.py:1060] (3/4) Exclude cut with ID 5172-29468-0015-19128-0_sp0.9 from training. Duration: 21.5055625 2022-12-22 15:29:02,126 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.995e+02 6.529e+02 8.437e+02 1.023e+03 2.036e+03, threshold=1.687e+03, percent-clipped=2.0 2022-12-22 15:29:04,813 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp1.1 from training. Duration: 20.72725 2022-12-22 15:29:08,977 INFO [train.py:894] (3/4) Epoch 4, batch 1800, loss[loss=0.3326, simple_loss=0.3894, pruned_loss=0.1379, over 18448.00 frames. ], tot_loss[loss=0.2828, simple_loss=0.3429, pruned_loss=0.1113, over 3712870.67 frames. ], batch size: 64, lr: 2.67e-02, grad_scale: 4.0 2022-12-22 15:29:12,619 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2022-12-22 15:29:34,319 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp0.9 from training. Duration: 26.32775 2022-12-22 15:29:37,400 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12339.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:29:38,842 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12340.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:29:49,272 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5567, 0.9681, 1.9711, 2.9364, 1.8234, 2.0666, 0.6999, 1.7577], device='cuda:3'), covar=tensor([0.2198, 0.2212, 0.1672, 0.0548, 0.1666, 0.1535, 0.2895, 0.1550], device='cuda:3'), in_proj_covar=tensor([0.0103, 0.0114, 0.0123, 0.0084, 0.0104, 0.0121, 0.0137, 0.0101], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002], device='cuda:3') 2022-12-22 15:30:01,987 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12355.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:30:08,378 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0 from training. Duration: 20.025 2022-12-22 15:30:11,445 WARNING [train.py:1060] (3/4) Exclude cut with ID 6709-74022-0004-86860-0_sp1.1 from training. Duration: 0.9409375 2022-12-22 15:30:11,456 WARNING [train.py:1060] (3/4) Exclude cut with ID 4757-1811-0023-62229-0_sp0.9 from training. Duration: 21.37775 2022-12-22 15:30:24,826 INFO [train.py:894] (3/4) Epoch 4, batch 1850, loss[loss=0.3159, simple_loss=0.3634, pruned_loss=0.1343, over 18680.00 frames. ], tot_loss[loss=0.2868, simple_loss=0.3442, pruned_loss=0.1147, over 3713714.96 frames. ], batch size: 48, lr: 2.67e-02, grad_scale: 4.0 2022-12-22 15:30:33,673 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0004-25974-0_sp0.9 from training. Duration: 21.17225 2022-12-22 15:30:33,686 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp0.9 from training. Duration: 27.511125 2022-12-22 15:31:04,865 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0 from training. Duration: 22.8 2022-12-22 15:31:09,242 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0 from training. Duration: 22.585 2022-12-22 15:31:14,783 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12403.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:31:33,079 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.783e+02 7.220e+02 8.554e+02 1.149e+03 2.507e+03, threshold=1.711e+03, percent-clipped=5.0 2022-12-22 15:31:36,395 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7811, 1.7504, 1.9751, 1.2853, 1.8938, 1.8044, 1.2565, 2.0612], device='cuda:3'), covar=tensor([0.0843, 0.1143, 0.1051, 0.1362, 0.0696, 0.0992, 0.2078, 0.0516], device='cuda:3'), in_proj_covar=tensor([0.0192, 0.0167, 0.0194, 0.0192, 0.0182, 0.0198, 0.0191, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 15:31:40,396 INFO [train.py:894] (3/4) Epoch 4, batch 1900, loss[loss=0.287, simple_loss=0.3537, pruned_loss=0.1101, over 18729.00 frames. ], tot_loss[loss=0.2902, simple_loss=0.346, pruned_loss=0.1172, over 3713709.92 frames. ], batch size: 52, lr: 2.66e-02, grad_scale: 4.0 2022-12-22 15:31:40,428 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0001-146967-0_sp0.9 from training. Duration: 22.0166875 2022-12-22 15:31:44,490 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2022-12-22 15:31:55,450 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp1.1 from training. Duration: 24.395375 2022-12-22 15:31:58,717 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12432.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:32:00,783 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2022-12-22 15:32:03,221 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp0.9 from training. Duration: 27.47775 2022-12-22 15:32:08,104 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp0.9 from training. Duration: 24.8833125 2022-12-22 15:32:11,025 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0 from training. Duration: 23.39 2022-12-22 15:32:16,825 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp0.9 from training. Duration: 28.72225 2022-12-22 15:32:19,920 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12446.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:32:26,344 WARNING [train.py:1060] (3/4) Exclude cut with ID 585-294811-0110-133686-0_sp0.9 from training. Duration: 20.8944375 2022-12-22 15:32:39,510 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0_sp0.9 from training. Duration: 23.8444375 2022-12-22 15:32:55,866 INFO [train.py:894] (3/4) Epoch 4, batch 1950, loss[loss=0.3204, simple_loss=0.3663, pruned_loss=0.1373, over 18552.00 frames. ], tot_loss[loss=0.2939, simple_loss=0.348, pruned_loss=0.1199, over 3714260.42 frames. ], batch size: 77, lr: 2.66e-02, grad_scale: 4.0 2022-12-22 15:32:59,297 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12472.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:33:02,044 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0 from training. Duration: 25.85 2022-12-22 15:33:02,054 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0 from training. Duration: 21.39 2022-12-22 15:33:12,670 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0 from training. Duration: 27.92 2022-12-22 15:33:40,846 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0039-130165-0_sp0.9 from training. Duration: 20.661125 2022-12-22 15:33:45,448 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12502.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:33:52,963 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12507.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:34:04,116 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 5.166e+02 7.819e+02 9.278e+02 1.167e+03 2.388e+03, threshold=1.856e+03, percent-clipped=8.0 2022-12-22 15:34:04,191 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0043-15874-0_sp0.9 from training. Duration: 20.07225 2022-12-22 15:34:12,016 INFO [train.py:894] (3/4) Epoch 4, batch 2000, loss[loss=0.2981, simple_loss=0.3342, pruned_loss=0.1309, over 18465.00 frames. ], tot_loss[loss=0.2948, simple_loss=0.3478, pruned_loss=0.1209, over 3713346.09 frames. ], batch size: 43, lr: 2.65e-02, grad_scale: 8.0 2022-12-22 15:34:12,064 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0 from training. Duration: 21.01 2022-12-22 15:34:12,187 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12520.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:34:58,656 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12550.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:35:22,244 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0 from training. Duration: 20.65 2022-12-22 15:35:27,885 INFO [train.py:894] (3/4) Epoch 4, batch 2050, loss[loss=0.262, simple_loss=0.317, pruned_loss=0.1035, over 18631.00 frames. ], tot_loss[loss=0.2938, simple_loss=0.347, pruned_loss=0.1203, over 3713337.82 frames. ], batch size: 45, lr: 2.65e-02, grad_scale: 8.0 2022-12-22 15:35:29,509 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0 from training. Duration: 21.46 2022-12-22 15:35:37,188 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12576.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:36:15,541 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.2598, 2.1446, 1.5146, 2.8584, 1.9308, 2.0290, 2.2076, 3.6749], device='cuda:3'), covar=tensor([0.1068, 0.1906, 0.1135, 0.1795, 0.2166, 0.0699, 0.1917, 0.0321], device='cuda:3'), in_proj_covar=tensor([0.0221, 0.0196, 0.0179, 0.0271, 0.0188, 0.0174, 0.0204, 0.0148], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 15:36:16,481 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0 from training. Duration: 0.92 2022-12-22 15:36:21,158 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12604.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:36:22,371 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0_sp0.9 from training. Duration: 23.7666875 2022-12-22 15:36:36,980 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.601e+02 6.890e+02 8.468e+02 1.080e+03 1.755e+03, threshold=1.694e+03, percent-clipped=0.0 2022-12-22 15:36:44,324 INFO [train.py:894] (3/4) Epoch 4, batch 2100, loss[loss=0.2664, simple_loss=0.333, pruned_loss=0.09988, over 18494.00 frames. ], tot_loss[loss=0.2937, simple_loss=0.3469, pruned_loss=0.1203, over 3712726.50 frames. ], batch size: 77, lr: 2.64e-02, grad_scale: 8.0 2022-12-22 15:36:51,466 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12624.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:36:57,498 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.54 vs. limit=5.0 2022-12-22 15:37:00,650 WARNING [train.py:1060] (3/4) Exclude cut with ID 8544-281189-0060-101339-0_sp0.9 from training. Duration: 20.861125 2022-12-22 15:37:13,524 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0_sp0.9 from training. Duration: 22.711125 2022-12-22 15:37:16,589 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12640.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:37:34,285 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12652.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:37:52,302 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp1.1 from training. Duration: 22.986375 2022-12-22 15:38:00,975 INFO [train.py:894] (3/4) Epoch 4, batch 2150, loss[loss=0.26, simple_loss=0.3122, pruned_loss=0.1039, over 18521.00 frames. ], tot_loss[loss=0.2939, simple_loss=0.3468, pruned_loss=0.1205, over 3713189.55 frames. ], batch size: 44, lr: 2.64e-02, grad_scale: 8.0 2022-12-22 15:38:05,036 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3222, 1.6108, 2.1559, 2.4153, 2.0986, 1.7904, 2.2273, 1.2770], device='cuda:3'), covar=tensor([0.0818, 0.1502, 0.1007, 0.1014, 0.0630, 0.0517, 0.1073, 0.0669], device='cuda:3'), in_proj_covar=tensor([0.0169, 0.0202, 0.0189, 0.0192, 0.0170, 0.0161, 0.0182, 0.0159], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 15:38:09,474 WARNING [train.py:1060] (3/4) Exclude cut with ID 8040-260924-0003-80960-0_sp0.9 from training. Duration: 22.07225 2022-12-22 15:38:12,530 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0045-26330-0_sp0.9 from training. Duration: 20.3055625 2022-12-22 15:38:15,552 WARNING [train.py:1060] (3/4) Exclude cut with ID 6356-271890-0060-94317-0_sp0.9 from training. Duration: 20.72225 2022-12-22 15:38:30,021 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12688.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:38:32,313 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2022-12-22 15:38:37,065 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp1.1 from training. Duration: 22.4818125 2022-12-22 15:39:02,222 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp0.9 from training. Duration: 25.0944375 2022-12-22 15:39:06,501 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0 from training. Duration: 21.515 2022-12-22 15:39:09,165 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.874e+02 7.222e+02 8.655e+02 1.105e+03 2.731e+03, threshold=1.731e+03, percent-clipped=2.0 2022-12-22 15:39:14,388 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp0.9 from training. Duration: 27.02225 2022-12-22 15:39:17,194 INFO [train.py:894] (3/4) Epoch 4, batch 2200, loss[loss=0.3329, simple_loss=0.3828, pruned_loss=0.1415, over 18695.00 frames. ], tot_loss[loss=0.2954, simple_loss=0.3477, pruned_loss=0.1215, over 3713991.40 frames. ], batch size: 60, lr: 2.64e-02, grad_scale: 8.0 2022-12-22 15:39:19,482 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0010-62480-0_sp0.9 from training. Duration: 22.22225 2022-12-22 15:39:26,862 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0085-44554-0_sp0.9 from training. Duration: 20.85 2022-12-22 15:39:37,180 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12732.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:39:57,134 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0 from training. Duration: 21.54 2022-12-22 15:40:01,899 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp1.1 from training. Duration: 20.5318125 2022-12-22 15:40:09,928 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0012-134311-0_sp0.9 from training. Duration: 21.9333125 2022-12-22 15:40:25,816 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12763.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:40:36,700 INFO [train.py:894] (3/4) Epoch 4, batch 2250, loss[loss=0.2933, simple_loss=0.3436, pruned_loss=0.1215, over 18442.00 frames. ], tot_loss[loss=0.2952, simple_loss=0.3476, pruned_loss=0.1214, over 3713509.02 frames. ], batch size: 50, lr: 2.63e-02, grad_scale: 8.0 2022-12-22 15:40:52,038 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8238, 1.8076, 1.4812, 2.6101, 2.0398, 4.6619, 1.5272, 2.1796], device='cuda:3'), covar=tensor([0.1190, 0.1989, 0.1522, 0.1007, 0.1625, 0.0167, 0.1622, 0.1707], device='cuda:3'), in_proj_covar=tensor([0.0087, 0.0093, 0.0091, 0.0087, 0.0106, 0.0077, 0.0097, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 15:40:53,270 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12780.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:40:58,837 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0025-130151-0_sp0.9 from training. Duration: 21.7944375 2022-12-22 15:41:10,042 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.75 vs. limit=5.0 2022-12-22 15:41:10,517 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0_sp0.9 from training. Duration: 22.4666875 2022-12-22 15:41:16,477 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0 from training. Duration: 21.635 2022-12-22 15:41:22,627 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0_sp0.9 from training. Duration: 24.038875 2022-12-22 15:41:25,678 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12802.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:41:47,264 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.686e+02 7.014e+02 8.865e+02 1.069e+03 2.354e+03, threshold=1.773e+03, percent-clipped=4.0 2022-12-22 15:41:55,104 INFO [train.py:894] (3/4) Epoch 4, batch 2300, loss[loss=0.3241, simple_loss=0.3719, pruned_loss=0.1381, over 18721.00 frames. ], tot_loss[loss=0.2973, simple_loss=0.349, pruned_loss=0.1228, over 3713538.29 frames. ], batch size: 60, lr: 2.63e-02, grad_scale: 8.0 2022-12-22 15:42:01,623 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12824.0, num_to_drop=1, layers_to_drop={3} 2022-12-22 15:42:05,543 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp1.1 from training. Duration: 21.786375 2022-12-22 15:42:15,552 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0 from training. Duration: 20.22 2022-12-22 15:42:20,452 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.3716, 1.2360, 1.3912, 0.5268, 0.9691, 1.5219, 1.3706, 1.2540], device='cuda:3'), covar=tensor([0.0710, 0.0319, 0.0345, 0.0497, 0.0528, 0.0289, 0.0314, 0.0602], device='cuda:3'), in_proj_covar=tensor([0.0105, 0.0115, 0.0084, 0.0109, 0.0108, 0.0078, 0.0117, 0.0094], device='cuda:3'), out_proj_covar=tensor([1.1320e-04, 1.2388e-04, 9.3082e-05, 1.1663e-04, 1.1558e-04, 8.5236e-05, 1.2933e-04, 1.0269e-04], device='cuda:3') 2022-12-22 15:43:11,115 INFO [train.py:894] (3/4) Epoch 4, batch 2350, loss[loss=0.2573, simple_loss=0.3136, pruned_loss=0.1005, over 18526.00 frames. ], tot_loss[loss=0.2953, simple_loss=0.3473, pruned_loss=0.1217, over 3713154.08 frames. ], batch size: 44, lr: 2.62e-02, grad_scale: 8.0 2022-12-22 15:44:20,057 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.811e+02 6.843e+02 8.099e+02 1.087e+03 2.395e+03, threshold=1.620e+03, percent-clipped=1.0 2022-12-22 15:44:21,489 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0 from training. Duration: 25.285 2022-12-22 15:44:27,503 INFO [train.py:894] (3/4) Epoch 4, batch 2400, loss[loss=0.2633, simple_loss=0.3323, pruned_loss=0.09711, over 18615.00 frames. ], tot_loss[loss=0.294, simple_loss=0.3466, pruned_loss=0.1207, over 3713344.51 frames. ], batch size: 53, lr: 2.62e-02, grad_scale: 8.0 2022-12-22 15:45:26,862 WARNING [train.py:1060] (3/4) Exclude cut with ID 811-130148-0001-63453-0_sp0.9 from training. Duration: 20.861125 2022-12-22 15:45:43,628 INFO [train.py:894] (3/4) Epoch 4, batch 2450, loss[loss=0.2809, simple_loss=0.3507, pruned_loss=0.1055, over 18488.00 frames. ], tot_loss[loss=0.2945, simple_loss=0.3473, pruned_loss=0.1208, over 3713539.39 frames. ], batch size: 64, lr: 2.61e-02, grad_scale: 8.0 2022-12-22 15:45:49,683 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0 from training. Duration: 20.88 2022-12-22 15:46:11,463 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2022-12-22 15:46:20,309 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0_sp0.9 from training. Duration: 23.4166875 2022-12-22 15:46:51,986 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.109e+02 7.565e+02 9.195e+02 1.170e+03 3.094e+03, threshold=1.839e+03, percent-clipped=13.0 2022-12-22 15:46:59,552 INFO [train.py:894] (3/4) Epoch 4, batch 2500, loss[loss=0.335, simple_loss=0.3809, pruned_loss=0.1445, over 18726.00 frames. ], tot_loss[loss=0.2955, simple_loss=0.348, pruned_loss=0.1215, over 3714741.75 frames. ], batch size: 54, lr: 2.61e-02, grad_scale: 8.0 2022-12-22 15:47:38,564 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0 from training. Duration: 21.24 2022-12-22 15:47:40,160 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0_sp0.9 from training. Duration: 23.9055625 2022-12-22 15:48:13,914 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp0.9 from training. Duration: 25.988875 2022-12-22 15:48:15,369 INFO [train.py:894] (3/4) Epoch 4, batch 2550, loss[loss=0.3435, simple_loss=0.3757, pruned_loss=0.1556, over 18502.00 frames. ], tot_loss[loss=0.2939, simple_loss=0.347, pruned_loss=0.1204, over 3715068.86 frames. ], batch size: 64, lr: 2.60e-02, grad_scale: 8.0 2022-12-22 15:48:16,265 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.84 vs. limit=5.0 2022-12-22 15:48:22,567 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0001-134300-0_sp0.9 from training. Duration: 20.67225 2022-12-22 15:49:05,483 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=13102.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:49:09,790 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0 from training. Duration: 20.34 2022-12-22 15:49:16,221 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.0456, 0.9362, 1.0730, 0.0538, 0.5900, 1.2328, 1.1330, 1.0481], device='cuda:3'), covar=tensor([0.0628, 0.0335, 0.0370, 0.0493, 0.0467, 0.0274, 0.0291, 0.0455], device='cuda:3'), in_proj_covar=tensor([0.0109, 0.0118, 0.0087, 0.0113, 0.0109, 0.0080, 0.0116, 0.0095], device='cuda:3'), out_proj_covar=tensor([1.1621e-04, 1.2734e-04, 9.4690e-05, 1.2100e-04, 1.1686e-04, 8.6861e-05, 1.2762e-04, 1.0367e-04], device='cuda:3') 2022-12-22 15:49:24,507 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.669e+02 7.177e+02 8.673e+02 1.022e+03 2.536e+03, threshold=1.735e+03, percent-clipped=2.0 2022-12-22 15:49:30,798 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=13119.0, num_to_drop=1, layers_to_drop={2} 2022-12-22 15:49:31,960 INFO [train.py:894] (3/4) Epoch 4, batch 2600, loss[loss=0.3029, simple_loss=0.3514, pruned_loss=0.1272, over 18481.00 frames. ], tot_loss[loss=0.2927, simple_loss=0.3464, pruned_loss=0.1195, over 3715193.97 frames. ], batch size: 54, lr: 2.60e-02, grad_scale: 8.0 2022-12-22 15:50:18,444 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=13150.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:50:22,580 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp0.9 from training. Duration: 25.061125 2022-12-22 15:50:34,830 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0 from training. Duration: 0.83 2022-12-22 15:50:48,033 INFO [train.py:894] (3/4) Epoch 4, batch 2650, loss[loss=0.3143, simple_loss=0.374, pruned_loss=0.1273, over 18634.00 frames. ], tot_loss[loss=0.293, simple_loss=0.3469, pruned_loss=0.1196, over 3715710.28 frames. ], batch size: 53, lr: 2.60e-02, grad_scale: 8.0 2022-12-22 15:50:59,212 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0 from training. Duration: 24.73 2022-12-22 15:51:10,935 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0 from training. Duration: 23.965 2022-12-22 15:51:21,667 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0030-146996-0_sp0.9 from training. Duration: 22.088875 2022-12-22 15:51:37,312 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0_sp0.9 from training. Duration: 23.6 2022-12-22 15:51:49,493 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8541, 1.6650, 2.1459, 1.3728, 2.1960, 2.2320, 1.4703, 2.3808], device='cuda:3'), covar=tensor([0.1403, 0.1668, 0.1463, 0.2157, 0.1097, 0.1279, 0.2517, 0.0767], device='cuda:3'), in_proj_covar=tensor([0.0195, 0.0173, 0.0196, 0.0190, 0.0184, 0.0203, 0.0196, 0.0176], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 15:51:58,021 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.369e+02 7.191e+02 8.861e+02 1.051e+03 2.028e+03, threshold=1.772e+03, percent-clipped=5.0 2022-12-22 15:52:06,066 INFO [train.py:894] (3/4) Epoch 4, batch 2700, loss[loss=0.2376, simple_loss=0.3079, pruned_loss=0.08371, over 18682.00 frames. ], tot_loss[loss=0.2906, simple_loss=0.3445, pruned_loss=0.1183, over 3715486.07 frames. ], batch size: 48, lr: 2.59e-02, grad_scale: 8.0 2022-12-22 15:53:19,710 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0 from training. Duration: 23.795 2022-12-22 15:53:22,788 INFO [train.py:894] (3/4) Epoch 4, batch 2750, loss[loss=0.3253, simple_loss=0.3742, pruned_loss=0.1382, over 18531.00 frames. ], tot_loss[loss=0.2902, simple_loss=0.3439, pruned_loss=0.1183, over 3715883.42 frames. ], batch size: 58, lr: 2.59e-02, grad_scale: 8.0 2022-12-22 15:53:38,001 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp1.1 from training. Duration: 21.5409375 2022-12-22 15:53:40,991 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp0.9 from training. Duration: 24.97775 2022-12-22 15:53:53,079 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0_sp0.9 from training. Duration: 23.3444375 2022-12-22 15:54:19,724 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0_sp0.9 from training. Duration: 23.2 2022-12-22 15:54:25,860 WARNING [train.py:1060] (3/4) Exclude cut with ID 5653-46179-0060-117930-0_sp0.9 from training. Duration: 21.17225 2022-12-22 15:54:31,516 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.064e+02 6.793e+02 8.291e+02 1.011e+03 1.777e+03, threshold=1.658e+03, percent-clipped=1.0 2022-12-22 15:54:39,276 INFO [train.py:894] (3/4) Epoch 4, batch 2800, loss[loss=0.3433, simple_loss=0.3854, pruned_loss=0.1506, over 18560.00 frames. ], tot_loss[loss=0.2916, simple_loss=0.345, pruned_loss=0.1191, over 3715355.98 frames. ], batch size: 99, lr: 2.58e-02, grad_scale: 8.0 2022-12-22 15:54:46,039 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp0.9 from training. Duration: 24.6555625 2022-12-22 15:55:10,060 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2719, 1.9051, 1.6789, 2.8199, 2.0803, 4.8126, 1.9021, 2.1135], device='cuda:3'), covar=tensor([0.1071, 0.1806, 0.1294, 0.0960, 0.1642, 0.0159, 0.1411, 0.1588], device='cuda:3'), in_proj_covar=tensor([0.0084, 0.0091, 0.0087, 0.0086, 0.0105, 0.0076, 0.0095, 0.0086], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 15:55:40,872 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0 from training. Duration: 20.44 2022-12-22 15:55:56,109 INFO [train.py:894] (3/4) Epoch 4, batch 2850, loss[loss=0.2964, simple_loss=0.3516, pruned_loss=0.1206, over 18555.00 frames. ], tot_loss[loss=0.292, simple_loss=0.3457, pruned_loss=0.1191, over 3715969.55 frames. ], batch size: 69, lr: 2.58e-02, grad_scale: 8.0 2022-12-22 15:55:56,193 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0_sp0.9 from training. Duration: 23.45 2022-12-22 15:56:25,700 WARNING [train.py:1060] (3/4) Exclude cut with ID 6945-60535-0076-12784-0_sp0.9 from training. Duration: 20.52225 2022-12-22 15:56:32,906 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0 from training. Duration: 22.19 2022-12-22 15:56:36,565 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.33 vs. limit=5.0 2022-12-22 15:56:43,414 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp1.1 from training. Duration: 25.3818125 2022-12-22 15:56:51,156 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2022-12-22 15:56:59,889 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp0.9 from training. Duration: 28.0944375 2022-12-22 15:57:03,922 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.958e+02 6.920e+02 8.531e+02 1.147e+03 2.504e+03, threshold=1.706e+03, percent-clipped=8.0 2022-12-22 15:57:05,502 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0_sp0.9 from training. Duration: 22.9444375 2022-12-22 15:57:10,566 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=13419.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:57:11,740 INFO [train.py:894] (3/4) Epoch 4, batch 2900, loss[loss=0.2935, simple_loss=0.3429, pruned_loss=0.122, over 18679.00 frames. ], tot_loss[loss=0.2926, simple_loss=0.3463, pruned_loss=0.1194, over 3715172.53 frames. ], batch size: 48, lr: 2.57e-02, grad_scale: 8.0 2022-12-22 15:57:13,351 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp1.1 from training. Duration: 21.6318125 2022-12-22 15:57:32,294 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0 from training. Duration: 23.695 2022-12-22 15:57:42,792 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6055, 3.4005, 3.4818, 1.8475, 3.3639, 2.5029, 0.9955, 2.3679], device='cuda:3'), covar=tensor([0.1845, 0.0875, 0.1465, 0.3394, 0.1177, 0.1417, 0.5130, 0.1882], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0099, 0.0148, 0.0116, 0.0105, 0.0099, 0.0144, 0.0110], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 15:57:56,611 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2022-12-22 15:57:56,982 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0 from training. Duration: 23.955 2022-12-22 15:58:09,701 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6411, 1.5494, 1.9876, 1.1176, 1.7617, 1.8409, 1.2337, 2.2073], device='cuda:3'), covar=tensor([0.0983, 0.1306, 0.1061, 0.1518, 0.0764, 0.1052, 0.2188, 0.0467], device='cuda:3'), in_proj_covar=tensor([0.0196, 0.0175, 0.0199, 0.0192, 0.0182, 0.0205, 0.0200, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 15:58:23,222 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=13467.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 15:58:27,991 INFO [train.py:894] (3/4) Epoch 4, batch 2950, loss[loss=0.2541, simple_loss=0.3058, pruned_loss=0.1013, over 18513.00 frames. ], tot_loss[loss=0.2919, simple_loss=0.3457, pruned_loss=0.1191, over 3715302.47 frames. ], batch size: 44, lr: 2.57e-02, grad_scale: 8.0 2022-12-22 15:58:34,230 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp0.9 from training. Duration: 26.438875 2022-12-22 15:59:19,153 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0021-26306-0_sp0.9 from training. Duration: 21.2444375 2022-12-22 15:59:19,179 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp0.9 from training. Duration: 31.02225 2022-12-22 15:59:31,019 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0 from training. Duration: 22.395 2022-12-22 15:59:36,265 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.113e+02 6.895e+02 8.488e+02 1.043e+03 2.580e+03, threshold=1.698e+03, percent-clipped=2.0 2022-12-22 15:59:44,459 INFO [train.py:894] (3/4) Epoch 4, batch 3000, loss[loss=0.3099, simple_loss=0.3628, pruned_loss=0.1285, over 18675.00 frames. ], tot_loss[loss=0.2901, simple_loss=0.3446, pruned_loss=0.1178, over 3715711.85 frames. ], batch size: 60, lr: 2.57e-02, grad_scale: 8.0 2022-12-22 15:59:44,459 INFO [train.py:919] (3/4) Computing validation loss 2022-12-22 15:59:55,854 INFO [train.py:928] (3/4) Epoch 4, validation: loss=0.2143, simple_loss=0.31, pruned_loss=0.05929, over 944034.00 frames. 2022-12-22 15:59:55,855 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24676MB 2022-12-22 15:59:58,851 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0 from training. Duration: 21.075 2022-12-22 16:00:04,616 WARNING [train.py:1060] (3/4) Exclude cut with ID 6482-98857-0025-147532-0_sp0.9 from training. Duration: 20.0055625 2022-12-22 16:00:04,630 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0037-132304-0_sp0.9 from training. Duration: 22.05 2022-12-22 16:00:04,647 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0 from training. Duration: 26.8349375 2022-12-22 16:00:09,253 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp1.1 from training. Duration: 22.1090625 2022-12-22 16:00:15,109 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp0.9 from training. Duration: 26.6166875 2022-12-22 16:00:15,797 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2022-12-22 16:00:30,884 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6993, 2.3047, 1.4506, 3.2979, 2.9073, 1.5058, 2.1299, 1.2525], device='cuda:3'), covar=tensor([0.1819, 0.1575, 0.1449, 0.0578, 0.1405, 0.1155, 0.1581, 0.1498], device='cuda:3'), in_proj_covar=tensor([0.0212, 0.0183, 0.0170, 0.0161, 0.0224, 0.0164, 0.0182, 0.0173], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 16:00:33,721 WARNING [train.py:1060] (3/4) Exclude cut with ID 2046-178027-0000-53705-0_sp0.9 from training. Duration: 20.3055625 2022-12-22 16:00:53,935 WARNING [train.py:1060] (3/4) Exclude cut with ID 7205-50138-0008-5373-0_sp0.9 from training. Duration: 20.7 2022-12-22 16:01:11,983 INFO [train.py:894] (3/4) Epoch 4, batch 3050, loss[loss=0.2997, simple_loss=0.348, pruned_loss=0.1256, over 18617.00 frames. ], tot_loss[loss=0.2918, simple_loss=0.3458, pruned_loss=0.1188, over 3715676.06 frames. ], batch size: 176, lr: 2.56e-02, grad_scale: 8.0 2022-12-22 16:01:36,237 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0 from training. Duration: 22.48 2022-12-22 16:01:51,692 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp0.9 from training. Duration: 29.816625 2022-12-22 16:02:12,350 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp1.1 from training. Duration: 22.7590625 2022-12-22 16:02:12,663 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.2931, 0.9654, 1.3825, 1.7761, 1.2892, 1.5737, 0.7847, 1.2789], device='cuda:3'), covar=tensor([0.1478, 0.1512, 0.1169, 0.0687, 0.1289, 0.1607, 0.1989, 0.1157], device='cuda:3'), in_proj_covar=tensor([0.0109, 0.0117, 0.0125, 0.0092, 0.0108, 0.0127, 0.0140, 0.0105], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 16:02:17,201 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0 from training. Duration: 22.555 2022-12-22 16:02:21,449 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.251e+02 7.437e+02 9.296e+02 1.089e+03 1.791e+03, threshold=1.859e+03, percent-clipped=1.0 2022-12-22 16:02:30,305 INFO [train.py:894] (3/4) Epoch 4, batch 3100, loss[loss=0.2971, simple_loss=0.3456, pruned_loss=0.1243, over 18616.00 frames. ], tot_loss[loss=0.2927, simple_loss=0.3466, pruned_loss=0.1195, over 3715131.39 frames. ], batch size: 51, lr: 2.56e-02, grad_scale: 8.0 2022-12-22 16:02:38,941 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0005-25975-0_sp0.9 from training. Duration: 21.688875 2022-12-22 16:03:16,228 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0_sp0.9 from training. Duration: 22.6 2022-12-22 16:03:44,457 INFO [train.py:894] (3/4) Epoch 4, batch 3150, loss[loss=0.2127, simple_loss=0.2819, pruned_loss=0.0717, over 18596.00 frames. ], tot_loss[loss=0.2905, simple_loss=0.345, pruned_loss=0.118, over 3715728.53 frames. ], batch size: 45, lr: 2.55e-02, grad_scale: 8.0 2022-12-22 16:03:53,697 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0 from training. Duration: 24.32 2022-12-22 16:03:59,315 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.87 vs. limit=5.0 2022-12-22 16:04:48,027 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3546, 2.0468, 1.3016, 2.3151, 1.9321, 1.5206, 1.7455, 2.6036], device='cuda:3'), covar=tensor([0.1371, 0.1835, 0.1323, 0.2061, 0.1827, 0.0920, 0.1989, 0.0490], device='cuda:3'), in_proj_covar=tensor([0.0232, 0.0210, 0.0189, 0.0289, 0.0199, 0.0180, 0.0214, 0.0154], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 16:04:53,196 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.606e+02 7.220e+02 8.309e+02 1.076e+03 2.757e+03, threshold=1.662e+03, percent-clipped=2.0 2022-12-22 16:04:53,258 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-276745-0093-13116-0_sp0.9 from training. Duration: 21.061125 2022-12-22 16:05:01,303 INFO [train.py:894] (3/4) Epoch 4, batch 3200, loss[loss=0.2564, simple_loss=0.3147, pruned_loss=0.09907, over 18673.00 frames. ], tot_loss[loss=0.2892, simple_loss=0.344, pruned_loss=0.1172, over 3716037.51 frames. ], batch size: 48, lr: 2.55e-02, grad_scale: 8.0 2022-12-22 16:05:07,403 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0024-15855-0_sp0.9 from training. Duration: 20.32225 2022-12-22 16:05:12,258 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7405, 3.7915, 4.0401, 1.8583, 3.8142, 2.9376, 0.8296, 2.7702], device='cuda:3'), covar=tensor([0.1571, 0.0861, 0.1157, 0.3413, 0.0996, 0.1117, 0.5384, 0.1659], device='cuda:3'), in_proj_covar=tensor([0.0126, 0.0101, 0.0152, 0.0118, 0.0108, 0.0100, 0.0147, 0.0112], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 16:05:21,088 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp1.1 from training. Duration: 0.7545625 2022-12-22 16:05:28,802 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=13737.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 16:05:35,472 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0_sp0.9 from training. Duration: 23.9333125 2022-12-22 16:05:52,031 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2022-12-22 16:06:04,910 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp1.1 from training. Duration: 20.17275 2022-12-22 16:06:12,686 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp1.1 from training. Duration: 20.436375 2022-12-22 16:06:16,907 INFO [train.py:894] (3/4) Epoch 4, batch 3250, loss[loss=0.3025, simple_loss=0.3567, pruned_loss=0.1242, over 18592.00 frames. ], tot_loss[loss=0.2889, simple_loss=0.3436, pruned_loss=0.1171, over 3715332.63 frames. ], batch size: 98, lr: 2.55e-02, grad_scale: 8.0 2022-12-22 16:06:59,671 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=13798.0, num_to_drop=1, layers_to_drop={3} 2022-12-22 16:07:25,856 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 5.468e+02 7.247e+02 9.411e+02 1.182e+03 2.246e+03, threshold=1.882e+03, percent-clipped=4.0 2022-12-22 16:07:33,607 INFO [train.py:894] (3/4) Epoch 4, batch 3300, loss[loss=0.2768, simple_loss=0.3444, pruned_loss=0.1047, over 18548.00 frames. ], tot_loss[loss=0.2889, simple_loss=0.3436, pruned_loss=0.1171, over 3715477.82 frames. ], batch size: 55, lr: 2.54e-02, grad_scale: 8.0 2022-12-22 16:07:34,939 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0_sp0.9 from training. Duration: 23.1055625 2022-12-22 16:07:37,752 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp1.1 from training. Duration: 23.5 2022-12-22 16:07:49,596 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp0.9 from training. Duration: 26.62775 2022-12-22 16:07:55,486 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.0828, 1.2206, 0.4406, 1.1004, 1.2541, 2.3959, 1.1959, 1.2817], device='cuda:3'), covar=tensor([0.1190, 0.1895, 0.1524, 0.1111, 0.1662, 0.0402, 0.1456, 0.1770], device='cuda:3'), in_proj_covar=tensor([0.0086, 0.0090, 0.0087, 0.0087, 0.0106, 0.0078, 0.0095, 0.0086], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 16:08:02,606 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0 from training. Duration: 21.105 2022-12-22 16:08:07,121 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0_sp0.9 from training. Duration: 24.411125 2022-12-22 16:08:35,057 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp1.1 from training. Duration: 21.263625 2022-12-22 16:08:40,712 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=2.22 vs. limit=2.0 2022-12-22 16:08:48,539 INFO [train.py:894] (3/4) Epoch 4, batch 3350, loss[loss=0.3088, simple_loss=0.3661, pruned_loss=0.1258, over 18469.00 frames. ], tot_loss[loss=0.289, simple_loss=0.3438, pruned_loss=0.1171, over 3715308.99 frames. ], batch size: 54, lr: 2.54e-02, grad_scale: 8.0 2022-12-22 16:09:04,022 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2022-12-22 16:09:06,202 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0 from training. Duration: 20.795 2022-12-22 16:09:15,218 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0 from training. Duration: 24.76 2022-12-22 16:09:15,235 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0_sp0.9 from training. Duration: 22.25 2022-12-22 16:09:39,805 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp1.1 from training. Duration: 20.5045625 2022-12-22 16:09:44,323 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4082, 1.6225, 1.5677, 1.6505, 1.5507, 3.4568, 1.6415, 2.0817], device='cuda:3'), covar=tensor([0.3676, 0.2008, 0.1938, 0.2059, 0.1448, 0.0223, 0.1600, 0.1081], device='cuda:3'), in_proj_covar=tensor([0.0153, 0.0129, 0.0146, 0.0129, 0.0122, 0.0098, 0.0113, 0.0110], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 16:09:57,419 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 5.151e+02 8.018e+02 9.433e+02 1.110e+03 2.180e+03, threshold=1.887e+03, percent-clipped=3.0 2022-12-22 16:10:06,399 INFO [train.py:894] (3/4) Epoch 4, batch 3400, loss[loss=0.2821, simple_loss=0.3394, pruned_loss=0.1124, over 18703.00 frames. ], tot_loss[loss=0.2885, simple_loss=0.3433, pruned_loss=0.1169, over 3715855.16 frames. ], batch size: 50, lr: 2.53e-02, grad_scale: 8.0 2022-12-22 16:11:18,207 INFO [train.py:894] (3/4) Epoch 4, batch 3450, loss[loss=0.2177, simple_loss=0.2879, pruned_loss=0.07369, over 18419.00 frames. ], tot_loss[loss=0.2887, simple_loss=0.3436, pruned_loss=0.1169, over 3716016.71 frames. ], batch size: 48, lr: 2.53e-02, grad_scale: 8.0 2022-12-22 16:11:50,458 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2022-12-22 16:12:17,824 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2022-12-22 16:12:27,012 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.97 vs. limit=2.0 2022-12-22 16:12:27,571 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.799e+02 6.955e+02 8.936e+02 1.193e+03 2.861e+03, threshold=1.787e+03, percent-clipped=3.0 2022-12-22 16:12:36,000 INFO [train.py:894] (3/4) Epoch 4, batch 3500, loss[loss=0.2899, simple_loss=0.3428, pruned_loss=0.1185, over 18605.00 frames. ], tot_loss[loss=0.2897, simple_loss=0.3443, pruned_loss=0.1176, over 3715774.70 frames. ], batch size: 165, lr: 2.53e-02, grad_scale: 8.0 2022-12-22 16:12:57,740 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp1.1 from training. Duration: 22.2954375 2022-12-22 16:13:07,758 INFO [train.py:894] (3/4) Epoch 5, batch 0, loss[loss=0.2955, simple_loss=0.3433, pruned_loss=0.1238, over 18398.00 frames. ], tot_loss[loss=0.2955, simple_loss=0.3433, pruned_loss=0.1238, over 18398.00 frames. ], batch size: 46, lr: 2.35e-02, grad_scale: 8.0 2022-12-22 16:13:07,758 INFO [train.py:919] (3/4) Computing validation loss 2022-12-22 16:13:19,098 INFO [train.py:928] (3/4) Epoch 5, validation: loss=0.2114, simple_loss=0.3085, pruned_loss=0.05719, over 944034.00 frames. 2022-12-22 16:13:19,099 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24676MB 2022-12-22 16:14:09,323 WARNING [train.py:1060] (3/4) Exclude cut with ID 298-126791-0067-24026-0_sp0.9 from training. Duration: 21.438875 2022-12-22 16:14:16,497 WARNING [train.py:1060] (3/4) Exclude cut with ID 5652-39938-0025-23684-0_sp0.9 from training. Duration: 22.2055625 2022-12-22 16:14:27,010 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.39 vs. limit=5.0 2022-12-22 16:14:29,310 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14072.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 16:14:34,721 INFO [train.py:894] (3/4) Epoch 5, batch 50, loss[loss=0.255, simple_loss=0.3315, pruned_loss=0.08922, over 18577.00 frames. ], tot_loss[loss=0.2562, simple_loss=0.3273, pruned_loss=0.09252, over 837068.18 frames. ], batch size: 49, lr: 2.35e-02, grad_scale: 8.0 2022-12-22 16:14:59,974 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14093.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 16:15:14,654 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14103.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 16:15:33,500 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.796e+02 5.002e+02 6.137e+02 7.910e+02 1.797e+03, threshold=1.227e+03, percent-clipped=1.0 2022-12-22 16:15:41,914 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14120.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 16:15:50,072 INFO [train.py:894] (3/4) Epoch 5, batch 100, loss[loss=0.2401, simple_loss=0.3299, pruned_loss=0.07511, over 18616.00 frames. ], tot_loss[loss=0.2516, simple_loss=0.3236, pruned_loss=0.08978, over 1476066.46 frames. ], batch size: 56, lr: 2.34e-02, grad_scale: 8.0 2022-12-22 16:16:01,148 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14133.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 16:16:30,444 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.2061, 1.4775, 1.7142, 0.4208, 0.9330, 2.1128, 1.7426, 1.5032], device='cuda:3'), covar=tensor([0.0757, 0.0357, 0.0393, 0.0520, 0.0473, 0.0241, 0.0259, 0.0559], device='cuda:3'), in_proj_covar=tensor([0.0108, 0.0118, 0.0082, 0.0114, 0.0109, 0.0081, 0.0113, 0.0096], device='cuda:3'), out_proj_covar=tensor([1.1324e-04, 1.2537e-04, 8.7732e-05, 1.1848e-04, 1.1273e-04, 8.6258e-05, 1.2164e-04, 1.0293e-04], device='cuda:3') 2022-12-22 16:16:46,704 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14164.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 16:17:01,006 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6017, 1.9843, 1.8951, 1.1599, 1.9663, 2.0823, 1.3649, 2.4062], device='cuda:3'), covar=tensor([0.1456, 0.1308, 0.1769, 0.2385, 0.1054, 0.1389, 0.2457, 0.0660], device='cuda:3'), in_proj_covar=tensor([0.0190, 0.0169, 0.0198, 0.0189, 0.0181, 0.0200, 0.0194, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 16:17:04,865 INFO [train.py:894] (3/4) Epoch 5, batch 150, loss[loss=0.2388, simple_loss=0.3208, pruned_loss=0.07837, over 18458.00 frames. ], tot_loss[loss=0.2499, simple_loss=0.3229, pruned_loss=0.08848, over 1971971.33 frames. ], batch size: 50, lr: 2.34e-02, grad_scale: 8.0 2022-12-22 16:17:07,654 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=5.02 vs. limit=5.0 2022-12-22 16:17:13,364 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14181.0, num_to_drop=1, layers_to_drop={3} 2022-12-22 16:17:17,274 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0 from training. Duration: 24.525 2022-12-22 16:17:50,597 WARNING [train.py:1060] (3/4) Exclude cut with ID 3699-47246-0007-3408-0_sp0.9 from training. Duration: 20.26675 2022-12-22 16:18:04,151 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.653e+02 5.430e+02 6.354e+02 8.496e+02 1.598e+03, threshold=1.271e+03, percent-clipped=4.0 2022-12-22 16:18:04,256 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp0.9 from training. Duration: 27.25 2022-12-22 16:18:21,218 INFO [train.py:894] (3/4) Epoch 5, batch 200, loss[loss=0.2149, simple_loss=0.285, pruned_loss=0.07244, over 18600.00 frames. ], tot_loss[loss=0.2473, simple_loss=0.3202, pruned_loss=0.08718, over 2357983.76 frames. ], batch size: 45, lr: 2.34e-02, grad_scale: 8.0 2022-12-22 16:18:39,845 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5915, 1.7814, 1.2170, 1.9411, 1.5732, 3.4606, 1.4895, 1.5390], device='cuda:3'), covar=tensor([0.1114, 0.1768, 0.1487, 0.1087, 0.1663, 0.0241, 0.1473, 0.1712], device='cuda:3'), in_proj_covar=tensor([0.0085, 0.0092, 0.0088, 0.0087, 0.0107, 0.0078, 0.0094, 0.0087], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 16:19:18,252 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0 from training. Duration: 21.68 2022-12-22 16:19:29,212 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0 from training. Duration: 21.6300625 2022-12-22 16:19:37,226 INFO [train.py:894] (3/4) Epoch 5, batch 250, loss[loss=0.2079, simple_loss=0.2808, pruned_loss=0.06746, over 18510.00 frames. ], tot_loss[loss=0.2451, simple_loss=0.3182, pruned_loss=0.08601, over 2658801.23 frames. ], batch size: 44, lr: 2.33e-02, grad_scale: 16.0 2022-12-22 16:19:52,194 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0_sp0.9 from training. Duration: 24.033375 2022-12-22 16:20:01,069 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6119, 1.8006, 1.6341, 1.7079, 1.4789, 3.6382, 1.7926, 2.3059], device='cuda:3'), covar=tensor([0.3442, 0.1779, 0.1880, 0.2021, 0.1457, 0.0158, 0.1488, 0.0966], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0127, 0.0142, 0.0127, 0.0120, 0.0095, 0.0114, 0.0108], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], device='cuda:3') 2022-12-22 16:20:10,183 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9091, 1.7676, 1.0667, 1.8630, 2.1707, 1.7485, 2.7155, 1.9227], device='cuda:3'), covar=tensor([0.0783, 0.1475, 0.2288, 0.1710, 0.1379, 0.0754, 0.0645, 0.0984], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0171, 0.0207, 0.0253, 0.0208, 0.0165, 0.0168, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 16:20:36,226 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.207e+02 5.412e+02 6.626e+02 8.496e+02 2.185e+03, threshold=1.325e+03, percent-clipped=4.0 2022-12-22 16:20:51,205 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0 from training. Duration: 22.905 2022-12-22 16:20:52,590 INFO [train.py:894] (3/4) Epoch 5, batch 300, loss[loss=0.2362, simple_loss=0.3119, pruned_loss=0.08027, over 18404.00 frames. ], tot_loss[loss=0.2443, simple_loss=0.3175, pruned_loss=0.08553, over 2892792.58 frames. ], batch size: 48, lr: 2.33e-02, grad_scale: 16.0 2022-12-22 16:20:52,666 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp1.1 from training. Duration: 23.4318125 2022-12-22 16:22:06,651 INFO [train.py:894] (3/4) Epoch 5, batch 350, loss[loss=0.2595, simple_loss=0.3391, pruned_loss=0.08989, over 18689.00 frames. ], tot_loss[loss=0.2481, simple_loss=0.3206, pruned_loss=0.0878, over 3075816.71 frames. ], batch size: 60, lr: 2.33e-02, grad_scale: 16.0 2022-12-22 16:22:31,820 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14393.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 16:22:48,879 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp1.1 from training. Duration: 20.82275 2022-12-22 16:22:50,645 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp0.9 from training. Duration: 25.45 2022-12-22 16:23:04,201 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14413.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 16:23:06,651 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.028e+02 5.505e+02 6.625e+02 8.750e+02 2.038e+03, threshold=1.325e+03, percent-clipped=6.0 2022-12-22 16:23:22,872 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4509, 1.8470, 1.0687, 2.1524, 2.2706, 1.3662, 1.7496, 1.0710], device='cuda:3'), covar=tensor([0.1812, 0.1501, 0.1543, 0.0833, 0.1346, 0.1233, 0.1359, 0.1539], device='cuda:3'), in_proj_covar=tensor([0.0211, 0.0182, 0.0172, 0.0161, 0.0220, 0.0164, 0.0175, 0.0173], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 16:23:23,907 INFO [train.py:894] (3/4) Epoch 5, batch 400, loss[loss=0.2745, simple_loss=0.342, pruned_loss=0.1035, over 18655.00 frames. ], tot_loss[loss=0.2504, simple_loss=0.3222, pruned_loss=0.08932, over 3218011.68 frames. ], batch size: 69, lr: 2.32e-02, grad_scale: 16.0 2022-12-22 16:23:27,384 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14428.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 16:23:46,117 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=14441.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 16:23:53,045 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0 from training. Duration: 25.775 2022-12-22 16:24:13,363 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14459.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 16:24:15,873 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0_sp0.9 from training. Duration: 22.25 2022-12-22 16:24:36,133 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14474.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 16:24:38,462 INFO [train.py:894] (3/4) Epoch 5, batch 450, loss[loss=0.2991, simple_loss=0.3647, pruned_loss=0.1168, over 18608.00 frames. ], tot_loss[loss=0.2516, simple_loss=0.3233, pruned_loss=0.08991, over 3327974.31 frames. ], batch size: 69, lr: 2.32e-02, grad_scale: 16.0 2022-12-22 16:24:38,685 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14476.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 16:24:42,734 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0 from training. Duration: 26.205 2022-12-22 16:24:49,191 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1244, 1.8581, 1.2605, 1.1438, 2.7111, 2.2195, 1.7388, 1.2315], device='cuda:3'), covar=tensor([0.0523, 0.0498, 0.0974, 0.1047, 0.0150, 0.0428, 0.0767, 0.1232], device='cuda:3'), in_proj_covar=tensor([0.0118, 0.0110, 0.0132, 0.0120, 0.0076, 0.0118, 0.0138, 0.0146], device='cuda:3'), out_proj_covar=tensor([1.4862e-04, 1.3928e-04, 1.6211e-04, 1.4809e-04, 9.6625e-05, 1.4326e-04, 1.7150e-04, 1.8022e-04], device='cuda:3') 2022-12-22 16:24:53,463 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0038, 1.0292, 1.4906, 1.3325, 1.6007, 1.9700, 2.2020, 1.2671], device='cuda:3'), covar=tensor([0.0345, 0.0520, 0.0674, 0.0406, 0.0282, 0.0296, 0.0263, 0.0454], device='cuda:3'), in_proj_covar=tensor([0.0071, 0.0106, 0.0119, 0.0115, 0.0100, 0.0086, 0.0075, 0.0111], device='cuda:3'), out_proj_covar=tensor([7.3021e-05, 1.0463e-04, 1.2315e-04, 1.1372e-04, 1.0428e-04, 8.4151e-05, 7.6145e-05, 1.1084e-04], device='cuda:3') 2022-12-22 16:24:54,855 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.3614, 3.8287, 3.7511, 4.0663, 3.9246, 3.9237, 4.4872, 1.5227], device='cuda:3'), covar=tensor([0.0630, 0.0446, 0.0504, 0.0511, 0.1448, 0.1019, 0.0501, 0.3694], device='cuda:3'), in_proj_covar=tensor([0.0211, 0.0159, 0.0151, 0.0143, 0.0218, 0.0175, 0.0175, 0.0209], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-22 16:24:54,960 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14487.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 16:25:00,446 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp0.9 from training. Duration: 30.1555625 2022-12-22 16:25:06,102 WARNING [train.py:1060] (3/4) Exclude cut with ID 1265-135635-0050-6781-0_sp0.9 from training. Duration: 21.8333125 2022-12-22 16:25:16,958 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp1.1 from training. Duration: 20.6545625 2022-12-22 16:25:21,576 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5766, 1.3465, 1.1984, 0.9302, 1.9025, 1.6087, 1.3961, 1.0484], device='cuda:3'), covar=tensor([0.0458, 0.0523, 0.0720, 0.0843, 0.0244, 0.0413, 0.0638, 0.1252], device='cuda:3'), in_proj_covar=tensor([0.0115, 0.0108, 0.0130, 0.0118, 0.0075, 0.0116, 0.0135, 0.0145], device='cuda:3'), out_proj_covar=tensor([1.4589e-04, 1.3712e-04, 1.5980e-04, 1.4561e-04, 9.5411e-05, 1.4099e-04, 1.6719e-04, 1.7845e-04], device='cuda:3') 2022-12-22 16:25:35,597 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.434e+02 5.649e+02 6.496e+02 8.214e+02 2.350e+03, threshold=1.299e+03, percent-clipped=4.0 2022-12-22 16:25:41,618 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.2535, 0.9730, 1.0156, 1.0804, 1.6369, 0.5660, 1.1646, 1.3208], device='cuda:3'), covar=tensor([0.1579, 0.1990, 0.2118, 0.1629, 0.2026, 0.1807, 0.1499, 0.1568], device='cuda:3'), in_proj_covar=tensor([0.0092, 0.0104, 0.0130, 0.0102, 0.0112, 0.0094, 0.0097, 0.0098], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 16:25:51,859 INFO [train.py:894] (3/4) Epoch 5, batch 500, loss[loss=0.2355, simple_loss=0.3097, pruned_loss=0.08065, over 18539.00 frames. ], tot_loss[loss=0.2526, simple_loss=0.3239, pruned_loss=0.09065, over 3411760.34 frames. ], batch size: 47, lr: 2.31e-02, grad_scale: 16.0 2022-12-22 16:25:59,345 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0045-39920-0_sp0.9 from training. Duration: 20.52225 2022-12-22 16:26:01,051 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5490, 0.9374, 2.1190, 2.9096, 1.8771, 2.2276, 0.6328, 1.6164], device='cuda:3'), covar=tensor([0.1979, 0.2295, 0.1408, 0.0517, 0.1529, 0.1536, 0.3113, 0.1789], device='cuda:3'), in_proj_covar=tensor([0.0104, 0.0114, 0.0119, 0.0092, 0.0102, 0.0122, 0.0136, 0.0105], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 16:26:17,878 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp0.9 from training. Duration: 29.1166875 2022-12-22 16:26:24,272 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14548.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 16:26:59,407 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.3968, 5.4071, 4.9470, 2.7804, 5.2322, 3.9747, 1.1907, 4.0726], device='cuda:3'), covar=tensor([0.1727, 0.0484, 0.1216, 0.3553, 0.0572, 0.0999, 0.5988, 0.1516], device='cuda:3'), in_proj_covar=tensor([0.0126, 0.0100, 0.0149, 0.0121, 0.0106, 0.0099, 0.0148, 0.0112], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 16:27:06,686 INFO [train.py:894] (3/4) Epoch 5, batch 550, loss[loss=0.2447, simple_loss=0.3196, pruned_loss=0.08496, over 18601.00 frames. ], tot_loss[loss=0.2533, simple_loss=0.3245, pruned_loss=0.09106, over 3478591.73 frames. ], batch size: 51, lr: 2.31e-02, grad_scale: 16.0 2022-12-22 16:27:20,192 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133211-0007-59831-0_sp0.9 from training. Duration: 21.388875 2022-12-22 16:27:56,659 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0 from training. Duration: 22.72 2022-12-22 16:27:57,904 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0_sp0.9 from training. Duration: 22.7444375 2022-12-22 16:28:07,472 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.606e+02 5.759e+02 6.795e+02 8.562e+02 2.141e+03, threshold=1.359e+03, percent-clipped=4.0 2022-12-22 16:28:23,084 INFO [train.py:894] (3/4) Epoch 5, batch 600, loss[loss=0.2264, simple_loss=0.2986, pruned_loss=0.0771, over 18700.00 frames. ], tot_loss[loss=0.2535, simple_loss=0.3251, pruned_loss=0.09093, over 3532097.96 frames. ], batch size: 46, lr: 2.31e-02, grad_scale: 8.0 2022-12-22 16:28:44,074 WARNING [train.py:1060] (3/4) Exclude cut with ID 4133-6541-0027-40495-0_sp1.1 from training. Duration: 0.9681875 2022-12-22 16:28:46,933 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0_sp0.9 from training. Duration: 22.3166875 2022-12-22 16:28:51,367 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133212-0015-59917-0_sp0.9 from training. Duration: 21.8166875 2022-12-22 16:29:38,230 INFO [train.py:894] (3/4) Epoch 5, batch 650, loss[loss=0.2427, simple_loss=0.3139, pruned_loss=0.08578, over 18677.00 frames. ], tot_loss[loss=0.253, simple_loss=0.3251, pruned_loss=0.09041, over 3572800.20 frames. ], batch size: 48, lr: 2.30e-02, grad_scale: 8.0 2022-12-22 16:30:34,671 WARNING [train.py:1060] (3/4) Exclude cut with ID 4957-30119-0041-23990-0_sp0.9 from training. Duration: 20.22775 2022-12-22 16:30:38,994 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.231e+02 5.539e+02 6.191e+02 8.462e+02 2.551e+03, threshold=1.238e+03, percent-clipped=2.0 2022-12-22 16:30:54,116 INFO [train.py:894] (3/4) Epoch 5, batch 700, loss[loss=0.2191, simple_loss=0.2842, pruned_loss=0.07702, over 18602.00 frames. ], tot_loss[loss=0.2528, simple_loss=0.3248, pruned_loss=0.09043, over 3603754.90 frames. ], batch size: 45, lr: 2.30e-02, grad_scale: 8.0 2022-12-22 16:30:54,312 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.8469, 3.6441, 3.7017, 1.9038, 3.5999, 2.7372, 0.9373, 2.7521], device='cuda:3'), covar=tensor([0.1830, 0.0923, 0.1231, 0.3710, 0.0904, 0.1176, 0.5609, 0.1803], device='cuda:3'), in_proj_covar=tensor([0.0121, 0.0096, 0.0143, 0.0115, 0.0103, 0.0096, 0.0141, 0.0107], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 16:30:57,194 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14728.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 16:31:15,051 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14740.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 16:31:19,712 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp1.1 from training. Duration: 24.67275 2022-12-22 16:31:31,785 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3364, 2.1391, 1.7056, 1.1013, 2.8266, 2.4852, 1.7610, 1.3557], device='cuda:3'), covar=tensor([0.0346, 0.0331, 0.0635, 0.0846, 0.0111, 0.0276, 0.0577, 0.1005], device='cuda:3'), in_proj_covar=tensor([0.0114, 0.0107, 0.0131, 0.0119, 0.0074, 0.0112, 0.0131, 0.0143], device='cuda:3'), out_proj_covar=tensor([1.4421e-04, 1.3528e-04, 1.6069e-04, 1.4675e-04, 9.4614e-05, 1.3664e-04, 1.6189e-04, 1.7583e-04], device='cuda:3') 2022-12-22 16:31:44,847 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14759.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 16:31:46,371 WARNING [train.py:1060] (3/4) Exclude cut with ID 3082-165428-0081-50734-0_sp0.9 from training. Duration: 21.8055625 2022-12-22 16:32:00,571 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14769.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 16:32:10,898 INFO [train.py:894] (3/4) Epoch 5, batch 750, loss[loss=0.2332, simple_loss=0.3082, pruned_loss=0.0791, over 18538.00 frames. ], tot_loss[loss=0.2512, simple_loss=0.3236, pruned_loss=0.08947, over 3628363.13 frames. ], batch size: 47, lr: 2.30e-02, grad_scale: 8.0 2022-12-22 16:32:11,100 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=14776.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 16:32:11,323 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14776.0, num_to_drop=1, layers_to_drop={2} 2022-12-22 16:32:25,227 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0_sp0.9 from training. Duration: 22.6666875 2022-12-22 16:32:27,424 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 2022-12-22 16:32:49,048 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14801.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 16:32:53,878 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.79 vs. limit=5.0 2022-12-22 16:32:56,373 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2208, 2.4043, 1.2360, 3.0271, 2.9088, 2.4474, 3.9889, 2.3076], device='cuda:3'), covar=tensor([0.0875, 0.1468, 0.2405, 0.1501, 0.1301, 0.0789, 0.0519, 0.0972], device='cuda:3'), in_proj_covar=tensor([0.0160, 0.0168, 0.0207, 0.0251, 0.0203, 0.0163, 0.0169, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 16:32:57,988 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=14807.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 16:33:11,565 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.661e+02 5.575e+02 6.678e+02 8.427e+02 1.558e+03, threshold=1.336e+03, percent-clipped=4.0 2022-12-22 16:33:19,210 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.11 vs. limit=5.0 2022-12-22 16:33:22,945 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=14824.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 16:33:25,648 INFO [train.py:894] (3/4) Epoch 5, batch 800, loss[loss=0.2673, simple_loss=0.3401, pruned_loss=0.09731, over 18489.00 frames. ], tot_loss[loss=0.2526, simple_loss=0.3244, pruned_loss=0.09042, over 3646971.31 frames. ], batch size: 77, lr: 2.29e-02, grad_scale: 8.0 2022-12-22 16:33:27,127 WARNING [train.py:1060] (3/4) Exclude cut with ID 2411-132532-0017-82279-0_sp1.1 from training. Duration: 0.9681875 2022-12-22 16:33:50,317 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0 from training. Duration: 22.485 2022-12-22 16:33:51,916 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14843.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 16:34:28,536 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp1.1 from training. Duration: 23.82275 2022-12-22 16:34:41,450 INFO [train.py:894] (3/4) Epoch 5, batch 850, loss[loss=0.2322, simple_loss=0.3088, pruned_loss=0.07775, over 18436.00 frames. ], tot_loss[loss=0.252, simple_loss=0.3236, pruned_loss=0.09019, over 3660729.11 frames. ], batch size: 48, lr: 2.29e-02, grad_scale: 8.0 2022-12-22 16:34:41,506 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0 from training. Duration: 20.77 2022-12-22 16:34:48,797 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0_sp0.9 from training. Duration: 24.088875 2022-12-22 16:35:19,470 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp1.1 from training. Duration: 20.4409375 2022-12-22 16:35:44,373 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.622e+02 5.143e+02 7.004e+02 8.582e+02 1.972e+03, threshold=1.401e+03, percent-clipped=5.0 2022-12-22 16:35:58,977 INFO [train.py:894] (3/4) Epoch 5, batch 900, loss[loss=0.2286, simple_loss=0.2965, pruned_loss=0.0804, over 18431.00 frames. ], tot_loss[loss=0.2512, simple_loss=0.3233, pruned_loss=0.0896, over 3673286.57 frames. ], batch size: 42, lr: 2.29e-02, grad_scale: 8.0 2022-12-22 16:36:36,134 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0_sp0.9 from training. Duration: 22.511125 2022-12-22 16:36:36,158 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0 from training. Duration: 20.675 2022-12-22 16:36:49,607 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.55 vs. limit=5.0 2022-12-22 16:37:14,681 INFO [train.py:894] (3/4) Epoch 5, batch 950, loss[loss=0.2265, simple_loss=0.3067, pruned_loss=0.07314, over 18428.00 frames. ], tot_loss[loss=0.2514, simple_loss=0.3241, pruned_loss=0.0893, over 3681809.50 frames. ], batch size: 48, lr: 2.28e-02, grad_scale: 8.0 2022-12-22 16:37:56,676 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8064, 0.8429, 1.5899, 1.5411, 1.8614, 1.6501, 1.4762, 1.1552], device='cuda:3'), covar=tensor([0.0892, 0.1439, 0.1160, 0.1057, 0.0679, 0.0487, 0.1025, 0.0648], device='cuda:3'), in_proj_covar=tensor([0.0183, 0.0216, 0.0194, 0.0210, 0.0189, 0.0172, 0.0198, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 16:38:16,018 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.793e+02 5.959e+02 7.329e+02 8.646e+02 1.591e+03, threshold=1.466e+03, percent-clipped=3.0 2022-12-22 16:38:16,112 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp0.9 from training. Duration: 24.9833125 2022-12-22 16:38:30,753 INFO [train.py:894] (3/4) Epoch 5, batch 1000, loss[loss=0.2505, simple_loss=0.3328, pruned_loss=0.0841, over 18656.00 frames. ], tot_loss[loss=0.2499, simple_loss=0.3224, pruned_loss=0.08867, over 3688337.50 frames. ], batch size: 98, lr: 2.28e-02, grad_scale: 8.0 2022-12-22 16:38:49,222 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0 from training. Duration: 27.14 2022-12-22 16:39:04,493 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0 from training. Duration: 22.44 2022-12-22 16:39:36,630 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15069.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 16:39:46,858 INFO [train.py:894] (3/4) Epoch 5, batch 1050, loss[loss=0.2658, simple_loss=0.3411, pruned_loss=0.09525, over 18589.00 frames. ], tot_loss[loss=0.2491, simple_loss=0.3219, pruned_loss=0.08819, over 3693677.72 frames. ], batch size: 98, lr: 2.28e-02, grad_scale: 8.0 2022-12-22 16:40:18,245 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=15096.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 16:40:22,538 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0060-62364-0_sp0.9 from training. Duration: 21.361125 2022-12-22 16:40:30,682 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp1.1 from training. Duration: 27.0318125 2022-12-22 16:40:39,293 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp0.9 from training. Duration: 28.638875 2022-12-22 16:40:47,974 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.160e+02 5.503e+02 6.877e+02 7.962e+02 1.573e+03, threshold=1.375e+03, percent-clipped=2.0 2022-12-22 16:40:49,397 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=15117.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 16:40:55,267 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0 from training. Duration: 20.4 2022-12-22 16:41:03,203 INFO [train.py:894] (3/4) Epoch 5, batch 1100, loss[loss=0.2283, simple_loss=0.3034, pruned_loss=0.07656, over 18514.00 frames. ], tot_loss[loss=0.2493, simple_loss=0.3224, pruned_loss=0.08814, over 3697910.61 frames. ], batch size: 47, lr: 2.27e-02, grad_scale: 8.0 2022-12-22 16:41:27,820 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0 from training. Duration: 20.025 2022-12-22 16:41:29,195 WARNING [train.py:1060] (3/4) Exclude cut with ID 2364-131735-0112-64612-0_sp0.9 from training. Duration: 20.488875 2022-12-22 16:41:29,444 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15143.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 16:41:33,533 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0 from training. Duration: 29.735 2022-12-22 16:42:17,695 INFO [train.py:894] (3/4) Epoch 5, batch 1150, loss[loss=0.2485, simple_loss=0.3289, pruned_loss=0.08408, over 18448.00 frames. ], tot_loss[loss=0.2478, simple_loss=0.3206, pruned_loss=0.08747, over 3701930.69 frames. ], batch size: 64, lr: 2.27e-02, grad_scale: 8.0 2022-12-22 16:42:40,955 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=15191.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 16:42:47,355 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=15195.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 16:42:55,204 WARNING [train.py:1060] (3/4) Exclude cut with ID 7276-92427-0014-12983-0_sp0.9 from training. Duration: 21.3055625 2022-12-22 16:42:56,542 WARNING [train.py:1060] (3/4) Exclude cut with ID 1025-75365-0008-79168-0_sp0.9 from training. Duration: 22.0666875 2022-12-22 16:43:18,667 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.144e+02 5.609e+02 7.018e+02 8.970e+02 1.678e+03, threshold=1.404e+03, percent-clipped=4.0 2022-12-22 16:43:33,725 INFO [train.py:894] (3/4) Epoch 5, batch 1200, loss[loss=0.2198, simple_loss=0.3045, pruned_loss=0.06752, over 18584.00 frames. ], tot_loss[loss=0.2472, simple_loss=0.32, pruned_loss=0.0872, over 3705077.82 frames. ], batch size: 78, lr: 2.27e-02, grad_scale: 8.0 2022-12-22 16:44:14,192 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.8627, 3.6129, 3.7116, 1.4608, 3.5963, 2.6817, 0.6978, 2.5163], device='cuda:3'), covar=tensor([0.1786, 0.0810, 0.1414, 0.4059, 0.0873, 0.1214, 0.6072, 0.1969], device='cuda:3'), in_proj_covar=tensor([0.0126, 0.0100, 0.0149, 0.0121, 0.0106, 0.0099, 0.0146, 0.0111], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 16:44:20,488 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=15256.0, num_to_drop=1, layers_to_drop={3} 2022-12-22 16:44:31,472 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.71 vs. limit=5.0 2022-12-22 16:44:41,529 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0 from training. Duration: 20.26 2022-12-22 16:44:51,037 INFO [train.py:894] (3/4) Epoch 5, batch 1250, loss[loss=0.2438, simple_loss=0.3201, pruned_loss=0.08381, over 18389.00 frames. ], tot_loss[loss=0.2474, simple_loss=0.3202, pruned_loss=0.08729, over 3707563.83 frames. ], batch size: 51, lr: 2.26e-02, grad_scale: 8.0 2022-12-22 16:44:56,006 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0030-9324-0_sp0.9 from training. Duration: 21.3444375 2022-12-22 16:45:18,905 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5400, 1.0914, 1.1175, 1.0841, 1.7349, 0.9540, 1.4526, 1.7007], device='cuda:3'), covar=tensor([0.1536, 0.2007, 0.2166, 0.1741, 0.1897, 0.1640, 0.1402, 0.1589], device='cuda:3'), in_proj_covar=tensor([0.0091, 0.0104, 0.0127, 0.0101, 0.0112, 0.0095, 0.0097, 0.0100], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], device='cuda:3') 2022-12-22 16:45:52,316 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.243e+02 5.439e+02 6.876e+02 8.678e+02 2.167e+03, threshold=1.375e+03, percent-clipped=2.0 2022-12-22 16:45:53,869 WARNING [train.py:1060] (3/4) Exclude cut with ID 497-129325-0061-62254-0_sp1.1 from training. Duration: 0.97725 2022-12-22 16:46:07,156 INFO [train.py:894] (3/4) Epoch 5, batch 1300, loss[loss=0.2592, simple_loss=0.3393, pruned_loss=0.08957, over 18450.00 frames. ], tot_loss[loss=0.2469, simple_loss=0.3197, pruned_loss=0.08703, over 3708802.28 frames. ], batch size: 54, lr: 2.26e-02, grad_scale: 8.0 2022-12-22 16:46:35,740 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0_sp0.9 from training. Duration: 22.97225 2022-12-22 16:47:01,384 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3619, 1.3786, 0.8747, 1.7991, 1.3210, 3.0411, 1.1402, 1.2729], device='cuda:3'), covar=tensor([0.1239, 0.1886, 0.1628, 0.1089, 0.1787, 0.0360, 0.1605, 0.1830], device='cuda:3'), in_proj_covar=tensor([0.0083, 0.0088, 0.0088, 0.0084, 0.0102, 0.0078, 0.0092, 0.0084], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 16:47:06,782 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0047-39922-0_sp0.9 from training. Duration: 21.97775 2022-12-22 16:47:16,281 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3258, 1.6545, 2.0158, 2.1936, 2.3115, 2.0529, 2.1880, 1.3645], device='cuda:3'), covar=tensor([0.0893, 0.1527, 0.1114, 0.1163, 0.0645, 0.0454, 0.1267, 0.0627], device='cuda:3'), in_proj_covar=tensor([0.0187, 0.0220, 0.0198, 0.0214, 0.0194, 0.0176, 0.0205, 0.0170], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 16:47:21,157 WARNING [train.py:1060] (3/4) Exclude cut with ID 1112-1043-0006-89194-0_sp0.9 from training. Duration: 21.8333125 2022-12-22 16:47:24,007 INFO [train.py:894] (3/4) Epoch 5, batch 1350, loss[loss=0.2833, simple_loss=0.3488, pruned_loss=0.1089, over 18713.00 frames. ], tot_loss[loss=0.2461, simple_loss=0.3191, pruned_loss=0.08654, over 3709535.51 frames. ], batch size: 78, lr: 2.26e-02, grad_scale: 8.0 2022-12-22 16:47:33,426 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0 from training. Duration: 20.47 2022-12-22 16:47:33,765 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6952, 1.9807, 2.0178, 2.1331, 1.9565, 4.8750, 2.4841, 3.0944], device='cuda:3'), covar=tensor([0.3321, 0.1774, 0.1781, 0.1741, 0.1287, 0.0072, 0.1192, 0.0805], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0128, 0.0140, 0.0126, 0.0118, 0.0096, 0.0108, 0.0108], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 16:47:54,643 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15396.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 16:48:01,328 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2022-12-22 16:48:05,884 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2022-12-22 16:48:09,530 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.2027, 5.1675, 4.8639, 2.8840, 5.2181, 4.0928, 1.3843, 3.8231], device='cuda:3'), covar=tensor([0.1464, 0.0550, 0.1044, 0.2697, 0.0539, 0.0764, 0.4720, 0.1238], device='cuda:3'), in_proj_covar=tensor([0.0121, 0.0098, 0.0143, 0.0117, 0.0105, 0.0097, 0.0141, 0.0107], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 16:48:14,170 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0091, 2.0337, 2.4231, 1.3477, 2.4639, 2.4291, 1.5914, 2.7029], device='cuda:3'), covar=tensor([0.1035, 0.1276, 0.1082, 0.1740, 0.0806, 0.1050, 0.2014, 0.0563], device='cuda:3'), in_proj_covar=tensor([0.0189, 0.0174, 0.0199, 0.0186, 0.0181, 0.0205, 0.0194, 0.0170], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 16:48:22,771 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2022-12-22 16:48:24,645 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.202e+02 5.523e+02 6.743e+02 8.396e+02 3.038e+03, threshold=1.349e+03, percent-clipped=3.0 2022-12-22 16:48:38,853 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0037-39912-0_sp0.9 from training. Duration: 20.67225 2022-12-22 16:48:40,243 INFO [train.py:894] (3/4) Epoch 5, batch 1400, loss[loss=0.2629, simple_loss=0.3352, pruned_loss=0.0953, over 18454.00 frames. ], tot_loss[loss=0.2469, simple_loss=0.3196, pruned_loss=0.0871, over 3708646.09 frames. ], batch size: 64, lr: 2.25e-02, grad_scale: 8.0 2022-12-22 16:48:57,606 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp0.9 from training. Duration: 25.2444375 2022-12-22 16:49:07,188 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=15444.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 16:49:21,217 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0021-76797-0_sp0.9 from training. Duration: 21.1445 2022-12-22 16:49:51,276 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5135, 1.5241, 1.1182, 1.9377, 1.6014, 3.2718, 1.3178, 1.6358], device='cuda:3'), covar=tensor([0.1203, 0.1882, 0.1541, 0.1064, 0.1491, 0.0294, 0.1463, 0.1699], device='cuda:3'), in_proj_covar=tensor([0.0083, 0.0090, 0.0087, 0.0085, 0.0103, 0.0077, 0.0092, 0.0085], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 16:49:55,326 INFO [train.py:894] (3/4) Epoch 5, batch 1450, loss[loss=0.2594, simple_loss=0.3365, pruned_loss=0.09112, over 18686.00 frames. ], tot_loss[loss=0.2476, simple_loss=0.3204, pruned_loss=0.08735, over 3710411.97 frames. ], batch size: 62, lr: 2.25e-02, grad_scale: 8.0 2022-12-22 16:50:37,715 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp0.9 from training. Duration: 33.038875 2022-12-22 16:50:46,295 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2022-12-22 16:50:56,238 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.500e+02 5.658e+02 7.097e+02 9.527e+02 2.232e+03, threshold=1.419e+03, percent-clipped=9.0 2022-12-22 16:51:07,792 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.9041, 1.0536, 0.7859, 1.2316, 1.3917, 0.9449, 0.9097, 0.8251], device='cuda:3'), covar=tensor([0.1235, 0.1273, 0.1015, 0.0638, 0.0889, 0.0859, 0.1064, 0.0983], device='cuda:3'), in_proj_covar=tensor([0.0212, 0.0186, 0.0175, 0.0165, 0.0224, 0.0168, 0.0181, 0.0171], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 16:51:11,645 INFO [train.py:894] (3/4) Epoch 5, batch 1500, loss[loss=0.2171, simple_loss=0.2988, pruned_loss=0.06767, over 18528.00 frames. ], tot_loss[loss=0.2487, simple_loss=0.3211, pruned_loss=0.08822, over 3710941.27 frames. ], batch size: 58, lr: 2.25e-02, grad_scale: 8.0 2022-12-22 16:51:16,207 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64291-0000-16059-0_sp0.9 from training. Duration: 20.0944375 2022-12-22 16:51:31,155 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp1.1 from training. Duration: 20.4 2022-12-22 16:51:38,534 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0 from training. Duration: 20.085 2022-12-22 16:51:45,682 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2022-12-22 16:51:49,087 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0_sp0.9 from training. Duration: 23.07775 2022-12-22 16:51:49,253 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=15551.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 16:52:07,813 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4724, 2.1862, 1.5310, 2.5510, 1.9086, 1.9324, 2.0778, 2.6909], device='cuda:3'), covar=tensor([0.1224, 0.1855, 0.1175, 0.1910, 0.1965, 0.0702, 0.1789, 0.0431], device='cuda:3'), in_proj_covar=tensor([0.0237, 0.0220, 0.0199, 0.0299, 0.0211, 0.0186, 0.0224, 0.0159], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 16:52:11,518 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 2022-12-22 16:52:27,538 INFO [train.py:894] (3/4) Epoch 5, batch 1550, loss[loss=0.2616, simple_loss=0.3368, pruned_loss=0.09317, over 18646.00 frames. ], tot_loss[loss=0.2485, simple_loss=0.3213, pruned_loss=0.08786, over 3712636.48 frames. ], batch size: 62, lr: 2.24e-02, grad_scale: 8.0 2022-12-22 16:52:28,202 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2022-12-22 16:52:39,066 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp0.9 from training. Duration: 24.9333125 2022-12-22 16:53:25,439 WARNING [train.py:1060] (3/4) Exclude cut with ID 5118-111612-0016-124680-0_sp0.9 from training. Duration: 20.388875 2022-12-22 16:53:28,047 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.175e+02 5.012e+02 6.250e+02 7.299e+02 1.571e+03, threshold=1.250e+03, percent-clipped=1.0 2022-12-22 16:53:30,928 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp1.1 from training. Duration: 20.3590625 2022-12-22 16:53:43,438 INFO [train.py:894] (3/4) Epoch 5, batch 1600, loss[loss=0.282, simple_loss=0.3456, pruned_loss=0.1092, over 18543.00 frames. ], tot_loss[loss=0.2469, simple_loss=0.3198, pruned_loss=0.08701, over 3713030.00 frames. ], batch size: 97, lr: 2.24e-02, grad_scale: 8.0 2022-12-22 16:53:51,547 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=2.06 vs. limit=2.0 2022-12-22 16:54:34,697 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4406, 2.2551, 1.4912, 2.5291, 1.8490, 1.8743, 2.0005, 2.9200], device='cuda:3'), covar=tensor([0.1269, 0.1604, 0.1177, 0.1823, 0.2078, 0.0703, 0.1700, 0.0354], device='cuda:3'), in_proj_covar=tensor([0.0234, 0.0217, 0.0198, 0.0299, 0.0210, 0.0185, 0.0221, 0.0158], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 16:54:40,087 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0_sp1.1 from training. Duration: 0.836375 2022-12-22 16:54:59,794 INFO [train.py:894] (3/4) Epoch 5, batch 1650, loss[loss=0.2397, simple_loss=0.3006, pruned_loss=0.08937, over 18418.00 frames. ], tot_loss[loss=0.2498, simple_loss=0.3213, pruned_loss=0.08912, over 3713440.73 frames. ], batch size: 42, lr: 2.24e-02, grad_scale: 8.0 2022-12-22 16:55:23,215 WARNING [train.py:1060] (3/4) Exclude cut with ID 8565-290391-0049-67394-0_sp0.9 from training. Duration: 21.3166875 2022-12-22 16:55:54,768 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0029-104863-0_sp0.9 from training. Duration: 22.1055625 2022-12-22 16:56:01,469 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.196e+02 6.902e+02 8.534e+02 1.056e+03 2.303e+03, threshold=1.707e+03, percent-clipped=12.0 2022-12-22 16:56:06,128 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp1.1 from training. Duration: 21.77725 2022-12-22 16:56:16,971 INFO [train.py:894] (3/4) Epoch 5, batch 1700, loss[loss=0.2411, simple_loss=0.317, pruned_loss=0.0826, over 18566.00 frames. ], tot_loss[loss=0.2546, simple_loss=0.3238, pruned_loss=0.09267, over 3713731.23 frames. ], batch size: 56, lr: 2.23e-02, grad_scale: 8.0 2022-12-22 16:56:26,145 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp0.9 from training. Duration: 27.8166875 2022-12-22 16:56:49,740 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp1.1 from training. Duration: 22.5090625 2022-12-22 16:56:57,478 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0 from training. Duration: 25.035 2022-12-22 16:57:17,469 WARNING [train.py:1060] (3/4) Exclude cut with ID 774-127930-0014-10412-0_sp1.1 from training. Duration: 0.95 2022-12-22 16:57:32,043 INFO [train.py:894] (3/4) Epoch 5, batch 1750, loss[loss=0.2659, simple_loss=0.3294, pruned_loss=0.1012, over 18587.00 frames. ], tot_loss[loss=0.26, simple_loss=0.3268, pruned_loss=0.09654, over 3713161.98 frames. ], batch size: 51, lr: 2.23e-02, grad_scale: 8.0 2022-12-22 16:57:35,609 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp0.9 from training. Duration: 0.92225 2022-12-22 16:58:04,892 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0 from training. Duration: 21.97 2022-12-22 16:58:22,933 WARNING [train.py:1060] (3/4) Exclude cut with ID 7492-105653-0055-62765-0_sp0.9 from training. Duration: 21.97225 2022-12-22 16:58:22,976 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp0.9 from training. Duration: 25.3333125 2022-12-22 16:58:32,746 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.623e+02 6.430e+02 8.235e+02 1.006e+03 2.020e+03, threshold=1.647e+03, percent-clipped=2.0 2022-12-22 16:58:36,830 WARNING [train.py:1060] (3/4) Exclude cut with ID 5172-29468-0015-19128-0_sp0.9 from training. Duration: 21.5055625 2022-12-22 16:58:37,726 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5427, 0.8955, 0.6184, 1.2712, 1.5797, 1.2462, 1.3282, 1.9070], device='cuda:3'), covar=tensor([0.1823, 0.2460, 0.2924, 0.1758, 0.2299, 0.1528, 0.1642, 0.1525], device='cuda:3'), in_proj_covar=tensor([0.0093, 0.0105, 0.0130, 0.0102, 0.0112, 0.0095, 0.0097, 0.0100], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], device='cuda:3') 2022-12-22 16:58:46,325 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp1.1 from training. Duration: 20.72725 2022-12-22 16:58:47,850 INFO [train.py:894] (3/4) Epoch 5, batch 1800, loss[loss=0.2584, simple_loss=0.3262, pruned_loss=0.09528, over 18504.00 frames. ], tot_loss[loss=0.264, simple_loss=0.3294, pruned_loss=0.09924, over 3712601.62 frames. ], batch size: 52, lr: 2.23e-02, grad_scale: 8.0 2022-12-22 16:59:17,090 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp0.9 from training. Duration: 26.32775 2022-12-22 16:59:26,855 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15851.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 16:59:52,614 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0 from training. Duration: 20.025 2022-12-22 16:59:57,198 WARNING [train.py:1060] (3/4) Exclude cut with ID 6709-74022-0004-86860-0_sp1.1 from training. Duration: 0.9409375 2022-12-22 16:59:57,209 WARNING [train.py:1060] (3/4) Exclude cut with ID 4757-1811-0023-62229-0_sp0.9 from training. Duration: 21.37775 2022-12-22 17:00:05,172 INFO [train.py:894] (3/4) Epoch 5, batch 1850, loss[loss=0.2581, simple_loss=0.3163, pruned_loss=0.09997, over 18390.00 frames. ], tot_loss[loss=0.2678, simple_loss=0.331, pruned_loss=0.1023, over 3712056.01 frames. ], batch size: 46, lr: 2.22e-02, grad_scale: 8.0 2022-12-22 17:00:15,739 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=15883.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:00:19,745 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0004-25974-0_sp0.9 from training. Duration: 21.17225 2022-12-22 17:00:19,759 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp0.9 from training. Duration: 27.511125 2022-12-22 17:00:35,202 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2022-12-22 17:00:39,110 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=15899.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:00:49,815 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0 from training. Duration: 22.8 2022-12-22 17:00:54,690 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0 from training. Duration: 22.585 2022-12-22 17:01:04,855 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.750e+02 6.712e+02 8.382e+02 1.076e+03 1.765e+03, threshold=1.676e+03, percent-clipped=1.0 2022-12-22 17:01:19,849 INFO [train.py:894] (3/4) Epoch 5, batch 1900, loss[loss=0.3048, simple_loss=0.3634, pruned_loss=0.1231, over 18384.00 frames. ], tot_loss[loss=0.2716, simple_loss=0.333, pruned_loss=0.1051, over 3713102.30 frames. ], batch size: 53, lr: 2.22e-02, grad_scale: 8.0 2022-12-22 17:01:26,182 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0001-146967-0_sp0.9 from training. Duration: 22.0166875 2022-12-22 17:01:43,590 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp1.1 from training. Duration: 24.395375 2022-12-22 17:01:47,463 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=15944.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:01:49,925 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp0.9 from training. Duration: 27.47775 2022-12-22 17:01:54,193 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp0.9 from training. Duration: 24.8833125 2022-12-22 17:01:57,098 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0 from training. Duration: 23.39 2022-12-22 17:02:01,788 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp0.9 from training. Duration: 28.72225 2022-12-22 17:02:12,786 WARNING [train.py:1060] (3/4) Exclude cut with ID 585-294811-0110-133686-0_sp0.9 from training. Duration: 20.8944375 2022-12-22 17:02:27,115 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0_sp0.9 from training. Duration: 23.8444375 2022-12-22 17:02:37,330 INFO [train.py:894] (3/4) Epoch 5, batch 1950, loss[loss=0.3318, simple_loss=0.3826, pruned_loss=0.1405, over 18667.00 frames. ], tot_loss[loss=0.2724, simple_loss=0.333, pruned_loss=0.1059, over 3713011.80 frames. ], batch size: 60, lr: 2.22e-02, grad_scale: 8.0 2022-12-22 17:02:52,280 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0 from training. Duration: 25.85 2022-12-22 17:02:52,290 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0 from training. Duration: 21.39 2022-12-22 17:03:03,360 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0 from training. Duration: 27.92 2022-12-22 17:03:03,593 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.8192, 4.3064, 4.3757, 4.5692, 4.4141, 4.3852, 4.8552, 2.3683], device='cuda:3'), covar=tensor([0.0515, 0.0385, 0.0410, 0.0552, 0.1062, 0.0740, 0.0343, 0.3165], device='cuda:3'), in_proj_covar=tensor([0.0226, 0.0172, 0.0161, 0.0153, 0.0235, 0.0188, 0.0185, 0.0216], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 17:03:18,673 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16001.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:03:34,531 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0039-130165-0_sp0.9 from training. Duration: 20.661125 2022-12-22 17:03:42,603 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.362e+02 7.148e+02 8.465e+02 9.808e+02 1.967e+03, threshold=1.693e+03, percent-clipped=1.0 2022-12-22 17:03:57,082 INFO [train.py:894] (3/4) Epoch 5, batch 2000, loss[loss=0.3182, simple_loss=0.3694, pruned_loss=0.1335, over 18465.00 frames. ], tot_loss[loss=0.2753, simple_loss=0.3345, pruned_loss=0.108, over 3714172.36 frames. ], batch size: 64, lr: 2.21e-02, grad_scale: 8.0 2022-12-22 17:03:58,598 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0043-15874-0_sp0.9 from training. Duration: 20.07225 2022-12-22 17:04:06,021 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0 from training. Duration: 21.01 2022-12-22 17:04:51,962 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16062.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:05:12,894 INFO [train.py:894] (3/4) Epoch 5, batch 2050, loss[loss=0.3215, simple_loss=0.3688, pruned_loss=0.1371, over 18572.00 frames. ], tot_loss[loss=0.2762, simple_loss=0.3347, pruned_loss=0.1088, over 3715055.04 frames. ], batch size: 78, lr: 2.21e-02, grad_scale: 8.0 2022-12-22 17:05:14,581 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0 from training. Duration: 20.65 2022-12-22 17:05:22,478 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0 from training. Duration: 21.46 2022-12-22 17:05:27,186 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16085.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:05:46,773 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.9666, 2.5942, 3.0243, 0.3146, 2.3512, 3.5687, 2.1349, 2.7600], device='cuda:3'), covar=tensor([0.0854, 0.0285, 0.0332, 0.0592, 0.0371, 0.0137, 0.0354, 0.0435], device='cuda:3'), in_proj_covar=tensor([0.0114, 0.0123, 0.0083, 0.0114, 0.0111, 0.0088, 0.0116, 0.0104], device='cuda:3'), out_proj_covar=tensor([1.1584e-04, 1.2716e-04, 8.5701e-05, 1.1523e-04, 1.1109e-04, 8.9849e-05, 1.2046e-04, 1.0725e-04], device='cuda:3') 2022-12-22 17:06:08,822 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0 from training. Duration: 0.92 2022-12-22 17:06:09,038 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.9273, 5.1811, 4.8719, 2.6446, 4.8671, 3.7701, 0.9849, 3.0060], device='cuda:3'), covar=tensor([0.1606, 0.0617, 0.1212, 0.3115, 0.0755, 0.0958, 0.5568, 0.1871], device='cuda:3'), in_proj_covar=tensor([0.0122, 0.0105, 0.0147, 0.0119, 0.0109, 0.0100, 0.0146, 0.0108], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 17:06:13,318 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0_sp0.9 from training. Duration: 23.7666875 2022-12-22 17:06:14,712 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.991e+02 7.417e+02 9.427e+02 1.233e+03 2.937e+03, threshold=1.885e+03, percent-clipped=9.0 2022-12-22 17:06:30,464 INFO [train.py:894] (3/4) Epoch 5, batch 2100, loss[loss=0.2507, simple_loss=0.3105, pruned_loss=0.09549, over 18536.00 frames. ], tot_loss[loss=0.2759, simple_loss=0.3342, pruned_loss=0.1088, over 3715230.42 frames. ], batch size: 47, lr: 2.21e-02, grad_scale: 8.0 2022-12-22 17:06:49,590 WARNING [train.py:1060] (3/4) Exclude cut with ID 8544-281189-0060-101339-0_sp0.9 from training. Duration: 20.861125 2022-12-22 17:07:00,136 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0_sp0.9 from training. Duration: 22.711125 2022-12-22 17:07:02,548 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16146.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:07:06,940 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16149.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:07:39,786 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp1.1 from training. Duration: 22.986375 2022-12-22 17:07:47,532 INFO [train.py:894] (3/4) Epoch 5, batch 2150, loss[loss=0.2476, simple_loss=0.3057, pruned_loss=0.09476, over 18526.00 frames. ], tot_loss[loss=0.2744, simple_loss=0.3326, pruned_loss=0.1081, over 3715686.89 frames. ], batch size: 47, lr: 2.20e-02, grad_scale: 8.0 2022-12-22 17:07:57,606 WARNING [train.py:1060] (3/4) Exclude cut with ID 8040-260924-0003-80960-0_sp0.9 from training. Duration: 22.07225 2022-12-22 17:08:03,013 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0045-26330-0_sp0.9 from training. Duration: 20.3055625 2022-12-22 17:08:04,484 WARNING [train.py:1060] (3/4) Exclude cut with ID 6356-271890-0060-94317-0_sp0.9 from training. Duration: 20.72225 2022-12-22 17:08:10,043 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2022-12-22 17:08:22,165 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0572, 1.2825, 1.8379, 1.8935, 2.0036, 1.7201, 1.8349, 1.2422], device='cuda:3'), covar=tensor([0.0861, 0.1545, 0.1051, 0.1049, 0.0640, 0.0480, 0.1150, 0.0639], device='cuda:3'), in_proj_covar=tensor([0.0193, 0.0224, 0.0204, 0.0216, 0.0196, 0.0182, 0.0210, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 17:08:23,093 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp1.1 from training. Duration: 22.4818125 2022-12-22 17:08:40,603 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16210.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:08:49,802 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.943e+02 6.646e+02 8.265e+02 9.985e+02 1.734e+03, threshold=1.653e+03, percent-clipped=0.0 2022-12-22 17:08:49,867 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp0.9 from training. Duration: 25.0944375 2022-12-22 17:08:54,277 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0 from training. Duration: 21.515 2022-12-22 17:09:00,067 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp0.9 from training. Duration: 27.02225 2022-12-22 17:09:04,589 INFO [train.py:894] (3/4) Epoch 5, batch 2200, loss[loss=0.2836, simple_loss=0.3452, pruned_loss=0.111, over 18585.00 frames. ], tot_loss[loss=0.274, simple_loss=0.3326, pruned_loss=0.1077, over 3715084.13 frames. ], batch size: 51, lr: 2.20e-02, grad_scale: 8.0 2022-12-22 17:09:04,621 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0010-62480-0_sp0.9 from training. Duration: 22.22225 2022-12-22 17:09:12,162 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0085-44554-0_sp0.9 from training. Duration: 20.85 2022-12-22 17:09:25,751 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16239.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:09:40,031 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3730, 2.5754, 1.6018, 1.0843, 3.1569, 2.7443, 2.2246, 1.5517], device='cuda:3'), covar=tensor([0.0453, 0.0373, 0.0759, 0.0916, 0.0085, 0.0335, 0.0577, 0.1027], device='cuda:3'), in_proj_covar=tensor([0.0121, 0.0111, 0.0134, 0.0124, 0.0076, 0.0112, 0.0139, 0.0151], device='cuda:3'), out_proj_covar=tensor([1.5164e-04, 1.4074e-04, 1.6485e-04, 1.5411e-04, 9.6326e-05, 1.3654e-04, 1.7024e-04, 1.8546e-04], device='cuda:3') 2022-12-22 17:09:46,986 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0 from training. Duration: 21.54 2022-12-22 17:09:51,390 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp1.1 from training. Duration: 20.5318125 2022-12-22 17:10:02,333 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0012-134311-0_sp0.9 from training. Duration: 21.9333125 2022-12-22 17:10:22,608 INFO [train.py:894] (3/4) Epoch 5, batch 2250, loss[loss=0.2845, simple_loss=0.3432, pruned_loss=0.1129, over 18577.00 frames. ], tot_loss[loss=0.2734, simple_loss=0.3319, pruned_loss=0.1075, over 3714601.07 frames. ], batch size: 99, lr: 2.20e-02, grad_scale: 8.0 2022-12-22 17:10:51,028 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0025-130151-0_sp0.9 from training. Duration: 21.7944375 2022-12-22 17:11:01,648 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0_sp0.9 from training. Duration: 22.4666875 2022-12-22 17:11:09,672 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0 from training. Duration: 21.635 2022-12-22 17:11:14,072 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0_sp0.9 from training. Duration: 24.038875 2022-12-22 17:11:22,130 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9581, 1.3180, 2.0374, 2.9965, 2.0251, 2.4798, 0.8666, 2.0117], device='cuda:3'), covar=tensor([0.1742, 0.2023, 0.1530, 0.0604, 0.1388, 0.1113, 0.2670, 0.1350], device='cuda:3'), in_proj_covar=tensor([0.0106, 0.0117, 0.0122, 0.0097, 0.0103, 0.0124, 0.0136, 0.0106], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 17:11:24,858 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.321e+02 6.545e+02 8.299e+02 1.005e+03 2.324e+03, threshold=1.660e+03, percent-clipped=4.0 2022-12-22 17:11:40,730 INFO [train.py:894] (3/4) Epoch 5, batch 2300, loss[loss=0.3205, simple_loss=0.3577, pruned_loss=0.1416, over 18548.00 frames. ], tot_loss[loss=0.2729, simple_loss=0.3313, pruned_loss=0.1073, over 3714931.20 frames. ], batch size: 184, lr: 2.20e-02, grad_scale: 8.0 2022-12-22 17:12:00,674 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp1.1 from training. Duration: 21.786375 2022-12-22 17:12:13,681 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0 from training. Duration: 20.22 2022-12-22 17:12:27,045 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16357.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:12:47,879 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16371.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:12:55,726 INFO [train.py:894] (3/4) Epoch 5, batch 2350, loss[loss=0.291, simple_loss=0.3452, pruned_loss=0.1184, over 18731.00 frames. ], tot_loss[loss=0.2738, simple_loss=0.3321, pruned_loss=0.1077, over 3715068.49 frames. ], batch size: 54, lr: 2.19e-02, grad_scale: 8.0 2022-12-22 17:13:01,590 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.34 vs. limit=5.0 2022-12-22 17:13:40,024 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2022-12-22 17:13:56,205 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.206e+02 6.539e+02 8.779e+02 1.071e+03 2.079e+03, threshold=1.756e+03, percent-clipped=3.0 2022-12-22 17:14:11,145 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.23 vs. limit=5.0 2022-12-22 17:14:11,765 INFO [train.py:894] (3/4) Epoch 5, batch 2400, loss[loss=0.2558, simple_loss=0.3184, pruned_loss=0.09661, over 18429.00 frames. ], tot_loss[loss=0.2731, simple_loss=0.3318, pruned_loss=0.1073, over 3714753.80 frames. ], batch size: 48, lr: 2.19e-02, grad_scale: 8.0 2022-12-22 17:14:13,332 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0 from training. Duration: 25.285 2022-12-22 17:14:21,701 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16432.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:14:34,638 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16441.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:15:17,014 WARNING [train.py:1060] (3/4) Exclude cut with ID 811-130148-0001-63453-0_sp0.9 from training. Duration: 20.861125 2022-12-22 17:15:27,562 INFO [train.py:894] (3/4) Epoch 5, batch 2450, loss[loss=0.2688, simple_loss=0.3286, pruned_loss=0.1045, over 18449.00 frames. ], tot_loss[loss=0.2721, simple_loss=0.3312, pruned_loss=0.1065, over 3713873.77 frames. ], batch size: 50, lr: 2.19e-02, grad_scale: 8.0 2022-12-22 17:15:39,023 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0 from training. Duration: 20.88 2022-12-22 17:16:10,320 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0_sp0.9 from training. Duration: 23.4166875 2022-12-22 17:16:12,275 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16505.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:16:28,830 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3091, 0.9161, 1.2670, 2.1262, 1.3444, 1.9980, 0.8019, 1.4568], device='cuda:3'), covar=tensor([0.1762, 0.1979, 0.1336, 0.0710, 0.1492, 0.1061, 0.2116, 0.1420], device='cuda:3'), in_proj_covar=tensor([0.0106, 0.0117, 0.0124, 0.0098, 0.0104, 0.0125, 0.0136, 0.0105], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 17:16:29,859 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.784e+02 7.165e+02 8.582e+02 1.131e+03 1.915e+03, threshold=1.716e+03, percent-clipped=1.0 2022-12-22 17:16:43,514 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5068, 1.4559, 1.2596, 0.9455, 1.8862, 1.6041, 1.3743, 1.1327], device='cuda:3'), covar=tensor([0.0372, 0.0466, 0.0637, 0.0701, 0.0204, 0.0350, 0.0572, 0.1123], device='cuda:3'), in_proj_covar=tensor([0.0117, 0.0112, 0.0132, 0.0120, 0.0072, 0.0112, 0.0136, 0.0149], device='cuda:3'), out_proj_covar=tensor([1.4654e-04, 1.4131e-04, 1.6233e-04, 1.4749e-04, 9.1408e-05, 1.3586e-04, 1.6778e-04, 1.8395e-04], device='cuda:3') 2022-12-22 17:16:45,983 INFO [train.py:894] (3/4) Epoch 5, batch 2500, loss[loss=0.2918, simple_loss=0.3349, pruned_loss=0.1244, over 18549.00 frames. ], tot_loss[loss=0.2726, simple_loss=0.3317, pruned_loss=0.1068, over 3714102.49 frames. ], batch size: 44, lr: 2.18e-02, grad_scale: 8.0 2022-12-22 17:17:06,140 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16539.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:17:27,429 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0 from training. Duration: 21.24 2022-12-22 17:17:27,446 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0_sp0.9 from training. Duration: 23.9055625 2022-12-22 17:17:59,922 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp0.9 from training. Duration: 25.988875 2022-12-22 17:18:02,740 INFO [train.py:894] (3/4) Epoch 5, batch 2550, loss[loss=0.2425, simple_loss=0.2899, pruned_loss=0.09755, over 18558.00 frames. ], tot_loss[loss=0.2719, simple_loss=0.3313, pruned_loss=0.1063, over 3713916.82 frames. ], batch size: 44, lr: 2.18e-02, grad_scale: 8.0 2022-12-22 17:18:08,944 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0001-134300-0_sp0.9 from training. Duration: 20.67225 2022-12-22 17:18:19,862 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=16587.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:18:56,661 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.86 vs. limit=2.0 2022-12-22 17:18:56,960 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0 from training. Duration: 20.34 2022-12-22 17:19:06,750 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.751e+02 6.596e+02 8.364e+02 1.015e+03 1.761e+03, threshold=1.673e+03, percent-clipped=1.0 2022-12-22 17:19:17,369 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.3419, 3.7543, 3.6766, 4.1783, 3.8113, 3.8544, 4.4731, 1.3544], device='cuda:3'), covar=tensor([0.0666, 0.0627, 0.0547, 0.0582, 0.1582, 0.1124, 0.0585, 0.4508], device='cuda:3'), in_proj_covar=tensor([0.0233, 0.0178, 0.0168, 0.0160, 0.0242, 0.0193, 0.0193, 0.0222], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 17:19:20,145 INFO [train.py:894] (3/4) Epoch 5, batch 2600, loss[loss=0.2966, simple_loss=0.3575, pruned_loss=0.1178, over 18575.00 frames. ], tot_loss[loss=0.2723, simple_loss=0.3315, pruned_loss=0.1065, over 3713889.66 frames. ], batch size: 56, lr: 2.18e-02, grad_scale: 8.0 2022-12-22 17:20:07,340 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16657.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:20:14,467 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp0.9 from training. Duration: 25.061125 2022-12-22 17:20:26,543 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0 from training. Duration: 0.83 2022-12-22 17:20:35,644 INFO [train.py:894] (3/4) Epoch 5, batch 2650, loss[loss=0.2932, simple_loss=0.3536, pruned_loss=0.1164, over 18489.00 frames. ], tot_loss[loss=0.2718, simple_loss=0.3311, pruned_loss=0.1063, over 3713948.18 frames. ], batch size: 64, lr: 2.17e-02, grad_scale: 8.0 2022-12-22 17:20:52,522 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0 from training. Duration: 24.73 2022-12-22 17:21:06,264 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0 from training. Duration: 23.965 2022-12-22 17:21:13,709 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0030-146996-0_sp0.9 from training. Duration: 22.088875 2022-12-22 17:21:14,118 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5426, 1.7412, 1.2876, 2.0296, 2.6619, 1.3515, 2.0366, 1.1240], device='cuda:3'), covar=tensor([0.2342, 0.2248, 0.2006, 0.1252, 0.1399, 0.1624, 0.1547, 0.2138], device='cuda:3'), in_proj_covar=tensor([0.0224, 0.0192, 0.0183, 0.0171, 0.0233, 0.0175, 0.0190, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 17:21:19,926 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=16705.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:21:31,206 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0_sp0.9 from training. Duration: 23.6 2022-12-22 17:21:37,440 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5775, 1.4300, 1.1404, 1.8838, 1.4322, 3.3378, 1.1929, 1.4783], device='cuda:3'), covar=tensor([0.0980, 0.1841, 0.1469, 0.0953, 0.1628, 0.0284, 0.1579, 0.1658], device='cuda:3'), in_proj_covar=tensor([0.0081, 0.0090, 0.0086, 0.0083, 0.0103, 0.0076, 0.0094, 0.0084], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 17:21:38,520 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.340e+02 6.537e+02 7.713e+02 9.830e+02 1.584e+03, threshold=1.543e+03, percent-clipped=0.0 2022-12-22 17:21:45,138 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6143, 2.1875, 1.4236, 2.6999, 1.9716, 1.8250, 2.1152, 2.8951], device='cuda:3'), covar=tensor([0.1239, 0.1969, 0.1299, 0.1841, 0.2027, 0.0752, 0.1812, 0.0425], device='cuda:3'), in_proj_covar=tensor([0.0248, 0.0229, 0.0204, 0.0313, 0.0218, 0.0193, 0.0232, 0.0165], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 17:21:51,812 INFO [train.py:894] (3/4) Epoch 5, batch 2700, loss[loss=0.2849, simple_loss=0.3299, pruned_loss=0.1199, over 18450.00 frames. ], tot_loss[loss=0.2721, simple_loss=0.3311, pruned_loss=0.1065, over 3713950.05 frames. ], batch size: 42, lr: 2.17e-02, grad_scale: 8.0 2022-12-22 17:21:53,442 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16727.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:22:15,922 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16741.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:23:06,367 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3563, 0.9076, 1.1942, 1.1313, 1.4015, 1.3755, 1.4301, 1.0041], device='cuda:3'), covar=tensor([0.0353, 0.0301, 0.0467, 0.0291, 0.0326, 0.0331, 0.0237, 0.0362], device='cuda:3'), in_proj_covar=tensor([0.0077, 0.0108, 0.0127, 0.0123, 0.0102, 0.0090, 0.0079, 0.0115], device='cuda:3'), out_proj_covar=tensor([7.7153e-05, 1.0299e-04, 1.2672e-04, 1.1844e-04, 1.0305e-04, 8.5613e-05, 7.7640e-05, 1.1219e-04], device='cuda:3') 2022-12-22 17:23:09,471 INFO [train.py:894] (3/4) Epoch 5, batch 2750, loss[loss=0.272, simple_loss=0.3396, pruned_loss=0.1022, over 18515.00 frames. ], tot_loss[loss=0.2723, simple_loss=0.3315, pruned_loss=0.1065, over 3714510.68 frames. ], batch size: 58, lr: 2.17e-02, grad_scale: 8.0 2022-12-22 17:23:12,387 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0 from training. Duration: 23.795 2022-12-22 17:23:29,376 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp1.1 from training. Duration: 21.5409375 2022-12-22 17:23:29,533 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=16789.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:23:32,241 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp0.9 from training. Duration: 24.97775 2022-12-22 17:23:38,359 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.72 vs. limit=5.0 2022-12-22 17:23:41,510 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0_sp0.9 from training. Duration: 23.3444375 2022-12-22 17:23:54,695 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16805.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:24:11,943 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.184e+02 6.794e+02 8.297e+02 1.071e+03 3.389e+03, threshold=1.659e+03, percent-clipped=6.0 2022-12-22 17:24:11,995 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0_sp0.9 from training. Duration: 23.2 2022-12-22 17:24:16,734 WARNING [train.py:1060] (3/4) Exclude cut with ID 5653-46179-0060-117930-0_sp0.9 from training. Duration: 21.17225 2022-12-22 17:24:25,954 INFO [train.py:894] (3/4) Epoch 5, batch 2800, loss[loss=0.2461, simple_loss=0.31, pruned_loss=0.09113, over 18676.00 frames. ], tot_loss[loss=0.2726, simple_loss=0.3314, pruned_loss=0.1069, over 3713512.43 frames. ], batch size: 48, lr: 2.17e-02, grad_scale: 8.0 2022-12-22 17:24:26,412 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.0430, 0.5795, 0.7925, 0.7815, 1.1603, 1.0269, 1.0284, 0.8931], device='cuda:3'), covar=tensor([0.0353, 0.0369, 0.0605, 0.0396, 0.0268, 0.0348, 0.0265, 0.0400], device='cuda:3'), in_proj_covar=tensor([0.0078, 0.0110, 0.0127, 0.0125, 0.0102, 0.0091, 0.0079, 0.0118], device='cuda:3'), out_proj_covar=tensor([7.7983e-05, 1.0486e-04, 1.2717e-04, 1.2000e-04, 1.0324e-04, 8.6712e-05, 7.6941e-05, 1.1446e-04], device='cuda:3') 2022-12-22 17:24:38,450 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp0.9 from training. Duration: 24.6555625 2022-12-22 17:24:47,880 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5272, 1.9802, 1.3820, 2.2988, 2.5195, 1.4387, 1.6973, 1.1593], device='cuda:3'), covar=tensor([0.1818, 0.1522, 0.1426, 0.0781, 0.1233, 0.1208, 0.1476, 0.1517], device='cuda:3'), in_proj_covar=tensor([0.0224, 0.0191, 0.0183, 0.0173, 0.0235, 0.0175, 0.0192, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 17:25:07,680 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=16853.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:25:30,133 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16868.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 17:25:35,486 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0 from training. Duration: 20.44 2022-12-22 17:25:41,410 INFO [train.py:894] (3/4) Epoch 5, batch 2850, loss[loss=0.2768, simple_loss=0.3422, pruned_loss=0.1057, over 18727.00 frames. ], tot_loss[loss=0.272, simple_loss=0.3311, pruned_loss=0.1064, over 3712878.27 frames. ], batch size: 54, lr: 2.16e-02, grad_scale: 8.0 2022-12-22 17:25:51,539 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0_sp0.9 from training. Duration: 23.45 2022-12-22 17:25:54,539 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7611, 1.8934, 1.9687, 1.8942, 2.0725, 4.7526, 2.4478, 3.1989], device='cuda:3'), covar=tensor([0.4414, 0.2681, 0.2236, 0.2428, 0.1335, 0.0172, 0.1476, 0.0872], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0124, 0.0137, 0.0125, 0.0115, 0.0099, 0.0108, 0.0104], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 17:26:05,280 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6965, 1.9713, 2.3931, 1.3856, 2.3118, 2.2975, 1.4740, 2.4723], device='cuda:3'), covar=tensor([0.1265, 0.1336, 0.1216, 0.1788, 0.0827, 0.1060, 0.2136, 0.0598], device='cuda:3'), in_proj_covar=tensor([0.0198, 0.0178, 0.0199, 0.0188, 0.0181, 0.0201, 0.0198, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 17:26:22,402 WARNING [train.py:1060] (3/4) Exclude cut with ID 6945-60535-0076-12784-0_sp0.9 from training. Duration: 20.52225 2022-12-22 17:26:27,174 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16906.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:26:28,186 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0 from training. Duration: 22.19 2022-12-22 17:26:38,833 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp1.1 from training. Duration: 25.3818125 2022-12-22 17:26:43,354 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.260e+02 6.499e+02 9.019e+02 1.112e+03 1.634e+03, threshold=1.804e+03, percent-clipped=0.0 2022-12-22 17:26:56,460 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp0.9 from training. Duration: 28.0944375 2022-12-22 17:26:58,678 INFO [train.py:894] (3/4) Epoch 5, batch 2900, loss[loss=0.2189, simple_loss=0.2846, pruned_loss=0.0766, over 18552.00 frames. ], tot_loss[loss=0.2709, simple_loss=0.3299, pruned_loss=0.1059, over 3713153.78 frames. ], batch size: 41, lr: 2.16e-02, grad_scale: 8.0 2022-12-22 17:27:01,526 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0_sp0.9 from training. Duration: 22.9444375 2022-12-22 17:27:03,727 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16929.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 17:27:08,554 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6661, 1.7239, 1.3386, 1.1184, 2.1913, 3.1311, 2.7194, 1.7572], device='cuda:3'), covar=tensor([0.0471, 0.0426, 0.0547, 0.0464, 0.0222, 0.0188, 0.0281, 0.0441], device='cuda:3'), in_proj_covar=tensor([0.0078, 0.0108, 0.0127, 0.0121, 0.0102, 0.0090, 0.0079, 0.0116], device='cuda:3'), out_proj_covar=tensor([7.8043e-05, 1.0216e-04, 1.2655e-04, 1.1638e-04, 1.0229e-04, 8.5010e-05, 7.7198e-05, 1.1194e-04], device='cuda:3') 2022-12-22 17:27:10,730 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp1.1 from training. Duration: 21.6318125 2022-12-22 17:27:27,469 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0 from training. Duration: 23.695 2022-12-22 17:27:34,203 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2022-12-22 17:27:53,665 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0 from training. Duration: 23.955 2022-12-22 17:27:57,204 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.4670, 2.9584, 2.4441, 1.3960, 2.4575, 2.4705, 2.0975, 2.2795], device='cuda:3'), covar=tensor([0.0604, 0.0493, 0.1409, 0.1572, 0.1435, 0.1131, 0.1315, 0.0964], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0170, 0.0200, 0.0194, 0.0191, 0.0177, 0.0191, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 17:27:58,715 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6539, 1.4795, 2.1073, 1.1923, 1.9967, 1.7637, 1.2217, 2.2421], device='cuda:3'), covar=tensor([0.1172, 0.1774, 0.1130, 0.1823, 0.0905, 0.1317, 0.2584, 0.0585], device='cuda:3'), in_proj_covar=tensor([0.0198, 0.0180, 0.0199, 0.0191, 0.0182, 0.0204, 0.0200, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 17:28:01,650 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16967.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 17:28:14,203 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.2111, 0.9891, 0.8709, 0.3265, 0.8465, 1.0496, 0.8365, 1.0738], device='cuda:3'), covar=tensor([0.0419, 0.0341, 0.0710, 0.0985, 0.0706, 0.0987, 0.1066, 0.0433], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0171, 0.0203, 0.0196, 0.0194, 0.0179, 0.0192, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 17:28:15,332 INFO [train.py:894] (3/4) Epoch 5, batch 2950, loss[loss=0.3399, simple_loss=0.3743, pruned_loss=0.1527, over 18666.00 frames. ], tot_loss[loss=0.2722, simple_loss=0.3311, pruned_loss=0.1067, over 3713890.21 frames. ], batch size: 182, lr: 2.16e-02, grad_scale: 8.0 2022-12-22 17:28:24,993 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp0.9 from training. Duration: 26.438875 2022-12-22 17:28:32,262 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.12 vs. limit=5.0 2022-12-22 17:29:08,773 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0021-26306-0_sp0.9 from training. Duration: 21.2444375 2022-12-22 17:29:08,806 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp0.9 from training. Duration: 31.02225 2022-12-22 17:29:10,775 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9674, 0.8942, 1.7829, 1.6330, 1.9537, 1.7708, 1.7036, 1.2328], device='cuda:3'), covar=tensor([0.0889, 0.1517, 0.1101, 0.1055, 0.0734, 0.0496, 0.1049, 0.0674], device='cuda:3'), in_proj_covar=tensor([0.0205, 0.0236, 0.0213, 0.0226, 0.0207, 0.0190, 0.0221, 0.0185], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 17:29:18,057 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.179e+02 6.420e+02 8.025e+02 1.086e+03 2.995e+03, threshold=1.605e+03, percent-clipped=4.0 2022-12-22 17:29:20,934 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0 from training. Duration: 22.395 2022-12-22 17:29:30,965 INFO [train.py:894] (3/4) Epoch 5, batch 3000, loss[loss=0.3355, simple_loss=0.3729, pruned_loss=0.149, over 18562.00 frames. ], tot_loss[loss=0.2721, simple_loss=0.3312, pruned_loss=0.1065, over 3713946.06 frames. ], batch size: 173, lr: 2.15e-02, grad_scale: 8.0 2022-12-22 17:29:30,965 INFO [train.py:919] (3/4) Computing validation loss 2022-12-22 17:29:42,386 INFO [train.py:928] (3/4) Epoch 5, validation: loss=0.2011, simple_loss=0.2985, pruned_loss=0.05191, over 944034.00 frames. 2022-12-22 17:29:42,387 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24676MB 2022-12-22 17:29:44,475 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17027.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:29:48,710 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0 from training. Duration: 21.075 2022-12-22 17:29:51,924 WARNING [train.py:1060] (3/4) Exclude cut with ID 6482-98857-0025-147532-0_sp0.9 from training. Duration: 20.0055625 2022-12-22 17:29:53,202 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0037-132304-0_sp0.9 from training. Duration: 22.05 2022-12-22 17:29:53,216 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0 from training. Duration: 26.8349375 2022-12-22 17:29:54,174 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.79 vs. limit=5.0 2022-12-22 17:29:56,359 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp1.1 from training. Duration: 22.1090625 2022-12-22 17:29:56,751 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4254, 1.1928, 1.0908, 2.0501, 1.4582, 3.2149, 1.1379, 1.3024], device='cuda:3'), covar=tensor([0.1068, 0.1973, 0.1438, 0.0898, 0.1673, 0.0286, 0.1529, 0.1672], device='cuda:3'), in_proj_covar=tensor([0.0081, 0.0090, 0.0085, 0.0083, 0.0102, 0.0076, 0.0093, 0.0086], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 17:30:03,120 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp0.9 from training. Duration: 26.6166875 2022-12-22 17:30:12,276 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6979, 2.4258, 1.7469, 3.2580, 2.7411, 1.7858, 2.2295, 1.3303], device='cuda:3'), covar=tensor([0.1910, 0.1624, 0.1458, 0.0724, 0.1796, 0.1183, 0.1626, 0.1544], device='cuda:3'), in_proj_covar=tensor([0.0224, 0.0192, 0.0187, 0.0176, 0.0240, 0.0179, 0.0195, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 17:30:21,264 WARNING [train.py:1060] (3/4) Exclude cut with ID 2046-178027-0000-53705-0_sp0.9 from training. Duration: 20.3055625 2022-12-22 17:30:43,979 WARNING [train.py:1060] (3/4) Exclude cut with ID 7205-50138-0008-5373-0_sp0.9 from training. Duration: 20.7 2022-12-22 17:30:57,713 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=17075.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:30:59,057 INFO [train.py:894] (3/4) Epoch 5, batch 3050, loss[loss=0.3019, simple_loss=0.3536, pruned_loss=0.1251, over 18556.00 frames. ], tot_loss[loss=0.2708, simple_loss=0.3302, pruned_loss=0.1057, over 3713971.58 frames. ], batch size: 69, lr: 2.15e-02, grad_scale: 8.0 2022-12-22 17:31:28,560 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0 from training. Duration: 22.48 2022-12-22 17:31:46,133 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp0.9 from training. Duration: 29.816625 2022-12-22 17:32:01,345 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.119e+02 6.435e+02 7.611e+02 9.477e+02 1.701e+03, threshold=1.522e+03, percent-clipped=2.0 2022-12-22 17:32:05,945 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp1.1 from training. Duration: 22.7590625 2022-12-22 17:32:10,300 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0 from training. Duration: 22.555 2022-12-22 17:32:14,810 INFO [train.py:894] (3/4) Epoch 5, batch 3100, loss[loss=0.2646, simple_loss=0.3296, pruned_loss=0.09984, over 18656.00 frames. ], tot_loss[loss=0.2708, simple_loss=0.3301, pruned_loss=0.1058, over 3713799.80 frames. ], batch size: 98, lr: 2.15e-02, grad_scale: 8.0 2022-12-22 17:32:31,968 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0005-25975-0_sp0.9 from training. Duration: 21.688875 2022-12-22 17:32:51,166 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5945, 2.3420, 1.4582, 2.9186, 3.0341, 1.5212, 2.1679, 1.3024], device='cuda:3'), covar=tensor([0.2036, 0.1599, 0.1623, 0.0760, 0.1582, 0.1324, 0.1576, 0.1640], device='cuda:3'), in_proj_covar=tensor([0.0224, 0.0192, 0.0185, 0.0175, 0.0240, 0.0178, 0.0192, 0.0176], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 17:33:04,145 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0_sp0.9 from training. Duration: 22.6 2022-12-22 17:33:30,780 INFO [train.py:894] (3/4) Epoch 5, batch 3150, loss[loss=0.311, simple_loss=0.3592, pruned_loss=0.1314, over 18543.00 frames. ], tot_loss[loss=0.2721, simple_loss=0.3315, pruned_loss=0.1064, over 3713367.37 frames. ], batch size: 69, lr: 2.15e-02, grad_scale: 8.0 2022-12-22 17:33:43,538 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0 from training. Duration: 24.32 2022-12-22 17:33:46,486 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8278, 2.1311, 0.9309, 2.1441, 1.9562, 1.5339, 3.0052, 1.8587], device='cuda:3'), covar=tensor([0.0981, 0.1263, 0.2655, 0.1891, 0.1786, 0.1018, 0.0767, 0.1161], device='cuda:3'), in_proj_covar=tensor([0.0165, 0.0174, 0.0214, 0.0262, 0.0213, 0.0172, 0.0182, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 17:33:50,961 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17189.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:34:33,625 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.890e+02 6.624e+02 8.139e+02 1.082e+03 2.327e+03, threshold=1.628e+03, percent-clipped=7.0 2022-12-22 17:34:37,353 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.8823, 2.4646, 2.6583, 0.4286, 2.2933, 3.2619, 2.0220, 2.6429], device='cuda:3'), covar=tensor([0.0711, 0.0347, 0.0551, 0.0478, 0.0467, 0.0161, 0.0422, 0.0462], device='cuda:3'), in_proj_covar=tensor([0.0120, 0.0131, 0.0093, 0.0117, 0.0116, 0.0093, 0.0127, 0.0114], device='cuda:3'), out_proj_covar=tensor([1.1899e-04, 1.3181e-04, 9.3957e-05, 1.1618e-04, 1.1286e-04, 9.3515e-05, 1.3006e-04, 1.1517e-04], device='cuda:3') 2022-12-22 17:34:39,938 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-276745-0093-13116-0_sp0.9 from training. Duration: 21.061125 2022-12-22 17:34:45,182 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17224.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 17:34:48,219 INFO [train.py:894] (3/4) Epoch 5, batch 3200, loss[loss=0.3181, simple_loss=0.3594, pruned_loss=0.1384, over 18621.00 frames. ], tot_loss[loss=0.2714, simple_loss=0.3306, pruned_loss=0.1061, over 3713344.21 frames. ], batch size: 179, lr: 2.14e-02, grad_scale: 8.0 2022-12-22 17:34:55,596 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0024-15855-0_sp0.9 from training. Duration: 20.32225 2022-12-22 17:35:07,914 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp1.1 from training. Duration: 0.7545625 2022-12-22 17:35:23,434 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0_sp0.9 from training. Duration: 23.9333125 2022-12-22 17:35:25,299 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17250.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:35:43,720 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17262.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 17:35:56,084 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp1.1 from training. Duration: 20.17275 2022-12-22 17:36:01,904 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp1.1 from training. Duration: 20.436375 2022-12-22 17:36:04,738 INFO [train.py:894] (3/4) Epoch 5, batch 3250, loss[loss=0.218, simple_loss=0.2815, pruned_loss=0.07728, over 18521.00 frames. ], tot_loss[loss=0.2716, simple_loss=0.3309, pruned_loss=0.1061, over 3713432.28 frames. ], batch size: 41, lr: 2.14e-02, grad_scale: 8.0 2022-12-22 17:36:13,239 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5061, 1.3090, 1.1901, 0.7796, 1.8423, 1.4981, 1.2082, 1.0059], device='cuda:3'), covar=tensor([0.0383, 0.0483, 0.0608, 0.0790, 0.0229, 0.0353, 0.0619, 0.1141], device='cuda:3'), in_proj_covar=tensor([0.0120, 0.0113, 0.0130, 0.0122, 0.0074, 0.0112, 0.0137, 0.0150], device='cuda:3'), out_proj_covar=tensor([1.4964e-04, 1.4209e-04, 1.5954e-04, 1.5014e-04, 9.5220e-05, 1.3599e-04, 1.6883e-04, 1.8530e-04], device='cuda:3') 2022-12-22 17:36:39,003 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2022-12-22 17:36:52,772 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8201, 1.0382, 0.5851, 1.3262, 1.9102, 1.2819, 1.4258, 1.7936], device='cuda:3'), covar=tensor([0.1568, 0.2274, 0.2856, 0.1614, 0.1902, 0.1546, 0.1517, 0.1637], device='cuda:3'), in_proj_covar=tensor([0.0096, 0.0110, 0.0131, 0.0103, 0.0116, 0.0097, 0.0101, 0.0103], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 17:37:07,496 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.347e+02 7.021e+02 8.751e+02 1.059e+03 1.719e+03, threshold=1.750e+03, percent-clipped=2.0 2022-12-22 17:37:20,743 INFO [train.py:894] (3/4) Epoch 5, batch 3300, loss[loss=0.2943, simple_loss=0.3506, pruned_loss=0.119, over 18601.00 frames. ], tot_loss[loss=0.2703, simple_loss=0.33, pruned_loss=0.1053, over 3713453.62 frames. ], batch size: 56, lr: 2.14e-02, grad_scale: 8.0 2022-12-22 17:37:27,047 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0_sp0.9 from training. Duration: 23.1055625 2022-12-22 17:37:28,378 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp1.1 from training. Duration: 23.5 2022-12-22 17:37:40,466 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp0.9 from training. Duration: 26.62775 2022-12-22 17:37:50,709 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0 from training. Duration: 21.105 2022-12-22 17:37:55,775 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0_sp0.9 from training. Duration: 24.411125 2022-12-22 17:38:15,827 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17362.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:38:23,006 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp1.1 from training. Duration: 21.263625 2022-12-22 17:38:37,318 INFO [train.py:894] (3/4) Epoch 5, batch 3350, loss[loss=0.2963, simple_loss=0.3478, pruned_loss=0.1225, over 18634.00 frames. ], tot_loss[loss=0.271, simple_loss=0.3305, pruned_loss=0.1058, over 3713757.33 frames. ], batch size: 180, lr: 2.13e-02, grad_scale: 8.0 2022-12-22 17:38:48,341 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17383.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 17:38:53,815 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0 from training. Duration: 20.795 2022-12-22 17:39:00,711 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5282, 1.5669, 1.1244, 1.7132, 1.6877, 1.4799, 2.1228, 1.7103], device='cuda:3'), covar=tensor([0.0912, 0.1294, 0.2215, 0.1371, 0.1571, 0.0828, 0.0854, 0.0955], device='cuda:3'), in_proj_covar=tensor([0.0168, 0.0177, 0.0218, 0.0266, 0.0217, 0.0174, 0.0186, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 17:39:05,995 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0 from training. Duration: 24.76 2022-12-22 17:39:06,019 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0_sp0.9 from training. Duration: 22.25 2022-12-22 17:39:32,743 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp1.1 from training. Duration: 20.5045625 2022-12-22 17:39:39,975 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.521e+02 6.812e+02 8.775e+02 1.065e+03 2.100e+03, threshold=1.755e+03, percent-clipped=2.0 2022-12-22 17:39:50,647 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17423.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:39:54,353 INFO [train.py:894] (3/4) Epoch 5, batch 3400, loss[loss=0.2668, simple_loss=0.3333, pruned_loss=0.1002, over 18470.00 frames. ], tot_loss[loss=0.2718, simple_loss=0.3312, pruned_loss=0.1062, over 3713951.35 frames. ], batch size: 50, lr: 2.13e-02, grad_scale: 8.0 2022-12-22 17:40:05,088 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2022-12-22 17:40:21,507 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17444.0, num_to_drop=1, layers_to_drop={2} 2022-12-22 17:41:06,283 INFO [train.py:894] (3/4) Epoch 5, batch 3450, loss[loss=0.2458, simple_loss=0.3064, pruned_loss=0.09262, over 18430.00 frames. ], tot_loss[loss=0.2709, simple_loss=0.3308, pruned_loss=0.1055, over 3714081.55 frames. ], batch size: 48, lr: 2.13e-02, grad_scale: 8.0 2022-12-22 17:41:34,165 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 2022-12-22 17:41:42,126 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3698, 1.7612, 1.2268, 2.1227, 2.4785, 1.3074, 1.6742, 1.0658], device='cuda:3'), covar=tensor([0.2038, 0.1691, 0.1515, 0.0880, 0.1146, 0.1217, 0.1432, 0.1522], device='cuda:3'), in_proj_covar=tensor([0.0229, 0.0196, 0.0187, 0.0175, 0.0240, 0.0180, 0.0192, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 17:42:04,841 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.179e+02 6.499e+02 8.083e+02 1.067e+03 3.667e+03, threshold=1.617e+03, percent-clipped=5.0 2022-12-22 17:42:16,636 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17524.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 17:42:19,511 INFO [train.py:894] (3/4) Epoch 5, batch 3500, loss[loss=0.4052, simple_loss=0.4143, pruned_loss=0.1981, over 18682.00 frames. ], tot_loss[loss=0.2722, simple_loss=0.3314, pruned_loss=0.1065, over 3714301.95 frames. ], batch size: 179, lr: 2.13e-02, grad_scale: 8.0 2022-12-22 17:42:20,106 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7939, 1.9407, 1.7238, 2.3141, 2.6925, 1.6562, 2.0319, 1.5661], device='cuda:3'), covar=tensor([0.1678, 0.1522, 0.1272, 0.0779, 0.1127, 0.1195, 0.1320, 0.1340], device='cuda:3'), in_proj_covar=tensor([0.0228, 0.0195, 0.0187, 0.0175, 0.0239, 0.0180, 0.0190, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 17:42:41,343 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp1.1 from training. Duration: 22.2954375 2022-12-22 17:42:51,005 INFO [train.py:894] (3/4) Epoch 6, batch 0, loss[loss=0.2653, simple_loss=0.3421, pruned_loss=0.09427, over 18462.00 frames. ], tot_loss[loss=0.2653, simple_loss=0.3421, pruned_loss=0.09427, over 18462.00 frames. ], batch size: 54, lr: 1.98e-02, grad_scale: 8.0 2022-12-22 17:42:51,006 INFO [train.py:919] (3/4) Computing validation loss 2022-12-22 17:43:02,448 INFO [train.py:928] (3/4) Epoch 6, validation: loss=0.2017, simple_loss=0.2993, pruned_loss=0.05207, over 944034.00 frames. 2022-12-22 17:43:02,449 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24676MB 2022-12-22 17:43:22,702 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17545.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:43:40,118 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3705, 0.8625, 0.8988, 1.0110, 1.5764, 0.6666, 1.0826, 1.3768], device='cuda:3'), covar=tensor([0.1900, 0.2600, 0.2420, 0.1922, 0.2300, 0.1896, 0.1814, 0.1774], device='cuda:3'), in_proj_covar=tensor([0.0096, 0.0109, 0.0132, 0.0103, 0.0116, 0.0097, 0.0100, 0.0103], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 17:43:49,454 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17562.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 17:43:52,180 WARNING [train.py:1060] (3/4) Exclude cut with ID 298-126791-0067-24026-0_sp0.9 from training. Duration: 21.438875 2022-12-22 17:43:58,012 WARNING [train.py:1060] (3/4) Exclude cut with ID 5652-39938-0025-23684-0_sp0.9 from training. Duration: 22.2055625 2022-12-22 17:44:04,560 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=17572.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 17:44:18,707 INFO [train.py:894] (3/4) Epoch 6, batch 50, loss[loss=0.2608, simple_loss=0.3356, pruned_loss=0.09296, over 18516.00 frames. ], tot_loss[loss=0.2479, simple_loss=0.3215, pruned_loss=0.08718, over 837679.29 frames. ], batch size: 64, lr: 1.98e-02, grad_scale: 8.0 2022-12-22 17:45:01,192 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=17610.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:45:11,045 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.839e+02 4.470e+02 5.623e+02 8.118e+02 1.969e+03, threshold=1.125e+03, percent-clipped=4.0 2022-12-22 17:45:34,097 INFO [train.py:894] (3/4) Epoch 6, batch 100, loss[loss=0.1866, simple_loss=0.2753, pruned_loss=0.04893, over 18700.00 frames. ], tot_loss[loss=0.2379, simple_loss=0.3129, pruned_loss=0.08146, over 1475900.70 frames. ], batch size: 50, lr: 1.98e-02, grad_scale: 8.0 2022-12-22 17:45:51,183 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.0586, 0.8190, 1.0717, 0.0577, 0.6951, 1.2801, 1.1227, 1.1476], device='cuda:3'), covar=tensor([0.0644, 0.0324, 0.0284, 0.0470, 0.0427, 0.0332, 0.0289, 0.0446], device='cuda:3'), in_proj_covar=tensor([0.0118, 0.0132, 0.0091, 0.0117, 0.0115, 0.0093, 0.0123, 0.0112], device='cuda:3'), out_proj_covar=tensor([1.1643e-04, 1.3169e-04, 9.0921e-05, 1.1488e-04, 1.1176e-04, 9.2235e-05, 1.2465e-04, 1.1244e-04], device='cuda:3') 2022-12-22 17:45:55,609 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9172, 2.0830, 1.0866, 2.1847, 2.2436, 1.7947, 3.1191, 2.1171], device='cuda:3'), covar=tensor([0.0824, 0.1461, 0.2506, 0.1945, 0.1458, 0.0854, 0.0728, 0.0985], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0175, 0.0216, 0.0264, 0.0213, 0.0174, 0.0181, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 17:46:51,281 INFO [train.py:894] (3/4) Epoch 6, batch 150, loss[loss=0.194, simple_loss=0.2689, pruned_loss=0.05953, over 18461.00 frames. ], tot_loss[loss=0.2363, simple_loss=0.3114, pruned_loss=0.08065, over 1970185.66 frames. ], batch size: 43, lr: 1.98e-02, grad_scale: 8.0 2022-12-22 17:47:00,278 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0 from training. Duration: 24.525 2022-12-22 17:47:17,567 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17700.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:47:34,225 WARNING [train.py:1060] (3/4) Exclude cut with ID 3699-47246-0007-3408-0_sp0.9 from training. Duration: 20.26675 2022-12-22 17:47:43,065 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.077e+02 5.092e+02 6.350e+02 7.853e+02 1.179e+03, threshold=1.270e+03, percent-clipped=2.0 2022-12-22 17:47:44,733 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17718.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:47:47,807 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp0.9 from training. Duration: 27.25 2022-12-22 17:48:03,258 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17730.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:48:06,101 INFO [train.py:894] (3/4) Epoch 6, batch 200, loss[loss=0.1905, simple_loss=0.2632, pruned_loss=0.05889, over 18401.00 frames. ], tot_loss[loss=0.2331, simple_loss=0.3086, pruned_loss=0.07886, over 2357082.62 frames. ], batch size: 42, lr: 1.97e-02, grad_scale: 8.0 2022-12-22 17:48:17,182 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17739.0, num_to_drop=1, layers_to_drop={3} 2022-12-22 17:48:49,635 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17761.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:48:59,696 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0 from training. Duration: 21.68 2022-12-22 17:49:10,504 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0 from training. Duration: 21.6300625 2022-12-22 17:49:21,683 INFO [train.py:894] (3/4) Epoch 6, batch 250, loss[loss=0.226, simple_loss=0.3028, pruned_loss=0.07456, over 18507.00 frames. ], tot_loss[loss=0.2321, simple_loss=0.3082, pruned_loss=0.07802, over 2657939.01 frames. ], batch size: 77, lr: 1.97e-02, grad_scale: 8.0 2022-12-22 17:49:32,192 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0_sp0.9 from training. Duration: 24.033375 2022-12-22 17:49:35,441 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17791.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:50:14,418 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.256e+02 5.169e+02 5.967e+02 7.942e+02 1.897e+03, threshold=1.193e+03, percent-clipped=6.0 2022-12-22 17:50:23,093 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7646, 1.0030, 0.7119, 1.2757, 1.7888, 1.3583, 1.5193, 1.8577], device='cuda:3'), covar=tensor([0.1783, 0.2552, 0.2925, 0.1780, 0.1986, 0.1585, 0.1635, 0.1796], device='cuda:3'), in_proj_covar=tensor([0.0093, 0.0108, 0.0129, 0.0098, 0.0112, 0.0094, 0.0099, 0.0101], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 17:50:29,899 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0 from training. Duration: 22.905 2022-12-22 17:50:31,947 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp1.1 from training. Duration: 23.4318125 2022-12-22 17:50:37,528 INFO [train.py:894] (3/4) Epoch 6, batch 300, loss[loss=0.2048, simple_loss=0.2749, pruned_loss=0.06731, over 18542.00 frames. ], tot_loss[loss=0.2323, simple_loss=0.3081, pruned_loss=0.07828, over 2891853.65 frames. ], batch size: 44, lr: 1.97e-02, grad_scale: 8.0 2022-12-22 17:50:57,118 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17845.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:51:41,061 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.20 vs. limit=5.0 2022-12-22 17:51:53,309 INFO [train.py:894] (3/4) Epoch 6, batch 350, loss[loss=0.2203, simple_loss=0.2923, pruned_loss=0.07411, over 18396.00 frames. ], tot_loss[loss=0.2331, simple_loss=0.3087, pruned_loss=0.07874, over 3073694.75 frames. ], batch size: 46, lr: 1.96e-02, grad_scale: 8.0 2022-12-22 17:52:09,717 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=17893.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:52:30,019 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp1.1 from training. Duration: 20.82275 2022-12-22 17:52:31,592 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp0.9 from training. Duration: 25.45 2022-12-22 17:52:46,761 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.444e+02 5.136e+02 6.255e+02 7.350e+02 1.917e+03, threshold=1.251e+03, percent-clipped=5.0 2022-12-22 17:53:10,028 INFO [train.py:894] (3/4) Epoch 6, batch 400, loss[loss=0.2519, simple_loss=0.3201, pruned_loss=0.09183, over 18573.00 frames. ], tot_loss[loss=0.2347, simple_loss=0.31, pruned_loss=0.07971, over 3215236.58 frames. ], batch size: 49, lr: 1.96e-02, grad_scale: 8.0 2022-12-22 17:53:32,452 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0 from training. Duration: 25.775 2022-12-22 17:53:34,789 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.98 vs. limit=5.0 2022-12-22 17:53:52,410 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0_sp0.9 from training. Duration: 22.25 2022-12-22 17:53:52,818 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17960.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:53:52,961 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0639, 1.3318, 1.8246, 1.8760, 2.0332, 1.7603, 1.8277, 1.2669], device='cuda:3'), covar=tensor([0.1058, 0.1672, 0.1196, 0.1227, 0.0775, 0.0516, 0.1411, 0.0671], device='cuda:3'), in_proj_covar=tensor([0.0211, 0.0241, 0.0216, 0.0234, 0.0217, 0.0195, 0.0231, 0.0188], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 17:53:54,787 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.94 vs. limit=5.0 2022-12-22 17:54:20,740 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0 from training. Duration: 26.205 2022-12-22 17:54:26,350 INFO [train.py:894] (3/4) Epoch 6, batch 450, loss[loss=0.2331, simple_loss=0.3079, pruned_loss=0.07916, over 18524.00 frames. ], tot_loss[loss=0.2359, simple_loss=0.3112, pruned_loss=0.08037, over 3325634.85 frames. ], batch size: 47, lr: 1.96e-02, grad_scale: 8.0 2022-12-22 17:54:28,619 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2022-12-22 17:54:37,924 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp0.9 from training. Duration: 30.1555625 2022-12-22 17:54:43,768 WARNING [train.py:1060] (3/4) Exclude cut with ID 1265-135635-0050-6781-0_sp0.9 from training. Duration: 21.8333125 2022-12-22 17:54:52,605 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp1.1 from training. Duration: 20.6545625 2022-12-22 17:55:21,989 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.240e+02 5.677e+02 6.582e+02 7.577e+02 1.434e+03, threshold=1.316e+03, percent-clipped=3.0 2022-12-22 17:55:23,947 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18018.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:55:28,393 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18021.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:55:41,509 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2022-12-22 17:55:41,932 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0045-39920-0_sp0.9 from training. Duration: 20.52225 2022-12-22 17:55:44,511 INFO [train.py:894] (3/4) Epoch 6, batch 500, loss[loss=0.2347, simple_loss=0.3239, pruned_loss=0.07271, over 18484.00 frames. ], tot_loss[loss=0.2367, simple_loss=0.3117, pruned_loss=0.08085, over 3410374.23 frames. ], batch size: 54, lr: 1.96e-02, grad_scale: 8.0 2022-12-22 17:55:55,082 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18039.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 17:56:02,869 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp0.9 from training. Duration: 29.1166875 2022-12-22 17:56:16,732 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3883, 2.7792, 2.8361, 1.2669, 3.0538, 2.4816, 1.7576, 3.6472], device='cuda:3'), covar=tensor([0.1289, 0.1276, 0.1471, 0.2326, 0.0868, 0.1394, 0.2194, 0.0436], device='cuda:3'), in_proj_covar=tensor([0.0198, 0.0181, 0.0199, 0.0191, 0.0180, 0.0205, 0.0197, 0.0173], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 17:56:21,311 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18056.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:56:36,512 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=18066.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:56:53,929 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8107, 1.6719, 2.2373, 1.3237, 2.1790, 1.9273, 1.3128, 2.7184], device='cuda:3'), covar=tensor([0.1172, 0.1487, 0.1153, 0.1689, 0.0837, 0.1092, 0.2122, 0.0409], device='cuda:3'), in_proj_covar=tensor([0.0196, 0.0177, 0.0197, 0.0188, 0.0178, 0.0202, 0.0194, 0.0170], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 17:56:59,289 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133211-0007-59831-0_sp0.9 from training. Duration: 21.388875 2022-12-22 17:57:00,624 INFO [train.py:894] (3/4) Epoch 6, batch 550, loss[loss=0.2662, simple_loss=0.351, pruned_loss=0.09068, over 18510.00 frames. ], tot_loss[loss=0.2372, simple_loss=0.3119, pruned_loss=0.08126, over 3476586.89 frames. ], batch size: 64, lr: 1.95e-02, grad_scale: 8.0 2022-12-22 17:57:06,788 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18086.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:57:07,187 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0849, 2.0608, 1.3158, 2.2672, 1.7455, 1.6361, 1.7003, 2.2149], device='cuda:3'), covar=tensor([0.1435, 0.2038, 0.1197, 0.1819, 0.2101, 0.0750, 0.1903, 0.0456], device='cuda:3'), in_proj_covar=tensor([0.0249, 0.0233, 0.0207, 0.0313, 0.0224, 0.0194, 0.0238, 0.0165], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 17:57:08,321 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=18087.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 17:57:37,420 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0 from training. Duration: 22.72 2022-12-22 17:57:38,938 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0_sp0.9 from training. Duration: 22.7444375 2022-12-22 17:57:54,189 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.507e+02 5.216e+02 6.559e+02 8.218e+02 2.484e+03, threshold=1.312e+03, percent-clipped=5.0 2022-12-22 17:58:17,195 INFO [train.py:894] (3/4) Epoch 6, batch 600, loss[loss=0.3032, simple_loss=0.3586, pruned_loss=0.124, over 18572.00 frames. ], tot_loss[loss=0.2348, simple_loss=0.3096, pruned_loss=0.07997, over 3528763.04 frames. ], batch size: 57, lr: 1.95e-02, grad_scale: 8.0 2022-12-22 17:58:17,791 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0422, 1.9137, 1.3021, 2.1832, 1.7125, 1.7029, 1.7456, 2.3408], device='cuda:3'), covar=tensor([0.1432, 0.2150, 0.1294, 0.1808, 0.2115, 0.0742, 0.1855, 0.0453], device='cuda:3'), in_proj_covar=tensor([0.0246, 0.0232, 0.0204, 0.0311, 0.0223, 0.0191, 0.0234, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 17:58:23,425 WARNING [train.py:1060] (3/4) Exclude cut with ID 4133-6541-0027-40495-0_sp1.1 from training. Duration: 0.9681875 2022-12-22 17:58:28,317 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0_sp0.9 from training. Duration: 22.3166875 2022-12-22 17:58:32,959 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133212-0015-59917-0_sp0.9 from training. Duration: 21.8166875 2022-12-22 17:59:21,339 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18174.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:59:22,888 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18175.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 17:59:32,802 INFO [train.py:894] (3/4) Epoch 6, batch 650, loss[loss=0.2266, simple_loss=0.3012, pruned_loss=0.07599, over 18696.00 frames. ], tot_loss[loss=0.2365, simple_loss=0.3111, pruned_loss=0.08095, over 3569824.48 frames. ], batch size: 50, lr: 1.95e-02, grad_scale: 8.0 2022-12-22 18:00:14,599 WARNING [train.py:1060] (3/4) Exclude cut with ID 4957-30119-0041-23990-0_sp0.9 from training. Duration: 20.22775 2022-12-22 18:00:21,810 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.97 vs. limit=2.0 2022-12-22 18:00:27,697 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.151e+02 5.694e+02 6.622e+02 8.175e+02 1.499e+03, threshold=1.324e+03, percent-clipped=1.0 2022-12-22 18:00:39,752 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2272, 1.4701, 2.2953, 4.3553, 2.9539, 2.4049, 0.6655, 2.6296], device='cuda:3'), covar=tensor([0.1703, 0.1872, 0.1509, 0.0345, 0.1054, 0.1536, 0.2719, 0.1085], device='cuda:3'), in_proj_covar=tensor([0.0104, 0.0115, 0.0125, 0.0100, 0.0105, 0.0125, 0.0133, 0.0105], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 18:00:49,610 INFO [train.py:894] (3/4) Epoch 6, batch 700, loss[loss=0.2725, simple_loss=0.3435, pruned_loss=0.1008, over 18577.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.3125, pruned_loss=0.08139, over 3601795.21 frames. ], batch size: 97, lr: 1.95e-02, grad_scale: 8.0 2022-12-22 18:00:54,954 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18235.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:00:56,453 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18236.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:01:00,401 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp1.1 from training. Duration: 24.67275 2022-12-22 18:01:27,509 WARNING [train.py:1060] (3/4) Exclude cut with ID 3082-165428-0081-50734-0_sp0.9 from training. Duration: 21.8055625 2022-12-22 18:01:37,021 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9744, 1.1544, 1.6860, 1.6538, 1.9311, 1.6931, 1.6614, 1.1514], device='cuda:3'), covar=tensor([0.0973, 0.1687, 0.1241, 0.1239, 0.0820, 0.0524, 0.1347, 0.0714], device='cuda:3'), in_proj_covar=tensor([0.0211, 0.0244, 0.0219, 0.0240, 0.0220, 0.0199, 0.0236, 0.0190], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 18:01:46,025 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6765, 3.8178, 3.8794, 1.7824, 3.8037, 2.9564, 0.8057, 2.8128], device='cuda:3'), covar=tensor([0.1629, 0.0651, 0.1149, 0.3323, 0.0895, 0.0885, 0.5405, 0.1346], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0103, 0.0144, 0.0114, 0.0105, 0.0097, 0.0143, 0.0107], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 18:02:05,110 INFO [train.py:894] (3/4) Epoch 6, batch 750, loss[loss=0.2518, simple_loss=0.3299, pruned_loss=0.08685, over 18578.00 frames. ], tot_loss[loss=0.2377, simple_loss=0.3127, pruned_loss=0.08133, over 3626531.64 frames. ], batch size: 57, lr: 1.94e-02, grad_scale: 8.0 2022-12-22 18:02:05,175 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0_sp0.9 from training. Duration: 22.6666875 2022-12-22 18:02:19,719 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8304, 2.0954, 1.0702, 2.4523, 2.0585, 1.8086, 3.0291, 1.8923], device='cuda:3'), covar=tensor([0.1175, 0.1545, 0.2887, 0.1895, 0.1763, 0.1083, 0.0758, 0.1414], device='cuda:3'), in_proj_covar=tensor([0.0169, 0.0178, 0.0218, 0.0265, 0.0214, 0.0173, 0.0186, 0.0186], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 18:02:39,311 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2022-12-22 18:02:57,207 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18316.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:02:58,471 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.968e+02 5.188e+02 6.272e+02 8.631e+02 1.940e+03, threshold=1.254e+03, percent-clipped=5.0 2022-12-22 18:03:09,586 WARNING [train.py:1060] (3/4) Exclude cut with ID 2411-132532-0017-82279-0_sp1.1 from training. Duration: 0.9681875 2022-12-22 18:03:16,297 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.8916, 2.1428, 1.5213, 2.8082, 1.9758, 2.0516, 2.1374, 3.5063], device='cuda:3'), covar=tensor([0.1425, 0.2488, 0.1324, 0.2363, 0.2343, 0.0782, 0.2360, 0.0375], device='cuda:3'), in_proj_covar=tensor([0.0251, 0.0236, 0.0207, 0.0311, 0.0227, 0.0195, 0.0236, 0.0166], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 18:03:20,392 INFO [train.py:894] (3/4) Epoch 6, batch 800, loss[loss=0.2446, simple_loss=0.3314, pruned_loss=0.07894, over 18551.00 frames. ], tot_loss[loss=0.2381, simple_loss=0.3132, pruned_loss=0.08148, over 3645249.43 frames. ], batch size: 55, lr: 1.94e-02, grad_scale: 8.0 2022-12-22 18:03:36,045 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0 from training. Duration: 22.485 2022-12-22 18:03:56,420 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18356.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:04:15,043 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp1.1 from training. Duration: 23.82275 2022-12-22 18:04:27,993 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0 from training. Duration: 20.77 2022-12-22 18:04:35,184 INFO [train.py:894] (3/4) Epoch 6, batch 850, loss[loss=0.2297, simple_loss=0.3162, pruned_loss=0.07158, over 18629.00 frames. ], tot_loss[loss=0.2387, simple_loss=0.3136, pruned_loss=0.08187, over 3660307.79 frames. ], batch size: 53, lr: 1.94e-02, grad_scale: 8.0 2022-12-22 18:04:35,199 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0_sp0.9 from training. Duration: 24.088875 2022-12-22 18:04:41,620 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18386.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:05:02,372 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp1.1 from training. Duration: 20.4409375 2022-12-22 18:05:09,302 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=18404.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:05:29,138 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.354e+02 5.112e+02 6.131e+02 8.014e+02 1.589e+03, threshold=1.226e+03, percent-clipped=3.0 2022-12-22 18:05:51,629 INFO [train.py:894] (3/4) Epoch 6, batch 900, loss[loss=0.2385, simple_loss=0.321, pruned_loss=0.07801, over 18674.00 frames. ], tot_loss[loss=0.2367, simple_loss=0.3124, pruned_loss=0.08053, over 3671966.17 frames. ], batch size: 60, lr: 1.94e-02, grad_scale: 8.0 2022-12-22 18:05:53,763 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3386, 1.9394, 1.4328, 0.4448, 1.4346, 1.8567, 1.3323, 1.6909], device='cuda:3'), covar=tensor([0.0494, 0.0401, 0.0976, 0.1364, 0.0928, 0.1200, 0.1425, 0.0622], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0169, 0.0195, 0.0189, 0.0189, 0.0172, 0.0189, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 18:05:54,919 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=18434.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:06:10,891 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2022-12-22 18:06:19,420 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2022-12-22 18:06:23,320 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0_sp0.9 from training. Duration: 22.511125 2022-12-22 18:06:23,346 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0 from training. Duration: 20.675 2022-12-22 18:06:45,172 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7166, 1.1042, 1.3755, 1.2579, 1.6779, 1.6300, 1.7077, 1.2009], device='cuda:3'), covar=tensor([0.0318, 0.0310, 0.0502, 0.0264, 0.0213, 0.0320, 0.0306, 0.0345], device='cuda:3'), in_proj_covar=tensor([0.0076, 0.0109, 0.0130, 0.0119, 0.0105, 0.0093, 0.0080, 0.0117], device='cuda:3'), out_proj_covar=tensor([7.4721e-05, 1.0270e-04, 1.2874e-04, 1.1243e-04, 1.0427e-04, 8.6626e-05, 7.6452e-05, 1.1147e-04], device='cuda:3') 2022-12-22 18:07:07,804 INFO [train.py:894] (3/4) Epoch 6, batch 950, loss[loss=0.2108, simple_loss=0.2747, pruned_loss=0.07342, over 18413.00 frames. ], tot_loss[loss=0.2371, simple_loss=0.3127, pruned_loss=0.08074, over 3681507.58 frames. ], batch size: 42, lr: 1.93e-02, grad_scale: 8.0 2022-12-22 18:07:20,675 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18490.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:07:47,174 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18508.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 18:08:00,283 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.265e+02 5.277e+02 6.294e+02 8.257e+02 2.016e+03, threshold=1.259e+03, percent-clipped=4.0 2022-12-22 18:08:06,519 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp0.9 from training. Duration: 24.9833125 2022-12-22 18:08:19,601 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18530.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:08:21,858 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18531.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:08:23,119 INFO [train.py:894] (3/4) Epoch 6, batch 1000, loss[loss=0.2252, simple_loss=0.3064, pruned_loss=0.07194, over 18505.00 frames. ], tot_loss[loss=0.2373, simple_loss=0.3128, pruned_loss=0.08087, over 3687558.68 frames. ], batch size: 52, lr: 1.93e-02, grad_scale: 8.0 2022-12-22 18:08:35,650 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0 from training. Duration: 27.14 2022-12-22 18:08:51,916 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18551.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:08:53,056 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0 from training. Duration: 22.44 2022-12-22 18:09:10,499 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.72 vs. limit=5.0 2022-12-22 18:09:18,869 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18569.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 18:09:36,970 INFO [train.py:894] (3/4) Epoch 6, batch 1050, loss[loss=0.3001, simple_loss=0.3541, pruned_loss=0.123, over 18556.00 frames. ], tot_loss[loss=0.2383, simple_loss=0.3134, pruned_loss=0.08158, over 3693274.65 frames. ], batch size: 173, lr: 1.93e-02, grad_scale: 8.0 2022-12-22 18:10:09,747 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0060-62364-0_sp0.9 from training. Duration: 21.361125 2022-12-22 18:10:15,389 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp1.1 from training. Duration: 27.0318125 2022-12-22 18:10:26,363 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp0.9 from training. Duration: 28.638875 2022-12-22 18:10:29,192 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18616.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:10:30,228 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.146e+02 5.294e+02 6.925e+02 9.252e+02 1.661e+03, threshold=1.385e+03, percent-clipped=6.0 2022-12-22 18:10:42,026 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0 from training. Duration: 20.4 2022-12-22 18:10:52,468 INFO [train.py:894] (3/4) Epoch 6, batch 1100, loss[loss=0.2295, simple_loss=0.2984, pruned_loss=0.08034, over 18538.00 frames. ], tot_loss[loss=0.2387, simple_loss=0.3134, pruned_loss=0.08198, over 3698210.42 frames. ], batch size: 44, lr: 1.93e-02, grad_scale: 16.0 2022-12-22 18:11:15,792 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0 from training. Duration: 20.025 2022-12-22 18:11:15,806 WARNING [train.py:1060] (3/4) Exclude cut with ID 2364-131735-0112-64612-0_sp0.9 from training. Duration: 20.488875 2022-12-22 18:11:21,871 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0 from training. Duration: 29.735 2022-12-22 18:11:40,273 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=18664.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:11:51,189 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2022-12-22 18:11:55,077 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18674.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:12:06,864 INFO [train.py:894] (3/4) Epoch 6, batch 1150, loss[loss=0.2373, simple_loss=0.3025, pruned_loss=0.08605, over 18557.00 frames. ], tot_loss[loss=0.2373, simple_loss=0.3122, pruned_loss=0.08126, over 3701484.17 frames. ], batch size: 49, lr: 1.93e-02, grad_scale: 16.0 2022-12-22 18:12:42,282 WARNING [train.py:1060] (3/4) Exclude cut with ID 7276-92427-0014-12983-0_sp0.9 from training. Duration: 21.3055625 2022-12-22 18:12:43,700 WARNING [train.py:1060] (3/4) Exclude cut with ID 1025-75365-0008-79168-0_sp0.9 from training. Duration: 22.0666875 2022-12-22 18:12:46,277 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2022-12-22 18:12:53,299 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2022-12-22 18:12:59,394 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.797e+02 5.337e+02 6.606e+02 8.749e+02 2.715e+03, threshold=1.321e+03, percent-clipped=4.0 2022-12-22 18:13:21,629 INFO [train.py:894] (3/4) Epoch 6, batch 1200, loss[loss=0.2149, simple_loss=0.2923, pruned_loss=0.0688, over 18578.00 frames. ], tot_loss[loss=0.2375, simple_loss=0.3123, pruned_loss=0.08133, over 3703982.05 frames. ], batch size: 49, lr: 1.92e-02, grad_scale: 16.0 2022-12-22 18:13:22,011 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0082, 2.0444, 2.2211, 1.4047, 2.3874, 2.3238, 1.5607, 2.6901], device='cuda:3'), covar=tensor([0.1060, 0.1416, 0.1464, 0.1809, 0.0764, 0.1057, 0.2255, 0.0427], device='cuda:3'), in_proj_covar=tensor([0.0197, 0.0182, 0.0201, 0.0190, 0.0181, 0.0202, 0.0201, 0.0173], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 18:13:26,760 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18735.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:13:50,706 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.65 vs. limit=5.0 2022-12-22 18:14:32,168 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0 from training. Duration: 20.26 2022-12-22 18:14:38,458 INFO [train.py:894] (3/4) Epoch 6, batch 1250, loss[loss=0.2467, simple_loss=0.3185, pruned_loss=0.08746, over 18683.00 frames. ], tot_loss[loss=0.2367, simple_loss=0.3119, pruned_loss=0.08074, over 3706513.23 frames. ], batch size: 69, lr: 1.92e-02, grad_scale: 16.0 2022-12-22 18:14:44,456 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0030-9324-0_sp0.9 from training. Duration: 21.3444375 2022-12-22 18:15:32,034 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.330e+02 4.900e+02 6.256e+02 8.296e+02 1.790e+03, threshold=1.251e+03, percent-clipped=3.0 2022-12-22 18:15:39,425 WARNING [train.py:1060] (3/4) Exclude cut with ID 497-129325-0061-62254-0_sp1.1 from training. Duration: 0.97725 2022-12-22 18:15:52,910 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18830.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:15:54,404 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18831.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:15:55,641 INFO [train.py:894] (3/4) Epoch 6, batch 1300, loss[loss=0.2359, simple_loss=0.313, pruned_loss=0.07939, over 18554.00 frames. ], tot_loss[loss=0.2369, simple_loss=0.3122, pruned_loss=0.08083, over 3707761.19 frames. ], batch size: 55, lr: 1.92e-02, grad_scale: 16.0 2022-12-22 18:16:16,782 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18846.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:16:22,830 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0_sp0.9 from training. Duration: 22.97225 2022-12-22 18:16:40,899 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4930, 3.3407, 3.3862, 1.5569, 3.2815, 2.2814, 0.9184, 2.3906], device='cuda:3'), covar=tensor([0.1975, 0.0932, 0.1581, 0.3825, 0.1049, 0.1291, 0.5445, 0.1810], device='cuda:3'), in_proj_covar=tensor([0.0126, 0.0104, 0.0144, 0.0115, 0.0107, 0.0098, 0.0142, 0.0108], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 18:16:43,828 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18864.0, num_to_drop=1, layers_to_drop={2} 2022-12-22 18:16:54,006 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0047-39922-0_sp0.9 from training. Duration: 21.97775 2022-12-22 18:16:59,252 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.3094, 3.7359, 3.6788, 4.1345, 3.7439, 3.8631, 4.4273, 1.5885], device='cuda:3'), covar=tensor([0.0718, 0.0553, 0.0571, 0.0673, 0.1658, 0.0874, 0.0538, 0.4475], device='cuda:3'), in_proj_covar=tensor([0.0235, 0.0171, 0.0168, 0.0161, 0.0240, 0.0191, 0.0189, 0.0226], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-22 18:17:05,348 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=18878.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:17:06,669 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=18879.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:17:07,950 WARNING [train.py:1060] (3/4) Exclude cut with ID 1112-1043-0006-89194-0_sp0.9 from training. Duration: 21.8333125 2022-12-22 18:17:10,448 INFO [train.py:894] (3/4) Epoch 6, batch 1350, loss[loss=0.2418, simple_loss=0.3184, pruned_loss=0.0826, over 18664.00 frames. ], tot_loss[loss=0.2355, simple_loss=0.3109, pruned_loss=0.08004, over 3709632.06 frames. ], batch size: 60, lr: 1.92e-02, grad_scale: 8.0 2022-12-22 18:17:12,430 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1979, 2.1557, 1.6792, 1.1042, 3.1435, 2.5640, 1.7550, 1.7411], device='cuda:3'), covar=tensor([0.0433, 0.0372, 0.0596, 0.0791, 0.0092, 0.0286, 0.0533, 0.0799], device='cuda:3'), in_proj_covar=tensor([0.0118, 0.0111, 0.0128, 0.0120, 0.0073, 0.0114, 0.0135, 0.0145], device='cuda:3'), out_proj_covar=tensor([1.4715e-04, 1.3903e-04, 1.5559e-04, 1.4857e-04, 9.4482e-05, 1.3860e-04, 1.6570e-04, 1.7852e-04], device='cuda:3') 2022-12-22 18:17:18,071 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0 from training. Duration: 20.47 2022-12-22 18:18:00,369 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.2818, 1.6350, 1.3273, 2.0650, 2.1608, 1.3660, 1.5375, 1.1393], device='cuda:3'), covar=tensor([0.1971, 0.1660, 0.1476, 0.0935, 0.1252, 0.1265, 0.1574, 0.1534], device='cuda:3'), in_proj_covar=tensor([0.0223, 0.0195, 0.0186, 0.0175, 0.0238, 0.0176, 0.0193, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 18:18:04,573 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.035e+02 5.335e+02 6.297e+02 7.950e+02 1.792e+03, threshold=1.259e+03, percent-clipped=3.0 2022-12-22 18:18:22,346 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0037-39912-0_sp0.9 from training. Duration: 20.67225 2022-12-22 18:18:24,070 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18930.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:18:26,873 INFO [train.py:894] (3/4) Epoch 6, batch 1400, loss[loss=0.2415, simple_loss=0.3224, pruned_loss=0.08032, over 18611.00 frames. ], tot_loss[loss=0.2365, simple_loss=0.3119, pruned_loss=0.08055, over 3710254.77 frames. ], batch size: 53, lr: 1.91e-02, grad_scale: 8.0 2022-12-22 18:18:40,702 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp0.9 from training. Duration: 25.2444375 2022-12-22 18:19:04,180 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0021-76797-0_sp0.9 from training. Duration: 21.1445 2022-12-22 18:19:42,955 INFO [train.py:894] (3/4) Epoch 6, batch 1450, loss[loss=0.2526, simple_loss=0.3329, pruned_loss=0.08614, over 18505.00 frames. ], tot_loss[loss=0.2369, simple_loss=0.3122, pruned_loss=0.08074, over 3711312.69 frames. ], batch size: 58, lr: 1.91e-02, grad_scale: 8.0 2022-12-22 18:19:57,331 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18991.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:20:19,030 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp0.9 from training. Duration: 33.038875 2022-12-22 18:20:37,423 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.579e+02 4.874e+02 5.865e+02 6.786e+02 3.300e+03, threshold=1.173e+03, percent-clipped=4.0 2022-12-22 18:20:56,539 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19030.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:20:58,007 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64291-0000-16059-0_sp0.9 from training. Duration: 20.0944375 2022-12-22 18:20:59,362 INFO [train.py:894] (3/4) Epoch 6, batch 1500, loss[loss=0.2514, simple_loss=0.3186, pruned_loss=0.09205, over 18467.00 frames. ], tot_loss[loss=0.2361, simple_loss=0.3119, pruned_loss=0.08015, over 3712843.01 frames. ], batch size: 54, lr: 1.91e-02, grad_scale: 8.0 2022-12-22 18:21:09,128 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2022-12-22 18:21:12,839 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp1.1 from training. Duration: 20.4 2022-12-22 18:21:20,098 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0 from training. Duration: 20.085 2022-12-22 18:21:30,216 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0_sp0.9 from training. Duration: 23.07775 2022-12-22 18:22:14,908 INFO [train.py:894] (3/4) Epoch 6, batch 1550, loss[loss=0.2108, simple_loss=0.2923, pruned_loss=0.06467, over 18425.00 frames. ], tot_loss[loss=0.2351, simple_loss=0.3111, pruned_loss=0.07959, over 3712503.34 frames. ], batch size: 48, lr: 1.91e-02, grad_scale: 8.0 2022-12-22 18:22:17,622 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp0.9 from training. Duration: 24.9333125 2022-12-22 18:22:19,902 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2022-12-22 18:23:04,228 WARNING [train.py:1060] (3/4) Exclude cut with ID 5118-111612-0016-124680-0_sp0.9 from training. Duration: 20.388875 2022-12-22 18:23:08,351 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.609e+02 5.320e+02 6.098e+02 7.401e+02 2.797e+03, threshold=1.220e+03, percent-clipped=4.0 2022-12-22 18:23:09,957 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp1.1 from training. Duration: 20.3590625 2022-12-22 18:23:25,899 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2780, 1.5212, 2.3062, 4.2371, 3.0772, 2.4430, 0.3517, 2.5252], device='cuda:3'), covar=tensor([0.1592, 0.2019, 0.1535, 0.0358, 0.1089, 0.1378, 0.3049, 0.1144], device='cuda:3'), in_proj_covar=tensor([0.0103, 0.0116, 0.0124, 0.0100, 0.0104, 0.0125, 0.0134, 0.0104], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 18:23:25,961 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19129.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 18:23:29,943 INFO [train.py:894] (3/4) Epoch 6, batch 1600, loss[loss=0.2153, simple_loss=0.3052, pruned_loss=0.0627, over 18462.00 frames. ], tot_loss[loss=0.2346, simple_loss=0.3108, pruned_loss=0.07916, over 3712841.76 frames. ], batch size: 54, lr: 1.90e-02, grad_scale: 8.0 2022-12-22 18:23:48,701 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6668, 2.1629, 1.4768, 2.4752, 1.8970, 2.1021, 1.9568, 2.8774], device='cuda:3'), covar=tensor([0.1221, 0.1988, 0.1266, 0.1913, 0.2209, 0.0711, 0.1920, 0.0386], device='cuda:3'), in_proj_covar=tensor([0.0250, 0.0236, 0.0204, 0.0313, 0.0225, 0.0196, 0.0237, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 18:23:51,481 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19146.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:24:17,849 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19164.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 18:24:20,482 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0_sp1.1 from training. Duration: 0.836375 2022-12-22 18:24:46,279 INFO [train.py:894] (3/4) Epoch 6, batch 1650, loss[loss=0.2297, simple_loss=0.3092, pruned_loss=0.07506, over 18583.00 frames. ], tot_loss[loss=0.2369, simple_loss=0.3119, pruned_loss=0.08096, over 3713056.13 frames. ], batch size: 57, lr: 1.90e-02, grad_scale: 8.0 2022-12-22 18:24:59,036 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19190.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 18:25:04,732 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=19194.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:25:06,008 WARNING [train.py:1060] (3/4) Exclude cut with ID 8565-290391-0049-67394-0_sp0.9 from training. Duration: 21.3166875 2022-12-22 18:25:23,026 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2022-12-22 18:25:31,116 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=19212.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 18:25:35,175 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0029-104863-0_sp0.9 from training. Duration: 22.1055625 2022-12-22 18:25:40,080 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.440e+02 5.588e+02 6.972e+02 9.071e+02 2.121e+03, threshold=1.394e+03, percent-clipped=5.0 2022-12-22 18:25:44,739 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp1.1 from training. Duration: 21.77725 2022-12-22 18:26:02,483 INFO [train.py:894] (3/4) Epoch 6, batch 1700, loss[loss=0.2454, simple_loss=0.3269, pruned_loss=0.08193, over 18592.00 frames. ], tot_loss[loss=0.2417, simple_loss=0.3143, pruned_loss=0.08459, over 3712058.55 frames. ], batch size: 97, lr: 1.90e-02, grad_scale: 8.0 2022-12-22 18:26:05,366 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp0.9 from training. Duration: 27.8166875 2022-12-22 18:26:28,984 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp1.1 from training. Duration: 22.5090625 2022-12-22 18:26:33,292 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0 from training. Duration: 25.035 2022-12-22 18:26:52,454 WARNING [train.py:1060] (3/4) Exclude cut with ID 774-127930-0014-10412-0_sp1.1 from training. Duration: 0.95 2022-12-22 18:27:10,375 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp0.9 from training. Duration: 0.92225 2022-12-22 18:27:14,374 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9031, 2.1805, 1.1643, 2.2646, 2.2486, 1.8780, 3.0936, 2.1958], device='cuda:3'), covar=tensor([0.0849, 0.1326, 0.2346, 0.1919, 0.1431, 0.0830, 0.0862, 0.0939], device='cuda:3'), in_proj_covar=tensor([0.0168, 0.0181, 0.0221, 0.0269, 0.0215, 0.0174, 0.0189, 0.0185], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 18:27:15,583 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0831, 1.2907, 2.0175, 3.9202, 2.7012, 2.5891, 0.7694, 2.5671], device='cuda:3'), covar=tensor([0.1830, 0.2194, 0.1804, 0.0549, 0.1318, 0.1323, 0.3047, 0.1290], device='cuda:3'), in_proj_covar=tensor([0.0105, 0.0117, 0.0125, 0.0104, 0.0104, 0.0126, 0.0136, 0.0107], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 18:27:18,392 INFO [train.py:894] (3/4) Epoch 6, batch 1750, loss[loss=0.2284, simple_loss=0.291, pruned_loss=0.08291, over 18691.00 frames. ], tot_loss[loss=0.2461, simple_loss=0.3172, pruned_loss=0.08753, over 3711473.87 frames. ], batch size: 48, lr: 1.90e-02, grad_scale: 8.0 2022-12-22 18:27:24,745 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19286.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:27:35,163 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0 from training. Duration: 21.97 2022-12-22 18:27:53,780 WARNING [train.py:1060] (3/4) Exclude cut with ID 7492-105653-0055-62765-0_sp0.9 from training. Duration: 21.97225 2022-12-22 18:27:55,027 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp0.9 from training. Duration: 25.3333125 2022-12-22 18:27:58,373 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2830, 2.0731, 1.6886, 0.9336, 2.7972, 2.5646, 1.6606, 1.7009], device='cuda:3'), covar=tensor([0.0291, 0.0355, 0.0482, 0.0769, 0.0114, 0.0213, 0.0545, 0.0730], device='cuda:3'), in_proj_covar=tensor([0.0116, 0.0111, 0.0129, 0.0121, 0.0074, 0.0113, 0.0134, 0.0145], device='cuda:3'), out_proj_covar=tensor([1.4424e-04, 1.3908e-04, 1.5703e-04, 1.4949e-04, 9.5037e-05, 1.3648e-04, 1.6430e-04, 1.7887e-04], device='cuda:3') 2022-12-22 18:28:06,717 WARNING [train.py:1060] (3/4) Exclude cut with ID 5172-29468-0015-19128-0_sp0.9 from training. Duration: 21.5055625 2022-12-22 18:28:12,634 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.198e+02 6.146e+02 7.824e+02 9.741e+02 2.877e+03, threshold=1.565e+03, percent-clipped=5.0 2022-12-22 18:28:16,312 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp1.1 from training. Duration: 20.72725 2022-12-22 18:28:31,561 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19330.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:28:34,155 INFO [train.py:894] (3/4) Epoch 6, batch 1800, loss[loss=0.2582, simple_loss=0.3122, pruned_loss=0.1021, over 18392.00 frames. ], tot_loss[loss=0.2493, simple_loss=0.3182, pruned_loss=0.09018, over 3712370.32 frames. ], batch size: 46, lr: 1.89e-02, grad_scale: 8.0 2022-12-22 18:28:49,182 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp0.9 from training. Duration: 26.32775 2022-12-22 18:29:20,895 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0 from training. Duration: 20.025 2022-12-22 18:29:27,440 WARNING [train.py:1060] (3/4) Exclude cut with ID 6709-74022-0004-86860-0_sp1.1 from training. Duration: 0.9409375 2022-12-22 18:29:27,448 WARNING [train.py:1060] (3/4) Exclude cut with ID 4757-1811-0023-62229-0_sp0.9 from training. Duration: 21.37775 2022-12-22 18:29:44,345 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=19378.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:29:46,994 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0004-25974-0_sp0.9 from training. Duration: 21.17225 2022-12-22 18:29:47,004 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp0.9 from training. Duration: 27.511125 2022-12-22 18:29:50,071 INFO [train.py:894] (3/4) Epoch 6, batch 1850, loss[loss=0.2588, simple_loss=0.3293, pruned_loss=0.09416, over 18671.00 frames. ], tot_loss[loss=0.2515, simple_loss=0.3187, pruned_loss=0.09218, over 3712144.77 frames. ], batch size: 60, lr: 1.89e-02, grad_scale: 8.0 2022-12-22 18:30:07,639 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19394.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:30:14,056 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2794, 2.0116, 1.9505, 1.4975, 1.9299, 2.7863, 2.5846, 1.9608], device='cuda:3'), covar=tensor([0.0583, 0.0360, 0.0567, 0.0388, 0.0392, 0.0297, 0.0433, 0.0474], device='cuda:3'), in_proj_covar=tensor([0.0081, 0.0112, 0.0133, 0.0121, 0.0106, 0.0098, 0.0084, 0.0119], device='cuda:3'), out_proj_covar=tensor([7.7879e-05, 1.0420e-04, 1.2982e-04, 1.1388e-04, 1.0375e-04, 8.9983e-05, 7.9972e-05, 1.1141e-04], device='cuda:3') 2022-12-22 18:30:22,006 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0 from training. Duration: 22.8 2022-12-22 18:30:26,471 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0 from training. Duration: 22.585 2022-12-22 18:30:44,799 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.043e+02 6.226e+02 7.501e+02 1.005e+03 2.297e+03, threshold=1.500e+03, percent-clipped=4.0 2022-12-22 18:30:57,163 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0001-146967-0_sp0.9 from training. Duration: 22.0166875 2022-12-22 18:31:05,992 INFO [train.py:894] (3/4) Epoch 6, batch 1900, loss[loss=0.2632, simple_loss=0.3092, pruned_loss=0.1086, over 18615.00 frames. ], tot_loss[loss=0.255, simple_loss=0.3203, pruned_loss=0.09481, over 3711942.21 frames. ], batch size: 45, lr: 1.89e-02, grad_scale: 8.0 2022-12-22 18:31:13,639 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp1.1 from training. Duration: 24.395375 2022-12-22 18:31:20,801 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp0.9 from training. Duration: 27.47775 2022-12-22 18:31:25,324 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp0.9 from training. Duration: 24.8833125 2022-12-22 18:31:30,621 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0 from training. Duration: 23.39 2022-12-22 18:31:35,150 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp0.9 from training. Duration: 28.72225 2022-12-22 18:31:41,348 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19455.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:31:43,874 WARNING [train.py:1060] (3/4) Exclude cut with ID 585-294811-0110-133686-0_sp0.9 from training. Duration: 20.8944375 2022-12-22 18:31:59,302 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0_sp0.9 from training. Duration: 23.8444375 2022-12-22 18:32:22,700 INFO [train.py:894] (3/4) Epoch 6, batch 1950, loss[loss=0.3259, simple_loss=0.3681, pruned_loss=0.1419, over 18617.00 frames. ], tot_loss[loss=0.2558, simple_loss=0.3204, pruned_loss=0.09562, over 3712281.42 frames. ], batch size: 177, lr: 1.89e-02, grad_scale: 8.0 2022-12-22 18:32:22,776 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0 from training. Duration: 25.85 2022-12-22 18:32:22,788 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0 from training. Duration: 21.39 2022-12-22 18:32:23,221 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5412, 2.0691, 1.5061, 2.5275, 1.8411, 1.8159, 2.1612, 2.7106], device='cuda:3'), covar=tensor([0.1321, 0.2131, 0.1361, 0.2145, 0.2286, 0.0778, 0.1803, 0.0465], device='cuda:3'), in_proj_covar=tensor([0.0254, 0.0239, 0.0206, 0.0319, 0.0225, 0.0197, 0.0237, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 18:32:27,124 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19485.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 18:32:32,820 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0 from training. Duration: 27.92 2022-12-22 18:33:01,916 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0039-130165-0_sp0.9 from training. Duration: 20.661125 2022-12-22 18:33:16,721 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.731e+02 6.433e+02 7.824e+02 1.097e+03 2.201e+03, threshold=1.565e+03, percent-clipped=9.0 2022-12-22 18:33:26,889 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0043-15874-0_sp0.9 from training. Duration: 20.07225 2022-12-22 18:33:32,666 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0 from training. Duration: 21.01 2022-12-22 18:33:37,262 INFO [train.py:894] (3/4) Epoch 6, batch 2000, loss[loss=0.2565, simple_loss=0.3191, pruned_loss=0.0969, over 18673.00 frames. ], tot_loss[loss=0.2591, simple_loss=0.3229, pruned_loss=0.09762, over 3713459.66 frames. ], batch size: 62, lr: 1.89e-02, grad_scale: 8.0 2022-12-22 18:34:40,540 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0 from training. Duration: 20.65 2022-12-22 18:34:47,555 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0 from training. Duration: 21.46 2022-12-22 18:34:53,398 INFO [train.py:894] (3/4) Epoch 6, batch 2050, loss[loss=0.2619, simple_loss=0.3283, pruned_loss=0.0978, over 18698.00 frames. ], tot_loss[loss=0.2603, simple_loss=0.3232, pruned_loss=0.09876, over 3714131.72 frames. ], batch size: 65, lr: 1.88e-02, grad_scale: 8.0 2022-12-22 18:35:00,336 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19586.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:35:34,096 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0 from training. Duration: 0.92 2022-12-22 18:35:41,637 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0_sp0.9 from training. Duration: 23.7666875 2022-12-22 18:35:48,883 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.006e+02 6.032e+02 7.436e+02 8.945e+02 2.067e+03, threshold=1.487e+03, percent-clipped=2.0 2022-12-22 18:35:49,334 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19618.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:36:09,628 INFO [train.py:894] (3/4) Epoch 6, batch 2100, loss[loss=0.2397, simple_loss=0.3176, pruned_loss=0.08092, over 18384.00 frames. ], tot_loss[loss=0.2603, simple_loss=0.3231, pruned_loss=0.09871, over 3713958.17 frames. ], batch size: 51, lr: 1.88e-02, grad_scale: 8.0 2022-12-22 18:36:13,549 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=19634.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:36:19,361 WARNING [train.py:1060] (3/4) Exclude cut with ID 8544-281189-0060-101339-0_sp0.9 from training. Duration: 20.861125 2022-12-22 18:36:30,117 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0_sp0.9 from training. Duration: 22.711125 2022-12-22 18:37:10,942 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp1.1 from training. Duration: 22.986375 2022-12-22 18:37:11,316 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6318, 2.4487, 1.9349, 1.0350, 1.9158, 2.0607, 1.6574, 2.0419], device='cuda:3'), covar=tensor([0.0591, 0.0400, 0.1181, 0.1445, 0.1157, 0.1109, 0.1293, 0.0814], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0173, 0.0199, 0.0192, 0.0196, 0.0177, 0.0193, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 18:37:21,950 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19679.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:37:25,761 INFO [train.py:894] (3/4) Epoch 6, batch 2150, loss[loss=0.2387, simple_loss=0.2931, pruned_loss=0.09218, over 18638.00 frames. ], tot_loss[loss=0.2624, simple_loss=0.3243, pruned_loss=0.1003, over 3715044.20 frames. ], batch size: 45, lr: 1.88e-02, grad_scale: 8.0 2022-12-22 18:37:28,762 WARNING [train.py:1060] (3/4) Exclude cut with ID 8040-260924-0003-80960-0_sp0.9 from training. Duration: 22.07225 2022-12-22 18:37:32,367 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0045-26330-0_sp0.9 from training. Duration: 20.3055625 2022-12-22 18:37:35,232 WARNING [train.py:1060] (3/4) Exclude cut with ID 6356-271890-0060-94317-0_sp0.9 from training. Duration: 20.72225 2022-12-22 18:37:56,546 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp1.1 from training. Duration: 22.4818125 2022-12-22 18:38:19,421 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2022-12-22 18:38:21,383 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.545e+02 5.984e+02 7.689e+02 9.435e+02 1.935e+03, threshold=1.538e+03, percent-clipped=2.0 2022-12-22 18:38:22,902 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp0.9 from training. Duration: 25.0944375 2022-12-22 18:38:26,093 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0 from training. Duration: 21.515 2022-12-22 18:38:31,979 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp0.9 from training. Duration: 27.02225 2022-12-22 18:38:36,951 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0010-62480-0_sp0.9 from training. Duration: 22.22225 2022-12-22 18:38:43,684 INFO [train.py:894] (3/4) Epoch 6, batch 2200, loss[loss=0.2837, simple_loss=0.3483, pruned_loss=0.1095, over 18615.00 frames. ], tot_loss[loss=0.2606, simple_loss=0.3226, pruned_loss=0.09929, over 3715232.89 frames. ], batch size: 53, lr: 1.88e-02, grad_scale: 8.0 2022-12-22 18:38:45,187 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0085-44554-0_sp0.9 from training. Duration: 20.85 2022-12-22 18:38:53,480 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2022-12-22 18:39:12,294 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19750.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:39:18,150 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0 from training. Duration: 21.54 2022-12-22 18:39:22,330 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp1.1 from training. Duration: 20.5318125 2022-12-22 18:39:30,486 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0012-134311-0_sp0.9 from training. Duration: 21.9333125 2022-12-22 18:39:40,568 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2022-12-22 18:39:59,931 INFO [train.py:894] (3/4) Epoch 6, batch 2250, loss[loss=0.2519, simple_loss=0.3188, pruned_loss=0.09251, over 18590.00 frames. ], tot_loss[loss=0.2612, simple_loss=0.3229, pruned_loss=0.09974, over 3714525.50 frames. ], batch size: 56, lr: 1.87e-02, grad_scale: 8.0 2022-12-22 18:40:04,894 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19785.0, num_to_drop=1, layers_to_drop={2} 2022-12-22 18:40:13,944 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2022-12-22 18:40:19,196 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0025-130151-0_sp0.9 from training. Duration: 21.7944375 2022-12-22 18:40:33,365 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0_sp0.9 from training. Duration: 22.4666875 2022-12-22 18:40:41,916 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0 from training. Duration: 21.635 2022-12-22 18:40:47,237 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0_sp0.9 from training. Duration: 24.038875 2022-12-22 18:40:53,007 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.367e+02 6.231e+02 7.152e+02 9.491e+02 2.745e+03, threshold=1.430e+03, percent-clipped=1.0 2022-12-22 18:41:14,829 INFO [train.py:894] (3/4) Epoch 6, batch 2300, loss[loss=0.3078, simple_loss=0.3569, pruned_loss=0.1294, over 18577.00 frames. ], tot_loss[loss=0.2617, simple_loss=0.3234, pruned_loss=0.09998, over 3714545.57 frames. ], batch size: 56, lr: 1.87e-02, grad_scale: 8.0 2022-12-22 18:41:17,424 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=19833.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 18:41:31,228 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp1.1 from training. Duration: 21.786375 2022-12-22 18:41:42,438 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0 from training. Duration: 20.22 2022-12-22 18:42:31,654 INFO [train.py:894] (3/4) Epoch 6, batch 2350, loss[loss=0.2282, simple_loss=0.2906, pruned_loss=0.08292, over 18548.00 frames. ], tot_loss[loss=0.2619, simple_loss=0.3235, pruned_loss=0.1001, over 3713518.95 frames. ], batch size: 41, lr: 1.87e-02, grad_scale: 8.0 2022-12-22 18:43:26,587 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.986e+02 6.436e+02 8.082e+02 9.463e+02 1.888e+03, threshold=1.616e+03, percent-clipped=4.0 2022-12-22 18:43:37,101 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2022-12-22 18:43:45,589 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0 from training. Duration: 25.285 2022-12-22 18:43:48,411 INFO [train.py:894] (3/4) Epoch 6, batch 2400, loss[loss=0.2213, simple_loss=0.2964, pruned_loss=0.07309, over 18438.00 frames. ], tot_loss[loss=0.2613, simple_loss=0.3234, pruned_loss=0.09964, over 3714139.02 frames. ], batch size: 48, lr: 1.87e-02, grad_scale: 8.0 2022-12-22 18:44:34,947 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8025, 1.3741, 1.4885, 1.3284, 1.4901, 1.8492, 2.0147, 1.2027], device='cuda:3'), covar=tensor([0.0413, 0.0340, 0.0457, 0.0305, 0.0243, 0.0314, 0.0260, 0.0368], device='cuda:3'), in_proj_covar=tensor([0.0081, 0.0111, 0.0134, 0.0119, 0.0102, 0.0096, 0.0081, 0.0117], device='cuda:3'), out_proj_covar=tensor([7.7365e-05, 1.0238e-04, 1.2961e-04, 1.1201e-04, 9.9055e-05, 8.7094e-05, 7.5908e-05, 1.0841e-04], device='cuda:3') 2022-12-22 18:44:49,860 WARNING [train.py:1060] (3/4) Exclude cut with ID 811-130148-0001-63453-0_sp0.9 from training. Duration: 20.861125 2022-12-22 18:44:53,623 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19974.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:45:05,737 INFO [train.py:894] (3/4) Epoch 6, batch 2450, loss[loss=0.2645, simple_loss=0.3248, pruned_loss=0.1021, over 18431.00 frames. ], tot_loss[loss=0.2609, simple_loss=0.3233, pruned_loss=0.09928, over 3713655.51 frames. ], batch size: 48, lr: 1.87e-02, grad_scale: 8.0 2022-12-22 18:45:11,558 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0 from training. Duration: 20.88 2022-12-22 18:45:22,821 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5836, 1.5920, 1.6275, 1.8707, 1.5791, 3.8235, 2.0444, 2.3598], device='cuda:3'), covar=tensor([0.3076, 0.2007, 0.1717, 0.1704, 0.1221, 0.0171, 0.1254, 0.0802], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0124, 0.0135, 0.0123, 0.0109, 0.0100, 0.0106, 0.0103], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2022-12-22 18:45:47,167 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0_sp0.9 from training. Duration: 23.4166875 2022-12-22 18:46:04,284 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.728e+02 6.177e+02 7.986e+02 9.402e+02 1.900e+03, threshold=1.597e+03, percent-clipped=3.0 2022-12-22 18:46:08,328 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20020.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:46:25,449 INFO [train.py:894] (3/4) Epoch 6, batch 2500, loss[loss=0.2434, simple_loss=0.318, pruned_loss=0.08442, over 18648.00 frames. ], tot_loss[loss=0.2598, simple_loss=0.3224, pruned_loss=0.09858, over 3713579.22 frames. ], batch size: 69, lr: 1.86e-02, grad_scale: 8.0 2022-12-22 18:46:32,312 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.7260, 2.5643, 3.1721, 1.2714, 2.4821, 3.4165, 1.9282, 2.9333], device='cuda:3'), covar=tensor([0.0815, 0.0477, 0.0266, 0.0336, 0.0428, 0.0225, 0.0479, 0.0457], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0137, 0.0094, 0.0118, 0.0124, 0.0100, 0.0128, 0.0117], device='cuda:3'), out_proj_covar=tensor([1.1654e-04, 1.3243e-04, 9.0929e-05, 1.1218e-04, 1.1752e-04, 9.6205e-05, 1.2535e-04, 1.1232e-04], device='cuda:3') 2022-12-22 18:46:52,644 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20050.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:47:07,527 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0 from training. Duration: 21.24 2022-12-22 18:47:08,997 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0_sp0.9 from training. Duration: 23.9055625 2022-12-22 18:47:40,011 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20081.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:47:41,028 INFO [train.py:894] (3/4) Epoch 6, batch 2550, loss[loss=0.2744, simple_loss=0.3424, pruned_loss=0.1032, over 18508.00 frames. ], tot_loss[loss=0.2587, simple_loss=0.3215, pruned_loss=0.09794, over 3713534.62 frames. ], batch size: 52, lr: 1.86e-02, grad_scale: 8.0 2022-12-22 18:47:42,961 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp0.9 from training. Duration: 25.988875 2022-12-22 18:47:43,466 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4056, 2.1518, 1.4349, 2.2916, 1.8951, 1.7321, 1.9915, 2.4679], device='cuda:3'), covar=tensor([0.1464, 0.2179, 0.1399, 0.2039, 0.2090, 0.0825, 0.1797, 0.0496], device='cuda:3'), in_proj_covar=tensor([0.0264, 0.0244, 0.0214, 0.0328, 0.0231, 0.0202, 0.0243, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 18:47:50,646 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0001-134300-0_sp0.9 from training. Duration: 20.67225 2022-12-22 18:47:52,397 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20089.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:48:06,656 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=20098.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:48:28,669 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20112.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:48:37,932 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.210e+02 7.103e+02 8.652e+02 9.800e+02 3.308e+03, threshold=1.730e+03, percent-clipped=5.0 2022-12-22 18:48:39,474 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0 from training. Duration: 20.34 2022-12-22 18:48:59,242 INFO [train.py:894] (3/4) Epoch 6, batch 2600, loss[loss=0.2615, simple_loss=0.3229, pruned_loss=0.1, over 18670.00 frames. ], tot_loss[loss=0.2581, simple_loss=0.3208, pruned_loss=0.09769, over 3713380.02 frames. ], batch size: 62, lr: 1.86e-02, grad_scale: 8.0 2022-12-22 18:49:25,879 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20150.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:49:40,738 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4929, 1.4657, 1.7765, 1.3182, 1.6819, 1.6890, 1.3385, 1.9845], device='cuda:3'), covar=tensor([0.0859, 0.1023, 0.1028, 0.1135, 0.0556, 0.0818, 0.1373, 0.0355], device='cuda:3'), in_proj_covar=tensor([0.0201, 0.0184, 0.0195, 0.0189, 0.0178, 0.0208, 0.0195, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 18:49:50,414 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2022-12-22 18:49:51,127 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp0.9 from training. Duration: 25.061125 2022-12-22 18:50:01,429 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20173.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:50:02,471 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0 from training. Duration: 0.83 2022-12-22 18:50:14,689 INFO [train.py:894] (3/4) Epoch 6, batch 2650, loss[loss=0.2726, simple_loss=0.3431, pruned_loss=0.1011, over 18720.00 frames. ], tot_loss[loss=0.2582, simple_loss=0.3209, pruned_loss=0.09775, over 3714607.52 frames. ], batch size: 65, lr: 1.86e-02, grad_scale: 8.0 2022-12-22 18:50:28,848 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0 from training. Duration: 24.73 2022-12-22 18:50:36,879 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2022-12-22 18:50:40,418 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0 from training. Duration: 23.965 2022-12-22 18:50:48,889 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0030-146996-0_sp0.9 from training. Duration: 22.088875 2022-12-22 18:50:49,894 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20204.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:51:04,975 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0_sp0.9 from training. Duration: 23.6 2022-12-22 18:51:10,962 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.129e+02 6.870e+02 8.496e+02 1.069e+03 1.947e+03, threshold=1.699e+03, percent-clipped=3.0 2022-12-22 18:51:31,617 INFO [train.py:894] (3/4) Epoch 6, batch 2700, loss[loss=0.2415, simple_loss=0.3048, pruned_loss=0.0891, over 18717.00 frames. ], tot_loss[loss=0.2588, simple_loss=0.3214, pruned_loss=0.09807, over 3713886.03 frames. ], batch size: 50, lr: 1.85e-02, grad_scale: 8.0 2022-12-22 18:52:13,865 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.7628, 2.8233, 2.6151, 1.2683, 2.3639, 2.3911, 1.8219, 2.0664], device='cuda:3'), covar=tensor([0.0535, 0.0520, 0.1244, 0.1756, 0.1573, 0.1190, 0.1490, 0.1075], device='cuda:3'), in_proj_covar=tensor([0.0155, 0.0179, 0.0207, 0.0197, 0.0202, 0.0185, 0.0202, 0.0186], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 18:52:22,539 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20265.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:52:36,236 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20274.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:52:48,471 INFO [train.py:894] (3/4) Epoch 6, batch 2750, loss[loss=0.2105, simple_loss=0.276, pruned_loss=0.07244, over 18609.00 frames. ], tot_loss[loss=0.2587, simple_loss=0.3214, pruned_loss=0.09803, over 3714738.97 frames. ], batch size: 45, lr: 1.85e-02, grad_scale: 8.0 2022-12-22 18:52:48,472 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0 from training. Duration: 23.795 2022-12-22 18:52:58,587 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2022-12-22 18:53:03,733 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp1.1 from training. Duration: 21.5409375 2022-12-22 18:53:06,906 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp0.9 from training. Duration: 24.97775 2022-12-22 18:53:19,713 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0_sp0.9 from training. Duration: 23.3444375 2022-12-22 18:53:44,340 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.398e+02 6.276e+02 7.992e+02 1.109e+03 2.849e+03, threshold=1.598e+03, percent-clipped=4.0 2022-12-22 18:53:45,843 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0_sp0.9 from training. Duration: 23.2 2022-12-22 18:53:51,273 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=20322.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:53:52,645 WARNING [train.py:1060] (3/4) Exclude cut with ID 5653-46179-0060-117930-0_sp0.9 from training. Duration: 21.17225 2022-12-22 18:54:05,730 INFO [train.py:894] (3/4) Epoch 6, batch 2800, loss[loss=0.2708, simple_loss=0.3285, pruned_loss=0.1066, over 18629.00 frames. ], tot_loss[loss=0.2582, simple_loss=0.321, pruned_loss=0.09766, over 3714536.70 frames. ], batch size: 53, lr: 1.85e-02, grad_scale: 8.0 2022-12-22 18:54:12,064 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp0.9 from training. Duration: 24.6555625 2022-12-22 18:55:06,165 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.0730, 0.8510, 1.0831, 0.0753, 0.8053, 1.2600, 1.1733, 1.1449], device='cuda:3'), covar=tensor([0.0640, 0.0323, 0.0295, 0.0415, 0.0418, 0.0384, 0.0265, 0.0489], device='cuda:3'), in_proj_covar=tensor([0.0120, 0.0135, 0.0092, 0.0116, 0.0121, 0.0098, 0.0124, 0.0117], device='cuda:3'), out_proj_covar=tensor([1.1223e-04, 1.2926e-04, 8.7930e-05, 1.0899e-04, 1.1401e-04, 9.4077e-05, 1.2092e-04, 1.1199e-04], device='cuda:3') 2022-12-22 18:55:08,625 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0 from training. Duration: 20.44 2022-12-22 18:55:13,449 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20376.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:55:17,758 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2022-12-22 18:55:22,168 INFO [train.py:894] (3/4) Epoch 6, batch 2850, loss[loss=0.2646, simple_loss=0.3296, pruned_loss=0.09977, over 18653.00 frames. ], tot_loss[loss=0.2577, simple_loss=0.3205, pruned_loss=0.0974, over 3714780.63 frames. ], batch size: 62, lr: 1.85e-02, grad_scale: 8.0 2022-12-22 18:55:23,891 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0_sp0.9 from training. Duration: 23.45 2022-12-22 18:55:31,542 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20388.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:55:53,495 WARNING [train.py:1060] (3/4) Exclude cut with ID 6945-60535-0076-12784-0_sp0.9 from training. Duration: 20.52225 2022-12-22 18:55:58,024 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.2656, 0.9746, 1.2395, 1.9494, 1.4023, 1.7286, 0.9362, 1.3573], device='cuda:3'), covar=tensor([0.1554, 0.1613, 0.1370, 0.0658, 0.1179, 0.1643, 0.1812, 0.1237], device='cuda:3'), in_proj_covar=tensor([0.0105, 0.0117, 0.0125, 0.0105, 0.0106, 0.0127, 0.0133, 0.0107], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 18:56:00,844 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0 from training. Duration: 22.19 2022-12-22 18:56:11,894 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp1.1 from training. Duration: 25.3818125 2022-12-22 18:56:18,484 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.122e+02 6.167e+02 7.330e+02 9.180e+02 5.006e+03, threshold=1.466e+03, percent-clipped=5.0 2022-12-22 18:56:18,954 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20418.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:56:29,226 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp0.9 from training. Duration: 28.0944375 2022-12-22 18:56:35,092 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0_sp0.9 from training. Duration: 22.9444375 2022-12-22 18:56:39,023 INFO [train.py:894] (3/4) Epoch 6, batch 2900, loss[loss=0.2552, simple_loss=0.3307, pruned_loss=0.08986, over 18587.00 frames. ], tot_loss[loss=0.257, simple_loss=0.3202, pruned_loss=0.09691, over 3714295.06 frames. ], batch size: 56, lr: 1.85e-02, grad_scale: 8.0 2022-12-22 18:56:41,971 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp1.1 from training. Duration: 21.6318125 2022-12-22 18:56:59,264 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0 from training. Duration: 23.695 2022-12-22 18:56:59,373 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20445.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:57:05,529 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20449.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:57:21,400 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9064, 1.8438, 2.0873, 1.3150, 2.2044, 2.2324, 1.4771, 2.5614], device='cuda:3'), covar=tensor([0.0931, 0.1231, 0.1230, 0.1603, 0.0649, 0.0970, 0.1994, 0.0435], device='cuda:3'), in_proj_covar=tensor([0.0200, 0.0187, 0.0195, 0.0190, 0.0178, 0.0206, 0.0196, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 18:57:25,494 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0 from training. Duration: 23.955 2022-12-22 18:57:33,894 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20468.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:57:50,581 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20479.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 18:57:54,441 INFO [train.py:894] (3/4) Epoch 6, batch 2950, loss[loss=0.2491, simple_loss=0.3144, pruned_loss=0.09188, over 18695.00 frames. ], tot_loss[loss=0.2572, simple_loss=0.3205, pruned_loss=0.09701, over 3714869.63 frames. ], batch size: 50, lr: 1.84e-02, grad_scale: 8.0 2022-12-22 18:58:01,538 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp0.9 from training. Duration: 26.438875 2022-12-22 18:58:46,204 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.26 vs. limit=5.0 2022-12-22 18:58:48,772 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0021-26306-0_sp0.9 from training. Duration: 21.2444375 2022-12-22 18:58:48,809 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp0.9 from training. Duration: 31.02225 2022-12-22 18:58:50,133 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.777e+02 5.957e+02 7.468e+02 9.949e+02 2.082e+03, threshold=1.494e+03, percent-clipped=7.0 2022-12-22 18:58:58,508 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0 from training. Duration: 22.395 2022-12-22 18:59:10,916 INFO [train.py:894] (3/4) Epoch 6, batch 3000, loss[loss=0.2487, simple_loss=0.3045, pruned_loss=0.09641, over 18492.00 frames. ], tot_loss[loss=0.2563, simple_loss=0.3197, pruned_loss=0.09644, over 3713795.48 frames. ], batch size: 43, lr: 1.84e-02, grad_scale: 8.0 2022-12-22 18:59:10,916 INFO [train.py:919] (3/4) Computing validation loss 2022-12-22 18:59:15,232 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6955, 1.6117, 1.7550, 1.5659, 1.9538, 2.4809, 2.1196, 1.7341], device='cuda:3'), covar=tensor([0.0616, 0.0432, 0.0454, 0.0369, 0.0355, 0.0285, 0.0354, 0.0373], device='cuda:3'), in_proj_covar=tensor([0.0079, 0.0112, 0.0133, 0.0120, 0.0103, 0.0095, 0.0081, 0.0116], device='cuda:3'), out_proj_covar=tensor([7.5134e-05, 1.0236e-04, 1.2773e-04, 1.1156e-04, 9.9697e-05, 8.5689e-05, 7.5732e-05, 1.0698e-04], device='cuda:3') 2022-12-22 18:59:22,156 INFO [train.py:928] (3/4) Epoch 6, validation: loss=0.1911, simple_loss=0.2898, pruned_loss=0.04619, over 944034.00 frames. 2022-12-22 18:59:22,157 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24676MB 2022-12-22 18:59:28,922 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0 from training. Duration: 21.075 2022-12-22 18:59:35,148 WARNING [train.py:1060] (3/4) Exclude cut with ID 6482-98857-0025-147532-0_sp0.9 from training. Duration: 20.0055625 2022-12-22 18:59:35,161 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0037-132304-0_sp0.9 from training. Duration: 22.05 2022-12-22 18:59:35,174 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0 from training. Duration: 26.8349375 2022-12-22 18:59:38,072 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp1.1 from training. Duration: 22.1090625 2022-12-22 18:59:44,133 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp0.9 from training. Duration: 26.6166875 2022-12-22 18:59:49,351 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5911, 1.3859, 0.8590, 1.3223, 1.8328, 1.3104, 1.6194, 1.9091], device='cuda:3'), covar=tensor([0.1766, 0.2231, 0.2533, 0.1753, 0.1937, 0.1662, 0.1407, 0.1555], device='cuda:3'), in_proj_covar=tensor([0.0096, 0.0107, 0.0128, 0.0103, 0.0115, 0.0096, 0.0100, 0.0101], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 19:00:02,762 WARNING [train.py:1060] (3/4) Exclude cut with ID 2046-178027-0000-53705-0_sp0.9 from training. Duration: 20.3055625 2022-12-22 19:00:06,095 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20560.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:00:26,392 WARNING [train.py:1060] (3/4) Exclude cut with ID 7205-50138-0008-5373-0_sp0.9 from training. Duration: 20.7 2022-12-22 19:00:40,409 INFO [train.py:894] (3/4) Epoch 6, batch 3050, loss[loss=0.2714, simple_loss=0.3484, pruned_loss=0.09724, over 18519.00 frames. ], tot_loss[loss=0.2559, simple_loss=0.3197, pruned_loss=0.09607, over 3714264.38 frames. ], batch size: 52, lr: 1.84e-02, grad_scale: 8.0 2022-12-22 19:00:49,785 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20588.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:01:10,633 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0 from training. Duration: 22.48 2022-12-22 19:01:26,412 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp0.9 from training. Duration: 29.816625 2022-12-22 19:01:36,229 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.602e+02 6.371e+02 7.791e+02 9.358e+02 2.068e+03, threshold=1.558e+03, percent-clipped=2.0 2022-12-22 19:01:46,701 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp1.1 from training. Duration: 22.7590625 2022-12-22 19:01:51,692 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0 from training. Duration: 22.555 2022-12-22 19:01:57,508 INFO [train.py:894] (3/4) Epoch 6, batch 3100, loss[loss=0.2113, simple_loss=0.2733, pruned_loss=0.07463, over 18616.00 frames. ], tot_loss[loss=0.2575, simple_loss=0.3209, pruned_loss=0.09705, over 3714505.42 frames. ], batch size: 45, lr: 1.84e-02, grad_scale: 8.0 2022-12-22 19:02:11,310 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0005-25975-0_sp0.9 from training. Duration: 21.688875 2022-12-22 19:02:23,639 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20649.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:02:47,504 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0_sp0.9 from training. Duration: 22.6 2022-12-22 19:03:04,000 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20676.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:03:12,819 INFO [train.py:894] (3/4) Epoch 6, batch 3150, loss[loss=0.2315, simple_loss=0.2983, pruned_loss=0.08235, over 18436.00 frames. ], tot_loss[loss=0.2583, simple_loss=0.3215, pruned_loss=0.09752, over 3715135.47 frames. ], batch size: 48, lr: 1.84e-02, grad_scale: 8.0 2022-12-22 19:03:23,685 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0 from training. Duration: 24.32 2022-12-22 19:03:54,363 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9648, 1.7757, 1.9865, 2.5588, 2.1845, 4.8143, 1.7118, 1.9409], device='cuda:3'), covar=tensor([0.0955, 0.1791, 0.1164, 0.0883, 0.1376, 0.0158, 0.1350, 0.1439], device='cuda:3'), in_proj_covar=tensor([0.0079, 0.0088, 0.0081, 0.0083, 0.0097, 0.0073, 0.0090, 0.0081], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 19:04:08,115 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.940e+02 6.143e+02 7.454e+02 8.942e+02 3.821e+03, threshold=1.491e+03, percent-clipped=3.0 2022-12-22 19:04:17,369 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=20724.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:04:20,121 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-276745-0093-13116-0_sp0.9 from training. Duration: 21.061125 2022-12-22 19:04:29,667 INFO [train.py:894] (3/4) Epoch 6, batch 3200, loss[loss=0.2232, simple_loss=0.2928, pruned_loss=0.07679, over 18725.00 frames. ], tot_loss[loss=0.2591, simple_loss=0.3224, pruned_loss=0.09796, over 3715360.37 frames. ], batch size: 41, lr: 1.83e-02, grad_scale: 8.0 2022-12-22 19:04:35,497 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0024-15855-0_sp0.9 from training. Duration: 20.32225 2022-12-22 19:04:48,092 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20744.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:04:49,788 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp1.1 from training. Duration: 0.7545625 2022-12-22 19:04:50,015 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20745.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:05:01,023 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0_sp0.9 from training. Duration: 23.9333125 2022-12-22 19:05:11,126 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2022-12-22 19:05:25,551 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20768.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:05:34,262 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20774.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:05:37,119 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp1.1 from training. Duration: 20.17275 2022-12-22 19:05:43,362 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp1.1 from training. Duration: 20.436375 2022-12-22 19:05:46,182 INFO [train.py:894] (3/4) Epoch 6, batch 3250, loss[loss=0.2196, simple_loss=0.2917, pruned_loss=0.07369, over 18472.00 frames. ], tot_loss[loss=0.2566, simple_loss=0.3204, pruned_loss=0.09644, over 3714113.78 frames. ], batch size: 50, lr: 1.83e-02, grad_scale: 4.0 2022-12-22 19:06:03,699 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=20793.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:06:40,259 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=20816.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:06:44,642 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.773e+02 6.050e+02 7.541e+02 9.973e+02 3.613e+03, threshold=1.508e+03, percent-clipped=8.0 2022-12-22 19:07:04,325 INFO [train.py:894] (3/4) Epoch 6, batch 3300, loss[loss=0.2518, simple_loss=0.3224, pruned_loss=0.0906, over 18517.00 frames. ], tot_loss[loss=0.255, simple_loss=0.3193, pruned_loss=0.09533, over 3714465.28 frames. ], batch size: 58, lr: 1.83e-02, grad_scale: 4.0 2022-12-22 19:07:05,586 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0_sp0.9 from training. Duration: 23.1055625 2022-12-22 19:07:07,000 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp1.1 from training. Duration: 23.5 2022-12-22 19:07:17,715 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp0.9 from training. Duration: 26.62775 2022-12-22 19:07:31,420 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0 from training. Duration: 21.105 2022-12-22 19:07:35,883 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0_sp0.9 from training. Duration: 24.411125 2022-12-22 19:07:46,366 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20860.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:08:05,300 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp1.1 from training. Duration: 21.263625 2022-12-22 19:08:19,487 INFO [train.py:894] (3/4) Epoch 6, batch 3350, loss[loss=0.2773, simple_loss=0.3349, pruned_loss=0.1099, over 18651.00 frames. ], tot_loss[loss=0.2537, simple_loss=0.3183, pruned_loss=0.09454, over 3714884.04 frames. ], batch size: 98, lr: 1.83e-02, grad_scale: 4.0 2022-12-22 19:08:35,950 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0 from training. Duration: 20.795 2022-12-22 19:08:46,162 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0 from training. Duration: 24.76 2022-12-22 19:08:46,181 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0_sp0.9 from training. Duration: 22.25 2022-12-22 19:09:00,344 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=20908.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:09:10,578 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp1.1 from training. Duration: 20.5045625 2022-12-22 19:09:16,851 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.707e+02 6.289e+02 7.876e+02 9.972e+02 1.794e+03, threshold=1.575e+03, percent-clipped=4.0 2022-12-22 19:09:37,393 INFO [train.py:894] (3/4) Epoch 6, batch 3400, loss[loss=0.2618, simple_loss=0.3117, pruned_loss=0.106, over 18486.00 frames. ], tot_loss[loss=0.2548, simple_loss=0.3184, pruned_loss=0.09555, over 3715284.58 frames. ], batch size: 43, lr: 1.83e-02, grad_scale: 4.0 2022-12-22 19:09:46,617 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3268, 1.7130, 1.1067, 2.1747, 2.2374, 1.4827, 1.2519, 1.1206], device='cuda:3'), covar=tensor([0.1859, 0.1518, 0.1546, 0.0785, 0.1088, 0.1138, 0.1579, 0.1445], device='cuda:3'), in_proj_covar=tensor([0.0223, 0.0193, 0.0189, 0.0174, 0.0237, 0.0177, 0.0195, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 19:09:55,198 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20944.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:10:50,670 INFO [train.py:894] (3/4) Epoch 6, batch 3450, loss[loss=0.3047, simple_loss=0.3583, pruned_loss=0.1256, over 18633.00 frames. ], tot_loss[loss=0.2546, simple_loss=0.3187, pruned_loss=0.0953, over 3715751.25 frames. ], batch size: 178, lr: 1.82e-02, grad_scale: 4.0 2022-12-22 19:11:40,339 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2022-12-22 19:11:44,686 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.589e+02 6.787e+02 7.956e+02 1.031e+03 2.201e+03, threshold=1.591e+03, percent-clipped=5.0 2022-12-22 19:12:04,664 INFO [train.py:894] (3/4) Epoch 6, batch 3500, loss[loss=0.3291, simple_loss=0.368, pruned_loss=0.145, over 18640.00 frames. ], tot_loss[loss=0.2574, simple_loss=0.3206, pruned_loss=0.09711, over 3715999.87 frames. ], batch size: 178, lr: 1.82e-02, grad_scale: 4.0 2022-12-22 19:12:25,800 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp1.1 from training. Duration: 22.2954375 2022-12-22 19:12:35,558 INFO [train.py:894] (3/4) Epoch 7, batch 0, loss[loss=0.249, simple_loss=0.3282, pruned_loss=0.08489, over 18580.00 frames. ], tot_loss[loss=0.249, simple_loss=0.3282, pruned_loss=0.08489, over 18580.00 frames. ], batch size: 78, lr: 1.71e-02, grad_scale: 8.0 2022-12-22 19:12:35,559 INFO [train.py:919] (3/4) Computing validation loss 2022-12-22 19:12:47,233 INFO [train.py:928] (3/4) Epoch 7, validation: loss=0.1926, simple_loss=0.291, pruned_loss=0.0471, over 944034.00 frames. 2022-12-22 19:12:47,234 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24676MB 2022-12-22 19:12:56,951 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21044.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:13:09,218 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21052.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:13:39,803 WARNING [train.py:1060] (3/4) Exclude cut with ID 298-126791-0067-24026-0_sp0.9 from training. Duration: 21.438875 2022-12-22 19:13:41,442 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21074.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:13:43,908 WARNING [train.py:1060] (3/4) Exclude cut with ID 5652-39938-0025-23684-0_sp0.9 from training. Duration: 22.2055625 2022-12-22 19:14:00,029 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4994, 1.4962, 1.4416, 1.8816, 1.6499, 3.5138, 1.5634, 1.5254], device='cuda:3'), covar=tensor([0.1049, 0.1778, 0.1215, 0.0991, 0.1398, 0.0185, 0.1309, 0.1506], device='cuda:3'), in_proj_covar=tensor([0.0081, 0.0090, 0.0084, 0.0084, 0.0099, 0.0075, 0.0092, 0.0083], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2022-12-22 19:14:01,170 INFO [train.py:894] (3/4) Epoch 7, batch 50, loss[loss=0.2169, simple_loss=0.3003, pruned_loss=0.06677, over 18582.00 frames. ], tot_loss[loss=0.2364, simple_loss=0.3109, pruned_loss=0.08094, over 839053.16 frames. ], batch size: 51, lr: 1.70e-02, grad_scale: 8.0 2022-12-22 19:14:08,010 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=21092.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:14:40,461 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21113.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:14:48,439 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21118.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:14:49,495 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.520e+02 5.183e+02 6.375e+02 7.921e+02 1.523e+03, threshold=1.275e+03, percent-clipped=0.0 2022-12-22 19:14:53,940 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=21122.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:15:17,780 INFO [train.py:894] (3/4) Epoch 7, batch 100, loss[loss=0.2378, simple_loss=0.304, pruned_loss=0.08583, over 18464.00 frames. ], tot_loss[loss=0.2311, simple_loss=0.3072, pruned_loss=0.07755, over 1476402.12 frames. ], batch size: 50, lr: 1.70e-02, grad_scale: 8.0 2022-12-22 19:16:20,288 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21179.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:16:33,904 INFO [train.py:894] (3/4) Epoch 7, batch 150, loss[loss=0.2471, simple_loss=0.3227, pruned_loss=0.08575, over 18727.00 frames. ], tot_loss[loss=0.2282, simple_loss=0.305, pruned_loss=0.07574, over 1972284.30 frames. ], batch size: 52, lr: 1.70e-02, grad_scale: 4.0 2022-12-22 19:16:41,825 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0 from training. Duration: 24.525 2022-12-22 19:17:16,292 WARNING [train.py:1060] (3/4) Exclude cut with ID 3699-47246-0007-3408-0_sp0.9 from training. Duration: 20.26675 2022-12-22 19:17:22,377 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.317e+02 5.094e+02 6.367e+02 7.897e+02 2.757e+03, threshold=1.273e+03, percent-clipped=4.0 2022-12-22 19:17:28,279 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp0.9 from training. Duration: 27.25 2022-12-22 19:17:38,079 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2022-12-22 19:17:49,906 INFO [train.py:894] (3/4) Epoch 7, batch 200, loss[loss=0.2016, simple_loss=0.2836, pruned_loss=0.05983, over 18724.00 frames. ], tot_loss[loss=0.2261, simple_loss=0.303, pruned_loss=0.0746, over 2358316.68 frames. ], batch size: 50, lr: 1.70e-02, grad_scale: 4.0 2022-12-22 19:18:00,280 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21244.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:18:41,640 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0 from training. Duration: 21.68 2022-12-22 19:18:42,476 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.77 vs. limit=5.0 2022-12-22 19:18:53,609 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0 from training. Duration: 21.6300625 2022-12-22 19:19:06,070 INFO [train.py:894] (3/4) Epoch 7, batch 250, loss[loss=0.2081, simple_loss=0.2946, pruned_loss=0.06074, over 18391.00 frames. ], tot_loss[loss=0.2231, simple_loss=0.3004, pruned_loss=0.07287, over 2658187.24 frames. ], batch size: 53, lr: 1.70e-02, grad_scale: 4.0 2022-12-22 19:19:13,123 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=21292.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:19:18,620 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0_sp0.9 from training. Duration: 24.033375 2022-12-22 19:19:43,247 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7921, 1.7980, 1.9612, 2.0189, 1.9100, 5.0075, 2.3142, 3.1145], device='cuda:3'), covar=tensor([0.3138, 0.1871, 0.1762, 0.1826, 0.1168, 0.0076, 0.1283, 0.0709], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0125, 0.0135, 0.0125, 0.0110, 0.0100, 0.0106, 0.0102], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2022-12-22 19:19:43,265 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21312.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:19:54,505 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.081e+02 4.850e+02 5.968e+02 7.648e+02 1.525e+03, threshold=1.194e+03, percent-clipped=4.0 2022-12-22 19:20:06,490 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21328.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:20:15,604 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0 from training. Duration: 22.905 2022-12-22 19:20:17,060 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp1.1 from training. Duration: 23.4318125 2022-12-22 19:20:21,444 INFO [train.py:894] (3/4) Epoch 7, batch 300, loss[loss=0.2634, simple_loss=0.3321, pruned_loss=0.09738, over 18584.00 frames. ], tot_loss[loss=0.224, simple_loss=0.3016, pruned_loss=0.07321, over 2892418.54 frames. ], batch size: 57, lr: 1.70e-02, grad_scale: 4.0 2022-12-22 19:21:14,040 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21373.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:21:36,820 INFO [train.py:894] (3/4) Epoch 7, batch 350, loss[loss=0.2228, simple_loss=0.2964, pruned_loss=0.07458, over 18671.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.303, pruned_loss=0.07385, over 3074276.73 frames. ], batch size: 48, lr: 1.69e-02, grad_scale: 4.0 2022-12-22 19:21:39,157 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21389.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:21:42,604 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2022-12-22 19:21:52,923 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2022-12-22 19:22:07,400 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21408.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:22:14,534 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp1.1 from training. Duration: 20.82275 2022-12-22 19:22:14,581 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp0.9 from training. Duration: 25.45 2022-12-22 19:22:25,364 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.397e+02 5.112e+02 6.223e+02 8.119e+02 3.223e+03, threshold=1.245e+03, percent-clipped=4.0 2022-12-22 19:22:53,013 INFO [train.py:894] (3/4) Epoch 7, batch 400, loss[loss=0.1877, simple_loss=0.2647, pruned_loss=0.05538, over 18625.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.3019, pruned_loss=0.07377, over 3215743.92 frames. ], batch size: 45, lr: 1.69e-02, grad_scale: 8.0 2022-12-22 19:22:55,091 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5167, 2.2132, 1.3405, 2.4156, 3.0613, 1.5931, 1.7543, 1.1701], device='cuda:3'), covar=tensor([0.1891, 0.1457, 0.1563, 0.0810, 0.0912, 0.1163, 0.1579, 0.1450], device='cuda:3'), in_proj_covar=tensor([0.0226, 0.0198, 0.0191, 0.0176, 0.0236, 0.0182, 0.0197, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 19:23:14,381 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0 from training. Duration: 25.775 2022-12-22 19:23:34,069 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([5.9885, 5.0485, 5.1713, 5.8567, 5.3903, 5.2244, 5.7868, 1.5578], device='cuda:3'), covar=tensor([0.0406, 0.0395, 0.0399, 0.0476, 0.0960, 0.0690, 0.0240, 0.4596], device='cuda:3'), in_proj_covar=tensor([0.0240, 0.0175, 0.0171, 0.0168, 0.0239, 0.0198, 0.0193, 0.0226], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-22 19:23:35,412 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0_sp0.9 from training. Duration: 22.25 2022-12-22 19:23:47,399 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21474.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:24:05,034 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0 from training. Duration: 26.205 2022-12-22 19:24:09,419 INFO [train.py:894] (3/4) Epoch 7, batch 450, loss[loss=0.2064, simple_loss=0.2878, pruned_loss=0.06256, over 18420.00 frames. ], tot_loss[loss=0.2258, simple_loss=0.303, pruned_loss=0.07432, over 3325486.24 frames. ], batch size: 48, lr: 1.69e-02, grad_scale: 8.0 2022-12-22 19:24:21,335 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp0.9 from training. Duration: 30.1555625 2022-12-22 19:24:25,431 WARNING [train.py:1060] (3/4) Exclude cut with ID 1265-135635-0050-6781-0_sp0.9 from training. Duration: 21.8333125 2022-12-22 19:24:34,970 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp1.1 from training. Duration: 20.6545625 2022-12-22 19:24:57,402 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.897e+02 4.831e+02 6.096e+02 7.649e+02 1.561e+03, threshold=1.219e+03, percent-clipped=2.0 2022-12-22 19:25:21,708 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0045-39920-0_sp0.9 from training. Duration: 20.52225 2022-12-22 19:25:22,128 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6695, 1.3109, 1.2923, 1.3799, 1.6989, 1.8401, 1.8278, 1.2774], device='cuda:3'), covar=tensor([0.0362, 0.0302, 0.0502, 0.0277, 0.0263, 0.0321, 0.0255, 0.0301], device='cuda:3'), in_proj_covar=tensor([0.0079, 0.0108, 0.0132, 0.0119, 0.0104, 0.0095, 0.0080, 0.0111], device='cuda:3'), out_proj_covar=tensor([7.4243e-05, 9.8264e-05, 1.2549e-04, 1.1022e-04, 9.9481e-05, 8.4850e-05, 7.3348e-05, 1.0137e-04], device='cuda:3') 2022-12-22 19:25:24,463 INFO [train.py:894] (3/4) Epoch 7, batch 500, loss[loss=0.2309, simple_loss=0.3211, pruned_loss=0.07028, over 18478.00 frames. ], tot_loss[loss=0.2275, simple_loss=0.3049, pruned_loss=0.07504, over 3412568.33 frames. ], batch size: 54, lr: 1.69e-02, grad_scale: 8.0 2022-12-22 19:25:39,318 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp0.9 from training. Duration: 29.1166875 2022-12-22 19:25:41,824 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7717, 2.2251, 1.7717, 2.6683, 2.9679, 1.8389, 2.2702, 1.5819], device='cuda:3'), covar=tensor([0.1769, 0.1588, 0.1400, 0.0810, 0.1331, 0.1065, 0.1483, 0.1333], device='cuda:3'), in_proj_covar=tensor([0.0226, 0.0199, 0.0192, 0.0176, 0.0236, 0.0180, 0.0197, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 19:26:38,662 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133211-0007-59831-0_sp0.9 from training. Duration: 21.388875 2022-12-22 19:26:40,249 INFO [train.py:894] (3/4) Epoch 7, batch 550, loss[loss=0.2141, simple_loss=0.2923, pruned_loss=0.06793, over 18423.00 frames. ], tot_loss[loss=0.2278, simple_loss=0.3049, pruned_loss=0.07538, over 3479095.96 frames. ], batch size: 48, lr: 1.69e-02, grad_scale: 8.0 2022-12-22 19:26:50,943 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21595.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:27:12,880 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.2399, 0.8930, 1.1687, 2.1472, 1.4918, 2.1117, 0.6931, 1.4783], device='cuda:3'), covar=tensor([0.1793, 0.1853, 0.1342, 0.0651, 0.1333, 0.1016, 0.2157, 0.1289], device='cuda:3'), in_proj_covar=tensor([0.0103, 0.0113, 0.0125, 0.0108, 0.0103, 0.0127, 0.0133, 0.0106], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 19:27:15,502 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0 from training. Duration: 22.72 2022-12-22 19:27:15,539 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0_sp0.9 from training. Duration: 22.7444375 2022-12-22 19:27:18,842 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21613.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:27:28,907 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.471e+02 5.026e+02 6.152e+02 7.291e+02 2.360e+03, threshold=1.230e+03, percent-clipped=3.0 2022-12-22 19:27:54,780 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.16 vs. limit=5.0 2022-12-22 19:27:57,046 INFO [train.py:894] (3/4) Epoch 7, batch 600, loss[loss=0.2156, simple_loss=0.3013, pruned_loss=0.06499, over 18597.00 frames. ], tot_loss[loss=0.2281, simple_loss=0.3051, pruned_loss=0.07556, over 3531239.70 frames. ], batch size: 51, lr: 1.68e-02, grad_scale: 8.0 2022-12-22 19:27:58,618 WARNING [train.py:1060] (3/4) Exclude cut with ID 4133-6541-0027-40495-0_sp1.1 from training. Duration: 0.9681875 2022-12-22 19:28:01,648 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0_sp0.9 from training. Duration: 22.3166875 2022-12-22 19:28:08,986 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133212-0015-59917-0_sp0.9 from training. Duration: 21.8166875 2022-12-22 19:28:09,633 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.87 vs. limit=5.0 2022-12-22 19:28:12,706 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2878, 3.6575, 3.5339, 1.4691, 3.6440, 2.7045, 0.4992, 2.5394], device='cuda:3'), covar=tensor([0.2177, 0.0791, 0.1402, 0.3747, 0.0815, 0.1075, 0.5850, 0.1574], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0106, 0.0141, 0.0115, 0.0111, 0.0099, 0.0139, 0.0108], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 19:28:21,567 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3497, 1.2520, 0.8879, 1.5494, 1.4316, 2.8538, 1.0944, 1.2892], device='cuda:3'), covar=tensor([0.1132, 0.2073, 0.1538, 0.1094, 0.1736, 0.0310, 0.1629, 0.1849], device='cuda:3'), in_proj_covar=tensor([0.0080, 0.0088, 0.0082, 0.0084, 0.0098, 0.0075, 0.0089, 0.0083], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 19:28:24,635 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21656.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:28:42,104 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21668.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:28:46,689 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6109, 1.0507, 0.5960, 1.1263, 1.8157, 0.8220, 1.2562, 1.4300], device='cuda:3'), covar=tensor([0.1700, 0.2351, 0.2695, 0.1783, 0.1851, 0.1855, 0.1498, 0.1801], device='cuda:3'), in_proj_covar=tensor([0.0093, 0.0106, 0.0124, 0.0099, 0.0110, 0.0092, 0.0097, 0.0099], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 19:28:50,957 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21674.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:29:07,183 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21684.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:29:13,218 INFO [train.py:894] (3/4) Epoch 7, batch 650, loss[loss=0.1962, simple_loss=0.271, pruned_loss=0.06068, over 18610.00 frames. ], tot_loss[loss=0.2282, simple_loss=0.3048, pruned_loss=0.0758, over 3570657.34 frames. ], batch size: 41, lr: 1.68e-02, grad_scale: 8.0 2022-12-22 19:29:36,894 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2022-12-22 19:29:45,247 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21708.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:29:56,834 WARNING [train.py:1060] (3/4) Exclude cut with ID 4957-30119-0041-23990-0_sp0.9 from training. Duration: 20.22775 2022-12-22 19:30:02,248 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.404e+02 4.992e+02 6.426e+02 7.914e+02 1.249e+03, threshold=1.285e+03, percent-clipped=1.0 2022-12-22 19:30:29,427 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3515, 2.0798, 1.4315, 2.3351, 1.7880, 1.7679, 1.9728, 2.4481], device='cuda:3'), covar=tensor([0.1346, 0.2158, 0.1370, 0.1873, 0.2264, 0.0798, 0.1916, 0.0500], device='cuda:3'), in_proj_covar=tensor([0.0258, 0.0243, 0.0213, 0.0319, 0.0230, 0.0202, 0.0243, 0.0171], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 19:30:30,282 INFO [train.py:894] (3/4) Epoch 7, batch 700, loss[loss=0.2411, simple_loss=0.327, pruned_loss=0.07763, over 18623.00 frames. ], tot_loss[loss=0.229, simple_loss=0.3057, pruned_loss=0.07612, over 3601952.89 frames. ], batch size: 62, lr: 1.68e-02, grad_scale: 8.0 2022-12-22 19:30:41,990 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp1.1 from training. Duration: 24.67275 2022-12-22 19:30:57,385 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=21756.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:31:07,754 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21763.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:31:10,211 WARNING [train.py:1060] (3/4) Exclude cut with ID 3082-165428-0081-50734-0_sp0.9 from training. Duration: 21.8055625 2022-12-22 19:31:24,508 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21774.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:31:45,977 INFO [train.py:894] (3/4) Epoch 7, batch 750, loss[loss=0.2244, simple_loss=0.3066, pruned_loss=0.07105, over 18603.00 frames. ], tot_loss[loss=0.2282, simple_loss=0.3056, pruned_loss=0.07542, over 3627315.72 frames. ], batch size: 98, lr: 1.68e-02, grad_scale: 8.0 2022-12-22 19:31:47,687 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0_sp0.9 from training. Duration: 22.6666875 2022-12-22 19:32:34,810 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.290e+02 4.818e+02 5.879e+02 7.436e+02 2.235e+03, threshold=1.176e+03, percent-clipped=3.0 2022-12-22 19:32:38,399 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=21822.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:32:42,035 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21824.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:32:49,478 WARNING [train.py:1060] (3/4) Exclude cut with ID 2411-132532-0017-82279-0_sp1.1 from training. Duration: 0.9681875 2022-12-22 19:33:02,681 INFO [train.py:894] (3/4) Epoch 7, batch 800, loss[loss=0.2221, simple_loss=0.3079, pruned_loss=0.06816, over 18639.00 frames. ], tot_loss[loss=0.2297, simple_loss=0.3068, pruned_loss=0.07632, over 3645999.16 frames. ], batch size: 53, lr: 1.68e-02, grad_scale: 8.0 2022-12-22 19:33:16,761 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0 from training. Duration: 22.485 2022-12-22 19:33:21,566 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21850.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:33:35,469 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2022-12-22 19:33:44,218 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2022-12-22 19:33:54,538 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp1.1 from training. Duration: 23.82275 2022-12-22 19:34:08,666 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0 from training. Duration: 20.77 2022-12-22 19:34:14,778 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0_sp0.9 from training. Duration: 24.088875 2022-12-22 19:34:19,351 INFO [train.py:894] (3/4) Epoch 7, batch 850, loss[loss=0.1945, simple_loss=0.2706, pruned_loss=0.05918, over 18472.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.3065, pruned_loss=0.07607, over 3661243.47 frames. ], batch size: 43, lr: 1.67e-02, grad_scale: 8.0 2022-12-22 19:34:33,861 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.9230, 2.3819, 1.8464, 0.9086, 1.8480, 2.2440, 1.7514, 2.0527], device='cuda:3'), covar=tensor([0.0466, 0.0467, 0.1187, 0.1626, 0.1382, 0.1075, 0.1226, 0.0752], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0170, 0.0195, 0.0188, 0.0194, 0.0176, 0.0189, 0.0176], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 19:34:46,489 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp1.1 from training. Duration: 20.4409375 2022-12-22 19:34:54,193 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21911.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:35:07,549 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.100e+02 5.077e+02 6.142e+02 7.793e+02 1.798e+03, threshold=1.228e+03, percent-clipped=5.0 2022-12-22 19:35:34,840 INFO [train.py:894] (3/4) Epoch 7, batch 900, loss[loss=0.2325, simple_loss=0.3116, pruned_loss=0.07673, over 18704.00 frames. ], tot_loss[loss=0.2284, simple_loss=0.3058, pruned_loss=0.07548, over 3672998.74 frames. ], batch size: 60, lr: 1.67e-02, grad_scale: 8.0 2022-12-22 19:35:54,493 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21951.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:36:04,153 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0_sp0.9 from training. Duration: 22.511125 2022-12-22 19:36:05,531 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0 from training. Duration: 20.675 2022-12-22 19:36:19,584 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21968.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:36:20,832 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21969.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:36:22,547 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21970.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:36:40,555 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2022-12-22 19:36:44,883 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21984.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:36:50,568 INFO [train.py:894] (3/4) Epoch 7, batch 950, loss[loss=0.2163, simple_loss=0.2957, pruned_loss=0.06847, over 18557.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.3051, pruned_loss=0.07503, over 3681954.27 frames. ], batch size: 49, lr: 1.67e-02, grad_scale: 8.0 2022-12-22 19:37:34,087 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2022-12-22 19:37:36,855 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22016.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:37:42,502 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.707e+02 4.665e+02 5.685e+02 7.143e+02 1.158e+03, threshold=1.137e+03, percent-clipped=0.0 2022-12-22 19:37:51,819 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp0.9 from training. Duration: 24.9833125 2022-12-22 19:37:59,649 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22031.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:38:01,076 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22032.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:38:09,056 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0382, 2.2773, 1.6300, 1.1062, 3.1677, 2.6538, 1.8280, 1.6786], device='cuda:3'), covar=tensor([0.0458, 0.0356, 0.0607, 0.0785, 0.0094, 0.0267, 0.0632, 0.0841], device='cuda:3'), in_proj_covar=tensor([0.0122, 0.0112, 0.0127, 0.0119, 0.0076, 0.0112, 0.0134, 0.0147], device='cuda:3'), out_proj_covar=tensor([1.5085e-04, 1.4023e-04, 1.5574e-04, 1.4707e-04, 9.6661e-05, 1.3525e-04, 1.6528e-04, 1.8157e-04], device='cuda:3') 2022-12-22 19:38:10,122 INFO [train.py:894] (3/4) Epoch 7, batch 1000, loss[loss=0.2259, simple_loss=0.3054, pruned_loss=0.07318, over 18592.00 frames. ], tot_loss[loss=0.2278, simple_loss=0.3056, pruned_loss=0.07502, over 3689326.30 frames. ], batch size: 51, lr: 1.67e-02, grad_scale: 8.0 2022-12-22 19:38:20,448 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0 from training. Duration: 27.14 2022-12-22 19:38:34,603 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0 from training. Duration: 22.44 2022-12-22 19:39:16,340 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2022-12-22 19:39:21,708 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4573, 1.2932, 1.3484, 1.2006, 0.9643, 2.1563, 0.9943, 1.4122], device='cuda:3'), covar=tensor([0.3365, 0.2054, 0.1920, 0.2063, 0.1325, 0.0373, 0.1548, 0.0936], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0124, 0.0135, 0.0124, 0.0109, 0.0098, 0.0105, 0.0102], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2022-12-22 19:39:22,811 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2022-12-22 19:39:26,394 INFO [train.py:894] (3/4) Epoch 7, batch 1050, loss[loss=0.2, simple_loss=0.2774, pruned_loss=0.06126, over 18419.00 frames. ], tot_loss[loss=0.2273, simple_loss=0.305, pruned_loss=0.07474, over 3694346.46 frames. ], batch size: 48, lr: 1.67e-02, grad_scale: 8.0 2022-12-22 19:39:53,359 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0060-62364-0_sp0.9 from training. Duration: 21.361125 2022-12-22 19:40:01,128 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp1.1 from training. Duration: 27.0318125 2022-12-22 19:40:08,710 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp0.9 from training. Duration: 28.638875 2022-12-22 19:40:13,884 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22119.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:40:15,138 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.278e+02 5.032e+02 6.199e+02 7.170e+02 1.495e+03, threshold=1.240e+03, percent-clipped=4.0 2022-12-22 19:40:25,147 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0 from training. Duration: 20.4 2022-12-22 19:40:29,949 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22130.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:40:42,005 INFO [train.py:894] (3/4) Epoch 7, batch 1100, loss[loss=0.2405, simple_loss=0.3234, pruned_loss=0.07884, over 18454.00 frames. ], tot_loss[loss=0.2275, simple_loss=0.305, pruned_loss=0.07495, over 3698327.66 frames. ], batch size: 64, lr: 1.67e-02, grad_scale: 8.0 2022-12-22 19:40:54,273 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3872, 2.6871, 2.5723, 1.5382, 2.7648, 2.6686, 1.8111, 3.4207], device='cuda:3'), covar=tensor([0.1143, 0.1194, 0.1459, 0.2060, 0.0828, 0.1243, 0.1916, 0.0429], device='cuda:3'), in_proj_covar=tensor([0.0200, 0.0186, 0.0195, 0.0188, 0.0179, 0.0206, 0.0196, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 19:40:58,309 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0 from training. Duration: 20.025 2022-12-22 19:40:59,726 WARNING [train.py:1060] (3/4) Exclude cut with ID 2364-131735-0112-64612-0_sp0.9 from training. Duration: 20.488875 2022-12-22 19:41:04,524 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0 from training. Duration: 29.735 2022-12-22 19:41:40,825 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1699, 1.4510, 0.7336, 1.7778, 2.2839, 1.5939, 2.0431, 2.2126], device='cuda:3'), covar=tensor([0.2158, 0.2968, 0.3312, 0.1937, 0.2279, 0.2208, 0.1753, 0.2289], device='cuda:3'), in_proj_covar=tensor([0.0093, 0.0106, 0.0124, 0.0100, 0.0111, 0.0094, 0.0098, 0.0100], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 19:41:40,867 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22176.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:41:58,850 INFO [train.py:894] (3/4) Epoch 7, batch 1150, loss[loss=0.205, simple_loss=0.2787, pruned_loss=0.06561, over 18502.00 frames. ], tot_loss[loss=0.226, simple_loss=0.3034, pruned_loss=0.07426, over 3701611.97 frames. ], batch size: 41, lr: 1.66e-02, grad_scale: 8.0 2022-12-22 19:42:03,752 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22191.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:42:11,860 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2022-12-22 19:42:24,316 WARNING [train.py:1060] (3/4) Exclude cut with ID 7276-92427-0014-12983-0_sp0.9 from training. Duration: 21.3055625 2022-12-22 19:42:26,479 WARNING [train.py:1060] (3/4) Exclude cut with ID 1025-75365-0008-79168-0_sp0.9 from training. Duration: 22.0666875 2022-12-22 19:42:26,578 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22206.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:42:40,383 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22215.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:42:47,745 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.336e+02 4.819e+02 6.126e+02 7.928e+02 2.801e+03, threshold=1.225e+03, percent-clipped=6.0 2022-12-22 19:43:14,471 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22237.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:43:14,482 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22237.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:43:15,638 INFO [train.py:894] (3/4) Epoch 7, batch 1200, loss[loss=0.211, simple_loss=0.2908, pruned_loss=0.06558, over 18636.00 frames. ], tot_loss[loss=0.2264, simple_loss=0.3038, pruned_loss=0.07447, over 3705130.74 frames. ], batch size: 53, lr: 1.66e-02, grad_scale: 8.0 2022-12-22 19:43:34,773 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22251.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:43:45,343 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.0707, 2.4648, 2.8882, 0.6697, 2.2701, 3.1010, 2.1157, 2.5304], device='cuda:3'), covar=tensor([0.0846, 0.0323, 0.0268, 0.0475, 0.0439, 0.0221, 0.0321, 0.0540], device='cuda:3'), in_proj_covar=tensor([0.0121, 0.0138, 0.0096, 0.0116, 0.0124, 0.0101, 0.0123, 0.0120], device='cuda:3'), out_proj_covar=tensor([1.1040e-04, 1.2868e-04, 8.9799e-05, 1.0649e-04, 1.1520e-04, 9.3678e-05, 1.1640e-04, 1.1269e-04], device='cuda:3') 2022-12-22 19:44:02,560 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22269.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:44:13,714 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22276.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:44:16,022 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0 from training. Duration: 20.26 2022-12-22 19:44:28,778 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0030-9324-0_sp0.9 from training. Duration: 21.3444375 2022-12-22 19:44:30,251 INFO [train.py:894] (3/4) Epoch 7, batch 1250, loss[loss=0.2778, simple_loss=0.3396, pruned_loss=0.108, over 18594.00 frames. ], tot_loss[loss=0.2259, simple_loss=0.3035, pruned_loss=0.07416, over 3706252.24 frames. ], batch size: 175, lr: 1.66e-02, grad_scale: 8.0 2022-12-22 19:44:44,932 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22298.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:44:46,118 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22299.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:45:14,636 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22317.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:45:18,808 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.340e+02 4.996e+02 6.161e+02 7.748e+02 1.648e+03, threshold=1.232e+03, percent-clipped=4.0 2022-12-22 19:45:25,238 WARNING [train.py:1060] (3/4) Exclude cut with ID 497-129325-0061-62254-0_sp1.1 from training. Duration: 0.97725 2022-12-22 19:45:28,512 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22326.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:45:45,641 INFO [train.py:894] (3/4) Epoch 7, batch 1300, loss[loss=0.231, simple_loss=0.3046, pruned_loss=0.07868, over 18516.00 frames. ], tot_loss[loss=0.2249, simple_loss=0.3032, pruned_loss=0.07327, over 3707677.15 frames. ], batch size: 47, lr: 1.66e-02, grad_scale: 8.0 2022-12-22 19:45:59,630 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2022-12-22 19:46:08,644 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0_sp0.9 from training. Duration: 22.97225 2022-12-22 19:46:39,091 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0047-39922-0_sp0.9 from training. Duration: 21.97775 2022-12-22 19:46:49,643 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.5879, 2.7895, 2.5673, 1.3157, 2.3270, 2.2624, 1.8598, 2.2853], device='cuda:3'), covar=tensor([0.0456, 0.0550, 0.1222, 0.1674, 0.1809, 0.1197, 0.1338, 0.0929], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0173, 0.0197, 0.0191, 0.0198, 0.0180, 0.0192, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 19:46:53,710 WARNING [train.py:1060] (3/4) Exclude cut with ID 1112-1043-0006-89194-0_sp0.9 from training. Duration: 21.8333125 2022-12-22 19:46:57,153 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22385.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:47:00,753 INFO [train.py:894] (3/4) Epoch 7, batch 1350, loss[loss=0.2628, simple_loss=0.3475, pruned_loss=0.08903, over 18470.00 frames. ], tot_loss[loss=0.2239, simple_loss=0.3024, pruned_loss=0.07266, over 3708901.42 frames. ], batch size: 54, lr: 1.66e-02, grad_scale: 8.0 2022-12-22 19:47:03,428 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0 from training. Duration: 20.47 2022-12-22 19:47:47,895 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22419.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:47:49,174 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.961e+02 4.936e+02 5.995e+02 7.767e+02 1.533e+03, threshold=1.199e+03, percent-clipped=4.0 2022-12-22 19:47:52,831 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.8223, 2.5771, 2.0227, 0.9265, 1.8211, 2.1477, 1.5355, 1.9952], device='cuda:3'), covar=tensor([0.0493, 0.0443, 0.1162, 0.1643, 0.1351, 0.1222, 0.1519, 0.0775], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0174, 0.0199, 0.0193, 0.0200, 0.0181, 0.0195, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 19:48:10,038 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0037-39912-0_sp0.9 from training. Duration: 20.67225 2022-12-22 19:48:15,714 INFO [train.py:894] (3/4) Epoch 7, batch 1400, loss[loss=0.1889, simple_loss=0.2654, pruned_loss=0.05617, over 18467.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.3014, pruned_loss=0.07225, over 3709743.87 frames. ], batch size: 43, lr: 1.66e-02, grad_scale: 8.0 2022-12-22 19:48:28,185 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp0.9 from training. Duration: 25.2444375 2022-12-22 19:48:28,572 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22446.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:48:29,998 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6847, 1.4728, 1.3216, 2.0062, 1.6741, 3.4590, 1.2513, 1.6008], device='cuda:3'), covar=tensor([0.0955, 0.1804, 0.1202, 0.0992, 0.1470, 0.0239, 0.1512, 0.1524], device='cuda:3'), in_proj_covar=tensor([0.0079, 0.0088, 0.0081, 0.0082, 0.0097, 0.0074, 0.0089, 0.0082], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2022-12-22 19:48:52,418 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0021-76797-0_sp0.9 from training. Duration: 21.1445 2022-12-22 19:49:01,417 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22467.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:49:29,874 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22486.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:49:32,421 INFO [train.py:894] (3/4) Epoch 7, batch 1450, loss[loss=0.2057, simple_loss=0.2943, pruned_loss=0.05854, over 18524.00 frames. ], tot_loss[loss=0.2241, simple_loss=0.3023, pruned_loss=0.073, over 3711314.57 frames. ], batch size: 55, lr: 1.65e-02, grad_scale: 8.0 2022-12-22 19:49:59,537 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22506.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:50:02,524 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8018, 1.4849, 1.1681, 0.2372, 1.1744, 1.5305, 1.1924, 1.6243], device='cuda:3'), covar=tensor([0.0562, 0.0483, 0.0878, 0.1537, 0.1005, 0.1486, 0.1634, 0.0537], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0173, 0.0197, 0.0192, 0.0199, 0.0181, 0.0193, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 19:50:10,387 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp0.9 from training. Duration: 33.038875 2022-12-22 19:50:21,667 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.790e+02 4.574e+02 5.514e+02 7.111e+02 1.060e+03, threshold=1.103e+03, percent-clipped=0.0 2022-12-22 19:50:39,587 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22532.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:50:46,394 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64291-0000-16059-0_sp0.9 from training. Duration: 20.0944375 2022-12-22 19:50:47,905 INFO [train.py:894] (3/4) Epoch 7, batch 1500, loss[loss=0.2159, simple_loss=0.3089, pruned_loss=0.0614, over 18515.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.3031, pruned_loss=0.07322, over 3712024.42 frames. ], batch size: 55, lr: 1.65e-02, grad_scale: 8.0 2022-12-22 19:51:01,908 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp1.1 from training. Duration: 20.4 2022-12-22 19:51:09,231 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0 from training. Duration: 20.085 2022-12-22 19:51:12,169 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22554.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:51:22,430 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0_sp0.9 from training. Duration: 23.07775 2022-12-22 19:51:39,104 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22571.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:52:03,709 INFO [train.py:894] (3/4) Epoch 7, batch 1550, loss[loss=0.2707, simple_loss=0.3418, pruned_loss=0.09983, over 18671.00 frames. ], tot_loss[loss=0.2249, simple_loss=0.3033, pruned_loss=0.07327, over 3711371.50 frames. ], batch size: 100, lr: 1.65e-02, grad_scale: 8.0 2022-12-22 19:52:05,463 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp0.9 from training. Duration: 24.9333125 2022-12-22 19:52:12,270 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22593.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:52:55,025 WARNING [train.py:1060] (3/4) Exclude cut with ID 5118-111612-0016-124680-0_sp0.9 from training. Duration: 20.388875 2022-12-22 19:52:56,542 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.322e+02 5.232e+02 6.595e+02 8.156e+02 1.283e+03, threshold=1.319e+03, percent-clipped=2.0 2022-12-22 19:53:02,292 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp1.1 from training. Duration: 20.3590625 2022-12-22 19:53:05,195 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22626.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:53:21,286 INFO [train.py:894] (3/4) Epoch 7, batch 1600, loss[loss=0.2236, simple_loss=0.3054, pruned_loss=0.07095, over 18579.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.303, pruned_loss=0.07309, over 3712542.43 frames. ], batch size: 51, lr: 1.65e-02, grad_scale: 8.0 2022-12-22 19:53:23,194 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22639.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:53:39,880 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5122, 1.0783, 1.7039, 3.0296, 2.0051, 2.1654, 0.4636, 1.8157], device='cuda:3'), covar=tensor([0.1801, 0.1918, 0.1551, 0.0437, 0.1220, 0.1435, 0.2614, 0.1459], device='cuda:3'), in_proj_covar=tensor([0.0105, 0.0114, 0.0126, 0.0108, 0.0104, 0.0130, 0.0131, 0.0108], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 19:54:06,755 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0_sp1.1 from training. Duration: 0.836375 2022-12-22 19:54:16,997 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22674.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:54:37,274 INFO [train.py:894] (3/4) Epoch 7, batch 1650, loss[loss=0.2509, simple_loss=0.3093, pruned_loss=0.09623, over 18396.00 frames. ], tot_loss[loss=0.2274, simple_loss=0.3048, pruned_loss=0.07502, over 3712889.65 frames. ], batch size: 46, lr: 1.65e-02, grad_scale: 8.0 2022-12-22 19:54:51,056 WARNING [train.py:1060] (3/4) Exclude cut with ID 8565-290391-0049-67394-0_sp0.9 from training. Duration: 21.3166875 2022-12-22 19:54:55,854 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22700.0, num_to_drop=1, layers_to_drop={2} 2022-12-22 19:55:00,907 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7094, 2.5238, 1.5631, 3.0965, 3.1823, 1.5695, 2.3725, 1.3563], device='cuda:3'), covar=tensor([0.1697, 0.1569, 0.1364, 0.0649, 0.1320, 0.1118, 0.1438, 0.1374], device='cuda:3'), in_proj_covar=tensor([0.0225, 0.0198, 0.0189, 0.0176, 0.0239, 0.0182, 0.0199, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 19:55:12,524 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2022-12-22 19:55:22,668 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0029-104863-0_sp0.9 from training. Duration: 22.1055625 2022-12-22 19:55:26,849 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.320e+02 5.398e+02 6.482e+02 7.925e+02 1.712e+03, threshold=1.296e+03, percent-clipped=2.0 2022-12-22 19:55:32,306 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp1.1 from training. Duration: 21.77725 2022-12-22 19:55:53,842 INFO [train.py:894] (3/4) Epoch 7, batch 1700, loss[loss=0.239, simple_loss=0.3071, pruned_loss=0.08541, over 18516.00 frames. ], tot_loss[loss=0.2303, simple_loss=0.3062, pruned_loss=0.07719, over 3712992.63 frames. ], batch size: 47, lr: 1.64e-02, grad_scale: 8.0 2022-12-22 19:55:53,879 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp0.9 from training. Duration: 27.8166875 2022-12-22 19:55:58,462 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22741.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:56:16,559 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp1.1 from training. Duration: 22.5090625 2022-12-22 19:56:24,478 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0 from training. Duration: 25.035 2022-12-22 19:56:32,526 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.83 vs. limit=5.0 2022-12-22 19:56:41,781 WARNING [train.py:1060] (3/4) Exclude cut with ID 774-127930-0014-10412-0_sp1.1 from training. Duration: 0.95 2022-12-22 19:56:43,618 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6070, 2.3225, 1.6773, 2.8669, 2.8438, 1.6298, 1.8606, 1.3161], device='cuda:3'), covar=tensor([0.1722, 0.1453, 0.1223, 0.0667, 0.1285, 0.1081, 0.1656, 0.1366], device='cuda:3'), in_proj_covar=tensor([0.0228, 0.0202, 0.0192, 0.0179, 0.0243, 0.0183, 0.0203, 0.0181], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 19:57:02,190 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp0.9 from training. Duration: 0.92225 2022-12-22 19:57:07,024 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22786.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:57:09,509 INFO [train.py:894] (3/4) Epoch 7, batch 1750, loss[loss=0.2408, simple_loss=0.3104, pruned_loss=0.08562, over 18697.00 frames. ], tot_loss[loss=0.2348, simple_loss=0.309, pruned_loss=0.08033, over 3713988.85 frames. ], batch size: 62, lr: 1.64e-02, grad_scale: 8.0 2022-12-22 19:57:18,616 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22794.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:57:27,840 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0 from training. Duration: 21.97 2022-12-22 19:57:48,439 WARNING [train.py:1060] (3/4) Exclude cut with ID 7492-105653-0055-62765-0_sp0.9 from training. Duration: 21.97225 2022-12-22 19:57:48,479 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp0.9 from training. Duration: 25.3333125 2022-12-22 19:57:58,564 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.602e+02 6.360e+02 7.801e+02 9.201e+02 2.109e+03, threshold=1.560e+03, percent-clipped=7.0 2022-12-22 19:58:00,214 WARNING [train.py:1060] (3/4) Exclude cut with ID 5172-29468-0015-19128-0_sp0.9 from training. Duration: 21.5055625 2022-12-22 19:58:01,043 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2022-12-22 19:58:06,203 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22825.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:58:07,445 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp1.1 from training. Duration: 20.72725 2022-12-22 19:58:17,322 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22832.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:58:20,134 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22834.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:58:26,162 INFO [train.py:894] (3/4) Epoch 7, batch 1800, loss[loss=0.23, simple_loss=0.3034, pruned_loss=0.07825, over 18503.00 frames. ], tot_loss[loss=0.2397, simple_loss=0.3116, pruned_loss=0.08385, over 3714193.99 frames. ], batch size: 52, lr: 1.64e-02, grad_scale: 8.0 2022-12-22 19:58:39,331 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp0.9 from training. Duration: 26.32775 2022-12-22 19:58:55,881 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22855.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:59:13,707 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0 from training. Duration: 20.025 2022-12-22 19:59:19,526 WARNING [train.py:1060] (3/4) Exclude cut with ID 6709-74022-0004-86860-0_sp1.1 from training. Duration: 0.9409375 2022-12-22 19:59:19,535 WARNING [train.py:1060] (3/4) Exclude cut with ID 4757-1811-0023-62229-0_sp0.9 from training. Duration: 21.37775 2022-12-22 19:59:20,640 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22871.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:59:33,387 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.30 vs. limit=5.0 2022-12-22 19:59:34,093 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22880.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:59:34,244 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4827, 1.0890, 1.8552, 2.7392, 1.9293, 2.2206, 0.6606, 1.7319], device='cuda:3'), covar=tensor([0.1810, 0.1969, 0.1465, 0.0644, 0.1341, 0.1286, 0.2519, 0.1513], device='cuda:3'), in_proj_covar=tensor([0.0106, 0.0115, 0.0128, 0.0111, 0.0107, 0.0130, 0.0132, 0.0110], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 19:59:42,806 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0004-25974-0_sp0.9 from training. Duration: 21.17225 2022-12-22 19:59:42,816 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp0.9 from training. Duration: 27.511125 2022-12-22 19:59:43,333 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22886.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 19:59:45,885 INFO [train.py:894] (3/4) Epoch 7, batch 1850, loss[loss=0.2508, simple_loss=0.3216, pruned_loss=0.09003, over 18641.00 frames. ], tot_loss[loss=0.2429, simple_loss=0.3134, pruned_loss=0.08616, over 3713875.57 frames. ], batch size: 69, lr: 1.64e-02, grad_scale: 8.0 2022-12-22 19:59:50,058 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3787, 2.4416, 1.6799, 3.0733, 2.6206, 1.4740, 1.8999, 1.1265], device='cuda:3'), covar=tensor([0.2038, 0.1670, 0.1384, 0.0708, 0.1813, 0.1315, 0.1822, 0.1537], device='cuda:3'), in_proj_covar=tensor([0.0227, 0.0201, 0.0192, 0.0179, 0.0242, 0.0182, 0.0201, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 19:59:54,123 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22893.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:00:15,936 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0 from training. Duration: 22.8 2022-12-22 20:00:20,726 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0 from training. Duration: 22.585 2022-12-22 20:00:34,570 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22919.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:00:35,824 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.355e+02 5.680e+02 7.296e+02 9.676e+02 2.623e+03, threshold=1.459e+03, percent-clipped=2.0 2022-12-22 20:00:51,927 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0001-146967-0_sp0.9 from training. Duration: 22.0166875 2022-12-22 20:01:02,714 INFO [train.py:894] (3/4) Epoch 7, batch 1900, loss[loss=0.2161, simple_loss=0.2776, pruned_loss=0.07729, over 18674.00 frames. ], tot_loss[loss=0.2451, simple_loss=0.3142, pruned_loss=0.08801, over 3714349.41 frames. ], batch size: 41, lr: 1.64e-02, grad_scale: 8.0 2022-12-22 20:01:07,072 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22941.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:01:08,624 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp1.1 from training. Duration: 24.395375 2022-12-22 20:01:16,815 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp0.9 from training. Duration: 27.47775 2022-12-22 20:01:19,732 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp0.9 from training. Duration: 24.8833125 2022-12-22 20:01:22,710 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0 from training. Duration: 23.39 2022-12-22 20:01:29,496 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp0.9 from training. Duration: 28.72225 2022-12-22 20:01:38,106 WARNING [train.py:1060] (3/4) Exclude cut with ID 585-294811-0110-133686-0_sp0.9 from training. Duration: 20.8944375 2022-12-22 20:01:52,335 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0_sp0.9 from training. Duration: 23.8444375 2022-12-22 20:02:17,685 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0 from training. Duration: 25.85 2022-12-22 20:02:17,693 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0 from training. Duration: 21.39 2022-12-22 20:02:18,903 INFO [train.py:894] (3/4) Epoch 7, batch 1950, loss[loss=0.2709, simple_loss=0.3387, pruned_loss=0.1016, over 18679.00 frames. ], tot_loss[loss=0.2464, simple_loss=0.3145, pruned_loss=0.08916, over 3713219.27 frames. ], batch size: 98, lr: 1.64e-02, grad_scale: 8.0 2022-12-22 20:02:28,406 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0 from training. Duration: 27.92 2022-12-22 20:02:31,602 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22995.0, num_to_drop=1, layers_to_drop={3} 2022-12-22 20:02:43,841 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4657, 1.0543, 1.5447, 2.5917, 1.7596, 2.2157, 0.6326, 1.5895], device='cuda:3'), covar=tensor([0.1821, 0.2008, 0.1572, 0.0606, 0.1383, 0.1111, 0.2481, 0.1639], device='cuda:3'), in_proj_covar=tensor([0.0106, 0.0115, 0.0128, 0.0111, 0.0106, 0.0129, 0.0133, 0.0109], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 20:02:58,670 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0039-130165-0_sp0.9 from training. Duration: 20.661125 2022-12-22 20:03:08,548 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.141e+02 5.951e+02 7.381e+02 9.078e+02 1.762e+03, threshold=1.476e+03, percent-clipped=3.0 2022-12-22 20:03:22,419 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0043-15874-0_sp0.9 from training. Duration: 20.07225 2022-12-22 20:03:31,666 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0 from training. Duration: 21.01 2022-12-22 20:03:35,799 INFO [train.py:894] (3/4) Epoch 7, batch 2000, loss[loss=0.2458, simple_loss=0.3089, pruned_loss=0.09129, over 18416.00 frames. ], tot_loss[loss=0.2475, simple_loss=0.3147, pruned_loss=0.09016, over 3713045.96 frames. ], batch size: 48, lr: 1.63e-02, grad_scale: 8.0 2022-12-22 20:03:41,230 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23041.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:03:51,568 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2022-12-22 20:04:38,383 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0 from training. Duration: 20.65 2022-12-22 20:04:48,202 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0 from training. Duration: 21.46 2022-12-22 20:04:52,429 INFO [train.py:894] (3/4) Epoch 7, batch 2050, loss[loss=0.2481, simple_loss=0.3165, pruned_loss=0.08981, over 18457.00 frames. ], tot_loss[loss=0.2472, simple_loss=0.3138, pruned_loss=0.09028, over 3712681.35 frames. ], batch size: 50, lr: 1.63e-02, grad_scale: 8.0 2022-12-22 20:04:54,122 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=23089.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:05:24,175 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23108.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:05:27,721 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2022-12-22 20:05:34,110 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0 from training. Duration: 0.92 2022-12-22 20:05:40,045 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0_sp0.9 from training. Duration: 23.7666875 2022-12-22 20:05:41,555 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.719e+02 5.993e+02 7.128e+02 8.967e+02 1.938e+03, threshold=1.426e+03, percent-clipped=4.0 2022-12-22 20:06:08,564 INFO [train.py:894] (3/4) Epoch 7, batch 2100, loss[loss=0.2451, simple_loss=0.3203, pruned_loss=0.08497, over 18544.00 frames. ], tot_loss[loss=0.2484, simple_loss=0.3152, pruned_loss=0.09085, over 3713743.82 frames. ], batch size: 95, lr: 1.63e-02, grad_scale: 8.0 2022-12-22 20:06:19,713 WARNING [train.py:1060] (3/4) Exclude cut with ID 8544-281189-0060-101339-0_sp0.9 from training. Duration: 20.861125 2022-12-22 20:06:27,338 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23150.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:06:28,906 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0_sp0.9 from training. Duration: 22.711125 2022-12-22 20:06:55,901 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23169.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:07:11,686 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp1.1 from training. Duration: 22.986375 2022-12-22 20:07:14,626 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23181.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:07:19,430 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.34 vs. limit=5.0 2022-12-22 20:07:25,726 INFO [train.py:894] (3/4) Epoch 7, batch 2150, loss[loss=0.2306, simple_loss=0.3001, pruned_loss=0.08059, over 18571.00 frames. ], tot_loss[loss=0.2485, simple_loss=0.3149, pruned_loss=0.09111, over 3714856.54 frames. ], batch size: 49, lr: 1.63e-02, grad_scale: 16.0 2022-12-22 20:07:27,243 WARNING [train.py:1060] (3/4) Exclude cut with ID 8040-260924-0003-80960-0_sp0.9 from training. Duration: 22.07225 2022-12-22 20:07:31,902 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0045-26330-0_sp0.9 from training. Duration: 20.3055625 2022-12-22 20:07:33,610 WARNING [train.py:1060] (3/4) Exclude cut with ID 6356-271890-0060-94317-0_sp0.9 from training. Duration: 20.72225 2022-12-22 20:07:52,350 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp1.1 from training. Duration: 22.4818125 2022-12-22 20:08:10,114 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23217.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:08:13,986 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.135e+02 6.077e+02 7.648e+02 9.193e+02 2.501e+03, threshold=1.530e+03, percent-clipped=6.0 2022-12-22 20:08:17,493 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp0.9 from training. Duration: 25.0944375 2022-12-22 20:08:21,951 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0 from training. Duration: 21.515 2022-12-22 20:08:27,938 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23229.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:08:29,035 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp0.9 from training. Duration: 27.02225 2022-12-22 20:08:34,028 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0010-62480-0_sp0.9 from training. Duration: 22.22225 2022-12-22 20:08:40,977 INFO [train.py:894] (3/4) Epoch 7, batch 2200, loss[loss=0.2457, simple_loss=0.3213, pruned_loss=0.085, over 18507.00 frames. ], tot_loss[loss=0.2495, simple_loss=0.3152, pruned_loss=0.09185, over 3714650.69 frames. ], batch size: 58, lr: 1.63e-02, grad_scale: 16.0 2022-12-22 20:08:42,570 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0085-44554-0_sp0.9 from training. Duration: 20.85 2022-12-22 20:09:14,865 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0 from training. Duration: 21.54 2022-12-22 20:09:20,600 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp1.1 from training. Duration: 20.5318125 2022-12-22 20:09:30,705 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0012-134311-0_sp0.9 from training. Duration: 21.9333125 2022-12-22 20:09:43,406 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23278.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:09:58,313 INFO [train.py:894] (3/4) Epoch 7, batch 2250, loss[loss=0.2456, simple_loss=0.3094, pruned_loss=0.0909, over 18406.00 frames. ], tot_loss[loss=0.249, simple_loss=0.3149, pruned_loss=0.09156, over 3714573.83 frames. ], batch size: 46, lr: 1.63e-02, grad_scale: 16.0 2022-12-22 20:10:01,569 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23290.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:10:09,208 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23295.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:10:18,260 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0025-130151-0_sp0.9 from training. Duration: 21.7944375 2022-12-22 20:10:30,219 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0_sp0.9 from training. Duration: 22.4666875 2022-12-22 20:10:37,755 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0 from training. Duration: 21.635 2022-12-22 20:10:46,147 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0_sp0.9 from training. Duration: 24.038875 2022-12-22 20:10:47,638 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.847e+02 6.548e+02 7.944e+02 9.504e+02 1.304e+03, threshold=1.589e+03, percent-clipped=0.0 2022-12-22 20:11:12,365 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2022-12-22 20:11:14,341 INFO [train.py:894] (3/4) Epoch 7, batch 2300, loss[loss=0.2777, simple_loss=0.3425, pruned_loss=0.1064, over 18475.00 frames. ], tot_loss[loss=0.2511, simple_loss=0.3166, pruned_loss=0.09282, over 3714719.18 frames. ], batch size: 54, lr: 1.62e-02, grad_scale: 16.0 2022-12-22 20:11:22,214 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=23343.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:11:27,061 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp1.1 from training. Duration: 21.786375 2022-12-22 20:11:37,239 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0 from training. Duration: 20.22 2022-12-22 20:12:31,221 INFO [train.py:894] (3/4) Epoch 7, batch 2350, loss[loss=0.2531, simple_loss=0.3199, pruned_loss=0.09319, over 18670.00 frames. ], tot_loss[loss=0.2514, simple_loss=0.3167, pruned_loss=0.09311, over 3714866.46 frames. ], batch size: 60, lr: 1.62e-02, grad_scale: 16.0 2022-12-22 20:12:42,472 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.0716, 1.4506, 1.6462, 0.4733, 1.0809, 1.8751, 1.5610, 1.5417], device='cuda:3'), covar=tensor([0.0545, 0.0238, 0.0276, 0.0305, 0.0371, 0.0305, 0.0210, 0.0454], device='cuda:3'), in_proj_covar=tensor([0.0126, 0.0142, 0.0101, 0.0119, 0.0128, 0.0106, 0.0127, 0.0126], device='cuda:3'), out_proj_covar=tensor([1.1366e-04, 1.3125e-04, 9.2502e-05, 1.0752e-04, 1.1658e-04, 9.7008e-05, 1.1844e-04, 1.1601e-04], device='cuda:3') 2022-12-22 20:12:56,988 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2022-12-22 20:13:11,193 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5350, 1.5151, 1.4402, 1.5733, 1.4948, 3.7201, 1.8118, 2.3586], device='cuda:3'), covar=tensor([0.4679, 0.2797, 0.2584, 0.2625, 0.1520, 0.0232, 0.1692, 0.1019], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0125, 0.0137, 0.0125, 0.0111, 0.0102, 0.0106, 0.0103], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2022-12-22 20:13:20,059 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.063e+02 6.275e+02 7.475e+02 9.020e+02 1.417e+03, threshold=1.495e+03, percent-clipped=0.0 2022-12-22 20:13:41,835 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0 from training. Duration: 25.285 2022-12-22 20:13:46,625 INFO [train.py:894] (3/4) Epoch 7, batch 2400, loss[loss=0.2528, simple_loss=0.3224, pruned_loss=0.09157, over 18510.00 frames. ], tot_loss[loss=0.2504, simple_loss=0.3156, pruned_loss=0.09262, over 3715096.66 frames. ], batch size: 52, lr: 1.62e-02, grad_scale: 16.0 2022-12-22 20:14:05,085 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23450.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:14:18,824 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2022-12-22 20:14:27,852 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23464.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:14:50,088 WARNING [train.py:1060] (3/4) Exclude cut with ID 811-130148-0001-63453-0_sp0.9 from training. Duration: 20.861125 2022-12-22 20:14:53,253 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23481.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:15:03,525 INFO [train.py:894] (3/4) Epoch 7, batch 2450, loss[loss=0.2596, simple_loss=0.3108, pruned_loss=0.1042, over 18480.00 frames. ], tot_loss[loss=0.2499, simple_loss=0.3149, pruned_loss=0.09247, over 3713707.54 frames. ], batch size: 43, lr: 1.62e-02, grad_scale: 16.0 2022-12-22 20:15:03,884 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23488.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:15:10,956 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0 from training. Duration: 20.88 2022-12-22 20:15:18,670 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=23498.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:15:43,971 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0_sp0.9 from training. Duration: 23.4166875 2022-12-22 20:15:52,507 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.340e+02 5.776e+02 7.206e+02 8.413e+02 2.583e+03, threshold=1.441e+03, percent-clipped=4.0 2022-12-22 20:16:05,570 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=23529.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:16:19,236 INFO [train.py:894] (3/4) Epoch 7, batch 2500, loss[loss=0.2528, simple_loss=0.3237, pruned_loss=0.09096, over 18671.00 frames. ], tot_loss[loss=0.2483, simple_loss=0.3137, pruned_loss=0.09144, over 3713167.24 frames. ], batch size: 65, lr: 1.62e-02, grad_scale: 16.0 2022-12-22 20:16:35,392 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23549.0, num_to_drop=1, layers_to_drop={3} 2022-12-22 20:16:35,643 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.99 vs. limit=5.0 2022-12-22 20:17:03,522 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0 from training. Duration: 21.24 2022-12-22 20:17:03,538 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0_sp0.9 from training. Duration: 23.9055625 2022-12-22 20:17:12,374 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23573.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:17:30,615 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23585.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:17:34,797 INFO [train.py:894] (3/4) Epoch 7, batch 2550, loss[loss=0.251, simple_loss=0.3256, pruned_loss=0.08825, over 18557.00 frames. ], tot_loss[loss=0.2484, simple_loss=0.3137, pruned_loss=0.09154, over 3713366.94 frames. ], batch size: 57, lr: 1.62e-02, grad_scale: 16.0 2022-12-22 20:17:36,315 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp0.9 from training. Duration: 25.988875 2022-12-22 20:17:45,020 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0001-134300-0_sp0.9 from training. Duration: 20.67225 2022-12-22 20:18:23,694 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.388e+02 6.581e+02 8.072e+02 9.783e+02 1.636e+03, threshold=1.614e+03, percent-clipped=3.0 2022-12-22 20:18:33,030 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0 from training. Duration: 20.34 2022-12-22 20:18:50,194 INFO [train.py:894] (3/4) Epoch 7, batch 2600, loss[loss=0.2637, simple_loss=0.3293, pruned_loss=0.09907, over 18600.00 frames. ], tot_loss[loss=0.2501, simple_loss=0.3149, pruned_loss=0.09262, over 3713174.49 frames. ], batch size: 98, lr: 1.61e-02, grad_scale: 16.0 2022-12-22 20:19:45,490 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp0.9 from training. Duration: 25.061125 2022-12-22 20:19:55,405 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0 from training. Duration: 0.83 2022-12-22 20:20:06,004 INFO [train.py:894] (3/4) Epoch 7, batch 2650, loss[loss=0.2172, simple_loss=0.2889, pruned_loss=0.07275, over 18424.00 frames. ], tot_loss[loss=0.2496, simple_loss=0.3147, pruned_loss=0.0923, over 3713540.21 frames. ], batch size: 48, lr: 1.61e-02, grad_scale: 16.0 2022-12-22 20:20:19,548 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.3722, 1.6968, 1.9844, 0.6650, 1.2178, 2.3068, 1.8366, 1.7561], device='cuda:3'), covar=tensor([0.0636, 0.0281, 0.0298, 0.0390, 0.0378, 0.0252, 0.0229, 0.0475], device='cuda:3'), in_proj_covar=tensor([0.0127, 0.0143, 0.0101, 0.0120, 0.0129, 0.0107, 0.0130, 0.0128], device='cuda:3'), out_proj_covar=tensor([1.1434e-04, 1.3153e-04, 9.1639e-05, 1.0857e-04, 1.1760e-04, 9.7792e-05, 1.2070e-04, 1.1729e-04], device='cuda:3') 2022-12-22 20:20:20,685 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0 from training. Duration: 24.73 2022-12-22 20:20:34,398 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0 from training. Duration: 23.965 2022-12-22 20:20:41,862 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0030-146996-0_sp0.9 from training. Duration: 22.088875 2022-12-22 20:20:55,656 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.767e+02 5.981e+02 7.717e+02 9.257e+02 2.451e+03, threshold=1.543e+03, percent-clipped=2.0 2022-12-22 20:20:58,602 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0_sp0.9 from training. Duration: 23.6 2022-12-22 20:21:22,064 INFO [train.py:894] (3/4) Epoch 7, batch 2700, loss[loss=0.3008, simple_loss=0.3483, pruned_loss=0.1267, over 18672.00 frames. ], tot_loss[loss=0.2494, simple_loss=0.3145, pruned_loss=0.0922, over 3712994.40 frames. ], batch size: 98, lr: 1.61e-02, grad_scale: 16.0 2022-12-22 20:22:01,211 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23764.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:22:36,697 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0 from training. Duration: 23.795 2022-12-22 20:22:38,056 INFO [train.py:894] (3/4) Epoch 7, batch 2750, loss[loss=0.2625, simple_loss=0.3347, pruned_loss=0.09518, over 18672.00 frames. ], tot_loss[loss=0.2493, simple_loss=0.3147, pruned_loss=0.09198, over 3713145.84 frames. ], batch size: 65, lr: 1.61e-02, grad_scale: 16.0 2022-12-22 20:22:55,009 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp1.1 from training. Duration: 21.5409375 2022-12-22 20:22:58,166 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp0.9 from training. Duration: 24.97775 2022-12-22 20:23:03,134 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4977, 1.7444, 1.6345, 1.7224, 2.1210, 2.6581, 2.5135, 1.8773], device='cuda:3'), covar=tensor([0.0339, 0.0363, 0.0474, 0.0292, 0.0260, 0.0328, 0.0341, 0.0302], device='cuda:3'), in_proj_covar=tensor([0.0080, 0.0113, 0.0135, 0.0120, 0.0102, 0.0097, 0.0085, 0.0112], device='cuda:3'), out_proj_covar=tensor([7.2994e-05, 1.0112e-04, 1.2594e-04, 1.0846e-04, 9.5716e-05, 8.4973e-05, 7.5943e-05, 9.9827e-05], device='cuda:3') 2022-12-22 20:23:08,759 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0_sp0.9 from training. Duration: 23.3444375 2022-12-22 20:23:15,230 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=23812.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:23:26,601 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.804e+02 6.137e+02 7.582e+02 9.579e+02 1.896e+03, threshold=1.516e+03, percent-clipped=2.0 2022-12-22 20:23:34,614 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0_sp0.9 from training. Duration: 23.2 2022-12-22 20:23:41,538 WARNING [train.py:1060] (3/4) Exclude cut with ID 5653-46179-0060-117930-0_sp0.9 from training. Duration: 21.17225 2022-12-22 20:23:56,092 INFO [train.py:894] (3/4) Epoch 7, batch 2800, loss[loss=0.2539, simple_loss=0.3219, pruned_loss=0.09289, over 18578.00 frames. ], tot_loss[loss=0.2479, simple_loss=0.3135, pruned_loss=0.09117, over 3713808.10 frames. ], batch size: 49, lr: 1.61e-02, grad_scale: 16.0 2022-12-22 20:24:02,367 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp0.9 from training. Duration: 24.6555625 2022-12-22 20:24:05,390 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23844.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 20:24:49,042 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23873.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:24:58,923 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0 from training. Duration: 20.44 2022-12-22 20:25:08,272 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23885.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:25:12,100 INFO [train.py:894] (3/4) Epoch 7, batch 2850, loss[loss=0.221, simple_loss=0.2819, pruned_loss=0.08007, over 18385.00 frames. ], tot_loss[loss=0.2486, simple_loss=0.3139, pruned_loss=0.09163, over 3714132.21 frames. ], batch size: 46, lr: 1.61e-02, grad_scale: 16.0 2022-12-22 20:25:13,871 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0_sp0.9 from training. Duration: 23.45 2022-12-22 20:25:25,806 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.2191, 1.5774, 1.0840, 1.9072, 2.1689, 1.3455, 1.4204, 1.0523], device='cuda:3'), covar=tensor([0.2103, 0.1803, 0.1768, 0.0941, 0.1242, 0.1194, 0.1667, 0.1534], device='cuda:3'), in_proj_covar=tensor([0.0230, 0.0201, 0.0195, 0.0179, 0.0244, 0.0183, 0.0203, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 20:25:44,914 WARNING [train.py:1060] (3/4) Exclude cut with ID 6945-60535-0076-12784-0_sp0.9 from training. Duration: 20.52225 2022-12-22 20:25:50,405 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2022-12-22 20:25:52,467 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0 from training. Duration: 22.19 2022-12-22 20:26:00,487 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.104e+02 6.657e+02 8.527e+02 1.024e+03 2.094e+03, threshold=1.705e+03, percent-clipped=3.0 2022-12-22 20:26:02,357 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=23921.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:26:03,620 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp1.1 from training. Duration: 25.3818125 2022-12-22 20:26:19,264 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp0.9 from training. Duration: 28.0944375 2022-12-22 20:26:20,855 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=23933.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:26:23,924 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0_sp0.9 from training. Duration: 22.9444375 2022-12-22 20:26:28,193 INFO [train.py:894] (3/4) Epoch 7, batch 2900, loss[loss=0.1911, simple_loss=0.2686, pruned_loss=0.05674, over 18408.00 frames. ], tot_loss[loss=0.2475, simple_loss=0.3134, pruned_loss=0.09082, over 3713822.08 frames. ], batch size: 48, lr: 1.60e-02, grad_scale: 16.0 2022-12-22 20:26:31,839 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp1.1 from training. Duration: 21.6318125 2022-12-22 20:26:48,811 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.1169, 1.2072, 1.6373, 0.2932, 1.0137, 1.8065, 1.5305, 1.4714], device='cuda:3'), covar=tensor([0.0576, 0.0322, 0.0225, 0.0351, 0.0357, 0.0278, 0.0227, 0.0462], device='cuda:3'), in_proj_covar=tensor([0.0126, 0.0142, 0.0099, 0.0120, 0.0126, 0.0107, 0.0129, 0.0127], device='cuda:3'), out_proj_covar=tensor([1.1303e-04, 1.2983e-04, 8.9327e-05, 1.0844e-04, 1.1374e-04, 9.7178e-05, 1.1925e-04, 1.1615e-04], device='cuda:3') 2022-12-22 20:26:49,765 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0 from training. Duration: 23.695 2022-12-22 20:27:15,090 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0 from training. Duration: 23.955 2022-12-22 20:27:44,170 INFO [train.py:894] (3/4) Epoch 7, batch 2950, loss[loss=0.2259, simple_loss=0.2945, pruned_loss=0.07864, over 18429.00 frames. ], tot_loss[loss=0.2472, simple_loss=0.3133, pruned_loss=0.09061, over 3712661.00 frames. ], batch size: 48, lr: 1.60e-02, grad_scale: 16.0 2022-12-22 20:27:50,109 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp0.9 from training. Duration: 26.438875 2022-12-22 20:28:36,258 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.403e+02 6.654e+02 7.554e+02 9.176e+02 1.909e+03, threshold=1.511e+03, percent-clipped=1.0 2022-12-22 20:28:40,032 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0021-26306-0_sp0.9 from training. Duration: 21.2444375 2022-12-22 20:28:41,450 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp0.9 from training. Duration: 31.02225 2022-12-22 20:28:49,979 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0 from training. Duration: 22.395 2022-12-22 20:29:02,651 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24037.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:29:03,979 INFO [train.py:894] (3/4) Epoch 7, batch 3000, loss[loss=0.252, simple_loss=0.3171, pruned_loss=0.09345, over 18658.00 frames. ], tot_loss[loss=0.2469, simple_loss=0.3132, pruned_loss=0.09028, over 3712862.45 frames. ], batch size: 62, lr: 1.60e-02, grad_scale: 16.0 2022-12-22 20:29:03,979 INFO [train.py:919] (3/4) Computing validation loss 2022-12-22 20:29:12,683 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.8934, 2.8123, 1.7399, 1.3821, 3.6664, 3.4381, 2.6485, 2.1822], device='cuda:3'), covar=tensor([0.0380, 0.0362, 0.0671, 0.0905, 0.0100, 0.0230, 0.0456, 0.0686], device='cuda:3'), in_proj_covar=tensor([0.0121, 0.0114, 0.0130, 0.0121, 0.0079, 0.0114, 0.0136, 0.0151], device='cuda:3'), out_proj_covar=tensor([1.4934e-04, 1.4163e-04, 1.5866e-04, 1.4864e-04, 9.9737e-05, 1.3705e-04, 1.6684e-04, 1.8534e-04], device='cuda:3') 2022-12-22 20:29:14,915 INFO [train.py:928] (3/4) Epoch 7, validation: loss=0.187, simple_loss=0.2863, pruned_loss=0.04389, over 944034.00 frames. 2022-12-22 20:29:14,916 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24676MB 2022-12-22 20:29:16,440 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0 from training. Duration: 21.075 2022-12-22 20:29:21,304 WARNING [train.py:1060] (3/4) Exclude cut with ID 6482-98857-0025-147532-0_sp0.9 from training. Duration: 20.0055625 2022-12-22 20:29:22,875 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0037-132304-0_sp0.9 from training. Duration: 22.05 2022-12-22 20:29:22,890 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0 from training. Duration: 26.8349375 2022-12-22 20:29:26,000 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp1.1 from training. Duration: 22.1090625 2022-12-22 20:29:33,714 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp0.9 from training. Duration: 26.6166875 2022-12-22 20:29:53,004 WARNING [train.py:1060] (3/4) Exclude cut with ID 2046-178027-0000-53705-0_sp0.9 from training. Duration: 20.3055625 2022-12-22 20:30:15,427 WARNING [train.py:1060] (3/4) Exclude cut with ID 7205-50138-0008-5373-0_sp0.9 from training. Duration: 20.7 2022-12-22 20:30:30,108 INFO [train.py:894] (3/4) Epoch 7, batch 3050, loss[loss=0.2567, simple_loss=0.3169, pruned_loss=0.09827, over 18389.00 frames. ], tot_loss[loss=0.2469, simple_loss=0.3133, pruned_loss=0.09023, over 3714115.95 frames. ], batch size: 46, lr: 1.60e-02, grad_scale: 16.0 2022-12-22 20:30:36,935 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.8094, 2.9954, 1.9107, 1.6688, 3.7879, 3.4358, 2.7126, 2.1228], device='cuda:3'), covar=tensor([0.0382, 0.0280, 0.0565, 0.0650, 0.0073, 0.0237, 0.0399, 0.0701], device='cuda:3'), in_proj_covar=tensor([0.0119, 0.0113, 0.0128, 0.0119, 0.0079, 0.0114, 0.0134, 0.0149], device='cuda:3'), out_proj_covar=tensor([1.4743e-04, 1.4105e-04, 1.5690e-04, 1.4728e-04, 9.9052e-05, 1.3644e-04, 1.6481e-04, 1.8309e-04], device='cuda:3') 2022-12-22 20:30:46,476 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24098.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:30:57,236 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0 from training. Duration: 22.48 2022-12-22 20:31:13,331 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp0.9 from training. Duration: 29.816625 2022-12-22 20:31:19,113 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.185e+02 6.628e+02 8.234e+02 1.025e+03 2.088e+03, threshold=1.647e+03, percent-clipped=1.0 2022-12-22 20:31:22,537 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2022-12-22 20:31:34,752 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp1.1 from training. Duration: 22.7590625 2022-12-22 20:31:39,151 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0 from training. Duration: 22.555 2022-12-22 20:31:46,791 INFO [train.py:894] (3/4) Epoch 7, batch 3100, loss[loss=0.2141, simple_loss=0.2887, pruned_loss=0.06971, over 18528.00 frames. ], tot_loss[loss=0.2452, simple_loss=0.3126, pruned_loss=0.08895, over 3713116.70 frames. ], batch size: 47, lr: 1.60e-02, grad_scale: 16.0 2022-12-22 20:31:56,177 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=24144.0, num_to_drop=1, layers_to_drop={2} 2022-12-22 20:32:00,119 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0005-25975-0_sp0.9 from training. Duration: 21.688875 2022-12-22 20:32:10,366 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0472, 2.1505, 1.2553, 2.3162, 2.4168, 2.0443, 3.0533, 2.0936], device='cuda:3'), covar=tensor([0.0789, 0.1374, 0.2408, 0.1764, 0.1432, 0.0769, 0.0893, 0.1084], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0184, 0.0224, 0.0271, 0.0219, 0.0173, 0.0196, 0.0185], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 20:32:34,314 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0_sp0.9 from training. Duration: 22.6 2022-12-22 20:33:02,558 INFO [train.py:894] (3/4) Epoch 7, batch 3150, loss[loss=0.2433, simple_loss=0.3142, pruned_loss=0.08617, over 18518.00 frames. ], tot_loss[loss=0.2441, simple_loss=0.3113, pruned_loss=0.0885, over 3713763.35 frames. ], batch size: 58, lr: 1.60e-02, grad_scale: 8.0 2022-12-22 20:33:09,124 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=24192.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:33:13,764 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0 from training. Duration: 24.32 2022-12-22 20:33:34,649 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7292, 1.0434, 1.6645, 2.9561, 2.0729, 2.2276, 0.4263, 1.8383], device='cuda:3'), covar=tensor([0.1779, 0.2065, 0.1659, 0.0532, 0.1306, 0.1341, 0.2806, 0.1550], device='cuda:3'), in_proj_covar=tensor([0.0103, 0.0116, 0.0125, 0.0112, 0.0105, 0.0128, 0.0131, 0.0109], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 20:33:53,928 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.219e+02 6.067e+02 7.224e+02 8.656e+02 2.684e+03, threshold=1.445e+03, percent-clipped=3.0 2022-12-22 20:33:59,128 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3131, 1.6734, 1.0968, 1.9063, 2.0040, 1.3299, 1.4039, 1.0408], device='cuda:3'), covar=tensor([0.2184, 0.1781, 0.1852, 0.1050, 0.1468, 0.1334, 0.1822, 0.1640], device='cuda:3'), in_proj_covar=tensor([0.0230, 0.0202, 0.0193, 0.0179, 0.0246, 0.0182, 0.0203, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 20:34:14,180 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-276745-0093-13116-0_sp0.9 from training. Duration: 21.061125 2022-12-22 20:34:20,626 INFO [train.py:894] (3/4) Epoch 7, batch 3200, loss[loss=0.2446, simple_loss=0.3159, pruned_loss=0.08668, over 18695.00 frames. ], tot_loss[loss=0.2426, simple_loss=0.3098, pruned_loss=0.08765, over 3712528.78 frames. ], batch size: 62, lr: 1.60e-02, grad_scale: 8.0 2022-12-22 20:34:28,840 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0024-15855-0_sp0.9 from training. Duration: 20.32225 2022-12-22 20:34:42,347 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp1.1 from training. Duration: 0.7545625 2022-12-22 20:34:57,888 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0_sp0.9 from training. Duration: 23.9333125 2022-12-22 20:35:27,772 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp1.1 from training. Duration: 20.17275 2022-12-22 20:35:33,868 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp1.1 from training. Duration: 20.436375 2022-12-22 20:35:39,168 INFO [train.py:894] (3/4) Epoch 7, batch 3250, loss[loss=0.234, simple_loss=0.3065, pruned_loss=0.08074, over 18561.00 frames. ], tot_loss[loss=0.2433, simple_loss=0.3105, pruned_loss=0.08802, over 3713013.27 frames. ], batch size: 49, lr: 1.59e-02, grad_scale: 8.0 2022-12-22 20:35:40,278 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2022-12-22 20:36:27,322 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.060e+02 5.677e+02 7.047e+02 8.775e+02 1.756e+03, threshold=1.409e+03, percent-clipped=1.0 2022-12-22 20:36:41,805 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6278, 1.7833, 1.2849, 1.8028, 1.6102, 1.5725, 1.5637, 1.7810], device='cuda:3'), covar=tensor([0.1986, 0.2256, 0.1633, 0.1944, 0.2377, 0.0993, 0.2042, 0.0707], device='cuda:3'), in_proj_covar=tensor([0.0271, 0.0257, 0.0221, 0.0336, 0.0245, 0.0210, 0.0254, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 20:36:53,152 INFO [train.py:894] (3/4) Epoch 7, batch 3300, loss[loss=0.208, simple_loss=0.2747, pruned_loss=0.07061, over 18545.00 frames. ], tot_loss[loss=0.2445, simple_loss=0.3114, pruned_loss=0.08881, over 3713450.36 frames. ], batch size: 44, lr: 1.59e-02, grad_scale: 8.0 2022-12-22 20:36:56,256 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0_sp0.9 from training. Duration: 23.1055625 2022-12-22 20:36:57,606 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp1.1 from training. Duration: 23.5 2022-12-22 20:37:07,961 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp0.9 from training. Duration: 26.62775 2022-12-22 20:37:22,404 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0 from training. Duration: 21.105 2022-12-22 20:37:24,176 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.9639, 2.4362, 2.9829, 0.8956, 2.3433, 3.5243, 2.1189, 2.5672], device='cuda:3'), covar=tensor([0.0791, 0.0360, 0.0271, 0.0423, 0.0399, 0.0175, 0.0339, 0.0523], device='cuda:3'), in_proj_covar=tensor([0.0128, 0.0142, 0.0097, 0.0120, 0.0127, 0.0106, 0.0128, 0.0123], device='cuda:3'), out_proj_covar=tensor([1.1440e-04, 1.2930e-04, 8.6918e-05, 1.0795e-04, 1.1413e-04, 9.6878e-05, 1.1775e-04, 1.1156e-04], device='cuda:3') 2022-12-22 20:37:26,393 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0_sp0.9 from training. Duration: 24.411125 2022-12-22 20:37:52,845 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp1.1 from training. Duration: 21.263625 2022-12-22 20:38:09,181 INFO [train.py:894] (3/4) Epoch 7, batch 3350, loss[loss=0.223, simple_loss=0.29, pruned_loss=0.07803, over 18399.00 frames. ], tot_loss[loss=0.244, simple_loss=0.3108, pruned_loss=0.08853, over 3713724.98 frames. ], batch size: 42, lr: 1.59e-02, grad_scale: 8.0 2022-12-22 20:38:15,992 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24393.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:38:23,466 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0 from training. Duration: 20.795 2022-12-22 20:38:35,008 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0 from training. Duration: 24.76 2022-12-22 20:38:35,019 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0_sp0.9 from training. Duration: 22.25 2022-12-22 20:38:47,739 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.4635, 4.1506, 4.1094, 2.5724, 3.9845, 3.0953, 2.0543, 3.0237], device='cuda:3'), covar=tensor([0.1786, 0.0919, 0.1121, 0.2566, 0.0829, 0.0915, 0.3361, 0.1464], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0116, 0.0151, 0.0122, 0.0118, 0.0104, 0.0144, 0.0113], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 20:39:00,411 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.754e+02 6.178e+02 8.136e+02 9.455e+02 2.840e+03, threshold=1.627e+03, percent-clipped=7.0 2022-12-22 20:39:00,443 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp1.1 from training. Duration: 20.5045625 2022-12-22 20:39:25,589 INFO [train.py:894] (3/4) Epoch 7, batch 3400, loss[loss=0.2141, simple_loss=0.2829, pruned_loss=0.07259, over 18493.00 frames. ], tot_loss[loss=0.2438, simple_loss=0.3106, pruned_loss=0.08851, over 3713082.58 frames. ], batch size: 43, lr: 1.59e-02, grad_scale: 8.0 2022-12-22 20:40:20,476 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6462, 1.1452, 1.8596, 3.0955, 2.0770, 2.2490, 0.9514, 1.8844], device='cuda:3'), covar=tensor([0.1785, 0.1994, 0.1531, 0.0616, 0.1377, 0.1278, 0.2538, 0.1427], device='cuda:3'), in_proj_covar=tensor([0.0106, 0.0116, 0.0126, 0.0115, 0.0106, 0.0131, 0.0132, 0.0111], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 20:40:37,525 INFO [train.py:894] (3/4) Epoch 7, batch 3450, loss[loss=0.206, simple_loss=0.2761, pruned_loss=0.06796, over 18708.00 frames. ], tot_loss[loss=0.2426, simple_loss=0.3092, pruned_loss=0.08797, over 3713431.97 frames. ], batch size: 46, lr: 1.59e-02, grad_scale: 8.0 2022-12-22 20:41:24,513 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.034e+02 6.206e+02 8.000e+02 1.008e+03 2.086e+03, threshold=1.600e+03, percent-clipped=2.0 2022-12-22 20:41:43,101 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.0495, 2.4092, 1.8314, 2.9904, 2.1857, 2.1172, 2.3539, 3.4764], device='cuda:3'), covar=tensor([0.1341, 0.2327, 0.1355, 0.2300, 0.2664, 0.0803, 0.2153, 0.0420], device='cuda:3'), in_proj_covar=tensor([0.0272, 0.0256, 0.0219, 0.0334, 0.0241, 0.0211, 0.0251, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 20:41:50,850 INFO [train.py:894] (3/4) Epoch 7, batch 3500, loss[loss=0.2571, simple_loss=0.3254, pruned_loss=0.09436, over 18562.00 frames. ], tot_loss[loss=0.2443, simple_loss=0.3106, pruned_loss=0.08904, over 3713526.30 frames. ], batch size: 77, lr: 1.59e-02, grad_scale: 8.0 2022-12-22 20:42:12,432 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp1.1 from training. Duration: 22.2954375 2022-12-22 20:42:24,843 INFO [train.py:894] (3/4) Epoch 8, batch 0, loss[loss=0.2107, simple_loss=0.2861, pruned_loss=0.06759, over 18447.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2861, pruned_loss=0.06759, over 18447.00 frames. ], batch size: 48, lr: 1.49e-02, grad_scale: 8.0 2022-12-22 20:42:24,843 INFO [train.py:919] (3/4) Computing validation loss 2022-12-22 20:42:35,896 INFO [train.py:928] (3/4) Epoch 8, validation: loss=0.1867, simple_loss=0.286, pruned_loss=0.04371, over 944034.00 frames. 2022-12-22 20:42:35,898 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24676MB 2022-12-22 20:42:40,709 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24547.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:43:02,054 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4727, 2.4834, 1.8526, 1.4154, 3.2440, 2.9394, 2.2370, 1.6615], device='cuda:3'), covar=tensor([0.0382, 0.0312, 0.0579, 0.0780, 0.0101, 0.0280, 0.0534, 0.0837], device='cuda:3'), in_proj_covar=tensor([0.0121, 0.0115, 0.0132, 0.0123, 0.0083, 0.0118, 0.0140, 0.0152], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-22 20:43:05,005 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.0284, 2.2953, 2.8582, 0.7238, 2.1931, 3.3719, 2.2425, 2.5050], device='cuda:3'), covar=tensor([0.0729, 0.0352, 0.0330, 0.0469, 0.0477, 0.0190, 0.0326, 0.0591], device='cuda:3'), in_proj_covar=tensor([0.0126, 0.0139, 0.0098, 0.0121, 0.0128, 0.0107, 0.0128, 0.0124], device='cuda:3'), out_proj_covar=tensor([1.1308e-04, 1.2678e-04, 8.8189e-05, 1.0758e-04, 1.1526e-04, 9.6919e-05, 1.1776e-04, 1.1235e-04], device='cuda:3') 2022-12-22 20:43:13,693 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24569.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:43:28,874 WARNING [train.py:1060] (3/4) Exclude cut with ID 298-126791-0067-24026-0_sp0.9 from training. Duration: 21.438875 2022-12-22 20:43:34,905 WARNING [train.py:1060] (3/4) Exclude cut with ID 5652-39938-0025-23684-0_sp0.9 from training. Duration: 22.2055625 2022-12-22 20:43:51,692 INFO [train.py:894] (3/4) Epoch 8, batch 50, loss[loss=0.2322, simple_loss=0.3102, pruned_loss=0.07714, over 18406.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.3006, pruned_loss=0.07058, over 837213.75 frames. ], batch size: 53, lr: 1.49e-02, grad_scale: 8.0 2022-12-22 20:44:14,263 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24608.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:44:33,407 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.554e+02 4.593e+02 5.489e+02 6.811e+02 1.722e+03, threshold=1.098e+03, percent-clipped=1.0 2022-12-22 20:44:47,107 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24630.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:45:07,744 INFO [train.py:894] (3/4) Epoch 8, batch 100, loss[loss=0.2156, simple_loss=0.2903, pruned_loss=0.07044, over 18451.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.2975, pruned_loss=0.06909, over 1474790.54 frames. ], batch size: 48, lr: 1.49e-02, grad_scale: 8.0 2022-12-22 20:45:52,671 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6816, 2.6605, 1.7141, 3.3366, 3.1938, 1.7369, 2.0645, 1.3371], device='cuda:3'), covar=tensor([0.1655, 0.1335, 0.1318, 0.0633, 0.1159, 0.1049, 0.1619, 0.1300], device='cuda:3'), in_proj_covar=tensor([0.0229, 0.0202, 0.0196, 0.0181, 0.0247, 0.0184, 0.0205, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 20:45:59,521 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5646, 1.9164, 1.2926, 2.2289, 2.5132, 1.4370, 1.6496, 1.1649], device='cuda:3'), covar=tensor([0.1638, 0.1467, 0.1480, 0.0835, 0.1233, 0.1119, 0.1620, 0.1375], device='cuda:3'), in_proj_covar=tensor([0.0228, 0.0202, 0.0195, 0.0180, 0.0246, 0.0183, 0.0204, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 20:46:21,713 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=24693.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:46:22,853 INFO [train.py:894] (3/4) Epoch 8, batch 150, loss[loss=0.2018, simple_loss=0.2718, pruned_loss=0.06593, over 18692.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.2963, pruned_loss=0.06891, over 1971106.95 frames. ], batch size: 46, lr: 1.49e-02, grad_scale: 8.0 2022-12-22 20:46:33,844 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0 from training. Duration: 24.525 2022-12-22 20:46:50,856 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2914, 2.3514, 1.6583, 2.7089, 2.4015, 2.1525, 3.0912, 2.4428], device='cuda:3'), covar=tensor([0.0649, 0.1304, 0.1940, 0.1252, 0.1159, 0.0643, 0.0670, 0.0785], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0188, 0.0230, 0.0274, 0.0224, 0.0177, 0.0198, 0.0190], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 20:47:04,371 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.865e+02 4.829e+02 5.972e+02 7.031e+02 2.033e+03, threshold=1.194e+03, percent-clipped=7.0 2022-12-22 20:47:05,955 WARNING [train.py:1060] (3/4) Exclude cut with ID 3699-47246-0007-3408-0_sp0.9 from training. Duration: 20.26675 2022-12-22 20:47:19,137 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp0.9 from training. Duration: 27.25 2022-12-22 20:47:35,596 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=24741.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:47:39,685 INFO [train.py:894] (3/4) Epoch 8, batch 200, loss[loss=0.1952, simple_loss=0.2668, pruned_loss=0.06176, over 18506.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2951, pruned_loss=0.06868, over 2358649.37 frames. ], batch size: 43, lr: 1.49e-02, grad_scale: 8.0 2022-12-22 20:47:56,170 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.22 vs. limit=5.0 2022-12-22 20:48:05,311 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2022-12-22 20:48:12,272 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24765.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:48:29,260 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2022-12-22 20:48:33,206 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0 from training. Duration: 21.68 2022-12-22 20:48:44,225 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0 from training. Duration: 21.6300625 2022-12-22 20:48:49,976 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2022-12-22 20:48:56,700 INFO [train.py:894] (3/4) Epoch 8, batch 250, loss[loss=0.2266, simple_loss=0.3139, pruned_loss=0.06963, over 18500.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2944, pruned_loss=0.06776, over 2659163.10 frames. ], batch size: 52, lr: 1.49e-02, grad_scale: 8.0 2022-12-22 20:49:09,665 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0_sp0.9 from training. Duration: 24.033375 2022-12-22 20:49:23,818 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.1794, 5.2627, 4.9423, 2.3985, 5.2397, 3.7789, 0.5889, 3.7700], device='cuda:3'), covar=tensor([0.1777, 0.0656, 0.1323, 0.3786, 0.0685, 0.0962, 0.6653, 0.1366], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0113, 0.0147, 0.0120, 0.0116, 0.0102, 0.0140, 0.0111], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 20:49:34,877 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7765, 1.4744, 1.3577, 2.0752, 1.7640, 3.6135, 1.5578, 1.4369], device='cuda:3'), covar=tensor([0.0788, 0.1653, 0.1193, 0.0840, 0.1337, 0.0179, 0.1283, 0.1458], device='cuda:3'), in_proj_covar=tensor([0.0077, 0.0086, 0.0079, 0.0080, 0.0095, 0.0074, 0.0087, 0.0080], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2022-12-22 20:49:37,318 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.500e+02 4.749e+02 6.003e+02 7.011e+02 1.781e+03, threshold=1.201e+03, percent-clipped=2.0 2022-12-22 20:49:45,231 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24826.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:49:56,815 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0190, 1.9495, 1.2404, 2.1780, 2.1971, 1.9267, 2.9887, 1.9839], device='cuda:3'), covar=tensor([0.0779, 0.1717, 0.2706, 0.1833, 0.1589, 0.0802, 0.0885, 0.1159], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0190, 0.0230, 0.0275, 0.0224, 0.0179, 0.0201, 0.0192], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 20:50:08,818 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0 from training. Duration: 22.905 2022-12-22 20:50:10,127 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp1.1 from training. Duration: 23.4318125 2022-12-22 20:50:13,378 INFO [train.py:894] (3/4) Epoch 8, batch 300, loss[loss=0.2197, simple_loss=0.2971, pruned_loss=0.07113, over 18376.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2928, pruned_loss=0.06685, over 2893028.16 frames. ], batch size: 51, lr: 1.49e-02, grad_scale: 8.0 2022-12-22 20:50:19,974 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24848.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:50:43,805 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9869, 1.9968, 1.8740, 1.1703, 2.2601, 2.0981, 1.4098, 2.6757], device='cuda:3'), covar=tensor([0.0941, 0.1332, 0.1462, 0.1790, 0.0695, 0.1020, 0.1963, 0.0455], device='cuda:3'), in_proj_covar=tensor([0.0200, 0.0196, 0.0199, 0.0190, 0.0181, 0.0208, 0.0204, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 20:51:29,563 INFO [train.py:894] (3/4) Epoch 8, batch 350, loss[loss=0.2221, simple_loss=0.2884, pruned_loss=0.07784, over 18640.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2945, pruned_loss=0.06783, over 3074894.95 frames. ], batch size: 41, lr: 1.48e-02, grad_scale: 8.0 2022-12-22 20:51:42,478 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24903.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:51:51,431 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24909.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:52:06,330 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp1.1 from training. Duration: 20.82275 2022-12-22 20:52:07,667 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp0.9 from training. Duration: 25.45 2022-12-22 20:52:08,981 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.882e+02 4.524e+02 5.830e+02 7.124e+02 1.429e+03, threshold=1.166e+03, percent-clipped=4.0 2022-12-22 20:52:15,105 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24925.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:52:45,008 INFO [train.py:894] (3/4) Epoch 8, batch 400, loss[loss=0.2018, simple_loss=0.2737, pruned_loss=0.06495, over 18620.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.2956, pruned_loss=0.06825, over 3215939.44 frames. ], batch size: 45, lr: 1.48e-02, grad_scale: 8.0 2022-12-22 20:52:59,757 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7773, 2.2575, 1.7581, 0.7876, 1.8968, 2.0330, 1.3546, 1.8687], device='cuda:3'), covar=tensor([0.0594, 0.0659, 0.1416, 0.1827, 0.1357, 0.1253, 0.1692, 0.0953], device='cuda:3'), in_proj_covar=tensor([0.0156, 0.0172, 0.0199, 0.0191, 0.0196, 0.0180, 0.0198, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 20:53:06,805 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0 from training. Duration: 25.775 2022-12-22 20:53:29,603 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0_sp0.9 from training. Duration: 22.25 2022-12-22 20:53:59,104 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0 from training. Duration: 26.205 2022-12-22 20:54:00,502 INFO [train.py:894] (3/4) Epoch 8, batch 450, loss[loss=0.2216, simple_loss=0.2967, pruned_loss=0.07327, over 18570.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.2967, pruned_loss=0.069, over 3326781.41 frames. ], batch size: 49, lr: 1.48e-02, grad_scale: 8.0 2022-12-22 20:54:02,343 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24995.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:54:16,700 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp0.9 from training. Duration: 30.1555625 2022-12-22 20:54:21,129 WARNING [train.py:1060] (3/4) Exclude cut with ID 1265-135635-0050-6781-0_sp0.9 from training. Duration: 21.8333125 2022-12-22 20:54:29,501 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp1.1 from training. Duration: 20.6545625 2022-12-22 20:54:41,152 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.289e+02 4.744e+02 5.823e+02 7.043e+02 1.775e+03, threshold=1.165e+03, percent-clipped=2.0 2022-12-22 20:55:11,415 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0045-39920-0_sp0.9 from training. Duration: 20.52225 2022-12-22 20:55:17,410 INFO [train.py:894] (3/4) Epoch 8, batch 500, loss[loss=0.2571, simple_loss=0.3286, pruned_loss=0.09282, over 18675.00 frames. ], tot_loss[loss=0.2186, simple_loss=0.2977, pruned_loss=0.06976, over 3412302.90 frames. ], batch size: 60, lr: 1.48e-02, grad_scale: 8.0 2022-12-22 20:55:19,066 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9741, 1.9235, 1.6636, 1.0756, 2.4319, 2.0722, 1.7267, 1.2580], device='cuda:3'), covar=tensor([0.0324, 0.0289, 0.0416, 0.0605, 0.0154, 0.0285, 0.0445, 0.0861], device='cuda:3'), in_proj_covar=tensor([0.0117, 0.0108, 0.0125, 0.0116, 0.0078, 0.0111, 0.0134, 0.0147], device='cuda:3'), out_proj_covar=tensor([1.4464e-04, 1.3355e-04, 1.5223e-04, 1.4294e-04, 9.7310e-05, 1.3320e-04, 1.6421e-04, 1.8046e-04], device='cuda:3') 2022-12-22 20:55:29,891 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5665, 1.4216, 1.2218, 0.6902, 1.8776, 1.5030, 1.3453, 1.0966], device='cuda:3'), covar=tensor([0.0328, 0.0361, 0.0499, 0.0689, 0.0198, 0.0345, 0.0459, 0.0848], device='cuda:3'), in_proj_covar=tensor([0.0118, 0.0108, 0.0126, 0.0117, 0.0078, 0.0112, 0.0135, 0.0148], device='cuda:3'), out_proj_covar=tensor([1.4586e-04, 1.3410e-04, 1.5338e-04, 1.4389e-04, 9.7825e-05, 1.3414e-04, 1.6513e-04, 1.8169e-04], device='cuda:3') 2022-12-22 20:55:32,370 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp0.9 from training. Duration: 29.1166875 2022-12-22 20:55:35,737 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25056.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 20:55:36,106 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.09 vs. limit=5.0 2022-12-22 20:56:31,293 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133211-0007-59831-0_sp0.9 from training. Duration: 21.388875 2022-12-22 20:56:34,190 INFO [train.py:894] (3/4) Epoch 8, batch 550, loss[loss=0.2361, simple_loss=0.3208, pruned_loss=0.07572, over 18479.00 frames. ], tot_loss[loss=0.2198, simple_loss=0.2988, pruned_loss=0.07038, over 3478311.12 frames. ], batch size: 54, lr: 1.48e-02, grad_scale: 8.0 2022-12-22 20:56:40,434 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8140, 1.5943, 1.5174, 1.5249, 1.5057, 1.9237, 2.1645, 1.3071], device='cuda:3'), covar=tensor([0.0377, 0.0274, 0.0420, 0.0287, 0.0245, 0.0296, 0.0233, 0.0291], device='cuda:3'), in_proj_covar=tensor([0.0083, 0.0113, 0.0137, 0.0125, 0.0106, 0.0101, 0.0084, 0.0114], device='cuda:3'), out_proj_covar=tensor([7.4976e-05, 1.0057e-04, 1.2759e-04, 1.1210e-04, 9.7009e-05, 8.7953e-05, 7.5352e-05, 1.0052e-04], device='cuda:3') 2022-12-22 20:56:47,989 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2022-12-22 20:57:09,735 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0 from training. Duration: 22.72 2022-12-22 20:57:11,222 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0_sp0.9 from training. Duration: 22.7444375 2022-12-22 20:57:14,378 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.133e+02 4.717e+02 6.040e+02 7.385e+02 1.198e+03, threshold=1.208e+03, percent-clipped=2.0 2022-12-22 20:57:14,684 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25121.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:57:50,258 INFO [train.py:894] (3/4) Epoch 8, batch 600, loss[loss=0.2685, simple_loss=0.3318, pruned_loss=0.1026, over 18649.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.2995, pruned_loss=0.07082, over 3530555.87 frames. ], batch size: 172, lr: 1.48e-02, grad_scale: 8.0 2022-12-22 20:57:54,371 WARNING [train.py:1060] (3/4) Exclude cut with ID 4133-6541-0027-40495-0_sp1.1 from training. Duration: 0.9681875 2022-12-22 20:57:58,876 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0_sp0.9 from training. Duration: 22.3166875 2022-12-22 20:58:03,243 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133212-0015-59917-0_sp0.9 from training. Duration: 21.8166875 2022-12-22 20:59:04,820 INFO [train.py:894] (3/4) Epoch 8, batch 650, loss[loss=0.1906, simple_loss=0.2721, pruned_loss=0.05451, over 18697.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.2995, pruned_loss=0.07061, over 3571405.90 frames. ], batch size: 46, lr: 1.48e-02, grad_scale: 8.0 2022-12-22 20:59:18,364 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25203.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:59:19,474 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25204.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 20:59:45,821 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.325e+02 5.178e+02 6.418e+02 7.876e+02 2.415e+03, threshold=1.284e+03, percent-clipped=3.0 2022-12-22 20:59:48,724 WARNING [train.py:1060] (3/4) Exclude cut with ID 4957-30119-0041-23990-0_sp0.9 from training. Duration: 20.22775 2022-12-22 20:59:51,934 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25225.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:00:06,495 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.64 vs. limit=5.0 2022-12-22 21:00:21,380 INFO [train.py:894] (3/4) Epoch 8, batch 700, loss[loss=0.273, simple_loss=0.3398, pruned_loss=0.1031, over 18642.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.2995, pruned_loss=0.07055, over 3603423.28 frames. ], batch size: 179, lr: 1.47e-02, grad_scale: 8.0 2022-12-22 21:00:27,608 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.8197, 4.6387, 4.5066, 2.4031, 4.5704, 3.7046, 1.2104, 3.2149], device='cuda:3'), covar=tensor([0.2154, 0.0874, 0.1188, 0.3254, 0.0684, 0.0875, 0.5552, 0.1483], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0118, 0.0152, 0.0123, 0.0119, 0.0106, 0.0148, 0.0114], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 21:00:30,224 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp1.1 from training. Duration: 24.67275 2022-12-22 21:00:31,775 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=25251.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:00:58,872 WARNING [train.py:1060] (3/4) Exclude cut with ID 3082-165428-0081-50734-0_sp0.9 from training. Duration: 21.8055625 2022-12-22 21:01:05,035 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=25273.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:01:14,632 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4154, 2.1717, 2.1921, 1.1092, 2.9387, 2.6149, 2.0984, 1.4500], device='cuda:3'), covar=tensor([0.0347, 0.0354, 0.0374, 0.0703, 0.0110, 0.0285, 0.0477, 0.0904], device='cuda:3'), in_proj_covar=tensor([0.0121, 0.0111, 0.0129, 0.0119, 0.0080, 0.0114, 0.0136, 0.0147], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-22 21:01:36,660 INFO [train.py:894] (3/4) Epoch 8, batch 750, loss[loss=0.228, simple_loss=0.3192, pruned_loss=0.06845, over 18585.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.2981, pruned_loss=0.07004, over 3627707.01 frames. ], batch size: 56, lr: 1.47e-02, grad_scale: 8.0 2022-12-22 21:01:38,018 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0_sp0.9 from training. Duration: 22.6666875 2022-12-22 21:02:17,918 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.591e+02 5.085e+02 6.172e+02 8.301e+02 1.456e+03, threshold=1.234e+03, percent-clipped=2.0 2022-12-22 21:02:40,197 WARNING [train.py:1060] (3/4) Exclude cut with ID 2411-132532-0017-82279-0_sp1.1 from training. Duration: 0.9681875 2022-12-22 21:02:53,248 INFO [train.py:894] (3/4) Epoch 8, batch 800, loss[loss=0.2178, simple_loss=0.296, pruned_loss=0.06975, over 18732.00 frames. ], tot_loss[loss=0.2205, simple_loss=0.2995, pruned_loss=0.07076, over 3647063.62 frames. ], batch size: 52, lr: 1.47e-02, grad_scale: 8.0 2022-12-22 21:03:04,122 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25351.0, num_to_drop=1, layers_to_drop={2} 2022-12-22 21:03:05,391 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0 from training. Duration: 22.485 2022-12-22 21:03:19,572 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8689, 1.9781, 1.9739, 1.3656, 1.9534, 2.0302, 1.4235, 2.4973], device='cuda:3'), covar=tensor([0.1047, 0.1297, 0.1276, 0.1643, 0.0862, 0.1064, 0.1986, 0.0449], device='cuda:3'), in_proj_covar=tensor([0.0197, 0.0197, 0.0199, 0.0187, 0.0182, 0.0205, 0.0203, 0.0176], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 21:03:43,873 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp1.1 from training. Duration: 23.82275 2022-12-22 21:03:56,223 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0 from training. Duration: 20.77 2022-12-22 21:04:03,667 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0_sp0.9 from training. Duration: 24.088875 2022-12-22 21:04:09,285 INFO [train.py:894] (3/4) Epoch 8, batch 850, loss[loss=0.1881, simple_loss=0.2757, pruned_loss=0.05022, over 18541.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.2987, pruned_loss=0.07055, over 3661540.49 frames. ], batch size: 47, lr: 1.47e-02, grad_scale: 8.0 2022-12-22 21:04:34,807 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp1.1 from training. Duration: 20.4409375 2022-12-22 21:04:51,068 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.528e+02 4.741e+02 5.831e+02 7.300e+02 1.630e+03, threshold=1.166e+03, percent-clipped=1.0 2022-12-22 21:04:51,317 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25421.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:05:01,968 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=25428.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 21:05:25,260 INFO [train.py:894] (3/4) Epoch 8, batch 900, loss[loss=0.1962, simple_loss=0.2743, pruned_loss=0.05898, over 18688.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.2993, pruned_loss=0.07069, over 3673706.87 frames. ], batch size: 46, lr: 1.47e-02, grad_scale: 8.0 2022-12-22 21:05:52,006 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2022-12-22 21:05:52,640 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0_sp0.9 from training. Duration: 22.511125 2022-12-22 21:05:54,120 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0 from training. Duration: 20.675 2022-12-22 21:06:03,832 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=25469.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:06:34,050 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25489.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 21:06:41,154 INFO [train.py:894] (3/4) Epoch 8, batch 950, loss[loss=0.1781, simple_loss=0.2648, pruned_loss=0.04569, over 18606.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.2991, pruned_loss=0.07062, over 3682651.28 frames. ], batch size: 45, lr: 1.47e-02, grad_scale: 8.0 2022-12-22 21:06:56,948 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25504.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:07:22,406 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.658e+02 4.840e+02 5.615e+02 6.973e+02 1.516e+03, threshold=1.123e+03, percent-clipped=4.0 2022-12-22 21:07:31,680 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp0.9 from training. Duration: 24.9833125 2022-12-22 21:07:56,195 INFO [train.py:894] (3/4) Epoch 8, batch 1000, loss[loss=0.2019, simple_loss=0.2818, pruned_loss=0.06104, over 18669.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.2991, pruned_loss=0.07006, over 3688730.49 frames. ], batch size: 48, lr: 1.47e-02, grad_scale: 8.0 2022-12-22 21:08:04,841 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0 from training. Duration: 27.14 2022-12-22 21:08:09,526 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=25552.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:08:19,635 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0 from training. Duration: 22.44 2022-12-22 21:09:11,419 INFO [train.py:894] (3/4) Epoch 8, batch 1050, loss[loss=0.2644, simple_loss=0.3369, pruned_loss=0.09593, over 18674.00 frames. ], tot_loss[loss=0.221, simple_loss=0.3001, pruned_loss=0.07094, over 3694395.14 frames. ], batch size: 62, lr: 1.46e-02, grad_scale: 8.0 2022-12-22 21:09:39,473 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0060-62364-0_sp0.9 from training. Duration: 21.361125 2022-12-22 21:09:45,440 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp1.1 from training. Duration: 27.0318125 2022-12-22 21:09:52,804 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.730e+02 4.743e+02 5.690e+02 7.080e+02 2.001e+03, threshold=1.138e+03, percent-clipped=6.0 2022-12-22 21:09:54,287 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp0.9 from training. Duration: 28.638875 2022-12-22 21:10:10,712 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0 from training. Duration: 20.4 2022-12-22 21:10:26,942 INFO [train.py:894] (3/4) Epoch 8, batch 1100, loss[loss=0.2127, simple_loss=0.299, pruned_loss=0.06315, over 18507.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.2999, pruned_loss=0.07087, over 3698452.80 frames. ], batch size: 52, lr: 1.46e-02, grad_scale: 8.0 2022-12-22 21:10:38,548 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25651.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 21:10:42,850 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0 from training. Duration: 20.025 2022-12-22 21:10:44,198 WARNING [train.py:1060] (3/4) Exclude cut with ID 2364-131735-0112-64612-0_sp0.9 from training. Duration: 20.488875 2022-12-22 21:10:46,144 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7906, 0.5972, 1.5323, 1.4382, 1.8105, 1.7490, 1.4357, 1.2732], device='cuda:3'), covar=tensor([0.1280, 0.2072, 0.1588, 0.1558, 0.1026, 0.0594, 0.1528, 0.0802], device='cuda:3'), in_proj_covar=tensor([0.0236, 0.0272, 0.0242, 0.0272, 0.0255, 0.0222, 0.0272, 0.0209], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 21:10:49,206 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0 from training. Duration: 29.735 2022-12-22 21:11:43,079 INFO [train.py:894] (3/4) Epoch 8, batch 1150, loss[loss=0.2432, simple_loss=0.3233, pruned_loss=0.08158, over 18469.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.2983, pruned_loss=0.0701, over 3701056.64 frames. ], batch size: 54, lr: 1.46e-02, grad_scale: 8.0 2022-12-22 21:11:51,124 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=25699.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:12:12,652 WARNING [train.py:1060] (3/4) Exclude cut with ID 7276-92427-0014-12983-0_sp0.9 from training. Duration: 21.3055625 2022-12-22 21:12:14,141 WARNING [train.py:1060] (3/4) Exclude cut with ID 1025-75365-0008-79168-0_sp0.9 from training. Duration: 22.0666875 2022-12-22 21:12:25,037 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.727e+02 4.430e+02 5.330e+02 7.922e+02 1.445e+03, threshold=1.066e+03, percent-clipped=6.0 2022-12-22 21:12:41,490 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7854, 1.0785, 0.6698, 1.1537, 2.0048, 0.9144, 1.2862, 1.7275], device='cuda:3'), covar=tensor([0.1586, 0.2237, 0.2601, 0.1645, 0.1664, 0.1816, 0.1460, 0.1558], device='cuda:3'), in_proj_covar=tensor([0.0089, 0.0103, 0.0123, 0.0099, 0.0111, 0.0094, 0.0098, 0.0097], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 21:12:59,957 INFO [train.py:894] (3/4) Epoch 8, batch 1200, loss[loss=0.2243, simple_loss=0.3028, pruned_loss=0.07293, over 18470.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.2983, pruned_loss=0.06992, over 3703781.54 frames. ], batch size: 54, lr: 1.46e-02, grad_scale: 8.0 2022-12-22 21:13:07,335 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3698, 1.3764, 0.9093, 1.6033, 1.4789, 2.8215, 1.0547, 1.4129], device='cuda:3'), covar=tensor([0.0992, 0.1765, 0.1349, 0.0943, 0.1524, 0.0286, 0.1560, 0.1630], device='cuda:3'), in_proj_covar=tensor([0.0079, 0.0086, 0.0079, 0.0081, 0.0095, 0.0075, 0.0089, 0.0081], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2022-12-22 21:13:11,672 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.0910, 3.5404, 3.5223, 4.0222, 3.6227, 3.6424, 4.2328, 1.0720], device='cuda:3'), covar=tensor([0.0703, 0.0603, 0.0600, 0.0673, 0.1327, 0.0964, 0.0494, 0.4808], device='cuda:3'), in_proj_covar=tensor([0.0262, 0.0181, 0.0184, 0.0183, 0.0251, 0.0212, 0.0207, 0.0232], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 21:14:01,357 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25784.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 21:14:02,580 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0 from training. Duration: 20.26 2022-12-22 21:14:15,487 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0030-9324-0_sp0.9 from training. Duration: 21.3444375 2022-12-22 21:14:16,944 INFO [train.py:894] (3/4) Epoch 8, batch 1250, loss[loss=0.2327, simple_loss=0.3191, pruned_loss=0.07316, over 18682.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.2981, pruned_loss=0.06965, over 3705780.61 frames. ], batch size: 62, lr: 1.46e-02, grad_scale: 8.0 2022-12-22 21:14:40,235 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.0606, 0.8993, 1.1062, 0.1336, 0.6381, 1.2275, 1.0704, 1.1447], device='cuda:3'), covar=tensor([0.0597, 0.0287, 0.0275, 0.0413, 0.0383, 0.0389, 0.0268, 0.0490], device='cuda:3'), in_proj_covar=tensor([0.0128, 0.0144, 0.0101, 0.0121, 0.0130, 0.0110, 0.0134, 0.0129], device='cuda:3'), out_proj_covar=tensor([1.1250e-04, 1.2944e-04, 8.9697e-05, 1.0617e-04, 1.1526e-04, 9.9502e-05, 1.2220e-04, 1.1669e-04], device='cuda:3') 2022-12-22 21:14:58,384 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.631e+02 4.769e+02 5.676e+02 7.211e+02 2.091e+03, threshold=1.135e+03, percent-clipped=3.0 2022-12-22 21:15:12,147 WARNING [train.py:1060] (3/4) Exclude cut with ID 497-129325-0061-62254-0_sp1.1 from training. Duration: 0.97725 2022-12-22 21:15:28,478 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2902, 1.9099, 1.9186, 2.1335, 2.1680, 2.1475, 2.0445, 1.3410], device='cuda:3'), covar=tensor([0.1506, 0.2378, 0.1699, 0.2008, 0.1343, 0.0722, 0.2343, 0.0988], device='cuda:3'), in_proj_covar=tensor([0.0235, 0.0273, 0.0243, 0.0275, 0.0256, 0.0224, 0.0276, 0.0210], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 21:15:33,889 INFO [train.py:894] (3/4) Epoch 8, batch 1300, loss[loss=0.1957, simple_loss=0.2738, pruned_loss=0.05881, over 18598.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.2983, pruned_loss=0.06975, over 3707943.14 frames. ], batch size: 45, lr: 1.46e-02, grad_scale: 8.0 2022-12-22 21:15:43,760 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.1298, 1.1910, 0.6454, 1.2966, 1.4167, 2.4005, 1.0746, 1.3347], device='cuda:3'), covar=tensor([0.1014, 0.1909, 0.1340, 0.0985, 0.1489, 0.0337, 0.1439, 0.1612], device='cuda:3'), in_proj_covar=tensor([0.0078, 0.0086, 0.0079, 0.0081, 0.0095, 0.0075, 0.0088, 0.0081], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2022-12-22 21:15:54,807 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0_sp0.9 from training. Duration: 22.97225 2022-12-22 21:15:55,133 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=25858.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 21:16:30,501 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0047-39922-0_sp0.9 from training. Duration: 21.97775 2022-12-22 21:16:43,473 WARNING [train.py:1060] (3/4) Exclude cut with ID 1112-1043-0006-89194-0_sp0.9 from training. Duration: 21.8333125 2022-12-22 21:16:49,568 INFO [train.py:894] (3/4) Epoch 8, batch 1350, loss[loss=0.2286, simple_loss=0.311, pruned_loss=0.07314, over 18582.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.2978, pruned_loss=0.06922, over 3708826.73 frames. ], batch size: 56, lr: 1.46e-02, grad_scale: 8.0 2022-12-22 21:16:54,769 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0 from training. Duration: 20.47 2022-12-22 21:17:28,342 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25919.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 21:17:30,637 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.888e+02 4.729e+02 5.749e+02 7.465e+02 1.908e+03, threshold=1.150e+03, percent-clipped=5.0 2022-12-22 21:18:00,895 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0037-39912-0_sp0.9 from training. Duration: 20.67225 2022-12-22 21:18:05,203 INFO [train.py:894] (3/4) Epoch 8, batch 1400, loss[loss=0.1924, simple_loss=0.2774, pruned_loss=0.05367, over 18577.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.2976, pruned_loss=0.06903, over 3710033.49 frames. ], batch size: 49, lr: 1.45e-02, grad_scale: 8.0 2022-12-22 21:18:20,477 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp0.9 from training. Duration: 25.2444375 2022-12-22 21:18:45,900 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0021-76797-0_sp0.9 from training. Duration: 21.1445 2022-12-22 21:19:12,086 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7286, 3.8777, 3.8289, 1.5851, 3.9613, 2.9059, 0.5246, 2.6974], device='cuda:3'), covar=tensor([0.1855, 0.0938, 0.1137, 0.3599, 0.0765, 0.1028, 0.5476, 0.1562], device='cuda:3'), in_proj_covar=tensor([0.0131, 0.0111, 0.0144, 0.0117, 0.0115, 0.0102, 0.0140, 0.0107], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 21:19:19,271 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0285, 1.3274, 0.9321, 1.4018, 2.0587, 1.4793, 1.6688, 2.1567], device='cuda:3'), covar=tensor([0.1396, 0.2046, 0.2361, 0.1425, 0.1555, 0.1418, 0.1398, 0.1315], device='cuda:3'), in_proj_covar=tensor([0.0088, 0.0101, 0.0121, 0.0097, 0.0109, 0.0092, 0.0096, 0.0096], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 21:19:20,511 INFO [train.py:894] (3/4) Epoch 8, batch 1450, loss[loss=0.2266, simple_loss=0.3106, pruned_loss=0.07131, over 18703.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.297, pruned_loss=0.06859, over 3711447.27 frames. ], batch size: 78, lr: 1.45e-02, grad_scale: 8.0 2022-12-22 21:20:02,267 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp0.9 from training. Duration: 33.038875 2022-12-22 21:20:02,753 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3823, 1.6383, 1.2460, 2.0605, 2.1851, 1.5034, 1.4252, 1.2731], device='cuda:3'), covar=tensor([0.1970, 0.1698, 0.1575, 0.0958, 0.1157, 0.1221, 0.1660, 0.1481], device='cuda:3'), in_proj_covar=tensor([0.0231, 0.0205, 0.0195, 0.0179, 0.0244, 0.0184, 0.0199, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 21:20:05,119 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.601e+02 4.725e+02 5.598e+02 6.886e+02 1.154e+03, threshold=1.120e+03, percent-clipped=1.0 2022-12-22 21:20:39,980 INFO [train.py:894] (3/4) Epoch 8, batch 1500, loss[loss=0.2348, simple_loss=0.3202, pruned_loss=0.07475, over 18567.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.296, pruned_loss=0.0677, over 3711400.33 frames. ], batch size: 96, lr: 1.45e-02, grad_scale: 8.0 2022-12-22 21:20:42,170 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64291-0000-16059-0_sp0.9 from training. Duration: 20.0944375 2022-12-22 21:20:53,704 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6744, 1.8426, 1.9114, 2.0757, 1.7749, 3.6249, 1.9800, 2.5639], device='cuda:3'), covar=tensor([0.2929, 0.1741, 0.1465, 0.1423, 0.1070, 0.0151, 0.1200, 0.0663], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0122, 0.0132, 0.0122, 0.0107, 0.0099, 0.0102, 0.0099], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2022-12-22 21:20:57,754 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp1.1 from training. Duration: 20.4 2022-12-22 21:21:07,572 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0 from training. Duration: 20.085 2022-12-22 21:21:17,835 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0_sp0.9 from training. Duration: 23.07775 2022-12-22 21:21:40,665 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26084.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 21:21:55,696 INFO [train.py:894] (3/4) Epoch 8, batch 1550, loss[loss=0.2245, simple_loss=0.2972, pruned_loss=0.0759, over 18390.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2953, pruned_loss=0.06756, over 3709757.05 frames. ], batch size: 46, lr: 1.45e-02, grad_scale: 8.0 2022-12-22 21:22:05,247 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp0.9 from training. Duration: 24.9333125 2022-12-22 21:22:08,761 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26102.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:22:36,911 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.030e+02 4.809e+02 6.108e+02 7.519e+02 1.643e+03, threshold=1.222e+03, percent-clipped=3.0 2022-12-22 21:22:51,604 WARNING [train.py:1060] (3/4) Exclude cut with ID 5118-111612-0016-124680-0_sp0.9 from training. Duration: 20.388875 2022-12-22 21:22:53,317 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=26132.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 21:22:57,936 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp1.1 from training. Duration: 20.3590625 2022-12-22 21:23:11,908 INFO [train.py:894] (3/4) Epoch 8, batch 1600, loss[loss=0.2193, simple_loss=0.2995, pruned_loss=0.06959, over 18575.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.295, pruned_loss=0.06709, over 3710661.46 frames. ], batch size: 51, lr: 1.45e-02, grad_scale: 8.0 2022-12-22 21:23:15,407 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7277, 1.8016, 1.1486, 1.8569, 1.9343, 1.6716, 2.5115, 1.9021], device='cuda:3'), covar=tensor([0.1075, 0.1717, 0.2810, 0.1800, 0.1645, 0.1073, 0.0955, 0.1328], device='cuda:3'), in_proj_covar=tensor([0.0169, 0.0188, 0.0227, 0.0273, 0.0220, 0.0177, 0.0201, 0.0190], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 21:23:41,483 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26163.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:23:48,628 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26168.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:23:54,867 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2022-12-22 21:24:07,990 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0_sp1.1 from training. Duration: 0.836375 2022-12-22 21:24:21,786 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.58 vs. limit=5.0 2022-12-22 21:24:27,780 INFO [train.py:894] (3/4) Epoch 8, batch 1650, loss[loss=0.2052, simple_loss=0.2907, pruned_loss=0.05989, over 18461.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.2968, pruned_loss=0.06931, over 3712035.74 frames. ], batch size: 50, lr: 1.45e-02, grad_scale: 16.0 2022-12-22 21:24:50,736 WARNING [train.py:1060] (3/4) Exclude cut with ID 8565-290391-0049-67394-0_sp0.9 from training. Duration: 21.3166875 2022-12-22 21:24:59,643 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26214.0, num_to_drop=1, layers_to_drop={2} 2022-12-22 21:25:09,422 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.083e+02 5.122e+02 6.330e+02 7.747e+02 1.750e+03, threshold=1.266e+03, percent-clipped=7.0 2022-12-22 21:25:20,536 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0029-104863-0_sp0.9 from training. Duration: 22.1055625 2022-12-22 21:25:22,350 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26229.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:25:31,646 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp1.1 from training. Duration: 21.77725 2022-12-22 21:25:45,446 INFO [train.py:894] (3/4) Epoch 8, batch 1700, loss[loss=0.2506, simple_loss=0.325, pruned_loss=0.08805, over 18713.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.2995, pruned_loss=0.07234, over 3712912.85 frames. ], batch size: 78, lr: 1.45e-02, grad_scale: 16.0 2022-12-22 21:25:51,171 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp0.9 from training. Duration: 27.8166875 2022-12-22 21:25:54,616 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6214, 1.6094, 1.6824, 1.6700, 1.4534, 3.7537, 1.6989, 2.4961], device='cuda:3'), covar=tensor([0.3462, 0.2064, 0.1910, 0.1987, 0.1316, 0.0184, 0.1478, 0.0800], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0122, 0.0132, 0.0123, 0.0108, 0.0100, 0.0102, 0.0099], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2022-12-22 21:25:54,656 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26250.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:26:06,790 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.92 vs. limit=5.0 2022-12-22 21:26:14,831 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp1.1 from training. Duration: 22.5090625 2022-12-22 21:26:20,986 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0 from training. Duration: 25.035 2022-12-22 21:26:21,745 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.33 vs. limit=5.0 2022-12-22 21:26:41,516 WARNING [train.py:1060] (3/4) Exclude cut with ID 774-127930-0014-10412-0_sp1.1 from training. Duration: 0.95 2022-12-22 21:26:58,099 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp0.9 from training. Duration: 0.92225 2022-12-22 21:27:02,528 INFO [train.py:894] (3/4) Epoch 8, batch 1750, loss[loss=0.2313, simple_loss=0.302, pruned_loss=0.08027, over 18372.00 frames. ], tot_loss[loss=0.2272, simple_loss=0.3025, pruned_loss=0.076, over 3713765.56 frames. ], batch size: 46, lr: 1.45e-02, grad_scale: 16.0 2022-12-22 21:27:23,849 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0 from training. Duration: 21.97 2022-12-22 21:27:28,962 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26311.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:27:43,954 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.748e+02 6.270e+02 7.569e+02 9.269e+02 2.899e+03, threshold=1.514e+03, percent-clipped=6.0 2022-12-22 21:27:43,979 WARNING [train.py:1060] (3/4) Exclude cut with ID 7492-105653-0055-62765-0_sp0.9 from training. Duration: 21.97225 2022-12-22 21:27:44,028 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp0.9 from training. Duration: 25.3333125 2022-12-22 21:27:55,252 WARNING [train.py:1060] (3/4) Exclude cut with ID 5172-29468-0015-19128-0_sp0.9 from training. Duration: 21.5055625 2022-12-22 21:27:56,155 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2022-12-22 21:28:06,085 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp1.1 from training. Duration: 20.72725 2022-12-22 21:28:18,995 INFO [train.py:894] (3/4) Epoch 8, batch 1800, loss[loss=0.2846, simple_loss=0.3435, pruned_loss=0.1129, over 18669.00 frames. ], tot_loss[loss=0.2306, simple_loss=0.3038, pruned_loss=0.07865, over 3713582.25 frames. ], batch size: 60, lr: 1.44e-02, grad_scale: 16.0 2022-12-22 21:28:37,671 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp0.9 from training. Duration: 26.32775 2022-12-22 21:29:11,050 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0 from training. Duration: 20.025 2022-12-22 21:29:16,129 WARNING [train.py:1060] (3/4) Exclude cut with ID 6709-74022-0004-86860-0_sp1.1 from training. Duration: 0.9409375 2022-12-22 21:29:17,710 WARNING [train.py:1060] (3/4) Exclude cut with ID 4757-1811-0023-62229-0_sp0.9 from training. Duration: 21.37775 2022-12-22 21:29:34,929 INFO [train.py:894] (3/4) Epoch 8, batch 1850, loss[loss=0.2322, simple_loss=0.2893, pruned_loss=0.08761, over 18602.00 frames. ], tot_loss[loss=0.2349, simple_loss=0.3063, pruned_loss=0.08173, over 3713355.10 frames. ], batch size: 45, lr: 1.44e-02, grad_scale: 16.0 2022-12-22 21:29:40,057 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0004-25974-0_sp0.9 from training. Duration: 21.17225 2022-12-22 21:29:40,075 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp0.9 from training. Duration: 27.511125 2022-12-22 21:30:05,983 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26414.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:30:12,240 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0 from training. Duration: 22.8 2022-12-22 21:30:17,100 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.061e+02 5.870e+02 7.240e+02 8.563e+02 1.417e+03, threshold=1.448e+03, percent-clipped=0.0 2022-12-22 21:30:17,157 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0 from training. Duration: 22.585 2022-12-22 21:30:47,607 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0001-146967-0_sp0.9 from training. Duration: 22.0166875 2022-12-22 21:30:48,479 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2022-12-22 21:30:53,354 INFO [train.py:894] (3/4) Epoch 8, batch 1900, loss[loss=0.261, simple_loss=0.3279, pruned_loss=0.09701, over 18458.00 frames. ], tot_loss[loss=0.2364, simple_loss=0.3069, pruned_loss=0.083, over 3713427.83 frames. ], batch size: 64, lr: 1.44e-02, grad_scale: 16.0 2022-12-22 21:31:04,426 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp1.1 from training. Duration: 24.395375 2022-12-22 21:31:11,913 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp0.9 from training. Duration: 27.47775 2022-12-22 21:31:14,969 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26458.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:31:16,294 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp0.9 from training. Duration: 24.8833125 2022-12-22 21:31:18,959 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0 from training. Duration: 23.39 2022-12-22 21:31:25,771 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp0.9 from training. Duration: 28.72225 2022-12-22 21:31:36,462 WARNING [train.py:1060] (3/4) Exclude cut with ID 585-294811-0110-133686-0_sp0.9 from training. Duration: 20.8944375 2022-12-22 21:31:41,177 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26475.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:31:44,691 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5348, 1.2150, 1.2031, 1.8827, 1.7838, 3.2061, 1.2558, 1.3905], device='cuda:3'), covar=tensor([0.1185, 0.2353, 0.1506, 0.1134, 0.1529, 0.0290, 0.1858, 0.2081], device='cuda:3'), in_proj_covar=tensor([0.0076, 0.0085, 0.0077, 0.0078, 0.0094, 0.0072, 0.0086, 0.0079], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2022-12-22 21:31:52,722 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0_sp0.9 from training. Duration: 23.8444375 2022-12-22 21:32:09,419 INFO [train.py:894] (3/4) Epoch 8, batch 1950, loss[loss=0.213, simple_loss=0.2721, pruned_loss=0.0769, over 18484.00 frames. ], tot_loss[loss=0.2365, simple_loss=0.3064, pruned_loss=0.08337, over 3714344.80 frames. ], batch size: 43, lr: 1.44e-02, grad_scale: 16.0 2022-12-22 21:32:15,778 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0 from training. Duration: 25.85 2022-12-22 21:32:15,791 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0 from training. Duration: 21.39 2022-12-22 21:32:27,303 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0 from training. Duration: 27.92 2022-12-22 21:32:39,769 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26514.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 21:32:48,274 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.02 vs. limit=5.0 2022-12-22 21:32:50,234 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.849e+02 6.560e+02 7.949e+02 9.276e+02 2.232e+03, threshold=1.590e+03, percent-clipped=3.0 2022-12-22 21:32:55,537 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0039-130165-0_sp0.9 from training. Duration: 20.661125 2022-12-22 21:32:55,653 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26524.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:33:17,685 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0043-15874-0_sp0.9 from training. Duration: 20.07225 2022-12-22 21:33:25,148 INFO [train.py:894] (3/4) Epoch 8, batch 2000, loss[loss=0.2446, simple_loss=0.3212, pruned_loss=0.08407, over 18480.00 frames. ], tot_loss[loss=0.2375, simple_loss=0.3068, pruned_loss=0.08405, over 3713823.01 frames. ], batch size: 64, lr: 1.44e-02, grad_scale: 16.0 2022-12-22 21:33:25,227 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0 from training. Duration: 21.01 2022-12-22 21:33:52,779 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=26562.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 21:34:36,315 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0 from training. Duration: 20.65 2022-12-22 21:34:40,619 INFO [train.py:894] (3/4) Epoch 8, batch 2050, loss[loss=0.3221, simple_loss=0.3559, pruned_loss=0.1441, over 18670.00 frames. ], tot_loss[loss=0.2391, simple_loss=0.3076, pruned_loss=0.08529, over 3714357.57 frames. ], batch size: 186, lr: 1.44e-02, grad_scale: 8.0 2022-12-22 21:34:41,978 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0 from training. Duration: 21.46 2022-12-22 21:34:58,834 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26606.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:35:23,329 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.180e+02 6.124e+02 7.302e+02 9.759e+02 1.688e+03, threshold=1.460e+03, percent-clipped=2.0 2022-12-22 21:35:28,052 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0 from training. Duration: 0.92 2022-12-22 21:35:36,046 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0_sp0.9 from training. Duration: 23.7666875 2022-12-22 21:35:56,672 INFO [train.py:894] (3/4) Epoch 8, batch 2100, loss[loss=0.2699, simple_loss=0.3387, pruned_loss=0.1005, over 18592.00 frames. ], tot_loss[loss=0.2411, simple_loss=0.3093, pruned_loss=0.08647, over 3715744.19 frames. ], batch size: 56, lr: 1.44e-02, grad_scale: 8.0 2022-12-22 21:36:02,695 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6079, 3.5528, 3.5450, 1.6625, 3.5616, 2.5611, 0.6081, 2.3794], device='cuda:3'), covar=tensor([0.1787, 0.0906, 0.1280, 0.3266, 0.0846, 0.1150, 0.4936, 0.1551], device='cuda:3'), in_proj_covar=tensor([0.0127, 0.0112, 0.0144, 0.0118, 0.0116, 0.0101, 0.0139, 0.0107], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 21:36:13,186 WARNING [train.py:1060] (3/4) Exclude cut with ID 8544-281189-0060-101339-0_sp0.9 from training. Duration: 20.861125 2022-12-22 21:36:23,031 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.50 vs. limit=5.0 2022-12-22 21:36:23,238 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0_sp0.9 from training. Duration: 22.711125 2022-12-22 21:37:05,878 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp1.1 from training. Duration: 22.986375 2022-12-22 21:37:11,693 INFO [train.py:894] (3/4) Epoch 8, batch 2150, loss[loss=0.2532, simple_loss=0.3218, pruned_loss=0.09226, over 18690.00 frames. ], tot_loss[loss=0.2396, simple_loss=0.3079, pruned_loss=0.08565, over 3716049.93 frames. ], batch size: 60, lr: 1.43e-02, grad_scale: 8.0 2022-12-22 21:37:22,902 WARNING [train.py:1060] (3/4) Exclude cut with ID 8040-260924-0003-80960-0_sp0.9 from training. Duration: 22.07225 2022-12-22 21:37:23,223 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7748, 1.1073, 0.8579, 1.2244, 1.8737, 1.1412, 1.4006, 1.8293], device='cuda:3'), covar=tensor([0.1430, 0.2162, 0.2289, 0.1432, 0.1784, 0.1486, 0.1423, 0.1443], device='cuda:3'), in_proj_covar=tensor([0.0091, 0.0104, 0.0122, 0.0097, 0.0113, 0.0093, 0.0098, 0.0098], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 21:37:27,588 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0045-26330-0_sp0.9 from training. Duration: 20.3055625 2022-12-22 21:37:30,345 WARNING [train.py:1060] (3/4) Exclude cut with ID 6356-271890-0060-94317-0_sp0.9 from training. Duration: 20.72225 2022-12-22 21:37:48,913 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp1.1 from training. Duration: 22.4818125 2022-12-22 21:37:54,907 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.474e+02 5.655e+02 7.658e+02 9.829e+02 2.063e+03, threshold=1.532e+03, percent-clipped=2.0 2022-12-22 21:38:13,673 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp0.9 from training. Duration: 25.0944375 2022-12-22 21:38:18,012 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0 from training. Duration: 21.515 2022-12-22 21:38:23,855 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp0.9 from training. Duration: 27.02225 2022-12-22 21:38:26,595 INFO [train.py:894] (3/4) Epoch 8, batch 2200, loss[loss=0.1882, simple_loss=0.2593, pruned_loss=0.05853, over 18476.00 frames. ], tot_loss[loss=0.2402, simple_loss=0.3085, pruned_loss=0.08591, over 3717385.54 frames. ], batch size: 43, lr: 1.43e-02, grad_scale: 8.0 2022-12-22 21:38:29,755 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0010-62480-0_sp0.9 from training. Duration: 22.22225 2022-12-22 21:38:36,310 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0085-44554-0_sp0.9 from training. Duration: 20.85 2022-12-22 21:38:49,184 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26758.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:39:07,953 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26770.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:39:11,963 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0 from training. Duration: 21.54 2022-12-22 21:39:16,228 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp1.1 from training. Duration: 20.5318125 2022-12-22 21:39:25,854 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0012-134311-0_sp0.9 from training. Duration: 21.9333125 2022-12-22 21:39:42,226 INFO [train.py:894] (3/4) Epoch 8, batch 2250, loss[loss=0.2757, simple_loss=0.3364, pruned_loss=0.1075, over 18714.00 frames. ], tot_loss[loss=0.2407, simple_loss=0.3086, pruned_loss=0.0864, over 3716543.56 frames. ], batch size: 69, lr: 1.43e-02, grad_scale: 8.0 2022-12-22 21:40:02,916 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=26806.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:40:09,808 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.83 vs. limit=5.0 2022-12-22 21:40:14,710 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0025-130151-0_sp0.9 from training. Duration: 21.7944375 2022-12-22 21:40:27,436 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.517e+02 5.710e+02 7.345e+02 8.844e+02 3.050e+03, threshold=1.469e+03, percent-clipped=4.0 2022-12-22 21:40:27,471 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0_sp0.9 from training. Duration: 22.4666875 2022-12-22 21:40:30,803 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26824.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:40:35,147 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0 from training. Duration: 21.635 2022-12-22 21:40:41,105 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0_sp0.9 from training. Duration: 24.038875 2022-12-22 21:41:01,015 INFO [train.py:894] (3/4) Epoch 8, batch 2300, loss[loss=0.2582, simple_loss=0.327, pruned_loss=0.09471, over 18725.00 frames. ], tot_loss[loss=0.2408, simple_loss=0.3085, pruned_loss=0.08655, over 3716337.02 frames. ], batch size: 54, lr: 1.43e-02, grad_scale: 8.0 2022-12-22 21:41:24,870 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp1.1 from training. Duration: 21.786375 2022-12-22 21:41:38,873 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0 from training. Duration: 20.22 2022-12-22 21:41:44,734 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=26872.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:42:18,273 INFO [train.py:894] (3/4) Epoch 8, batch 2350, loss[loss=0.2136, simple_loss=0.281, pruned_loss=0.07309, over 18531.00 frames. ], tot_loss[loss=0.2409, simple_loss=0.3083, pruned_loss=0.08674, over 3715632.14 frames. ], batch size: 47, lr: 1.43e-02, grad_scale: 8.0 2022-12-22 21:42:34,250 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26904.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 21:42:37,166 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26906.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:42:52,860 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2022-12-22 21:43:00,840 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.511e+02 5.486e+02 6.533e+02 8.294e+02 1.980e+03, threshold=1.307e+03, percent-clipped=2.0 2022-12-22 21:43:34,059 INFO [train.py:894] (3/4) Epoch 8, batch 2400, loss[loss=0.212, simple_loss=0.2712, pruned_loss=0.07636, over 18588.00 frames. ], tot_loss[loss=0.241, simple_loss=0.3083, pruned_loss=0.08683, over 3715179.92 frames. ], batch size: 41, lr: 1.43e-02, grad_scale: 8.0 2022-12-22 21:43:39,098 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0 from training. Duration: 25.285 2022-12-22 21:43:43,557 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26950.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:43:49,119 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=26954.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:44:06,515 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26965.0, num_to_drop=1, layers_to_drop={3} 2022-12-22 21:44:27,124 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7445, 1.1504, 1.5066, 1.5640, 1.9250, 1.8635, 1.9384, 1.1912], device='cuda:3'), covar=tensor([0.0322, 0.0276, 0.0432, 0.0236, 0.0180, 0.0289, 0.0201, 0.0269], device='cuda:3'), in_proj_covar=tensor([0.0079, 0.0113, 0.0137, 0.0121, 0.0105, 0.0102, 0.0087, 0.0112], device='cuda:3'), out_proj_covar=tensor([7.0202e-05, 9.9345e-05, 1.2544e-04, 1.0655e-04, 9.5128e-05, 8.7102e-05, 7.6331e-05, 9.7220e-05], device='cuda:3') 2022-12-22 21:44:41,622 WARNING [train.py:1060] (3/4) Exclude cut with ID 811-130148-0001-63453-0_sp0.9 from training. Duration: 20.861125 2022-12-22 21:44:50,475 INFO [train.py:894] (3/4) Epoch 8, batch 2450, loss[loss=0.2723, simple_loss=0.3353, pruned_loss=0.1047, over 18562.00 frames. ], tot_loss[loss=0.2408, simple_loss=0.3086, pruned_loss=0.08651, over 3714775.16 frames. ], batch size: 56, lr: 1.43e-02, grad_scale: 8.0 2022-12-22 21:45:04,462 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0 from training. Duration: 20.88 2022-12-22 21:45:17,532 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27011.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:45:33,581 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.757e+02 5.947e+02 7.436e+02 9.112e+02 3.064e+03, threshold=1.487e+03, percent-clipped=6.0 2022-12-22 21:45:35,103 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0_sp0.9 from training. Duration: 23.4166875 2022-12-22 21:45:45,870 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4968, 1.7871, 1.9228, 2.1676, 2.2817, 2.0253, 2.2882, 1.4106], device='cuda:3'), covar=tensor([0.1425, 0.2374, 0.1681, 0.1977, 0.1145, 0.0681, 0.1971, 0.0872], device='cuda:3'), in_proj_covar=tensor([0.0240, 0.0274, 0.0245, 0.0276, 0.0260, 0.0226, 0.0277, 0.0215], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 21:46:07,398 INFO [train.py:894] (3/4) Epoch 8, batch 2500, loss[loss=0.2283, simple_loss=0.3019, pruned_loss=0.07732, over 18725.00 frames. ], tot_loss[loss=0.2412, simple_loss=0.309, pruned_loss=0.08666, over 3715284.24 frames. ], batch size: 52, lr: 1.43e-02, grad_scale: 8.0 2022-12-22 21:46:43,537 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3706, 1.8935, 1.2729, 2.4135, 2.5993, 1.4265, 1.3262, 1.0728], device='cuda:3'), covar=tensor([0.1940, 0.1547, 0.1538, 0.0745, 0.1281, 0.1180, 0.1939, 0.1480], device='cuda:3'), in_proj_covar=tensor([0.0232, 0.0202, 0.0194, 0.0179, 0.0245, 0.0183, 0.0203, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 21:46:46,343 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27070.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:46:53,002 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0 from training. Duration: 21.24 2022-12-22 21:46:53,022 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0_sp0.9 from training. Duration: 23.9055625 2022-12-22 21:47:16,191 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3980, 0.9939, 0.6811, 0.9954, 1.7416, 0.6806, 1.0634, 1.3529], device='cuda:3'), covar=tensor([0.1750, 0.2260, 0.2407, 0.1744, 0.1964, 0.1910, 0.1609, 0.1660], device='cuda:3'), in_proj_covar=tensor([0.0091, 0.0103, 0.0121, 0.0097, 0.0113, 0.0092, 0.0097, 0.0097], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 21:47:23,372 INFO [train.py:894] (3/4) Epoch 8, batch 2550, loss[loss=0.2055, simple_loss=0.2846, pruned_loss=0.06321, over 18578.00 frames. ], tot_loss[loss=0.24, simple_loss=0.3081, pruned_loss=0.08601, over 3715475.63 frames. ], batch size: 98, lr: 1.42e-02, grad_scale: 8.0 2022-12-22 21:47:27,689 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp0.9 from training. Duration: 25.988875 2022-12-22 21:47:37,396 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0001-134300-0_sp0.9 from training. Duration: 20.67225 2022-12-22 21:48:00,712 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=27118.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:48:06,475 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.238e+02 5.522e+02 7.040e+02 8.791e+02 1.415e+03, threshold=1.408e+03, percent-clipped=0.0 2022-12-22 21:48:25,586 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0 from training. Duration: 20.34 2022-12-22 21:48:40,631 INFO [train.py:894] (3/4) Epoch 8, batch 2600, loss[loss=0.2554, simple_loss=0.3208, pruned_loss=0.09497, over 18474.00 frames. ], tot_loss[loss=0.2388, simple_loss=0.3071, pruned_loss=0.08526, over 3715463.08 frames. ], batch size: 50, lr: 1.42e-02, grad_scale: 8.0 2022-12-22 21:48:42,264 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2022-12-22 21:48:47,655 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=27148.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:49:39,835 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp0.9 from training. Duration: 25.061125 2022-12-22 21:49:49,882 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0 from training. Duration: 0.83 2022-12-22 21:49:57,939 INFO [train.py:894] (3/4) Epoch 8, batch 2650, loss[loss=0.2361, simple_loss=0.3011, pruned_loss=0.08555, over 18392.00 frames. ], tot_loss[loss=0.2379, simple_loss=0.3064, pruned_loss=0.08473, over 3714661.06 frames. ], batch size: 46, lr: 1.42e-02, grad_scale: 8.0 2022-12-22 21:50:16,675 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0 from training. Duration: 24.73 2022-12-22 21:50:21,913 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27209.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:50:24,973 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8955, 0.9621, 1.5806, 1.5687, 1.8363, 1.7170, 1.5947, 1.2997], device='cuda:3'), covar=tensor([0.1438, 0.2176, 0.1736, 0.1748, 0.1191, 0.0717, 0.1814, 0.0853], device='cuda:3'), in_proj_covar=tensor([0.0240, 0.0273, 0.0247, 0.0276, 0.0260, 0.0226, 0.0280, 0.0214], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 21:50:31,791 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0 from training. Duration: 23.965 2022-12-22 21:50:33,949 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7065, 1.7514, 2.0079, 1.9786, 1.6075, 4.9951, 2.2107, 2.8096], device='cuda:3'), covar=tensor([0.3259, 0.1967, 0.1706, 0.1875, 0.1345, 0.0087, 0.1356, 0.0820], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0120, 0.0133, 0.0123, 0.0107, 0.0101, 0.0103, 0.0098], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2022-12-22 21:50:41,442 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.270e+02 5.404e+02 7.023e+02 8.570e+02 1.922e+03, threshold=1.405e+03, percent-clipped=4.0 2022-12-22 21:50:41,523 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0030-146996-0_sp0.9 from training. Duration: 22.088875 2022-12-22 21:50:43,199 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3822, 1.1209, 0.9827, 1.5606, 1.5520, 2.9732, 1.0147, 1.3663], device='cuda:3'), covar=tensor([0.0871, 0.1842, 0.1222, 0.0941, 0.1451, 0.0236, 0.1507, 0.1524], device='cuda:3'), in_proj_covar=tensor([0.0076, 0.0085, 0.0076, 0.0078, 0.0095, 0.0072, 0.0086, 0.0079], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2022-12-22 21:50:57,198 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0_sp0.9 from training. Duration: 23.6 2022-12-22 21:51:15,442 INFO [train.py:894] (3/4) Epoch 8, batch 2700, loss[loss=0.2085, simple_loss=0.2731, pruned_loss=0.07188, over 18428.00 frames. ], tot_loss[loss=0.237, simple_loss=0.3056, pruned_loss=0.08415, over 3713969.62 frames. ], batch size: 42, lr: 1.42e-02, grad_scale: 8.0 2022-12-22 21:51:39,785 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27260.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 21:52:32,324 INFO [train.py:894] (3/4) Epoch 8, batch 2750, loss[loss=0.2514, simple_loss=0.3079, pruned_loss=0.09743, over 18395.00 frames. ], tot_loss[loss=0.2361, simple_loss=0.3047, pruned_loss=0.08378, over 3713888.29 frames. ], batch size: 46, lr: 1.42e-02, grad_scale: 8.0 2022-12-22 21:52:36,736 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0 from training. Duration: 23.795 2022-12-22 21:52:50,752 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27306.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:52:53,685 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp1.1 from training. Duration: 21.5409375 2022-12-22 21:52:55,167 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp0.9 from training. Duration: 24.97775 2022-12-22 21:53:06,740 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0_sp0.9 from training. Duration: 23.3444375 2022-12-22 21:53:15,434 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.310e+02 5.953e+02 7.157e+02 9.092e+02 1.661e+03, threshold=1.431e+03, percent-clipped=6.0 2022-12-22 21:53:33,681 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0_sp0.9 from training. Duration: 23.2 2022-12-22 21:53:39,358 WARNING [train.py:1060] (3/4) Exclude cut with ID 5653-46179-0060-117930-0_sp0.9 from training. Duration: 21.17225 2022-12-22 21:53:41,006 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3735, 0.8564, 1.1104, 2.0669, 1.4472, 2.1212, 0.5510, 1.5856], device='cuda:3'), covar=tensor([0.2188, 0.2893, 0.1943, 0.1019, 0.1730, 0.1205, 0.2654, 0.1838], device='cuda:3'), in_proj_covar=tensor([0.0105, 0.0115, 0.0128, 0.0119, 0.0105, 0.0132, 0.0130, 0.0107], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 21:53:47,800 INFO [train.py:894] (3/4) Epoch 8, batch 2800, loss[loss=0.2136, simple_loss=0.2732, pruned_loss=0.07701, over 18457.00 frames. ], tot_loss[loss=0.2372, simple_loss=0.3055, pruned_loss=0.08445, over 3714337.63 frames. ], batch size: 43, lr: 1.42e-02, grad_scale: 8.0 2022-12-22 21:54:00,097 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp0.9 from training. Duration: 24.6555625 2022-12-22 21:54:53,116 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0 from training. Duration: 20.44 2022-12-22 21:55:04,966 INFO [train.py:894] (3/4) Epoch 8, batch 2850, loss[loss=0.2392, simple_loss=0.3088, pruned_loss=0.08474, over 18726.00 frames. ], tot_loss[loss=0.2362, simple_loss=0.3045, pruned_loss=0.08396, over 3714523.90 frames. ], batch size: 52, lr: 1.42e-02, grad_scale: 8.0 2022-12-22 21:55:07,915 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0_sp0.9 from training. Duration: 23.45 2022-12-22 21:55:13,343 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.59 vs. limit=5.0 2022-12-22 21:55:38,563 WARNING [train.py:1060] (3/4) Exclude cut with ID 6945-60535-0076-12784-0_sp0.9 from training. Duration: 20.52225 2022-12-22 21:55:47,122 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0 from training. Duration: 22.19 2022-12-22 21:55:48,476 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.802e+02 6.087e+02 7.386e+02 9.087e+02 2.096e+03, threshold=1.477e+03, percent-clipped=5.0 2022-12-22 21:55:55,788 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp1.1 from training. Duration: 25.3818125 2022-12-22 21:56:13,653 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp0.9 from training. Duration: 28.0944375 2022-12-22 21:56:21,332 INFO [train.py:894] (3/4) Epoch 8, batch 2900, loss[loss=0.2229, simple_loss=0.2977, pruned_loss=0.07402, over 18710.00 frames. ], tot_loss[loss=0.2359, simple_loss=0.3045, pruned_loss=0.0836, over 3712967.05 frames. ], batch size: 69, lr: 1.42e-02, grad_scale: 8.0 2022-12-22 21:56:21,393 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0_sp0.9 from training. Duration: 22.9444375 2022-12-22 21:56:26,185 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6993, 2.1578, 1.7864, 0.7825, 1.7133, 1.9932, 1.2952, 1.8646], device='cuda:3'), covar=tensor([0.0609, 0.0754, 0.1554, 0.1962, 0.1519, 0.1559, 0.2211, 0.0993], device='cuda:3'), in_proj_covar=tensor([0.0159, 0.0175, 0.0201, 0.0193, 0.0206, 0.0186, 0.0200, 0.0190], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 21:56:28,851 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp1.1 from training. Duration: 21.6318125 2022-12-22 21:56:48,848 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0 from training. Duration: 23.695 2022-12-22 21:56:58,623 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.1851, 2.7327, 2.7221, 3.1117, 2.7807, 2.8708, 3.2935, 1.0766], device='cuda:3'), covar=tensor([0.1101, 0.0827, 0.0854, 0.1000, 0.2076, 0.1232, 0.0848, 0.4474], device='cuda:3'), in_proj_covar=tensor([0.0277, 0.0186, 0.0192, 0.0195, 0.0268, 0.0226, 0.0223, 0.0241], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 21:57:14,398 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0 from training. Duration: 23.955 2022-12-22 21:57:17,437 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5631, 0.9424, 1.9263, 3.2980, 2.1247, 2.1815, 0.6079, 2.1201], device='cuda:3'), covar=tensor([0.2215, 0.2644, 0.1923, 0.0676, 0.1594, 0.1704, 0.3147, 0.1607], device='cuda:3'), in_proj_covar=tensor([0.0104, 0.0113, 0.0127, 0.0118, 0.0104, 0.0130, 0.0131, 0.0107], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 21:57:36,463 INFO [train.py:894] (3/4) Epoch 8, batch 2950, loss[loss=0.2518, simple_loss=0.2957, pruned_loss=0.1039, over 18508.00 frames. ], tot_loss[loss=0.2363, simple_loss=0.305, pruned_loss=0.08382, over 3713519.40 frames. ], batch size: 44, lr: 1.41e-02, grad_scale: 8.0 2022-12-22 21:57:46,479 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp0.9 from training. Duration: 26.438875 2022-12-22 21:57:53,229 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27504.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 21:58:19,629 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.835e+02 5.842e+02 7.166e+02 9.533e+02 2.194e+03, threshold=1.433e+03, percent-clipped=8.0 2022-12-22 21:58:27,484 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0021-26306-0_sp0.9 from training. Duration: 21.2444375 2022-12-22 21:58:29,133 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp0.9 from training. Duration: 31.02225 2022-12-22 21:58:33,428 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6530, 0.6414, 1.4431, 1.3372, 1.6398, 1.6034, 1.4265, 1.2373], device='cuda:3'), covar=tensor([0.1317, 0.2126, 0.1692, 0.1573, 0.1201, 0.0656, 0.1561, 0.0839], device='cuda:3'), in_proj_covar=tensor([0.0239, 0.0272, 0.0246, 0.0274, 0.0258, 0.0224, 0.0277, 0.0212], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 21:58:38,846 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0 from training. Duration: 22.395 2022-12-22 21:58:54,002 INFO [train.py:894] (3/4) Epoch 8, batch 3000, loss[loss=0.2281, simple_loss=0.2857, pruned_loss=0.0852, over 18530.00 frames. ], tot_loss[loss=0.2377, simple_loss=0.306, pruned_loss=0.08466, over 3715017.83 frames. ], batch size: 44, lr: 1.41e-02, grad_scale: 8.0 2022-12-22 21:58:54,003 INFO [train.py:919] (3/4) Computing validation loss 2022-12-22 21:59:05,190 INFO [train.py:928] (3/4) Epoch 8, validation: loss=0.1857, simple_loss=0.2837, pruned_loss=0.04382, over 944034.00 frames. 2022-12-22 21:59:05,190 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24676MB 2022-12-22 21:59:08,088 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0 from training. Duration: 21.075 2022-12-22 21:59:14,139 WARNING [train.py:1060] (3/4) Exclude cut with ID 6482-98857-0025-147532-0_sp0.9 from training. Duration: 20.0055625 2022-12-22 21:59:14,152 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0037-132304-0_sp0.9 from training. Duration: 22.05 2022-12-22 21:59:14,166 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0 from training. Duration: 26.8349375 2022-12-22 21:59:17,095 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp1.1 from training. Duration: 22.1090625 2022-12-22 21:59:24,297 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp0.9 from training. Duration: 26.6166875 2022-12-22 21:59:29,695 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27560.0, num_to_drop=1, layers_to_drop={2} 2022-12-22 21:59:42,656 WARNING [train.py:1060] (3/4) Exclude cut with ID 2046-178027-0000-53705-0_sp0.9 from training. Duration: 20.3055625 2022-12-22 22:00:05,960 WARNING [train.py:1060] (3/4) Exclude cut with ID 7205-50138-0008-5373-0_sp0.9 from training. Duration: 20.7 2022-12-22 22:00:21,700 INFO [train.py:894] (3/4) Epoch 8, batch 3050, loss[loss=0.2097, simple_loss=0.2763, pruned_loss=0.07159, over 18401.00 frames. ], tot_loss[loss=0.2378, simple_loss=0.3058, pruned_loss=0.08488, over 3714817.74 frames. ], batch size: 42, lr: 1.41e-02, grad_scale: 8.0 2022-12-22 22:00:39,833 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27606.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:00:43,302 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=27608.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 22:00:51,736 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0 from training. Duration: 22.48 2022-12-22 22:00:59,897 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2022-12-22 22:01:03,207 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.695e+02 5.726e+02 6.987e+02 9.097e+02 1.949e+03, threshold=1.397e+03, percent-clipped=1.0 2022-12-22 22:01:06,776 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp0.9 from training. Duration: 29.816625 2022-12-22 22:01:27,374 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp1.1 from training. Duration: 22.7590625 2022-12-22 22:01:31,866 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0 from training. Duration: 22.555 2022-12-22 22:01:37,859 INFO [train.py:894] (3/4) Epoch 8, batch 3100, loss[loss=0.2441, simple_loss=0.3271, pruned_loss=0.08056, over 18581.00 frames. ], tot_loss[loss=0.2369, simple_loss=0.3054, pruned_loss=0.08427, over 3714512.85 frames. ], batch size: 57, lr: 1.41e-02, grad_scale: 8.0 2022-12-22 22:01:53,054 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0005-25975-0_sp0.9 from training. Duration: 21.688875 2022-12-22 22:01:53,215 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=27654.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:02:27,542 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0_sp0.9 from training. Duration: 22.6 2022-12-22 22:02:54,056 INFO [train.py:894] (3/4) Epoch 8, batch 3150, loss[loss=0.2498, simple_loss=0.324, pruned_loss=0.08778, over 18531.00 frames. ], tot_loss[loss=0.2374, simple_loss=0.3056, pruned_loss=0.08459, over 3715052.86 frames. ], batch size: 64, lr: 1.41e-02, grad_scale: 8.0 2022-12-22 22:03:04,804 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0 from training. Duration: 24.32 2022-12-22 22:03:11,621 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3662, 2.1147, 1.5146, 2.3147, 1.8291, 1.8167, 1.9848, 2.3749], device='cuda:3'), covar=tensor([0.1540, 0.2300, 0.1483, 0.2316, 0.2449, 0.0898, 0.2051, 0.0594], device='cuda:3'), in_proj_covar=tensor([0.0269, 0.0258, 0.0219, 0.0337, 0.0243, 0.0210, 0.0253, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 22:03:25,349 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=27714.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 22:03:36,992 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.801e+02 5.956e+02 7.094e+02 8.866e+02 2.469e+03, threshold=1.419e+03, percent-clipped=7.0 2022-12-22 22:03:42,600 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0741, 1.9410, 1.3479, 2.1032, 1.6548, 1.7479, 1.8089, 2.1159], device='cuda:3'), covar=tensor([0.1733, 0.2294, 0.1593, 0.2150, 0.2433, 0.0932, 0.2095, 0.0648], device='cuda:3'), in_proj_covar=tensor([0.0269, 0.0258, 0.0219, 0.0335, 0.0243, 0.0210, 0.0254, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 22:03:45,462 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.3929, 2.3816, 1.7496, 3.1265, 2.1274, 2.2928, 2.6208, 3.5678], device='cuda:3'), covar=tensor([0.1358, 0.2542, 0.1490, 0.2512, 0.3053, 0.0808, 0.2225, 0.0435], device='cuda:3'), in_proj_covar=tensor([0.0269, 0.0258, 0.0219, 0.0335, 0.0242, 0.0209, 0.0254, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 22:03:46,889 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4659, 2.1087, 1.3936, 2.1061, 2.6484, 2.3572, 2.4008, 2.5733], device='cuda:3'), covar=tensor([0.1188, 0.1661, 0.2274, 0.1157, 0.1451, 0.1228, 0.1000, 0.1094], device='cuda:3'), in_proj_covar=tensor([0.0091, 0.0103, 0.0123, 0.0098, 0.0114, 0.0092, 0.0096, 0.0096], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 22:04:05,115 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-276745-0093-13116-0_sp0.9 from training. Duration: 21.061125 2022-12-22 22:04:10,929 INFO [train.py:894] (3/4) Epoch 8, batch 3200, loss[loss=0.2189, simple_loss=0.2967, pruned_loss=0.07055, over 18582.00 frames. ], tot_loss[loss=0.2372, simple_loss=0.3056, pruned_loss=0.08437, over 3714424.97 frames. ], batch size: 51, lr: 1.41e-02, grad_scale: 8.0 2022-12-22 22:04:19,065 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0024-15855-0_sp0.9 from training. Duration: 20.32225 2022-12-22 22:04:31,516 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp1.1 from training. Duration: 0.7545625 2022-12-22 22:04:46,421 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0_sp0.9 from training. Duration: 23.9333125 2022-12-22 22:04:59,810 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27775.0, num_to_drop=1, layers_to_drop={2} 2022-12-22 22:05:17,612 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp1.1 from training. Duration: 20.17275 2022-12-22 22:05:22,895 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8738, 1.4925, 0.9928, 1.5652, 1.9722, 1.6638, 1.8571, 2.0883], device='cuda:3'), covar=tensor([0.1562, 0.2118, 0.2699, 0.1520, 0.1885, 0.1576, 0.1353, 0.1445], device='cuda:3'), in_proj_covar=tensor([0.0091, 0.0103, 0.0123, 0.0098, 0.0114, 0.0092, 0.0097, 0.0096], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 22:05:24,048 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp1.1 from training. Duration: 20.436375 2022-12-22 22:05:28,398 INFO [train.py:894] (3/4) Epoch 8, batch 3250, loss[loss=0.2626, simple_loss=0.3215, pruned_loss=0.1019, over 18680.00 frames. ], tot_loss[loss=0.2374, simple_loss=0.3059, pruned_loss=0.08447, over 3714205.65 frames. ], batch size: 178, lr: 1.41e-02, grad_scale: 8.0 2022-12-22 22:05:43,076 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4283, 2.1778, 1.6237, 2.2446, 1.8190, 1.8373, 1.9302, 2.5245], device='cuda:3'), covar=tensor([0.1551, 0.2197, 0.1406, 0.2251, 0.2705, 0.0877, 0.2100, 0.0539], device='cuda:3'), in_proj_covar=tensor([0.0267, 0.0255, 0.0218, 0.0334, 0.0241, 0.0208, 0.0253, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 22:05:44,361 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27804.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:06:11,578 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.550e+02 6.210e+02 7.285e+02 9.598e+02 1.821e+03, threshold=1.457e+03, percent-clipped=2.0 2022-12-22 22:06:43,542 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0_sp0.9 from training. Duration: 23.1055625 2022-12-22 22:06:44,987 INFO [train.py:894] (3/4) Epoch 8, batch 3300, loss[loss=0.2203, simple_loss=0.2769, pruned_loss=0.08183, over 18406.00 frames. ], tot_loss[loss=0.2366, simple_loss=0.3051, pruned_loss=0.08406, over 3714259.56 frames. ], batch size: 42, lr: 1.41e-02, grad_scale: 8.0 2022-12-22 22:06:45,767 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp1.1 from training. Duration: 23.5 2022-12-22 22:06:56,523 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp0.9 from training. Duration: 26.62775 2022-12-22 22:06:58,424 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=27852.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:06:58,851 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0740, 1.2030, 1.7589, 1.7066, 1.9792, 1.8885, 1.7389, 1.3964], device='cuda:3'), covar=tensor([0.1379, 0.2237, 0.1695, 0.1770, 0.1190, 0.0662, 0.1896, 0.0872], device='cuda:3'), in_proj_covar=tensor([0.0241, 0.0274, 0.0247, 0.0277, 0.0259, 0.0224, 0.0279, 0.0213], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 22:07:10,209 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0 from training. Duration: 21.105 2022-12-22 22:07:15,214 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0_sp0.9 from training. Duration: 24.411125 2022-12-22 22:07:42,980 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp1.1 from training. Duration: 21.263625 2022-12-22 22:08:02,678 INFO [train.py:894] (3/4) Epoch 8, batch 3350, loss[loss=0.2592, simple_loss=0.3193, pruned_loss=0.09957, over 18472.00 frames. ], tot_loss[loss=0.2365, simple_loss=0.3054, pruned_loss=0.08382, over 3714596.82 frames. ], batch size: 50, lr: 1.40e-02, grad_scale: 8.0 2022-12-22 22:08:16,292 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0 from training. Duration: 20.795 2022-12-22 22:08:27,244 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0 from training. Duration: 24.76 2022-12-22 22:08:27,267 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0_sp0.9 from training. Duration: 22.25 2022-12-22 22:08:45,535 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.385e+02 6.137e+02 7.435e+02 8.806e+02 1.482e+03, threshold=1.487e+03, percent-clipped=1.0 2022-12-22 22:08:52,549 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp1.1 from training. Duration: 20.5045625 2022-12-22 22:09:20,022 INFO [train.py:894] (3/4) Epoch 8, batch 3400, loss[loss=0.2381, simple_loss=0.3098, pruned_loss=0.08326, over 18597.00 frames. ], tot_loss[loss=0.2349, simple_loss=0.3039, pruned_loss=0.08293, over 3714923.31 frames. ], batch size: 57, lr: 1.40e-02, grad_scale: 8.0 2022-12-22 22:09:32,058 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5110, 1.4869, 1.5545, 1.6705, 1.2755, 3.6960, 1.8363, 2.1901], device='cuda:3'), covar=tensor([0.3357, 0.2083, 0.1996, 0.1932, 0.1404, 0.0178, 0.1416, 0.0892], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0119, 0.0132, 0.0122, 0.0105, 0.0101, 0.0100, 0.0098], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2022-12-22 22:10:01,211 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2022-12-22 22:10:32,710 INFO [train.py:894] (3/4) Epoch 8, batch 3450, loss[loss=0.2091, simple_loss=0.2698, pruned_loss=0.07424, over 18716.00 frames. ], tot_loss[loss=0.2351, simple_loss=0.3039, pruned_loss=0.0832, over 3715739.26 frames. ], batch size: 41, lr: 1.40e-02, grad_scale: 8.0 2022-12-22 22:10:37,956 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2022-12-22 22:11:16,876 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.454e+02 5.907e+02 7.440e+02 9.598e+02 2.157e+03, threshold=1.488e+03, percent-clipped=3.0 2022-12-22 22:11:45,749 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2022-12-22 22:11:49,489 INFO [train.py:894] (3/4) Epoch 8, batch 3500, loss[loss=0.2837, simple_loss=0.3355, pruned_loss=0.1159, over 18605.00 frames. ], tot_loss[loss=0.2367, simple_loss=0.3052, pruned_loss=0.08411, over 3716006.83 frames. ], batch size: 174, lr: 1.40e-02, grad_scale: 8.0 2022-12-22 22:12:11,339 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp1.1 from training. Duration: 22.2954375 2022-12-22 22:12:22,581 INFO [train.py:894] (3/4) Epoch 9, batch 0, loss[loss=0.2467, simple_loss=0.3199, pruned_loss=0.08677, over 18457.00 frames. ], tot_loss[loss=0.2467, simple_loss=0.3199, pruned_loss=0.08677, over 18457.00 frames. ], batch size: 64, lr: 1.33e-02, grad_scale: 8.0 2022-12-22 22:12:22,581 INFO [train.py:919] (3/4) Computing validation loss 2022-12-22 22:12:33,731 INFO [train.py:928] (3/4) Epoch 9, validation: loss=0.1816, simple_loss=0.2807, pruned_loss=0.04127, over 944034.00 frames. 2022-12-22 22:12:33,732 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24676MB 2022-12-22 22:12:56,265 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.06 vs. limit=5.0 2022-12-22 22:13:06,155 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28070.0, num_to_drop=1, layers_to_drop={2} 2022-12-22 22:13:26,230 WARNING [train.py:1060] (3/4) Exclude cut with ID 298-126791-0067-24026-0_sp0.9 from training. Duration: 21.438875 2022-12-22 22:13:32,551 WARNING [train.py:1060] (3/4) Exclude cut with ID 5652-39938-0025-23684-0_sp0.9 from training. Duration: 22.2055625 2022-12-22 22:13:52,623 INFO [train.py:894] (3/4) Epoch 9, batch 50, loss[loss=0.212, simple_loss=0.301, pruned_loss=0.06154, over 18470.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.2964, pruned_loss=0.06875, over 837845.90 frames. ], batch size: 54, lr: 1.32e-02, grad_scale: 8.0 2022-12-22 22:13:59,926 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2022-12-22 22:14:26,945 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.901e+02 4.305e+02 5.351e+02 6.877e+02 1.696e+03, threshold=1.070e+03, percent-clipped=2.0 2022-12-22 22:14:46,542 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5869, 2.3908, 1.5325, 0.9065, 2.9749, 2.7108, 2.1317, 1.6710], device='cuda:3'), covar=tensor([0.0313, 0.0328, 0.0650, 0.0842, 0.0140, 0.0311, 0.0483, 0.0789], device='cuda:3'), in_proj_covar=tensor([0.0117, 0.0116, 0.0130, 0.0119, 0.0082, 0.0116, 0.0135, 0.0150], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-22 22:15:02,486 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0953, 1.9030, 1.4661, 0.7679, 2.3574, 2.0614, 1.7213, 1.3737], device='cuda:3'), covar=tensor([0.0305, 0.0354, 0.0547, 0.0782, 0.0204, 0.0330, 0.0499, 0.0865], device='cuda:3'), in_proj_covar=tensor([0.0117, 0.0115, 0.0129, 0.0119, 0.0082, 0.0116, 0.0134, 0.0150], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-22 22:15:07,906 INFO [train.py:894] (3/4) Epoch 9, batch 100, loss[loss=0.2477, simple_loss=0.3208, pruned_loss=0.08731, over 18680.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2925, pruned_loss=0.06538, over 1475126.44 frames. ], batch size: 60, lr: 1.32e-02, grad_scale: 8.0 2022-12-22 22:16:22,822 INFO [train.py:894] (3/4) Epoch 9, batch 150, loss[loss=0.2064, simple_loss=0.2908, pruned_loss=0.06095, over 18530.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.291, pruned_loss=0.06484, over 1971450.19 frames. ], batch size: 58, lr: 1.32e-02, grad_scale: 8.0 2022-12-22 22:16:42,107 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0 from training. Duration: 24.525 2022-12-22 22:16:57,975 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.622e+02 4.291e+02 5.287e+02 6.560e+02 1.269e+03, threshold=1.057e+03, percent-clipped=2.0 2022-12-22 22:17:09,754 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28229.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:17:13,942 WARNING [train.py:1060] (3/4) Exclude cut with ID 3699-47246-0007-3408-0_sp0.9 from training. Duration: 20.26675 2022-12-22 22:17:27,236 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp0.9 from training. Duration: 27.25 2022-12-22 22:17:41,454 INFO [train.py:894] (3/4) Epoch 9, batch 200, loss[loss=0.2187, simple_loss=0.2939, pruned_loss=0.07179, over 18528.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2871, pruned_loss=0.06323, over 2357446.18 frames. ], batch size: 47, lr: 1.32e-02, grad_scale: 8.0 2022-12-22 22:17:49,000 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28255.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 22:18:38,217 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28287.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:18:42,376 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28290.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:18:44,747 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0 from training. Duration: 21.68 2022-12-22 22:18:55,179 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0 from training. Duration: 21.6300625 2022-12-22 22:18:57,279 INFO [train.py:894] (3/4) Epoch 9, batch 250, loss[loss=0.1979, simple_loss=0.2852, pruned_loss=0.05533, over 18561.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2873, pruned_loss=0.06305, over 2657666.44 frames. ], batch size: 55, lr: 1.32e-02, grad_scale: 8.0 2022-12-22 22:19:20,937 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0_sp0.9 from training. Duration: 24.033375 2022-12-22 22:19:22,798 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28316.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 22:19:31,402 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.623e+02 3.974e+02 4.954e+02 5.974e+02 1.449e+03, threshold=9.907e+02, percent-clipped=4.0 2022-12-22 22:20:11,011 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28348.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:20:13,529 INFO [train.py:894] (3/4) Epoch 9, batch 300, loss[loss=0.1925, simple_loss=0.2742, pruned_loss=0.05542, over 18554.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2866, pruned_loss=0.06208, over 2891305.21 frames. ], batch size: 47, lr: 1.32e-02, grad_scale: 8.0 2022-12-22 22:20:15,298 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0 from training. Duration: 22.905 2022-12-22 22:20:16,871 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp1.1 from training. Duration: 23.4318125 2022-12-22 22:20:45,321 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28370.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 22:21:03,394 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7373, 0.9770, 0.7249, 1.2087, 1.9926, 1.1080, 1.1935, 1.3979], device='cuda:3'), covar=tensor([0.2193, 0.3590, 0.3430, 0.2399, 0.2261, 0.2533, 0.2598, 0.2882], device='cuda:3'), in_proj_covar=tensor([0.0091, 0.0103, 0.0122, 0.0098, 0.0114, 0.0093, 0.0097, 0.0097], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 22:21:26,036 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7543, 1.5652, 1.3023, 2.0967, 1.7341, 3.6178, 1.5609, 1.7719], device='cuda:3'), covar=tensor([0.0888, 0.1859, 0.1213, 0.0846, 0.1492, 0.0219, 0.1288, 0.1511], device='cuda:3'), in_proj_covar=tensor([0.0078, 0.0086, 0.0078, 0.0077, 0.0097, 0.0073, 0.0088, 0.0081], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-22 22:21:29,907 INFO [train.py:894] (3/4) Epoch 9, batch 350, loss[loss=0.2189, simple_loss=0.3004, pruned_loss=0.06869, over 18698.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2874, pruned_loss=0.0626, over 3073754.73 frames. ], batch size: 65, lr: 1.32e-02, grad_scale: 8.0 2022-12-22 22:21:58,765 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=28418.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 22:22:01,800 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5813, 2.1265, 1.5917, 2.3705, 2.8446, 1.5444, 1.7619, 1.1894], device='cuda:3'), covar=tensor([0.1849, 0.1503, 0.1455, 0.0869, 0.1142, 0.1116, 0.1702, 0.1518], device='cuda:3'), in_proj_covar=tensor([0.0234, 0.0206, 0.0201, 0.0184, 0.0249, 0.0186, 0.0205, 0.0187], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 22:22:04,378 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.510e+02 4.571e+02 5.481e+02 6.652e+02 2.001e+03, threshold=1.096e+03, percent-clipped=7.0 2022-12-22 22:22:16,110 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp1.1 from training. Duration: 20.82275 2022-12-22 22:22:17,501 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp0.9 from training. Duration: 25.45 2022-12-22 22:22:46,142 INFO [train.py:894] (3/4) Epoch 9, batch 400, loss[loss=0.2491, simple_loss=0.3311, pruned_loss=0.0835, over 18520.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2894, pruned_loss=0.06411, over 3215936.27 frames. ], batch size: 64, lr: 1.32e-02, grad_scale: 8.0 2022-12-22 22:23:02,452 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.95 vs. limit=5.0 2022-12-22 22:23:15,016 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0 from training. Duration: 25.775 2022-12-22 22:23:35,645 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0_sp0.9 from training. Duration: 22.25 2022-12-22 22:23:41,918 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28486.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:23:51,276 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28492.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:24:03,934 INFO [train.py:894] (3/4) Epoch 9, batch 450, loss[loss=0.2259, simple_loss=0.2875, pruned_loss=0.08219, over 18507.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2911, pruned_loss=0.06534, over 3325469.83 frames. ], batch size: 43, lr: 1.32e-02, grad_scale: 8.0 2022-12-22 22:24:04,000 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0 from training. Duration: 26.205 2022-12-22 22:24:20,858 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp0.9 from training. Duration: 30.1555625 2022-12-22 22:24:22,987 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2022-12-22 22:24:26,216 WARNING [train.py:1060] (3/4) Exclude cut with ID 1265-135635-0050-6781-0_sp0.9 from training. Duration: 21.8333125 2022-12-22 22:24:35,450 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp1.1 from training. Duration: 20.6545625 2022-12-22 22:24:36,882 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.819e+02 4.572e+02 5.901e+02 7.346e+02 1.900e+03, threshold=1.180e+03, percent-clipped=4.0 2022-12-22 22:24:41,469 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5090, 1.8798, 1.4400, 2.2842, 2.5192, 1.5057, 1.4591, 1.1920], device='cuda:3'), covar=tensor([0.1960, 0.1700, 0.1635, 0.0904, 0.1261, 0.1269, 0.1879, 0.1588], device='cuda:3'), in_proj_covar=tensor([0.0235, 0.0207, 0.0201, 0.0185, 0.0248, 0.0186, 0.0205, 0.0187], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 22:24:51,630 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.40 vs. limit=5.0 2022-12-22 22:24:54,185 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4628, 0.9106, 1.1944, 1.1732, 1.6356, 1.5102, 1.5770, 1.0548], device='cuda:3'), covar=tensor([0.0271, 0.0255, 0.0432, 0.0235, 0.0174, 0.0312, 0.0202, 0.0290], device='cuda:3'), in_proj_covar=tensor([0.0086, 0.0119, 0.0140, 0.0123, 0.0108, 0.0105, 0.0090, 0.0119], device='cuda:3'), out_proj_covar=tensor([7.5985e-05, 1.0319e-04, 1.2737e-04, 1.0734e-04, 9.7573e-05, 8.8934e-05, 7.8063e-05, 1.0248e-04], device='cuda:3') 2022-12-22 22:25:08,560 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.93 vs. limit=2.0 2022-12-22 22:25:14,235 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28547.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:25:18,610 INFO [train.py:894] (3/4) Epoch 9, batch 500, loss[loss=0.2269, simple_loss=0.3136, pruned_loss=0.07012, over 18650.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2923, pruned_loss=0.06617, over 3409832.09 frames. ], batch size: 60, lr: 1.31e-02, grad_scale: 8.0 2022-12-22 22:25:18,694 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0045-39920-0_sp0.9 from training. Duration: 20.52225 2022-12-22 22:25:23,635 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28553.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:25:31,522 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3022, 1.5739, 2.4048, 4.3782, 3.3110, 2.7249, 0.8659, 2.9860], device='cuda:3'), covar=tensor([0.1639, 0.1910, 0.1550, 0.0371, 0.0960, 0.1147, 0.2479, 0.1066], device='cuda:3'), in_proj_covar=tensor([0.0103, 0.0113, 0.0125, 0.0119, 0.0104, 0.0131, 0.0129, 0.0108], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 22:25:38,759 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp0.9 from training. Duration: 29.1166875 2022-12-22 22:26:12,604 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28585.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:26:36,467 INFO [train.py:894] (3/4) Epoch 9, batch 550, loss[loss=0.2363, simple_loss=0.3135, pruned_loss=0.0795, over 18530.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2931, pruned_loss=0.06659, over 3477811.48 frames. ], batch size: 58, lr: 1.31e-02, grad_scale: 16.0 2022-12-22 22:26:41,220 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133211-0007-59831-0_sp0.9 from training. Duration: 21.388875 2022-12-22 22:26:53,836 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28611.0, num_to_drop=1, layers_to_drop={3} 2022-12-22 22:27:10,200 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.016e+02 4.589e+02 5.717e+02 7.953e+02 1.813e+03, threshold=1.143e+03, percent-clipped=5.0 2022-12-22 22:27:17,441 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0 from training. Duration: 22.72 2022-12-22 22:27:18,870 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0_sp0.9 from training. Duration: 22.7444375 2022-12-22 22:27:42,203 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28643.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:27:53,952 INFO [train.py:894] (3/4) Epoch 9, batch 600, loss[loss=0.1748, simple_loss=0.2533, pruned_loss=0.04813, over 18475.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2932, pruned_loss=0.0667, over 3529980.88 frames. ], batch size: 43, lr: 1.31e-02, grad_scale: 16.0 2022-12-22 22:28:00,850 WARNING [train.py:1060] (3/4) Exclude cut with ID 4133-6541-0027-40495-0_sp1.1 from training. Duration: 0.9681875 2022-12-22 22:28:05,043 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0_sp0.9 from training. Duration: 22.3166875 2022-12-22 22:28:09,577 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133212-0015-59917-0_sp0.9 from training. Duration: 21.8166875 2022-12-22 22:29:01,060 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2022-12-22 22:29:09,494 INFO [train.py:894] (3/4) Epoch 9, batch 650, loss[loss=0.1937, simple_loss=0.2722, pruned_loss=0.0576, over 18663.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.2939, pruned_loss=0.06702, over 3570467.44 frames. ], batch size: 46, lr: 1.31e-02, grad_scale: 16.0 2022-12-22 22:29:40,978 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4187, 1.0147, 1.6053, 2.7811, 1.9055, 2.1541, 0.5372, 1.8587], device='cuda:3'), covar=tensor([0.1897, 0.1942, 0.1566, 0.0650, 0.1176, 0.1211, 0.2629, 0.1438], device='cuda:3'), in_proj_covar=tensor([0.0103, 0.0112, 0.0123, 0.0118, 0.0102, 0.0129, 0.0127, 0.0108], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 22:29:46,464 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.349e+02 4.652e+02 5.570e+02 7.061e+02 1.295e+03, threshold=1.114e+03, percent-clipped=4.0 2022-12-22 22:29:56,849 WARNING [train.py:1060] (3/4) Exclude cut with ID 4957-30119-0041-23990-0_sp0.9 from training. Duration: 20.22775 2022-12-22 22:30:29,677 INFO [train.py:894] (3/4) Epoch 9, batch 700, loss[loss=0.275, simple_loss=0.3435, pruned_loss=0.1032, over 18583.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2932, pruned_loss=0.06664, over 3602182.07 frames. ], batch size: 56, lr: 1.31e-02, grad_scale: 16.0 2022-12-22 22:30:41,322 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp1.1 from training. Duration: 24.67275 2022-12-22 22:31:04,986 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2899, 1.7216, 2.1940, 4.4380, 3.1398, 2.5682, 0.6186, 2.7839], device='cuda:3'), covar=tensor([0.1500, 0.1656, 0.1426, 0.0388, 0.0913, 0.1111, 0.2528, 0.1011], device='cuda:3'), in_proj_covar=tensor([0.0104, 0.0113, 0.0124, 0.0118, 0.0103, 0.0131, 0.0129, 0.0109], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 22:31:08,179 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.2644, 2.6074, 3.2548, 0.8800, 2.6645, 3.3892, 2.3507, 2.8829], device='cuda:3'), covar=tensor([0.0748, 0.0376, 0.0399, 0.0503, 0.0429, 0.0252, 0.0368, 0.0503], device='cuda:3'), in_proj_covar=tensor([0.0130, 0.0150, 0.0107, 0.0129, 0.0136, 0.0118, 0.0136, 0.0139], device='cuda:3'), out_proj_covar=tensor([1.1184e-04, 1.3195e-04, 9.1385e-05, 1.0993e-04, 1.1729e-04, 1.0339e-04, 1.2076e-04, 1.2203e-04], device='cuda:3') 2022-12-22 22:31:09,780 WARNING [train.py:1060] (3/4) Exclude cut with ID 3082-165428-0081-50734-0_sp0.9 from training. Duration: 21.8055625 2022-12-22 22:31:23,153 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.0574, 3.4271, 2.8845, 1.6051, 2.7929, 2.8530, 2.5390, 2.6205], device='cuda:3'), covar=tensor([0.0418, 0.0462, 0.1186, 0.1503, 0.1518, 0.0938, 0.1011, 0.0849], device='cuda:3'), in_proj_covar=tensor([0.0157, 0.0175, 0.0197, 0.0190, 0.0203, 0.0182, 0.0194, 0.0186], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 22:31:23,179 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.2693, 2.7361, 3.3851, 0.9011, 2.7087, 3.6429, 2.4438, 3.1781], device='cuda:3'), covar=tensor([0.0671, 0.0382, 0.0311, 0.0476, 0.0413, 0.0271, 0.0367, 0.0422], device='cuda:3'), in_proj_covar=tensor([0.0130, 0.0150, 0.0107, 0.0129, 0.0136, 0.0118, 0.0136, 0.0139], device='cuda:3'), out_proj_covar=tensor([1.1209e-04, 1.3194e-04, 9.1807e-05, 1.0985e-04, 1.1741e-04, 1.0337e-04, 1.2115e-04, 1.2173e-04], device='cuda:3') 2022-12-22 22:31:45,293 INFO [train.py:894] (3/4) Epoch 9, batch 750, loss[loss=0.2629, simple_loss=0.3313, pruned_loss=0.09727, over 18731.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2934, pruned_loss=0.0668, over 3625779.95 frames. ], batch size: 65, lr: 1.31e-02, grad_scale: 16.0 2022-12-22 22:31:45,379 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0_sp0.9 from training. Duration: 22.6666875 2022-12-22 22:32:18,225 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.703e+02 4.340e+02 5.308e+02 6.932e+02 2.206e+03, threshold=1.062e+03, percent-clipped=3.0 2022-12-22 22:32:48,937 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28842.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:32:50,211 WARNING [train.py:1060] (3/4) Exclude cut with ID 2411-132532-0017-82279-0_sp1.1 from training. Duration: 0.9681875 2022-12-22 22:32:58,379 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28848.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:33:01,174 INFO [train.py:894] (3/4) Epoch 9, batch 800, loss[loss=0.189, simple_loss=0.2685, pruned_loss=0.05474, over 18665.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2924, pruned_loss=0.06617, over 3646074.99 frames. ], batch size: 48, lr: 1.31e-02, grad_scale: 16.0 2022-12-22 22:33:16,216 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0 from training. Duration: 22.485 2022-12-22 22:33:18,047 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28861.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:33:54,450 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28885.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:33:55,765 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp1.1 from training. Duration: 23.82275 2022-12-22 22:34:09,447 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0 from training. Duration: 20.77 2022-12-22 22:34:17,218 INFO [train.py:894] (3/4) Epoch 9, batch 850, loss[loss=0.2078, simple_loss=0.2952, pruned_loss=0.06023, over 18703.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2919, pruned_loss=0.06583, over 3661115.25 frames. ], batch size: 60, lr: 1.31e-02, grad_scale: 16.0 2022-12-22 22:34:17,316 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0_sp0.9 from training. Duration: 24.088875 2022-12-22 22:34:33,680 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28911.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 22:34:46,524 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp1.1 from training. Duration: 20.4409375 2022-12-22 22:34:49,304 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.759e+02 4.128e+02 5.297e+02 6.595e+02 1.319e+03, threshold=1.059e+03, percent-clipped=1.0 2022-12-22 22:34:50,460 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28922.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:35:05,811 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=28933.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:35:21,034 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28943.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:35:28,981 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7716, 1.6294, 1.4330, 1.5222, 1.7861, 2.0754, 2.0635, 1.4080], device='cuda:3'), covar=tensor([0.0317, 0.0250, 0.0436, 0.0241, 0.0178, 0.0268, 0.0257, 0.0256], device='cuda:3'), in_proj_covar=tensor([0.0084, 0.0117, 0.0138, 0.0121, 0.0107, 0.0104, 0.0088, 0.0114], device='cuda:3'), out_proj_covar=tensor([7.4491e-05, 1.0096e-04, 1.2581e-04, 1.0512e-04, 9.6039e-05, 8.7429e-05, 7.6192e-05, 9.8090e-05], device='cuda:3') 2022-12-22 22:35:31,621 INFO [train.py:894] (3/4) Epoch 9, batch 900, loss[loss=0.1995, simple_loss=0.2769, pruned_loss=0.06105, over 18476.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2921, pruned_loss=0.06564, over 3671991.05 frames. ], batch size: 43, lr: 1.31e-02, grad_scale: 16.0 2022-12-22 22:35:42,676 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7297, 0.6114, 1.5446, 1.3859, 1.7980, 1.7520, 1.3132, 1.3228], device='cuda:3'), covar=tensor([0.1480, 0.2296, 0.1750, 0.1797, 0.1241, 0.0720, 0.1892, 0.0906], device='cuda:3'), in_proj_covar=tensor([0.0246, 0.0277, 0.0250, 0.0285, 0.0265, 0.0226, 0.0285, 0.0215], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 22:35:43,964 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3239, 1.2239, 1.7496, 1.0428, 1.4256, 1.5137, 1.2114, 1.7010], device='cuda:3'), covar=tensor([0.0881, 0.1543, 0.0871, 0.1096, 0.0713, 0.0768, 0.1818, 0.0506], device='cuda:3'), in_proj_covar=tensor([0.0196, 0.0187, 0.0192, 0.0185, 0.0177, 0.0204, 0.0198, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 22:35:44,983 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=28959.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 22:36:04,323 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0_sp0.9 from training. Duration: 22.511125 2022-12-22 22:36:04,352 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0 from training. Duration: 20.675 2022-12-22 22:36:33,550 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=28991.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:36:47,763 INFO [train.py:894] (3/4) Epoch 9, batch 950, loss[loss=0.189, simple_loss=0.2767, pruned_loss=0.05066, over 18508.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2918, pruned_loss=0.06516, over 3680511.10 frames. ], batch size: 52, lr: 1.30e-02, grad_scale: 16.0 2022-12-22 22:36:51,407 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29002.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:36:53,135 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.5078, 1.8111, 2.1273, 0.8965, 1.4961, 2.3568, 1.9326, 1.8213], device='cuda:3'), covar=tensor([0.0579, 0.0341, 0.0315, 0.0367, 0.0311, 0.0262, 0.0204, 0.0488], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0152, 0.0106, 0.0128, 0.0135, 0.0118, 0.0137, 0.0140], device='cuda:3'), out_proj_covar=tensor([1.1345e-04, 1.3318e-04, 9.0962e-05, 1.0957e-04, 1.1659e-04, 1.0310e-04, 1.2111e-04, 1.2269e-04], device='cuda:3') 2022-12-22 22:37:21,573 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.190e+02 4.506e+02 5.344e+02 6.616e+02 1.256e+03, threshold=1.069e+03, percent-clipped=2.0 2022-12-22 22:37:43,312 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp0.9 from training. Duration: 24.9833125 2022-12-22 22:38:04,950 INFO [train.py:894] (3/4) Epoch 9, batch 1000, loss[loss=0.204, simple_loss=0.293, pruned_loss=0.05753, over 18494.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2924, pruned_loss=0.06539, over 3687712.33 frames. ], batch size: 52, lr: 1.30e-02, grad_scale: 16.0 2022-12-22 22:38:08,409 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4523, 1.9444, 1.8557, 1.6385, 2.0904, 2.8717, 2.5470, 2.1391], device='cuda:3'), covar=tensor([0.0302, 0.0343, 0.0406, 0.0298, 0.0244, 0.0365, 0.0359, 0.0257], device='cuda:3'), in_proj_covar=tensor([0.0084, 0.0116, 0.0138, 0.0121, 0.0107, 0.0103, 0.0088, 0.0115], device='cuda:3'), out_proj_covar=tensor([7.4368e-05, 1.0081e-04, 1.2499e-04, 1.0472e-04, 9.5943e-05, 8.6936e-05, 7.6944e-05, 9.8509e-05], device='cuda:3') 2022-12-22 22:38:15,813 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0 from training. Duration: 27.14 2022-12-22 22:38:25,186 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29063.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:38:29,319 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.0260, 4.2048, 3.9564, 1.7788, 4.0543, 3.1191, 0.6699, 2.8116], device='cuda:3'), covar=tensor([0.1767, 0.0865, 0.1232, 0.3776, 0.0876, 0.0985, 0.5720, 0.1486], device='cuda:3'), in_proj_covar=tensor([0.0131, 0.0114, 0.0145, 0.0116, 0.0120, 0.0101, 0.0140, 0.0107], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 22:38:30,606 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0 from training. Duration: 22.44 2022-12-22 22:38:51,254 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6245, 1.5858, 1.1816, 1.8733, 1.7074, 3.2555, 1.3939, 1.5532], device='cuda:3'), covar=tensor([0.0841, 0.1674, 0.1184, 0.0823, 0.1349, 0.0212, 0.1237, 0.1419], device='cuda:3'), in_proj_covar=tensor([0.0075, 0.0084, 0.0077, 0.0078, 0.0094, 0.0073, 0.0086, 0.0079], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2022-12-22 22:38:53,706 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2022-12-22 22:39:22,244 INFO [train.py:894] (3/4) Epoch 9, batch 1050, loss[loss=0.215, simple_loss=0.2824, pruned_loss=0.07378, over 18617.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2912, pruned_loss=0.06493, over 3693056.11 frames. ], batch size: 41, lr: 1.30e-02, grad_scale: 16.0 2022-12-22 22:39:27,790 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2022-12-22 22:39:51,687 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0060-62364-0_sp0.9 from training. Duration: 21.361125 2022-12-22 22:39:55,944 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.054e+02 4.752e+02 5.475e+02 6.695e+02 1.514e+03, threshold=1.095e+03, percent-clipped=3.0 2022-12-22 22:39:57,395 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp1.1 from training. Duration: 27.0318125 2022-12-22 22:40:06,958 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp0.9 from training. Duration: 28.638875 2022-12-22 22:40:20,745 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5782, 1.7639, 1.7607, 1.7860, 1.2257, 3.5776, 1.8508, 2.2686], device='cuda:3'), covar=tensor([0.3197, 0.1766, 0.1751, 0.1799, 0.1364, 0.0156, 0.1291, 0.0788], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0120, 0.0132, 0.0123, 0.0106, 0.0099, 0.0100, 0.0098], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-22 22:40:23,195 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0 from training. Duration: 20.4 2022-12-22 22:40:26,964 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29142.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:40:35,806 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29148.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:40:38,493 INFO [train.py:894] (3/4) Epoch 9, batch 1100, loss[loss=0.2278, simple_loss=0.3106, pruned_loss=0.07255, over 18558.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2911, pruned_loss=0.06448, over 3697990.26 frames. ], batch size: 77, lr: 1.30e-02, grad_scale: 16.0 2022-12-22 22:40:41,999 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.9886, 2.1527, 1.6161, 2.8750, 2.0467, 2.2083, 2.2454, 3.4472], device='cuda:3'), covar=tensor([0.1737, 0.2809, 0.1742, 0.2596, 0.3110, 0.1010, 0.2930, 0.0553], device='cuda:3'), in_proj_covar=tensor([0.0266, 0.0257, 0.0219, 0.0331, 0.0240, 0.0207, 0.0255, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 22:40:55,841 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0 from training. Duration: 20.025 2022-12-22 22:40:55,856 WARNING [train.py:1060] (3/4) Exclude cut with ID 2364-131735-0112-64612-0_sp0.9 from training. Duration: 20.488875 2022-12-22 22:41:02,631 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0 from training. Duration: 29.735 2022-12-22 22:41:40,809 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=29190.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:41:49,306 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=29196.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:41:54,966 INFO [train.py:894] (3/4) Epoch 9, batch 1150, loss[loss=0.1917, simple_loss=0.2901, pruned_loss=0.04666, over 18470.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2906, pruned_loss=0.06402, over 3700724.72 frames. ], batch size: 54, lr: 1.30e-02, grad_scale: 16.0 2022-12-22 22:42:05,694 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4727, 1.7817, 1.8284, 1.5496, 2.2007, 2.8682, 2.5334, 2.0128], device='cuda:3'), covar=tensor([0.0308, 0.0328, 0.0382, 0.0310, 0.0207, 0.0274, 0.0323, 0.0281], device='cuda:3'), in_proj_covar=tensor([0.0084, 0.0116, 0.0139, 0.0121, 0.0107, 0.0104, 0.0088, 0.0116], device='cuda:3'), out_proj_covar=tensor([7.4233e-05, 1.0061e-04, 1.2606e-04, 1.0513e-04, 9.5671e-05, 8.7034e-05, 7.6740e-05, 9.9329e-05], device='cuda:3') 2022-12-22 22:42:21,053 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29217.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:42:25,297 WARNING [train.py:1060] (3/4) Exclude cut with ID 7276-92427-0014-12983-0_sp0.9 from training. Duration: 21.3055625 2022-12-22 22:42:26,905 WARNING [train.py:1060] (3/4) Exclude cut with ID 1025-75365-0008-79168-0_sp0.9 from training. Duration: 22.0666875 2022-12-22 22:42:28,688 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.563e+02 4.647e+02 6.106e+02 7.932e+02 1.840e+03, threshold=1.221e+03, percent-clipped=6.0 2022-12-22 22:42:48,474 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2502, 2.7465, 2.7283, 1.2651, 2.8015, 1.9497, 0.6290, 1.8050], device='cuda:3'), covar=tensor([0.2220, 0.1252, 0.1632, 0.3781, 0.1254, 0.1375, 0.5119, 0.1860], device='cuda:3'), in_proj_covar=tensor([0.0129, 0.0114, 0.0144, 0.0116, 0.0120, 0.0100, 0.0139, 0.0106], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 22:42:50,100 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29236.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:43:11,132 INFO [train.py:894] (3/4) Epoch 9, batch 1200, loss[loss=0.1836, simple_loss=0.2587, pruned_loss=0.05428, over 18511.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2914, pruned_loss=0.0645, over 3702896.36 frames. ], batch size: 44, lr: 1.30e-02, grad_scale: 16.0 2022-12-22 22:44:15,642 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0 from training. Duration: 20.26 2022-12-22 22:44:23,538 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29297.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:44:27,742 INFO [train.py:894] (3/4) Epoch 9, batch 1250, loss[loss=0.1996, simple_loss=0.2885, pruned_loss=0.05532, over 18729.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2916, pruned_loss=0.06486, over 3705980.10 frames. ], batch size: 52, lr: 1.30e-02, grad_scale: 8.0 2022-12-22 22:44:29,301 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0030-9324-0_sp0.9 from training. Duration: 21.3444375 2022-12-22 22:45:03,216 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.891e+02 4.534e+02 5.383e+02 6.900e+02 1.545e+03, threshold=1.077e+03, percent-clipped=4.0 2022-12-22 22:45:25,692 WARNING [train.py:1060] (3/4) Exclude cut with ID 497-129325-0061-62254-0_sp1.1 from training. Duration: 0.97725 2022-12-22 22:45:43,222 INFO [train.py:894] (3/4) Epoch 9, batch 1300, loss[loss=0.2047, simple_loss=0.2776, pruned_loss=0.06588, over 18456.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2905, pruned_loss=0.06426, over 3707283.50 frames. ], batch size: 43, lr: 1.30e-02, grad_scale: 8.0 2022-12-22 22:45:55,772 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29358.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:46:08,700 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0_sp0.9 from training. Duration: 22.97225 2022-12-22 22:46:39,071 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0047-39922-0_sp0.9 from training. Duration: 21.97775 2022-12-22 22:46:53,510 WARNING [train.py:1060] (3/4) Exclude cut with ID 1112-1043-0006-89194-0_sp0.9 from training. Duration: 21.8333125 2022-12-22 22:46:59,058 INFO [train.py:894] (3/4) Epoch 9, batch 1350, loss[loss=0.1964, simple_loss=0.2768, pruned_loss=0.05795, over 18541.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2909, pruned_loss=0.06432, over 3708326.89 frames. ], batch size: 47, lr: 1.30e-02, grad_scale: 8.0 2022-12-22 22:47:03,867 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0 from training. Duration: 20.47 2022-12-22 22:47:35,243 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.581e+02 4.720e+02 6.038e+02 7.473e+02 1.219e+03, threshold=1.208e+03, percent-clipped=1.0 2022-12-22 22:48:11,277 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0037-39912-0_sp0.9 from training. Duration: 20.67225 2022-12-22 22:48:15,734 INFO [train.py:894] (3/4) Epoch 9, batch 1400, loss[loss=0.1639, simple_loss=0.245, pruned_loss=0.04139, over 18656.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2898, pruned_loss=0.06398, over 3709740.88 frames. ], batch size: 41, lr: 1.30e-02, grad_scale: 8.0 2022-12-22 22:48:31,020 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp0.9 from training. Duration: 25.2444375 2022-12-22 22:48:46,986 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4759, 2.2725, 1.7303, 2.2491, 1.9069, 1.9556, 2.1234, 2.6563], device='cuda:3'), covar=tensor([0.1439, 0.2335, 0.1268, 0.2350, 0.2467, 0.0824, 0.2009, 0.0545], device='cuda:3'), in_proj_covar=tensor([0.0269, 0.0261, 0.0221, 0.0334, 0.0244, 0.0209, 0.0258, 0.0181], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 22:48:50,607 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2022-12-22 22:48:54,652 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2022-12-22 22:48:55,303 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0021-76797-0_sp0.9 from training. Duration: 21.1445 2022-12-22 22:49:21,520 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0206, 1.7213, 1.7461, 1.0452, 2.4285, 2.1016, 1.8204, 1.4907], device='cuda:3'), covar=tensor([0.0314, 0.0422, 0.0428, 0.0680, 0.0175, 0.0294, 0.0421, 0.0831], device='cuda:3'), in_proj_covar=tensor([0.0121, 0.0117, 0.0128, 0.0122, 0.0083, 0.0115, 0.0136, 0.0152], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-22 22:49:31,477 INFO [train.py:894] (3/4) Epoch 9, batch 1450, loss[loss=0.177, simple_loss=0.2653, pruned_loss=0.04437, over 18389.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.291, pruned_loss=0.06445, over 3710088.91 frames. ], batch size: 46, lr: 1.29e-02, grad_scale: 8.0 2022-12-22 22:49:57,130 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29517.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:50:05,922 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.999e+02 4.337e+02 5.237e+02 6.782e+02 1.195e+03, threshold=1.047e+03, percent-clipped=0.0 2022-12-22 22:50:10,143 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp0.9 from training. Duration: 33.038875 2022-12-22 22:50:46,420 INFO [train.py:894] (3/4) Epoch 9, batch 1500, loss[loss=0.1826, simple_loss=0.2649, pruned_loss=0.05018, over 18706.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2917, pruned_loss=0.06451, over 3710776.28 frames. ], batch size: 50, lr: 1.29e-02, grad_scale: 8.0 2022-12-22 22:50:47,855 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64291-0000-16059-0_sp0.9 from training. Duration: 20.0944375 2022-12-22 22:50:48,166 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29551.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:51:02,521 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp1.1 from training. Duration: 20.4 2022-12-22 22:51:09,786 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=29565.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:51:11,175 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0 from training. Duration: 20.085 2022-12-22 22:51:21,711 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0_sp0.9 from training. Duration: 23.07775 2022-12-22 22:51:49,979 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29592.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:51:58,044 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0330, 1.7190, 1.8119, 1.1109, 2.4468, 2.1245, 1.7618, 1.3470], device='cuda:3'), covar=tensor([0.0325, 0.0406, 0.0430, 0.0659, 0.0167, 0.0319, 0.0459, 0.0901], device='cuda:3'), in_proj_covar=tensor([0.0120, 0.0116, 0.0129, 0.0120, 0.0083, 0.0116, 0.0135, 0.0152], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-22 22:52:01,851 INFO [train.py:894] (3/4) Epoch 9, batch 1550, loss[loss=0.2358, simple_loss=0.315, pruned_loss=0.07827, over 18691.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2929, pruned_loss=0.06489, over 3711888.52 frames. ], batch size: 77, lr: 1.29e-02, grad_scale: 8.0 2022-12-22 22:52:08,036 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp0.9 from training. Duration: 24.9333125 2022-12-22 22:52:20,477 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29612.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:52:37,793 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.815e+02 4.732e+02 5.774e+02 7.173e+02 1.545e+03, threshold=1.155e+03, percent-clipped=6.0 2022-12-22 22:52:51,488 WARNING [train.py:1060] (3/4) Exclude cut with ID 5118-111612-0016-124680-0_sp0.9 from training. Duration: 20.388875 2022-12-22 22:52:57,634 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp1.1 from training. Duration: 20.3590625 2022-12-22 22:53:17,333 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2022-12-22 22:53:19,293 INFO [train.py:894] (3/4) Epoch 9, batch 1600, loss[loss=0.209, simple_loss=0.2934, pruned_loss=0.06232, over 18582.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2939, pruned_loss=0.06522, over 3712633.09 frames. ], batch size: 51, lr: 1.29e-02, grad_scale: 8.0 2022-12-22 22:53:31,307 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29658.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:53:37,795 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29662.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:53:48,116 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6958, 1.8060, 1.8411, 2.1279, 1.3487, 4.9427, 2.2095, 2.4378], device='cuda:3'), covar=tensor([0.3134, 0.1905, 0.1783, 0.1638, 0.1380, 0.0088, 0.1323, 0.0906], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0121, 0.0132, 0.0123, 0.0106, 0.0100, 0.0101, 0.0098], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-22 22:54:07,611 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0_sp1.1 from training. Duration: 0.836375 2022-12-22 22:54:35,185 INFO [train.py:894] (3/4) Epoch 9, batch 1650, loss[loss=0.2193, simple_loss=0.305, pruned_loss=0.06681, over 18529.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.2937, pruned_loss=0.06567, over 3712398.08 frames. ], batch size: 97, lr: 1.29e-02, grad_scale: 8.0 2022-12-22 22:54:43,978 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=29706.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:54:46,756 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5531, 2.0292, 1.3987, 2.3359, 2.4444, 1.4341, 1.5840, 1.2061], device='cuda:3'), covar=tensor([0.1980, 0.1569, 0.1573, 0.0922, 0.1350, 0.1286, 0.1806, 0.1694], device='cuda:3'), in_proj_covar=tensor([0.0233, 0.0208, 0.0198, 0.0188, 0.0250, 0.0186, 0.0207, 0.0188], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 22:54:52,752 WARNING [train.py:1060] (3/4) Exclude cut with ID 8565-290391-0049-67394-0_sp0.9 from training. Duration: 21.3166875 2022-12-22 22:55:09,547 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.853e+02 4.793e+02 6.223e+02 8.275e+02 2.950e+03, threshold=1.245e+03, percent-clipped=9.0 2022-12-22 22:55:09,965 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29723.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 22:55:24,101 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0029-104863-0_sp0.9 from training. Duration: 22.1055625 2022-12-22 22:55:34,795 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp1.1 from training. Duration: 21.77725 2022-12-22 22:55:50,794 INFO [train.py:894] (3/4) Epoch 9, batch 1700, loss[loss=0.2287, simple_loss=0.3039, pruned_loss=0.07677, over 18522.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2955, pruned_loss=0.06861, over 3711394.93 frames. ], batch size: 52, lr: 1.29e-02, grad_scale: 8.0 2022-12-22 22:55:53,970 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp0.9 from training. Duration: 27.8166875 2022-12-22 22:56:19,190 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp1.1 from training. Duration: 22.5090625 2022-12-22 22:56:25,292 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0 from training. Duration: 25.035 2022-12-22 22:56:43,166 WARNING [train.py:1060] (3/4) Exclude cut with ID 774-127930-0014-10412-0_sp1.1 from training. Duration: 0.95 2022-12-22 22:57:01,255 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp0.9 from training. Duration: 0.92225 2022-12-22 22:57:07,349 INFO [train.py:894] (3/4) Epoch 9, batch 1750, loss[loss=0.1992, simple_loss=0.2803, pruned_loss=0.05904, over 18423.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.2981, pruned_loss=0.07139, over 3712832.33 frames. ], batch size: 48, lr: 1.29e-02, grad_scale: 8.0 2022-12-22 22:57:11,819 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29802.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:57:28,894 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0 from training. Duration: 21.97 2022-12-22 22:57:39,178 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4988, 1.2116, 1.6576, 2.5112, 1.7652, 2.1580, 1.0191, 1.7086], device='cuda:3'), covar=tensor([0.1783, 0.1721, 0.1329, 0.0692, 0.1335, 0.1121, 0.1945, 0.1346], device='cuda:3'), in_proj_covar=tensor([0.0104, 0.0115, 0.0128, 0.0123, 0.0104, 0.0130, 0.0129, 0.0109], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 22:57:42,904 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.618e+02 5.406e+02 6.621e+02 8.145e+02 1.955e+03, threshold=1.324e+03, percent-clipped=2.0 2022-12-22 22:57:48,790 WARNING [train.py:1060] (3/4) Exclude cut with ID 7492-105653-0055-62765-0_sp0.9 from training. Duration: 21.97225 2022-12-22 22:57:50,044 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp0.9 from training. Duration: 25.3333125 2022-12-22 22:57:51,374 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2022-12-22 22:57:54,038 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.92 vs. limit=5.0 2022-12-22 22:58:00,098 WARNING [train.py:1060] (3/4) Exclude cut with ID 5172-29468-0015-19128-0_sp0.9 from training. Duration: 21.5055625 2022-12-22 22:58:10,961 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp1.1 from training. Duration: 20.72725 2022-12-22 22:58:23,414 INFO [train.py:894] (3/4) Epoch 9, batch 1800, loss[loss=0.2503, simple_loss=0.321, pruned_loss=0.08975, over 18514.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.2992, pruned_loss=0.07334, over 3712946.07 frames. ], batch size: 58, lr: 1.29e-02, grad_scale: 8.0 2022-12-22 22:58:36,663 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2022-12-22 22:58:43,645 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29863.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:58:45,340 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp0.9 from training. Duration: 26.32775 2022-12-22 22:59:17,143 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0 from training. Duration: 20.025 2022-12-22 22:59:22,043 WARNING [train.py:1060] (3/4) Exclude cut with ID 6709-74022-0004-86860-0_sp1.1 from training. Duration: 0.9409375 2022-12-22 22:59:22,057 WARNING [train.py:1060] (3/4) Exclude cut with ID 4757-1811-0023-62229-0_sp0.9 from training. Duration: 21.37775 2022-12-22 22:59:26,657 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29892.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 22:59:30,712 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2022-12-22 22:59:39,030 INFO [train.py:894] (3/4) Epoch 9, batch 1850, loss[loss=0.2506, simple_loss=0.3153, pruned_loss=0.09299, over 18732.00 frames. ], tot_loss[loss=0.2254, simple_loss=0.3, pruned_loss=0.07536, over 3712875.66 frames. ], batch size: 52, lr: 1.29e-02, grad_scale: 8.0 2022-12-22 22:59:42,546 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0004-25974-0_sp0.9 from training. Duration: 21.17225 2022-12-22 22:59:42,560 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp0.9 from training. Duration: 27.511125 2022-12-22 22:59:50,555 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29907.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:00:11,323 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2022-12-22 23:00:14,056 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0 from training. Duration: 22.8 2022-12-22 23:00:15,337 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.806e+02 5.568e+02 6.649e+02 8.317e+02 1.675e+03, threshold=1.330e+03, percent-clipped=4.0 2022-12-22 23:00:18,526 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0 from training. Duration: 22.585 2022-12-22 23:00:27,649 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.1811, 2.2359, 1.8152, 2.9852, 2.2081, 2.2555, 2.3559, 3.6610], device='cuda:3'), covar=tensor([0.1583, 0.2709, 0.1504, 0.2509, 0.3006, 0.0911, 0.2571, 0.0436], device='cuda:3'), in_proj_covar=tensor([0.0270, 0.0258, 0.0220, 0.0334, 0.0243, 0.0206, 0.0256, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 23:00:42,408 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=29940.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:00:51,768 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0001-146967-0_sp0.9 from training. Duration: 22.0166875 2022-12-22 23:00:58,028 INFO [train.py:894] (3/4) Epoch 9, batch 1900, loss[loss=0.1985, simple_loss=0.2688, pruned_loss=0.06408, over 18409.00 frames. ], tot_loss[loss=0.227, simple_loss=0.3007, pruned_loss=0.07664, over 3713061.74 frames. ], batch size: 48, lr: 1.28e-02, grad_scale: 8.0 2022-12-22 23:01:10,435 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp1.1 from training. Duration: 24.395375 2022-12-22 23:01:10,760 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9128, 1.3248, 0.6423, 1.3837, 2.0308, 1.3925, 1.5541, 1.9152], device='cuda:3'), covar=tensor([0.1544, 0.2152, 0.2713, 0.1627, 0.1750, 0.1629, 0.1549, 0.1578], device='cuda:3'), in_proj_covar=tensor([0.0090, 0.0101, 0.0121, 0.0096, 0.0113, 0.0091, 0.0097, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 23:01:17,962 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp0.9 from training. Duration: 27.47775 2022-12-22 23:01:21,042 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp0.9 from training. Duration: 24.8833125 2022-12-22 23:01:24,081 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0 from training. Duration: 23.39 2022-12-22 23:01:30,066 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp0.9 from training. Duration: 28.72225 2022-12-22 23:01:40,343 WARNING [train.py:1060] (3/4) Exclude cut with ID 585-294811-0110-133686-0_sp0.9 from training. Duration: 20.8944375 2022-12-22 23:01:42,311 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9955, 1.8578, 1.6367, 1.7450, 1.9081, 2.1216, 2.2416, 1.8645], device='cuda:3'), covar=tensor([0.0272, 0.0250, 0.0321, 0.0203, 0.0221, 0.0282, 0.0244, 0.0270], device='cuda:3'), in_proj_covar=tensor([0.0081, 0.0113, 0.0136, 0.0120, 0.0104, 0.0103, 0.0088, 0.0115], device='cuda:3'), out_proj_covar=tensor([7.1633e-05, 9.7124e-05, 1.2311e-04, 1.0348e-04, 9.2425e-05, 8.6929e-05, 7.5970e-05, 9.8834e-05], device='cuda:3') 2022-12-22 23:01:53,205 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0_sp0.9 from training. Duration: 23.8444375 2022-12-22 23:02:15,626 INFO [train.py:894] (3/4) Epoch 9, batch 1950, loss[loss=0.1969, simple_loss=0.2627, pruned_loss=0.06559, over 18495.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.3018, pruned_loss=0.07834, over 3712873.89 frames. ], batch size: 43, lr: 1.28e-02, grad_scale: 8.0 2022-12-22 23:02:22,811 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0 from training. Duration: 25.85 2022-12-22 23:02:22,820 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0 from training. Duration: 21.39 2022-12-22 23:02:34,538 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0 from training. Duration: 27.92 2022-12-22 23:02:46,603 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30018.0, num_to_drop=1, layers_to_drop={3} 2022-12-22 23:02:53,456 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.837e+02 5.184e+02 6.235e+02 8.086e+02 1.649e+03, threshold=1.247e+03, percent-clipped=1.0 2022-12-22 23:03:04,100 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0039-130165-0_sp0.9 from training. Duration: 20.661125 2022-12-22 23:03:27,761 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0043-15874-0_sp0.9 from training. Duration: 20.07225 2022-12-22 23:03:35,306 INFO [train.py:894] (3/4) Epoch 9, batch 2000, loss[loss=0.2605, simple_loss=0.3283, pruned_loss=0.09636, over 18632.00 frames. ], tot_loss[loss=0.2297, simple_loss=0.302, pruned_loss=0.07877, over 3713107.95 frames. ], batch size: 69, lr: 1.28e-02, grad_scale: 8.0 2022-12-22 23:03:36,198 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0 from training. Duration: 21.01 2022-12-22 23:03:47,293 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4759, 3.6395, 3.6980, 1.6560, 3.5830, 2.8190, 0.8670, 2.5127], device='cuda:3'), covar=tensor([0.2096, 0.1110, 0.1358, 0.3548, 0.1054, 0.1081, 0.5070, 0.1477], device='cuda:3'), in_proj_covar=tensor([0.0131, 0.0115, 0.0147, 0.0117, 0.0121, 0.0102, 0.0140, 0.0106], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 23:04:43,453 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0 from training. Duration: 20.65 2022-12-22 23:04:44,633 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7630, 1.7414, 1.2422, 1.7194, 1.5692, 1.5783, 1.4614, 1.5951], device='cuda:3'), covar=tensor([0.1849, 0.2307, 0.1584, 0.2045, 0.2482, 0.0888, 0.2283, 0.0748], device='cuda:3'), in_proj_covar=tensor([0.0271, 0.0260, 0.0223, 0.0334, 0.0246, 0.0208, 0.0258, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 23:04:51,254 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0 from training. Duration: 21.46 2022-12-22 23:04:52,760 INFO [train.py:894] (3/4) Epoch 9, batch 2050, loss[loss=0.1998, simple_loss=0.2736, pruned_loss=0.06296, over 18445.00 frames. ], tot_loss[loss=0.2315, simple_loss=0.3027, pruned_loss=0.08011, over 3713941.01 frames. ], batch size: 48, lr: 1.28e-02, grad_scale: 8.0 2022-12-22 23:05:27,809 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.734e+02 6.048e+02 7.232e+02 9.558e+02 2.262e+03, threshold=1.446e+03, percent-clipped=10.0 2022-12-22 23:05:31,122 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30125.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:05:36,335 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0 from training. Duration: 0.92 2022-12-22 23:05:42,592 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0_sp0.9 from training. Duration: 23.7666875 2022-12-22 23:05:44,345 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3451, 1.9283, 1.7160, 1.6748, 1.9112, 2.7340, 2.5005, 1.8808], device='cuda:3'), covar=tensor([0.0266, 0.0293, 0.0426, 0.0265, 0.0306, 0.0237, 0.0269, 0.0289], device='cuda:3'), in_proj_covar=tensor([0.0084, 0.0116, 0.0140, 0.0122, 0.0107, 0.0105, 0.0088, 0.0118], device='cuda:3'), out_proj_covar=tensor([7.3585e-05, 9.9916e-05, 1.2657e-04, 1.0520e-04, 9.5184e-05, 8.7961e-05, 7.5658e-05, 1.0096e-04], device='cuda:3') 2022-12-22 23:06:08,366 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6187, 1.1208, 2.0667, 3.1996, 2.1569, 2.3427, 0.9729, 2.0941], device='cuda:3'), covar=tensor([0.2176, 0.2449, 0.1728, 0.0751, 0.1406, 0.1447, 0.2789, 0.1464], device='cuda:3'), in_proj_covar=tensor([0.0104, 0.0115, 0.0128, 0.0123, 0.0104, 0.0131, 0.0130, 0.0110], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 23:06:09,356 INFO [train.py:894] (3/4) Epoch 9, batch 2100, loss[loss=0.2038, simple_loss=0.2731, pruned_loss=0.06725, over 18607.00 frames. ], tot_loss[loss=0.2313, simple_loss=0.3023, pruned_loss=0.08021, over 3714364.89 frames. ], batch size: 45, lr: 1.28e-02, grad_scale: 8.0 2022-12-22 23:06:14,688 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.91 vs. limit=2.0 2022-12-22 23:06:17,709 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.95 vs. limit=5.0 2022-12-22 23:06:19,885 WARNING [train.py:1060] (3/4) Exclude cut with ID 8544-281189-0060-101339-0_sp0.9 from training. Duration: 20.861125 2022-12-22 23:06:21,296 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30158.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:06:29,710 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0_sp0.9 from training. Duration: 22.711125 2022-12-22 23:06:48,882 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.1142, 0.6702, 0.8930, 1.2404, 1.2178, 1.1405, 1.1228, 0.9572], device='cuda:3'), covar=tensor([0.0231, 0.0241, 0.0473, 0.0181, 0.0196, 0.0256, 0.0230, 0.0243], device='cuda:3'), in_proj_covar=tensor([0.0083, 0.0116, 0.0141, 0.0121, 0.0106, 0.0105, 0.0088, 0.0117], device='cuda:3'), out_proj_covar=tensor([7.2887e-05, 9.9741e-05, 1.2702e-04, 1.0492e-04, 9.4777e-05, 8.7730e-05, 7.5319e-05, 1.0065e-04], device='cuda:3') 2022-12-22 23:07:04,529 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30186.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:07:13,172 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp1.1 from training. Duration: 22.986375 2022-12-22 23:07:24,961 INFO [train.py:894] (3/4) Epoch 9, batch 2150, loss[loss=0.2345, simple_loss=0.3071, pruned_loss=0.08091, over 18547.00 frames. ], tot_loss[loss=0.2324, simple_loss=0.3027, pruned_loss=0.0811, over 3714786.36 frames. ], batch size: 55, lr: 1.28e-02, grad_scale: 8.0 2022-12-22 23:07:29,422 WARNING [train.py:1060] (3/4) Exclude cut with ID 8040-260924-0003-80960-0_sp0.9 from training. Duration: 22.07225 2022-12-22 23:07:33,799 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0045-26330-0_sp0.9 from training. Duration: 20.3055625 2022-12-22 23:07:35,446 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30207.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:07:36,648 WARNING [train.py:1060] (3/4) Exclude cut with ID 6356-271890-0060-94317-0_sp0.9 from training. Duration: 20.72225 2022-12-22 23:07:38,579 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30209.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:07:42,867 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5438, 1.5223, 1.7227, 1.5803, 1.3072, 3.8055, 1.6721, 2.1403], device='cuda:3'), covar=tensor([0.3298, 0.2019, 0.1838, 0.1877, 0.1401, 0.0173, 0.1502, 0.0961], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0121, 0.0132, 0.0122, 0.0106, 0.0101, 0.0100, 0.0098], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-22 23:07:55,147 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp1.1 from training. Duration: 22.4818125 2022-12-22 23:07:59,369 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.960e+02 5.109e+02 5.943e+02 7.734e+02 1.669e+03, threshold=1.189e+03, percent-clipped=3.0 2022-12-22 23:08:20,705 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp0.9 from training. Duration: 25.0944375 2022-12-22 23:08:25,241 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0 from training. Duration: 21.515 2022-12-22 23:08:32,233 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp0.9 from training. Duration: 27.02225 2022-12-22 23:08:36,597 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0010-62480-0_sp0.9 from training. Duration: 22.22225 2022-12-22 23:08:40,701 INFO [train.py:894] (3/4) Epoch 9, batch 2200, loss[loss=0.2309, simple_loss=0.3079, pruned_loss=0.07691, over 18679.00 frames. ], tot_loss[loss=0.2326, simple_loss=0.3027, pruned_loss=0.08129, over 3714448.21 frames. ], batch size: 60, lr: 1.28e-02, grad_scale: 8.0 2022-12-22 23:08:42,146 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0085-44554-0_sp0.9 from training. Duration: 20.85 2022-12-22 23:08:48,429 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=30255.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:09:11,378 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30270.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:09:13,053 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.8604, 2.9959, 2.5831, 1.3870, 2.5393, 2.4856, 2.3409, 2.2763], device='cuda:3'), covar=tensor([0.0514, 0.0553, 0.1286, 0.1675, 0.1641, 0.1240, 0.1283, 0.1132], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0181, 0.0202, 0.0192, 0.0206, 0.0190, 0.0201, 0.0190], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 23:09:17,747 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0 from training. Duration: 21.54 2022-12-22 23:09:22,796 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp1.1 from training. Duration: 20.5318125 2022-12-22 23:09:32,472 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0012-134311-0_sp0.9 from training. Duration: 21.9333125 2022-12-22 23:09:57,727 INFO [train.py:894] (3/4) Epoch 9, batch 2250, loss[loss=0.2314, simple_loss=0.2859, pruned_loss=0.08848, over 18398.00 frames. ], tot_loss[loss=0.2323, simple_loss=0.302, pruned_loss=0.0813, over 3714254.35 frames. ], batch size: 46, lr: 1.28e-02, grad_scale: 8.0 2022-12-22 23:10:20,981 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0025-130151-0_sp0.9 from training. Duration: 21.7944375 2022-12-22 23:10:25,704 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30318.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 23:10:33,674 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.779e+02 5.688e+02 7.340e+02 9.327e+02 2.181e+03, threshold=1.468e+03, percent-clipped=8.0 2022-12-22 23:10:33,790 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0_sp0.9 from training. Duration: 22.4666875 2022-12-22 23:10:41,174 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0 from training. Duration: 21.635 2022-12-22 23:10:47,511 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0_sp0.9 from training. Duration: 24.038875 2022-12-22 23:11:14,220 INFO [train.py:894] (3/4) Epoch 9, batch 2300, loss[loss=0.2252, simple_loss=0.3008, pruned_loss=0.07478, over 18720.00 frames. ], tot_loss[loss=0.2326, simple_loss=0.3022, pruned_loss=0.08147, over 3714371.02 frames. ], batch size: 52, lr: 1.28e-02, grad_scale: 8.0 2022-12-22 23:11:28,624 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp1.1 from training. Duration: 21.786375 2022-12-22 23:11:30,409 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30360.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:11:39,396 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=30366.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:11:42,264 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0 from training. Duration: 20.22 2022-12-22 23:12:29,575 INFO [train.py:894] (3/4) Epoch 9, batch 2350, loss[loss=0.2532, simple_loss=0.3075, pruned_loss=0.09943, over 18472.00 frames. ], tot_loss[loss=0.234, simple_loss=0.3031, pruned_loss=0.08247, over 3713751.08 frames. ], batch size: 50, lr: 1.28e-02, grad_scale: 8.0 2022-12-22 23:13:03,083 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30421.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:13:06,427 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.215e+02 5.140e+02 6.090e+02 7.966e+02 3.738e+03, threshold=1.218e+03, percent-clipped=4.0 2022-12-22 23:13:14,401 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2572, 2.0972, 2.0288, 2.6303, 2.0894, 4.7720, 1.9598, 2.1998], device='cuda:3'), covar=tensor([0.0864, 0.1640, 0.0988, 0.0863, 0.1413, 0.0192, 0.1250, 0.1368], device='cuda:3'), in_proj_covar=tensor([0.0075, 0.0084, 0.0076, 0.0075, 0.0093, 0.0072, 0.0085, 0.0077], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2022-12-22 23:13:14,509 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6685, 2.2001, 2.0609, 1.4878, 2.2252, 3.0318, 3.0515, 2.1071], device='cuda:3'), covar=tensor([0.0370, 0.0278, 0.0398, 0.0295, 0.0216, 0.0279, 0.0247, 0.0268], device='cuda:3'), in_proj_covar=tensor([0.0083, 0.0114, 0.0140, 0.0120, 0.0105, 0.0103, 0.0087, 0.0116], device='cuda:3'), out_proj_covar=tensor([7.3037e-05, 9.7968e-05, 1.2638e-04, 1.0337e-04, 9.3023e-05, 8.6596e-05, 7.4737e-05, 9.8874e-05], device='cuda:3') 2022-12-22 23:13:35,177 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2022-12-22 23:13:40,555 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0 from training. Duration: 25.285 2022-12-22 23:13:46,558 INFO [train.py:894] (3/4) Epoch 9, batch 2400, loss[loss=0.1987, simple_loss=0.2688, pruned_loss=0.06431, over 18538.00 frames. ], tot_loss[loss=0.2349, simple_loss=0.3039, pruned_loss=0.08296, over 3714509.85 frames. ], batch size: 47, lr: 1.27e-02, grad_scale: 8.0 2022-12-22 23:13:59,198 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30458.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:14:34,763 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30481.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:14:46,339 WARNING [train.py:1060] (3/4) Exclude cut with ID 811-130148-0001-63453-0_sp0.9 from training. Duration: 20.861125 2022-12-22 23:15:03,532 INFO [train.py:894] (3/4) Epoch 9, batch 2450, loss[loss=0.1963, simple_loss=0.2762, pruned_loss=0.05818, over 18371.00 frames. ], tot_loss[loss=0.2337, simple_loss=0.303, pruned_loss=0.08223, over 3715246.25 frames. ], batch size: 46, lr: 1.27e-02, grad_scale: 8.0 2022-12-22 23:15:11,086 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0 from training. Duration: 20.88 2022-12-22 23:15:12,707 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=30506.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:15:39,225 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.497e+02 5.250e+02 6.536e+02 8.030e+02 2.143e+03, threshold=1.307e+03, percent-clipped=5.0 2022-12-22 23:15:44,750 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0_sp0.9 from training. Duration: 23.4166875 2022-12-22 23:16:20,843 INFO [train.py:894] (3/4) Epoch 9, batch 2500, loss[loss=0.2048, simple_loss=0.2883, pruned_loss=0.0607, over 18466.00 frames. ], tot_loss[loss=0.2329, simple_loss=0.3022, pruned_loss=0.08181, over 3714897.22 frames. ], batch size: 50, lr: 1.27e-02, grad_scale: 8.0 2022-12-22 23:16:44,687 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30565.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:16:57,358 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0 from training. Duration: 21.24 2022-12-22 23:16:58,673 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0_sp0.9 from training. Duration: 23.9055625 2022-12-22 23:17:33,711 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp0.9 from training. Duration: 25.988875 2022-12-22 23:17:35,706 INFO [train.py:894] (3/4) Epoch 9, batch 2550, loss[loss=0.226, simple_loss=0.3032, pruned_loss=0.07442, over 18543.00 frames. ], tot_loss[loss=0.2325, simple_loss=0.3019, pruned_loss=0.08151, over 3714933.47 frames. ], batch size: 69, lr: 1.27e-02, grad_scale: 8.0 2022-12-22 23:17:41,023 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0001-134300-0_sp0.9 from training. Duration: 20.67225 2022-12-22 23:18:10,864 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.281e+02 5.961e+02 7.202e+02 9.301e+02 2.076e+03, threshold=1.440e+03, percent-clipped=8.0 2022-12-22 23:18:27,889 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0 from training. Duration: 20.34 2022-12-22 23:18:53,715 INFO [train.py:894] (3/4) Epoch 9, batch 2600, loss[loss=0.2448, simple_loss=0.3143, pruned_loss=0.08768, over 18485.00 frames. ], tot_loss[loss=0.2325, simple_loss=0.3022, pruned_loss=0.08138, over 3713720.31 frames. ], batch size: 52, lr: 1.27e-02, grad_scale: 8.0 2022-12-22 23:19:02,046 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30655.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:19:40,531 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp0.9 from training. Duration: 25.061125 2022-12-22 23:19:53,581 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0 from training. Duration: 0.83 2022-12-22 23:20:11,069 INFO [train.py:894] (3/4) Epoch 9, batch 2650, loss[loss=0.2553, simple_loss=0.3304, pruned_loss=0.09005, over 18576.00 frames. ], tot_loss[loss=0.2306, simple_loss=0.3008, pruned_loss=0.08017, over 3713665.41 frames. ], batch size: 57, lr: 1.27e-02, grad_scale: 8.0 2022-12-22 23:20:18,770 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0 from training. Duration: 24.73 2022-12-22 23:20:31,073 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0 from training. Duration: 23.965 2022-12-22 23:20:35,388 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30716.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:20:35,624 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30716.0, num_to_drop=1, layers_to_drop={3} 2022-12-22 23:20:41,070 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0030-146996-0_sp0.9 from training. Duration: 22.088875 2022-12-22 23:20:45,425 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.321e+02 5.727e+02 6.778e+02 8.704e+02 1.736e+03, threshold=1.356e+03, percent-clipped=1.0 2022-12-22 23:20:54,603 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0_sp0.9 from training. Duration: 23.6 2022-12-22 23:21:27,280 INFO [train.py:894] (3/4) Epoch 9, batch 2700, loss[loss=0.2057, simple_loss=0.273, pruned_loss=0.06927, over 18453.00 frames. ], tot_loss[loss=0.2312, simple_loss=0.3016, pruned_loss=0.08046, over 3713672.97 frames. ], batch size: 43, lr: 1.27e-02, grad_scale: 8.0 2022-12-22 23:22:14,299 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30781.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:22:40,582 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0 from training. Duration: 23.795 2022-12-22 23:22:43,534 INFO [train.py:894] (3/4) Epoch 9, batch 2750, loss[loss=0.2529, simple_loss=0.3153, pruned_loss=0.0952, over 18551.00 frames. ], tot_loss[loss=0.2304, simple_loss=0.3011, pruned_loss=0.07991, over 3714035.33 frames. ], batch size: 78, lr: 1.27e-02, grad_scale: 8.0 2022-12-22 23:22:56,901 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp1.1 from training. Duration: 21.5409375 2022-12-22 23:22:59,742 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp0.9 from training. Duration: 24.97775 2022-12-22 23:23:11,930 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0_sp0.9 from training. Duration: 23.3444375 2022-12-22 23:23:13,554 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5204, 3.5273, 3.5459, 1.4735, 3.7338, 2.8604, 0.9136, 2.6523], device='cuda:3'), covar=tensor([0.1923, 0.1163, 0.1322, 0.3788, 0.0821, 0.1001, 0.5106, 0.1484], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0118, 0.0147, 0.0118, 0.0120, 0.0102, 0.0140, 0.0108], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 23:23:17,517 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 4.042e+02 6.530e+02 8.093e+02 1.038e+03 2.043e+03, threshold=1.619e+03, percent-clipped=4.0 2022-12-22 23:23:27,014 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=30829.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:23:40,731 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0_sp0.9 from training. Duration: 23.2 2022-12-22 23:23:47,481 WARNING [train.py:1060] (3/4) Exclude cut with ID 5653-46179-0060-117930-0_sp0.9 from training. Duration: 21.17225 2022-12-22 23:23:59,698 INFO [train.py:894] (3/4) Epoch 9, batch 2800, loss[loss=0.1912, simple_loss=0.2616, pruned_loss=0.06043, over 18580.00 frames. ], tot_loss[loss=0.2302, simple_loss=0.3006, pruned_loss=0.07991, over 3714975.26 frames. ], batch size: 41, lr: 1.27e-02, grad_scale: 8.0 2022-12-22 23:24:07,359 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp0.9 from training. Duration: 24.6555625 2022-12-22 23:24:12,603 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2022-12-22 23:24:22,105 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30865.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:25:06,020 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0 from training. Duration: 20.44 2022-12-22 23:25:15,762 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.4764, 1.8110, 2.0241, 0.7113, 1.2229, 2.1261, 1.7549, 1.7721], device='cuda:3'), covar=tensor([0.0588, 0.0260, 0.0284, 0.0351, 0.0328, 0.0341, 0.0225, 0.0415], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0151, 0.0105, 0.0127, 0.0134, 0.0119, 0.0139, 0.0139], device='cuda:3'), out_proj_covar=tensor([1.1327e-04, 1.3005e-04, 8.8645e-05, 1.0623e-04, 1.1394e-04, 1.0184e-04, 1.2124e-04, 1.1898e-04], device='cuda:3') 2022-12-22 23:25:16,620 INFO [train.py:894] (3/4) Epoch 9, batch 2850, loss[loss=0.2369, simple_loss=0.3086, pruned_loss=0.08265, over 18705.00 frames. ], tot_loss[loss=0.2279, simple_loss=0.2983, pruned_loss=0.07877, over 3714375.29 frames. ], batch size: 78, lr: 1.27e-02, grad_scale: 8.0 2022-12-22 23:25:18,122 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0_sp0.9 from training. Duration: 23.45 2022-12-22 23:25:21,446 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30903.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:25:37,115 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=30913.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:25:47,918 WARNING [train.py:1060] (3/4) Exclude cut with ID 6945-60535-0076-12784-0_sp0.9 from training. Duration: 20.52225 2022-12-22 23:25:52,783 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.339e+02 5.387e+02 6.998e+02 8.710e+02 1.625e+03, threshold=1.400e+03, percent-clipped=1.0 2022-12-22 23:25:55,183 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0 from training. Duration: 22.19 2022-12-22 23:25:57,008 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30925.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:26:06,214 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp1.1 from training. Duration: 25.3818125 2022-12-22 23:26:10,686 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30934.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:26:22,093 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp0.9 from training. Duration: 28.0944375 2022-12-22 23:26:29,366 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0_sp0.9 from training. Duration: 22.9444375 2022-12-22 23:26:33,879 INFO [train.py:894] (3/4) Epoch 9, batch 2900, loss[loss=0.1946, simple_loss=0.2706, pruned_loss=0.05929, over 18672.00 frames. ], tot_loss[loss=0.2286, simple_loss=0.2987, pruned_loss=0.07924, over 3714498.48 frames. ], batch size: 46, lr: 1.26e-02, grad_scale: 8.0 2022-12-22 23:26:36,749 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp1.1 from training. Duration: 21.6318125 2022-12-22 23:26:55,040 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0 from training. Duration: 23.695 2022-12-22 23:26:55,414 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30964.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:27:09,338 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4417, 1.4379, 1.5441, 1.5718, 1.4615, 3.3684, 1.4948, 2.1287], device='cuda:3'), covar=tensor([0.4029, 0.2442, 0.2208, 0.2301, 0.1364, 0.0223, 0.1530, 0.0883], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0121, 0.0133, 0.0123, 0.0106, 0.0101, 0.0100, 0.0097], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-22 23:27:21,648 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5954, 2.2854, 1.7404, 0.7546, 1.7638, 1.9224, 1.4058, 2.0015], device='cuda:3'), covar=tensor([0.0531, 0.0508, 0.1191, 0.1565, 0.1161, 0.1308, 0.1570, 0.0723], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0177, 0.0204, 0.0191, 0.0205, 0.0188, 0.0200, 0.0188], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 23:27:22,684 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0 from training. Duration: 23.955 2022-12-22 23:27:28,787 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30986.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:27:41,683 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30995.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:27:48,926 INFO [train.py:894] (3/4) Epoch 9, batch 2950, loss[loss=0.2279, simple_loss=0.306, pruned_loss=0.07487, over 18616.00 frames. ], tot_loss[loss=0.2291, simple_loss=0.2993, pruned_loss=0.07946, over 3714760.18 frames. ], batch size: 69, lr: 1.26e-02, grad_scale: 8.0 2022-12-22 23:27:54,432 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp0.9 from training. Duration: 26.438875 2022-12-22 23:28:07,242 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31011.0, num_to_drop=1, layers_to_drop={3} 2022-12-22 23:28:07,426 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31011.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 23:28:15,941 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31016.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:28:27,060 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.802e+02 5.929e+02 7.348e+02 8.660e+02 3.219e+03, threshold=1.470e+03, percent-clipped=1.0 2022-12-22 23:28:39,432 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0021-26306-0_sp0.9 from training. Duration: 21.2444375 2022-12-22 23:28:41,025 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp0.9 from training. Duration: 31.02225 2022-12-22 23:28:49,577 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0 from training. Duration: 22.395 2022-12-22 23:29:06,548 INFO [train.py:894] (3/4) Epoch 9, batch 3000, loss[loss=0.206, simple_loss=0.2761, pruned_loss=0.068, over 18431.00 frames. ], tot_loss[loss=0.2298, simple_loss=0.2997, pruned_loss=0.07997, over 3715179.90 frames. ], batch size: 48, lr: 1.26e-02, grad_scale: 8.0 2022-12-22 23:29:06,548 INFO [train.py:919] (3/4) Computing validation loss 2022-12-22 23:29:17,603 INFO [train.py:928] (3/4) Epoch 9, validation: loss=0.1783, simple_loss=0.2774, pruned_loss=0.03963, over 944034.00 frames. 2022-12-22 23:29:17,604 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24676MB 2022-12-22 23:29:17,666 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0 from training. Duration: 21.075 2022-12-22 23:29:23,227 WARNING [train.py:1060] (3/4) Exclude cut with ID 6482-98857-0025-147532-0_sp0.9 from training. Duration: 20.0055625 2022-12-22 23:29:23,931 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0037-132304-0_sp0.9 from training. Duration: 22.05 2022-12-22 23:29:23,947 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0 from training. Duration: 26.8349375 2022-12-22 23:29:26,694 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp1.1 from training. Duration: 22.1090625 2022-12-22 23:29:34,421 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp0.9 from training. Duration: 26.6166875 2022-12-22 23:29:38,760 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=31064.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:29:50,724 WARNING [train.py:1060] (3/4) Exclude cut with ID 2046-178027-0000-53705-0_sp0.9 from training. Duration: 20.3055625 2022-12-22 23:29:51,108 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31072.0, num_to_drop=1, layers_to_drop={2} 2022-12-22 23:30:13,571 WARNING [train.py:1060] (3/4) Exclude cut with ID 7205-50138-0008-5373-0_sp0.9 from training. Duration: 20.7 2022-12-22 23:30:34,678 INFO [train.py:894] (3/4) Epoch 9, batch 3050, loss[loss=0.1971, simple_loss=0.2732, pruned_loss=0.06052, over 18709.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.2993, pruned_loss=0.07962, over 3715528.69 frames. ], batch size: 50, lr: 1.26e-02, grad_scale: 8.0 2022-12-22 23:30:56,893 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0 from training. Duration: 22.48 2022-12-22 23:31:08,707 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.995e+02 5.870e+02 7.181e+02 8.516e+02 1.347e+03, threshold=1.436e+03, percent-clipped=0.0 2022-12-22 23:31:13,023 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp0.9 from training. Duration: 29.816625 2022-12-22 23:31:24,743 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2022-12-22 23:31:33,853 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp1.1 from training. Duration: 22.7590625 2022-12-22 23:31:38,481 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0 from training. Duration: 22.555 2022-12-22 23:31:50,881 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3228, 2.0816, 1.7108, 1.1090, 3.0932, 2.5947, 2.1821, 1.8201], device='cuda:3'), covar=tensor([0.0351, 0.0377, 0.0487, 0.0689, 0.0121, 0.0281, 0.0382, 0.0718], device='cuda:3'), in_proj_covar=tensor([0.0117, 0.0116, 0.0126, 0.0118, 0.0083, 0.0115, 0.0134, 0.0147], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-22 23:31:51,843 INFO [train.py:894] (3/4) Epoch 9, batch 3100, loss[loss=0.2071, simple_loss=0.2752, pruned_loss=0.06951, over 18606.00 frames. ], tot_loss[loss=0.2291, simple_loss=0.2995, pruned_loss=0.07933, over 3715107.52 frames. ], batch size: 45, lr: 1.26e-02, grad_scale: 8.0 2022-12-22 23:32:00,424 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0005-25975-0_sp0.9 from training. Duration: 21.688875 2022-12-22 23:32:24,523 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3390, 1.7458, 1.5877, 1.5177, 2.1165, 2.6872, 2.6731, 1.9453], device='cuda:3'), covar=tensor([0.0427, 0.0317, 0.0463, 0.0284, 0.0250, 0.0313, 0.0234, 0.0335], device='cuda:3'), in_proj_covar=tensor([0.0083, 0.0114, 0.0139, 0.0121, 0.0104, 0.0103, 0.0088, 0.0115], device='cuda:3'), out_proj_covar=tensor([7.2382e-05, 9.7587e-05, 1.2484e-04, 1.0374e-04, 9.1668e-05, 8.5671e-05, 7.5308e-05, 9.7572e-05], device='cuda:3') 2022-12-22 23:32:36,480 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0_sp0.9 from training. Duration: 22.6 2022-12-22 23:33:08,578 INFO [train.py:894] (3/4) Epoch 9, batch 3150, loss[loss=0.2172, simple_loss=0.2966, pruned_loss=0.06888, over 18599.00 frames. ], tot_loss[loss=0.2284, simple_loss=0.2988, pruned_loss=0.07907, over 3714138.33 frames. ], batch size: 78, lr: 1.26e-02, grad_scale: 8.0 2022-12-22 23:33:13,590 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0 from training. Duration: 24.32 2022-12-22 23:33:44,244 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.945e+02 5.767e+02 6.869e+02 8.584e+02 1.481e+03, threshold=1.374e+03, percent-clipped=1.0 2022-12-22 23:34:14,794 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-276745-0093-13116-0_sp0.9 from training. Duration: 21.061125 2022-12-22 23:34:28,090 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0024-15855-0_sp0.9 from training. Duration: 20.32225 2022-12-22 23:34:29,493 INFO [train.py:894] (3/4) Epoch 9, batch 3200, loss[loss=0.2258, simple_loss=0.2982, pruned_loss=0.07667, over 18494.00 frames. ], tot_loss[loss=0.2281, simple_loss=0.2985, pruned_loss=0.07892, over 3715071.24 frames. ], batch size: 52, lr: 1.26e-02, grad_scale: 8.0 2022-12-22 23:34:32,211 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2022-12-22 23:34:41,509 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp1.1 from training. Duration: 0.7545625 2022-12-22 23:34:43,087 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31259.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:34:55,863 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0_sp0.9 from training. Duration: 23.9333125 2022-12-22 23:35:17,404 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31281.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:35:29,397 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp1.1 from training. Duration: 20.17275 2022-12-22 23:35:31,159 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31290.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:35:35,204 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp1.1 from training. Duration: 20.436375 2022-12-22 23:35:45,274 INFO [train.py:894] (3/4) Epoch 9, batch 3250, loss[loss=0.2508, simple_loss=0.3126, pruned_loss=0.09451, over 18473.00 frames. ], tot_loss[loss=0.2296, simple_loss=0.2999, pruned_loss=0.07962, over 3715608.76 frames. ], batch size: 54, lr: 1.26e-02, grad_scale: 16.0 2022-12-22 23:35:48,938 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6155, 2.2493, 1.4919, 2.4461, 1.9672, 1.9361, 2.0872, 2.4548], device='cuda:3'), covar=tensor([0.1720, 0.2465, 0.1671, 0.2572, 0.2809, 0.0970, 0.2325, 0.0676], device='cuda:3'), in_proj_covar=tensor([0.0276, 0.0263, 0.0225, 0.0341, 0.0248, 0.0214, 0.0260, 0.0185], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 23:36:01,455 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31311.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:36:19,275 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.404e+02 5.691e+02 7.000e+02 8.164e+02 1.618e+03, threshold=1.400e+03, percent-clipped=2.0 2022-12-22 23:36:57,750 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0_sp0.9 from training. Duration: 23.1055625 2022-12-22 23:36:59,100 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp1.1 from training. Duration: 23.5 2022-12-22 23:37:00,500 INFO [train.py:894] (3/4) Epoch 9, batch 3300, loss[loss=0.2695, simple_loss=0.3304, pruned_loss=0.1043, over 18691.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.2999, pruned_loss=0.07935, over 3714963.76 frames. ], batch size: 69, lr: 1.26e-02, grad_scale: 16.0 2022-12-22 23:37:09,695 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp0.9 from training. Duration: 26.62775 2022-12-22 23:37:14,212 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=31359.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:37:21,736 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0 from training. Duration: 21.105 2022-12-22 23:37:26,574 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0_sp0.9 from training. Duration: 24.411125 2022-12-22 23:37:26,714 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31367.0, num_to_drop=1, layers_to_drop={2} 2022-12-22 23:37:54,881 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp1.1 from training. Duration: 21.263625 2022-12-22 23:38:17,130 INFO [train.py:894] (3/4) Epoch 9, batch 3350, loss[loss=0.2291, simple_loss=0.3041, pruned_loss=0.07699, over 18616.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.2991, pruned_loss=0.07868, over 3714677.45 frames. ], batch size: 69, lr: 1.26e-02, grad_scale: 16.0 2022-12-22 23:38:27,942 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0 from training. Duration: 20.795 2022-12-22 23:38:31,950 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31409.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:38:38,862 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0 from training. Duration: 24.76 2022-12-22 23:38:38,888 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0_sp0.9 from training. Duration: 22.25 2022-12-22 23:38:52,964 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.333e+02 5.442e+02 7.006e+02 9.337e+02 1.866e+03, threshold=1.401e+03, percent-clipped=5.0 2022-12-22 23:39:04,059 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.6753, 3.9241, 4.0373, 4.5593, 4.1133, 4.0621, 4.7289, 1.3933], device='cuda:3'), covar=tensor([0.0661, 0.0617, 0.0595, 0.0675, 0.1519, 0.1124, 0.0535, 0.4825], device='cuda:3'), in_proj_covar=tensor([0.0296, 0.0197, 0.0204, 0.0205, 0.0280, 0.0232, 0.0236, 0.0254], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 23:39:06,431 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp1.1 from training. Duration: 20.5045625 2022-12-22 23:39:32,947 INFO [train.py:894] (3/4) Epoch 9, batch 3400, loss[loss=0.2236, simple_loss=0.2994, pruned_loss=0.07393, over 18685.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.2976, pruned_loss=0.07756, over 3714888.85 frames. ], batch size: 62, lr: 1.25e-02, grad_scale: 16.0 2022-12-22 23:40:01,532 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31470.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:40:23,171 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4847, 2.1518, 1.4790, 2.6208, 1.9014, 1.8901, 2.0792, 2.5684], device='cuda:3'), covar=tensor([0.1646, 0.2603, 0.1567, 0.2276, 0.2644, 0.0921, 0.2239, 0.0624], device='cuda:3'), in_proj_covar=tensor([0.0271, 0.0261, 0.0222, 0.0335, 0.0243, 0.0210, 0.0257, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 23:40:32,679 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31491.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 23:40:45,261 INFO [train.py:894] (3/4) Epoch 9, batch 3450, loss[loss=0.2073, simple_loss=0.284, pruned_loss=0.06531, over 18687.00 frames. ], tot_loss[loss=0.2268, simple_loss=0.298, pruned_loss=0.07775, over 3715579.89 frames. ], batch size: 50, lr: 1.25e-02, grad_scale: 16.0 2022-12-22 23:41:18,005 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9847, 1.7259, 2.1668, 1.3011, 2.3704, 2.2710, 1.5560, 2.6931], device='cuda:3'), covar=tensor([0.1034, 0.1555, 0.1257, 0.1670, 0.0713, 0.1049, 0.1985, 0.0454], device='cuda:3'), in_proj_covar=tensor([0.0194, 0.0197, 0.0199, 0.0187, 0.0180, 0.0206, 0.0205, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 23:41:20,629 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.481e+02 5.213e+02 6.398e+02 7.592e+02 2.358e+03, threshold=1.280e+03, percent-clipped=3.0 2022-12-22 23:42:01,854 INFO [train.py:894] (3/4) Epoch 9, batch 3500, loss[loss=0.274, simple_loss=0.3287, pruned_loss=0.1096, over 18574.00 frames. ], tot_loss[loss=0.2277, simple_loss=0.2987, pruned_loss=0.0784, over 3715250.42 frames. ], batch size: 176, lr: 1.25e-02, grad_scale: 16.0 2022-12-22 23:42:05,278 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31552.0, num_to_drop=1, layers_to_drop={2} 2022-12-22 23:42:22,997 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp1.1 from training. Duration: 22.2954375 2022-12-22 23:42:34,706 INFO [train.py:894] (3/4) Epoch 10, batch 0, loss[loss=0.2304, simple_loss=0.3122, pruned_loss=0.07433, over 18612.00 frames. ], tot_loss[loss=0.2304, simple_loss=0.3122, pruned_loss=0.07433, over 18612.00 frames. ], batch size: 57, lr: 1.19e-02, grad_scale: 16.0 2022-12-22 23:42:34,707 INFO [train.py:919] (3/4) Computing validation loss 2022-12-22 23:42:45,648 INFO [train.py:928] (3/4) Epoch 10, validation: loss=0.1774, simple_loss=0.277, pruned_loss=0.03887, over 944034.00 frames. 2022-12-22 23:42:45,649 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24676MB 2022-12-22 23:42:50,486 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31559.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:43:25,407 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31581.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:43:36,799 WARNING [train.py:1060] (3/4) Exclude cut with ID 298-126791-0067-24026-0_sp0.9 from training. Duration: 21.438875 2022-12-22 23:43:38,516 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31590.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:43:41,212 WARNING [train.py:1060] (3/4) Exclude cut with ID 5652-39938-0025-23684-0_sp0.9 from training. Duration: 22.2055625 2022-12-22 23:43:53,981 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.09 vs. limit=5.0 2022-12-22 23:44:02,467 INFO [train.py:894] (3/4) Epoch 10, batch 50, loss[loss=0.2169, simple_loss=0.2982, pruned_loss=0.06776, over 18429.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2924, pruned_loss=0.06569, over 837645.61 frames. ], batch size: 48, lr: 1.19e-02, grad_scale: 16.0 2022-12-22 23:44:03,982 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=31607.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:44:07,034 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.6451, 3.9482, 4.0100, 4.5859, 4.2566, 4.1631, 4.7880, 1.5094], device='cuda:3'), covar=tensor([0.0514, 0.0517, 0.0436, 0.0512, 0.1028, 0.0845, 0.0339, 0.4165], device='cuda:3'), in_proj_covar=tensor([0.0286, 0.0193, 0.0198, 0.0200, 0.0270, 0.0226, 0.0229, 0.0247], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 23:44:11,514 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5044, 1.0534, 1.8471, 2.5860, 1.7604, 2.3191, 0.9980, 1.9080], device='cuda:3'), covar=tensor([0.2013, 0.2590, 0.1642, 0.0857, 0.1570, 0.1442, 0.2516, 0.1553], device='cuda:3'), in_proj_covar=tensor([0.0101, 0.0113, 0.0126, 0.0120, 0.0101, 0.0129, 0.0128, 0.0107], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 23:44:28,597 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.286e+02 4.881e+02 5.764e+02 7.674e+02 1.540e+03, threshold=1.153e+03, percent-clipped=4.0 2022-12-22 23:44:37,726 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=31629.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:44:50,674 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=31638.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:44:51,023 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0860, 2.1539, 1.2919, 2.3525, 2.3099, 1.9012, 3.1070, 1.9800], device='cuda:3'), covar=tensor([0.0808, 0.1557, 0.2650, 0.1766, 0.1486, 0.0895, 0.0803, 0.1169], device='cuda:3'), in_proj_covar=tensor([0.0173, 0.0189, 0.0231, 0.0276, 0.0221, 0.0180, 0.0202, 0.0191], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 23:45:17,537 INFO [train.py:894] (3/4) Epoch 10, batch 100, loss[loss=0.2293, simple_loss=0.3103, pruned_loss=0.07419, over 18456.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2891, pruned_loss=0.06332, over 1474603.11 frames. ], batch size: 64, lr: 1.19e-02, grad_scale: 16.0 2022-12-22 23:45:35,745 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31667.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 23:46:23,259 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3071, 2.5463, 2.8426, 1.4459, 3.1672, 2.8624, 2.1927, 3.4874], device='cuda:3'), covar=tensor([0.1290, 0.1543, 0.1362, 0.2262, 0.0726, 0.1194, 0.1801, 0.0467], device='cuda:3'), in_proj_covar=tensor([0.0196, 0.0197, 0.0202, 0.0191, 0.0182, 0.0208, 0.0206, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 23:46:30,455 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31704.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:46:33,007 INFO [train.py:894] (3/4) Epoch 10, batch 150, loss[loss=0.2137, simple_loss=0.298, pruned_loss=0.06475, over 18618.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2881, pruned_loss=0.06284, over 1970889.02 frames. ], batch size: 56, lr: 1.19e-02, grad_scale: 16.0 2022-12-22 23:46:36,043 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0 from training. Duration: 24.525 2022-12-22 23:46:47,729 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=31715.0, num_to_drop=1, layers_to_drop={0} 2022-12-22 23:46:58,977 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.574e+02 4.062e+02 5.113e+02 6.573e+02 1.615e+03, threshold=1.023e+03, percent-clipped=2.0 2022-12-22 23:47:09,483 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2022-12-22 23:47:12,907 WARNING [train.py:1060] (3/4) Exclude cut with ID 3699-47246-0007-3408-0_sp0.9 from training. Duration: 20.26675 2022-12-22 23:47:25,894 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp0.9 from training. Duration: 27.25 2022-12-22 23:47:48,634 INFO [train.py:894] (3/4) Epoch 10, batch 200, loss[loss=0.1581, simple_loss=0.2435, pruned_loss=0.03641, over 18590.00 frames. ], tot_loss[loss=0.206, simple_loss=0.287, pruned_loss=0.06246, over 2358170.28 frames. ], batch size: 41, lr: 1.19e-02, grad_scale: 16.0 2022-12-22 23:48:02,762 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31765.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:48:03,005 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31765.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:48:37,906 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31788.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:48:40,571 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0 from training. Duration: 21.68 2022-12-22 23:48:52,713 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0 from training. Duration: 21.6300625 2022-12-22 23:49:04,722 INFO [train.py:894] (3/4) Epoch 10, batch 250, loss[loss=0.1797, simple_loss=0.2551, pruned_loss=0.05217, over 18425.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2852, pruned_loss=0.0615, over 2658524.06 frames. ], batch size: 42, lr: 1.19e-02, grad_scale: 16.0 2022-12-22 23:49:12,161 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2022-12-22 23:49:17,739 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0_sp0.9 from training. Duration: 24.033375 2022-12-22 23:49:30,996 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.762e+02 3.941e+02 4.740e+02 5.956e+02 1.524e+03, threshold=9.481e+02, percent-clipped=1.0 2022-12-22 23:49:31,311 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31823.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:50:07,725 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31847.0, num_to_drop=1, layers_to_drop={3} 2022-12-22 23:50:10,623 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31849.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:50:12,728 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2022-12-22 23:50:16,230 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0 from training. Duration: 22.905 2022-12-22 23:50:17,688 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp1.1 from training. Duration: 23.4318125 2022-12-22 23:50:21,130 INFO [train.py:894] (3/4) Epoch 10, batch 300, loss[loss=0.2194, simple_loss=0.2916, pruned_loss=0.07358, over 18467.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2837, pruned_loss=0.06005, over 2892545.69 frames. ], batch size: 50, lr: 1.19e-02, grad_scale: 16.0 2022-12-22 23:50:47,878 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5746, 4.0367, 3.9122, 1.7510, 4.1149, 3.0293, 0.5388, 2.8329], device='cuda:3'), covar=tensor([0.2206, 0.0845, 0.1248, 0.3774, 0.0808, 0.1039, 0.5964, 0.1645], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0118, 0.0146, 0.0117, 0.0121, 0.0103, 0.0141, 0.0108], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 23:50:55,778 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31878.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:51:04,483 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31884.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:51:11,248 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7402, 1.6127, 1.7712, 1.6847, 1.5015, 3.5749, 1.6446, 2.2912], device='cuda:3'), covar=tensor([0.3034, 0.1935, 0.1773, 0.1887, 0.1267, 0.0164, 0.1555, 0.0802], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0121, 0.0129, 0.0122, 0.0104, 0.0099, 0.0098, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-22 23:51:37,213 INFO [train.py:894] (3/4) Epoch 10, batch 350, loss[loss=0.1785, simple_loss=0.2687, pruned_loss=0.04419, over 18520.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2839, pruned_loss=0.05967, over 3074381.37 frames. ], batch size: 55, lr: 1.18e-02, grad_scale: 16.0 2022-12-22 23:51:48,715 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5320, 1.1566, 1.4727, 2.7676, 2.0597, 2.0966, 0.7586, 1.9919], device='cuda:3'), covar=tensor([0.1692, 0.1847, 0.1686, 0.0603, 0.1137, 0.1313, 0.2380, 0.1199], device='cuda:3'), in_proj_covar=tensor([0.0103, 0.0116, 0.0129, 0.0123, 0.0103, 0.0133, 0.0130, 0.0110], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-22 23:51:51,992 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31915.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:52:03,207 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.016e+02 4.446e+02 5.363e+02 6.750e+02 1.599e+03, threshold=1.073e+03, percent-clipped=7.0 2022-12-22 23:52:13,334 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp1.1 from training. Duration: 20.82275 2022-12-22 23:52:14,636 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp0.9 from training. Duration: 25.45 2022-12-22 23:52:27,807 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31939.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:52:37,139 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5061, 1.9629, 1.6098, 0.6777, 1.5976, 1.9505, 1.4817, 1.8107], device='cuda:3'), covar=tensor([0.0499, 0.0612, 0.1101, 0.1553, 0.1087, 0.1211, 0.1482, 0.0697], device='cuda:3'), in_proj_covar=tensor([0.0164, 0.0181, 0.0204, 0.0195, 0.0207, 0.0190, 0.0203, 0.0189], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-22 23:52:45,186 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31950.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:52:53,564 INFO [train.py:894] (3/4) Epoch 10, batch 400, loss[loss=0.2481, simple_loss=0.3219, pruned_loss=0.08717, over 18463.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2853, pruned_loss=0.06041, over 3216977.95 frames. ], batch size: 54, lr: 1.18e-02, grad_scale: 16.0 2022-12-22 23:53:13,547 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0 from training. Duration: 25.775 2022-12-22 23:53:24,102 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31976.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:53:34,154 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0_sp0.9 from training. Duration: 22.25 2022-12-22 23:54:06,307 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0 from training. Duration: 26.205 2022-12-22 23:54:12,709 INFO [train.py:894] (3/4) Epoch 10, batch 450, loss[loss=0.2132, simple_loss=0.2821, pruned_loss=0.07214, over 18406.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2876, pruned_loss=0.06208, over 3327118.77 frames. ], batch size: 42, lr: 1.18e-02, grad_scale: 16.0 2022-12-22 23:54:21,037 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=32011.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:54:23,846 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp0.9 from training. Duration: 30.1555625 2022-12-22 23:54:28,395 WARNING [train.py:1060] (3/4) Exclude cut with ID 1265-135635-0050-6781-0_sp0.9 from training. Duration: 21.8333125 2022-12-22 23:54:38,590 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.601e+02 4.140e+02 5.222e+02 6.476e+02 1.497e+03, threshold=1.044e+03, percent-clipped=4.0 2022-12-22 23:54:40,154 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp1.1 from training. Duration: 20.6545625 2022-12-22 23:55:22,843 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0045-39920-0_sp0.9 from training. Duration: 20.52225 2022-12-22 23:55:28,525 INFO [train.py:894] (3/4) Epoch 10, batch 500, loss[loss=0.2154, simple_loss=0.3012, pruned_loss=0.06484, over 18653.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2882, pruned_loss=0.06236, over 3413048.35 frames. ], batch size: 69, lr: 1.18e-02, grad_scale: 16.0 2022-12-22 23:55:33,941 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2022-12-22 23:55:35,969 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32060.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:55:43,756 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32065.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:55:44,903 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp0.9 from training. Duration: 29.1166875 2022-12-22 23:56:45,399 INFO [train.py:894] (3/4) Epoch 10, batch 550, loss[loss=0.2124, simple_loss=0.2978, pruned_loss=0.06346, over 18553.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2887, pruned_loss=0.06253, over 3479231.73 frames. ], batch size: 55, lr: 1.18e-02, grad_scale: 16.0 2022-12-22 23:56:45,504 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133211-0007-59831-0_sp0.9 from training. Duration: 21.388875 2022-12-22 23:56:47,220 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.4586, 1.8797, 2.2108, 0.7624, 1.2117, 2.3539, 2.0272, 1.7938], device='cuda:3'), covar=tensor([0.0647, 0.0293, 0.0248, 0.0361, 0.0319, 0.0266, 0.0207, 0.0481], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0152, 0.0105, 0.0126, 0.0133, 0.0119, 0.0139, 0.0140], device='cuda:3'), out_proj_covar=tensor([1.1198e-04, 1.2992e-04, 8.8396e-05, 1.0402e-04, 1.1189e-04, 1.0122e-04, 1.2009e-04, 1.1884e-04], device='cuda:3') 2022-12-22 23:56:56,146 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=32113.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:57:10,609 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.007e+02 4.171e+02 5.313e+02 6.192e+02 1.458e+03, threshold=1.063e+03, percent-clipped=4.0 2022-12-22 23:57:23,947 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0 from training. Duration: 22.72 2022-12-22 23:57:25,347 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0_sp0.9 from training. Duration: 22.7444375 2022-12-22 23:57:42,848 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32144.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:57:47,295 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32147.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 23:58:00,338 INFO [train.py:894] (3/4) Epoch 10, batch 600, loss[loss=0.1787, simple_loss=0.2607, pruned_loss=0.04832, over 18418.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2894, pruned_loss=0.06301, over 3531275.35 frames. ], batch size: 42, lr: 1.18e-02, grad_scale: 16.0 2022-12-22 23:58:10,355 WARNING [train.py:1060] (3/4) Exclude cut with ID 4133-6541-0027-40495-0_sp1.1 from training. Duration: 0.9681875 2022-12-22 23:58:11,625 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0_sp0.9 from training. Duration: 22.3166875 2022-12-22 23:58:17,760 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133212-0015-59917-0_sp0.9 from training. Duration: 21.8166875 2022-12-22 23:58:22,778 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6616, 2.4175, 1.7068, 1.2332, 3.3327, 2.8258, 2.1994, 1.7262], device='cuda:3'), covar=tensor([0.0291, 0.0335, 0.0545, 0.0704, 0.0102, 0.0283, 0.0477, 0.0791], device='cuda:3'), in_proj_covar=tensor([0.0117, 0.0115, 0.0126, 0.0119, 0.0085, 0.0116, 0.0134, 0.0148], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-22 23:58:35,999 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32179.0, num_to_drop=0, layers_to_drop=set() 2022-12-22 23:59:00,491 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=32195.0, num_to_drop=1, layers_to_drop={1} 2022-12-22 23:59:17,369 INFO [train.py:894] (3/4) Epoch 10, batch 650, loss[loss=0.1879, simple_loss=0.2836, pruned_loss=0.04607, over 18725.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2905, pruned_loss=0.0637, over 3571455.38 frames. ], batch size: 65, lr: 1.18e-02, grad_scale: 16.0 2022-12-22 23:59:43,718 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.998e+02 4.804e+02 5.575e+02 7.166e+02 1.378e+03, threshold=1.115e+03, percent-clipped=4.0 2022-12-22 23:59:56,245 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0887, 1.3513, 1.8218, 1.8043, 2.1215, 1.9322, 1.8249, 1.5279], device='cuda:3'), covar=tensor([0.1629, 0.2515, 0.1896, 0.2065, 0.1357, 0.0810, 0.2289, 0.1000], device='cuda:3'), in_proj_covar=tensor([0.0248, 0.0284, 0.0256, 0.0290, 0.0272, 0.0233, 0.0296, 0.0219], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 00:00:00,274 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32234.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:00:04,201 WARNING [train.py:1060] (3/4) Exclude cut with ID 4957-30119-0041-23990-0_sp0.9 from training. Duration: 20.22775 2022-12-23 00:00:35,986 INFO [train.py:894] (3/4) Epoch 10, batch 700, loss[loss=0.2307, simple_loss=0.3105, pruned_loss=0.07542, over 18622.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2908, pruned_loss=0.06355, over 3603080.31 frames. ], batch size: 53, lr: 1.18e-02, grad_scale: 16.0 2022-12-23 00:00:50,416 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp1.1 from training. Duration: 24.67275 2022-12-23 00:00:57,942 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32271.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:01:17,465 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.1332, 1.2936, 0.6205, 1.2240, 1.4348, 2.4205, 1.2240, 1.4409], device='cuda:3'), covar=tensor([0.0879, 0.1625, 0.1252, 0.0904, 0.1435, 0.0292, 0.1284, 0.1383], device='cuda:3'), in_proj_covar=tensor([0.0074, 0.0082, 0.0076, 0.0075, 0.0092, 0.0071, 0.0086, 0.0077], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2022-12-23 00:01:18,628 WARNING [train.py:1060] (3/4) Exclude cut with ID 3082-165428-0081-50734-0_sp0.9 from training. Duration: 21.8055625 2022-12-23 00:01:50,524 INFO [train.py:894] (3/4) Epoch 10, batch 750, loss[loss=0.2221, simple_loss=0.308, pruned_loss=0.06808, over 18663.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2905, pruned_loss=0.06384, over 3626567.51 frames. ], batch size: 98, lr: 1.18e-02, grad_scale: 16.0 2022-12-23 00:01:50,709 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32306.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:01:57,743 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0_sp0.9 from training. Duration: 22.6666875 2022-12-23 00:02:15,111 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.578e+02 4.144e+02 4.884e+02 6.281e+02 1.414e+03, threshold=9.769e+02, percent-clipped=2.0 2022-12-23 00:02:59,161 WARNING [train.py:1060] (3/4) Exclude cut with ID 2411-132532-0017-82279-0_sp1.1 from training. Duration: 0.9681875 2022-12-23 00:03:05,509 INFO [train.py:894] (3/4) Epoch 10, batch 800, loss[loss=0.2102, simple_loss=0.2901, pruned_loss=0.06514, over 18394.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2903, pruned_loss=0.06375, over 3646113.84 frames. ], batch size: 46, lr: 1.18e-02, grad_scale: 16.0 2022-12-23 00:03:12,438 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32360.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:03:26,883 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0 from training. Duration: 22.485 2022-12-23 00:04:03,867 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp1.1 from training. Duration: 23.82275 2022-12-23 00:04:17,770 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0 from training. Duration: 20.77 2022-12-23 00:04:22,598 INFO [train.py:894] (3/4) Epoch 10, batch 850, loss[loss=0.2175, simple_loss=0.2926, pruned_loss=0.0712, over 18681.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2906, pruned_loss=0.0633, over 3661633.28 frames. ], batch size: 60, lr: 1.18e-02, grad_scale: 16.0 2022-12-23 00:04:25,621 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=32408.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:04:27,080 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0_sp0.9 from training. Duration: 24.088875 2022-12-23 00:04:48,423 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.428e+02 4.593e+02 5.340e+02 6.790e+02 1.639e+03, threshold=1.068e+03, percent-clipped=5.0 2022-12-23 00:04:55,574 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp1.1 from training. Duration: 20.4409375 2022-12-23 00:05:13,008 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3334, 1.7671, 1.4271, 2.0646, 2.2888, 1.3790, 1.2648, 1.0982], device='cuda:3'), covar=tensor([0.2078, 0.1823, 0.1537, 0.1024, 0.1397, 0.1271, 0.2085, 0.1673], device='cuda:3'), in_proj_covar=tensor([0.0232, 0.0207, 0.0198, 0.0186, 0.0248, 0.0185, 0.0206, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 00:05:20,371 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32444.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:05:39,524 INFO [train.py:894] (3/4) Epoch 10, batch 900, loss[loss=0.2291, simple_loss=0.3192, pruned_loss=0.06947, over 18511.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2909, pruned_loss=0.06335, over 3673869.35 frames. ], batch size: 52, lr: 1.17e-02, grad_scale: 16.0 2022-12-23 00:06:00,234 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2022-12-23 00:06:12,523 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0_sp0.9 from training. Duration: 22.511125 2022-12-23 00:06:13,840 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0 from training. Duration: 20.675 2022-12-23 00:06:14,126 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32479.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:06:33,725 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=32492.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:06:34,421 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2022-12-23 00:06:37,323 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.27 vs. limit=5.0 2022-12-23 00:06:55,148 INFO [train.py:894] (3/4) Epoch 10, batch 950, loss[loss=0.2014, simple_loss=0.2939, pruned_loss=0.05446, over 18590.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2908, pruned_loss=0.06326, over 3682960.93 frames. ], batch size: 51, lr: 1.17e-02, grad_scale: 16.0 2022-12-23 00:07:19,173 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.034e+02 4.294e+02 5.171e+02 6.457e+02 1.266e+03, threshold=1.034e+03, percent-clipped=2.0 2022-12-23 00:07:25,137 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=32527.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:07:36,846 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32534.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:07:50,262 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp0.9 from training. Duration: 24.9833125 2022-12-23 00:08:10,205 INFO [train.py:894] (3/4) Epoch 10, batch 1000, loss[loss=0.2456, simple_loss=0.3221, pruned_loss=0.08457, over 18628.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2906, pruned_loss=0.06328, over 3689273.78 frames. ], batch size: 60, lr: 1.17e-02, grad_scale: 16.0 2022-12-23 00:08:20,644 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0 from training. Duration: 27.14 2022-12-23 00:08:28,044 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.8972, 3.8532, 3.7728, 2.0612, 3.7906, 2.8592, 1.1313, 2.6729], device='cuda:3'), covar=tensor([0.1893, 0.1057, 0.1171, 0.3305, 0.0880, 0.0998, 0.4745, 0.1609], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0118, 0.0147, 0.0119, 0.0122, 0.0103, 0.0142, 0.0109], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 00:08:32,477 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32571.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:08:35,455 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0 from training. Duration: 22.44 2022-12-23 00:08:50,738 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=32582.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:09:27,553 INFO [train.py:894] (3/4) Epoch 10, batch 1050, loss[loss=0.1808, simple_loss=0.2622, pruned_loss=0.04966, over 18423.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2903, pruned_loss=0.06326, over 3694851.69 frames. ], batch size: 48, lr: 1.17e-02, grad_scale: 8.0 2022-12-23 00:09:27,978 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32606.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:09:47,491 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=32619.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:09:51,929 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0060-62364-0_sp0.9 from training. Duration: 21.361125 2022-12-23 00:09:54,782 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.674e+02 4.245e+02 5.244e+02 7.030e+02 1.488e+03, threshold=1.049e+03, percent-clipped=6.0 2022-12-23 00:09:57,873 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp1.1 from training. Duration: 27.0318125 2022-12-23 00:10:08,596 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp0.9 from training. Duration: 28.638875 2022-12-23 00:10:24,086 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0 from training. Duration: 20.4 2022-12-23 00:10:24,888 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.05 vs. limit=5.0 2022-12-23 00:10:40,848 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=32654.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:10:43,585 INFO [train.py:894] (3/4) Epoch 10, batch 1100, loss[loss=0.2257, simple_loss=0.3041, pruned_loss=0.07363, over 18513.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2895, pruned_loss=0.06279, over 3698205.85 frames. ], batch size: 58, lr: 1.17e-02, grad_scale: 8.0 2022-12-23 00:10:59,680 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0 from training. Duration: 20.025 2022-12-23 00:10:59,700 WARNING [train.py:1060] (3/4) Exclude cut with ID 2364-131735-0112-64612-0_sp0.9 from training. Duration: 20.488875 2022-12-23 00:11:04,311 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0 from training. Duration: 29.735 2022-12-23 00:11:11,216 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6478, 1.4453, 1.3985, 2.0216, 1.6131, 3.5867, 1.4347, 1.5162], device='cuda:3'), covar=tensor([0.0888, 0.1725, 0.1104, 0.0875, 0.1375, 0.0193, 0.1294, 0.1556], device='cuda:3'), in_proj_covar=tensor([0.0075, 0.0083, 0.0077, 0.0075, 0.0093, 0.0073, 0.0087, 0.0078], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 00:12:00,082 INFO [train.py:894] (3/4) Epoch 10, batch 1150, loss[loss=0.2491, simple_loss=0.3115, pruned_loss=0.09332, over 18682.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2885, pruned_loss=0.06236, over 3702465.73 frames. ], batch size: 178, lr: 1.17e-02, grad_scale: 8.0 2022-12-23 00:12:27,862 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.652e+02 4.092e+02 4.850e+02 6.073e+02 1.087e+03, threshold=9.700e+02, percent-clipped=2.0 2022-12-23 00:12:27,969 WARNING [train.py:1060] (3/4) Exclude cut with ID 7276-92427-0014-12983-0_sp0.9 from training. Duration: 21.3055625 2022-12-23 00:12:29,914 WARNING [train.py:1060] (3/4) Exclude cut with ID 1025-75365-0008-79168-0_sp0.9 from training. Duration: 22.0666875 2022-12-23 00:13:16,602 INFO [train.py:894] (3/4) Epoch 10, batch 1200, loss[loss=0.1956, simple_loss=0.2775, pruned_loss=0.05689, over 18670.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2877, pruned_loss=0.06177, over 3704362.69 frames. ], batch size: 48, lr: 1.17e-02, grad_scale: 8.0 2022-12-23 00:13:44,027 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7710, 1.1533, 0.8898, 1.3016, 2.0094, 1.1011, 1.4752, 1.6562], device='cuda:3'), covar=tensor([0.1671, 0.2397, 0.2655, 0.1678, 0.1680, 0.1827, 0.1601, 0.1739], device='cuda:3'), in_proj_covar=tensor([0.0089, 0.0099, 0.0119, 0.0096, 0.0113, 0.0090, 0.0097, 0.0095], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 00:14:05,514 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3632, 3.2896, 3.1759, 1.3884, 3.3130, 2.5510, 0.4570, 2.1952], device='cuda:3'), covar=tensor([0.2308, 0.1126, 0.1654, 0.4022, 0.1006, 0.1125, 0.6061, 0.1755], device='cuda:3'), in_proj_covar=tensor([0.0131, 0.0118, 0.0144, 0.0118, 0.0121, 0.0102, 0.0142, 0.0108], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 00:14:22,877 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0 from training. Duration: 20.26 2022-12-23 00:14:33,122 INFO [train.py:894] (3/4) Epoch 10, batch 1250, loss[loss=0.244, simple_loss=0.3217, pruned_loss=0.08311, over 18603.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2878, pruned_loss=0.06154, over 3706319.47 frames. ], batch size: 51, lr: 1.17e-02, grad_scale: 8.0 2022-12-23 00:14:36,131 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0030-9324-0_sp0.9 from training. Duration: 21.3444375 2022-12-23 00:14:36,403 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6358, 1.3116, 1.7170, 2.8004, 1.9612, 2.3885, 1.2449, 1.9059], device='cuda:3'), covar=tensor([0.1711, 0.1621, 0.1424, 0.0564, 0.1106, 0.1400, 0.1930, 0.1230], device='cuda:3'), in_proj_covar=tensor([0.0103, 0.0113, 0.0129, 0.0123, 0.0102, 0.0133, 0.0129, 0.0109], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 00:15:00,107 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.696e+02 4.479e+02 5.169e+02 6.782e+02 1.244e+03, threshold=1.034e+03, percent-clipped=4.0 2022-12-23 00:15:24,603 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2022-12-23 00:15:36,297 WARNING [train.py:1060] (3/4) Exclude cut with ID 497-129325-0061-62254-0_sp1.1 from training. Duration: 0.97725 2022-12-23 00:15:49,586 INFO [train.py:894] (3/4) Epoch 10, batch 1300, loss[loss=0.1844, simple_loss=0.2631, pruned_loss=0.05287, over 18532.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2865, pruned_loss=0.06082, over 3708210.99 frames. ], batch size: 44, lr: 1.17e-02, grad_scale: 8.0 2022-12-23 00:16:17,485 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0_sp0.9 from training. Duration: 22.97225 2022-12-23 00:16:48,479 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0047-39922-0_sp0.9 from training. Duration: 21.97775 2022-12-23 00:16:54,732 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6055, 1.6297, 1.6829, 1.7089, 1.3918, 3.4280, 1.8292, 2.2797], device='cuda:3'), covar=tensor([0.3708, 0.2309, 0.2009, 0.2092, 0.1435, 0.0220, 0.1484, 0.0870], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0121, 0.0130, 0.0121, 0.0104, 0.0098, 0.0098, 0.0095], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 00:17:02,955 WARNING [train.py:1060] (3/4) Exclude cut with ID 1112-1043-0006-89194-0_sp0.9 from training. Duration: 21.8333125 2022-12-23 00:17:05,854 INFO [train.py:894] (3/4) Epoch 10, batch 1350, loss[loss=0.1704, simple_loss=0.2479, pruned_loss=0.04646, over 18683.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2871, pruned_loss=0.06137, over 3710450.38 frames. ], batch size: 46, lr: 1.17e-02, grad_scale: 8.0 2022-12-23 00:17:12,910 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0 from training. Duration: 20.47 2022-12-23 00:17:33,577 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.207e+02 4.434e+02 5.433e+02 6.652e+02 1.215e+03, threshold=1.087e+03, percent-clipped=3.0 2022-12-23 00:18:12,588 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.88 vs. limit=5.0 2022-12-23 00:18:19,007 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0037-39912-0_sp0.9 from training. Duration: 20.67225 2022-12-23 00:18:22,023 INFO [train.py:894] (3/4) Epoch 10, batch 1400, loss[loss=0.1817, simple_loss=0.2662, pruned_loss=0.04863, over 18665.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2867, pruned_loss=0.06142, over 3711474.01 frames. ], batch size: 48, lr: 1.17e-02, grad_scale: 8.0 2022-12-23 00:18:38,031 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=32966.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:18:39,270 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp0.9 from training. Duration: 25.2444375 2022-12-23 00:19:03,551 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0021-76797-0_sp0.9 from training. Duration: 21.1445 2022-12-23 00:19:37,772 INFO [train.py:894] (3/4) Epoch 10, batch 1450, loss[loss=0.2305, simple_loss=0.3114, pruned_loss=0.07482, over 18511.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2864, pruned_loss=0.06095, over 3711359.10 frames. ], batch size: 68, lr: 1.16e-02, grad_scale: 8.0 2022-12-23 00:19:41,340 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6582, 2.1132, 1.4598, 2.5042, 1.9606, 1.9368, 1.9758, 2.6064], device='cuda:3'), covar=tensor([0.1545, 0.2564, 0.1557, 0.2260, 0.2874, 0.0935, 0.2429, 0.0643], device='cuda:3'), in_proj_covar=tensor([0.0272, 0.0263, 0.0219, 0.0331, 0.0244, 0.0208, 0.0258, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 00:19:59,864 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.50 vs. limit=5.0 2022-12-23 00:20:05,660 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.915e+02 4.368e+02 5.270e+02 6.859e+02 1.163e+03, threshold=1.054e+03, percent-clipped=2.0 2022-12-23 00:20:10,985 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33027.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:20:16,527 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp0.9 from training. Duration: 33.038875 2022-12-23 00:20:16,824 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5383, 1.3937, 1.1657, 1.9617, 1.6283, 3.3839, 1.2168, 1.4725], device='cuda:3'), covar=tensor([0.0962, 0.1840, 0.1262, 0.0916, 0.1456, 0.0216, 0.1475, 0.1638], device='cuda:3'), in_proj_covar=tensor([0.0076, 0.0084, 0.0078, 0.0077, 0.0094, 0.0073, 0.0088, 0.0080], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 00:20:53,387 INFO [train.py:894] (3/4) Epoch 10, batch 1500, loss[loss=0.1785, simple_loss=0.2578, pruned_loss=0.0496, over 18619.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2866, pruned_loss=0.06078, over 3712239.30 frames. ], batch size: 45, lr: 1.16e-02, grad_scale: 8.0 2022-12-23 00:20:53,470 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64291-0000-16059-0_sp0.9 from training. Duration: 20.0944375 2022-12-23 00:21:09,543 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp1.1 from training. Duration: 20.4 2022-12-23 00:21:14,105 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33069.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:21:17,343 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0 from training. Duration: 20.085 2022-12-23 00:21:27,307 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0_sp0.9 from training. Duration: 23.07775 2022-12-23 00:22:07,064 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.89 vs. limit=5.0 2022-12-23 00:22:08,910 INFO [train.py:894] (3/4) Epoch 10, batch 1550, loss[loss=0.2219, simple_loss=0.3028, pruned_loss=0.07048, over 18563.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2872, pruned_loss=0.06085, over 3712738.76 frames. ], batch size: 77, lr: 1.16e-02, grad_scale: 8.0 2022-12-23 00:22:12,892 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp0.9 from training. Duration: 24.9333125 2022-12-23 00:22:20,901 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8789, 0.9908, 0.9167, 1.2558, 1.9309, 1.0283, 1.6132, 1.6610], device='cuda:3'), covar=tensor([0.1374, 0.2259, 0.2338, 0.1571, 0.1883, 0.1728, 0.1413, 0.1666], device='cuda:3'), in_proj_covar=tensor([0.0090, 0.0101, 0.0120, 0.0097, 0.0113, 0.0090, 0.0097, 0.0096], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 00:22:37,934 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.803e+02 4.567e+02 5.574e+02 6.708e+02 1.321e+03, threshold=1.115e+03, percent-clipped=4.0 2022-12-23 00:22:48,020 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33130.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:22:57,549 WARNING [train.py:1060] (3/4) Exclude cut with ID 5118-111612-0016-124680-0_sp0.9 from training. Duration: 20.388875 2022-12-23 00:23:03,649 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp1.1 from training. Duration: 20.3590625 2022-12-23 00:23:26,028 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33155.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:23:27,242 INFO [train.py:894] (3/4) Epoch 10, batch 1600, loss[loss=0.2179, simple_loss=0.3036, pruned_loss=0.06612, over 18666.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2864, pruned_loss=0.06031, over 3713512.13 frames. ], batch size: 60, lr: 1.16e-02, grad_scale: 8.0 2022-12-23 00:24:10,625 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0_sp1.1 from training. Duration: 0.836375 2022-12-23 00:24:43,958 INFO [train.py:894] (3/4) Epoch 10, batch 1650, loss[loss=0.2525, simple_loss=0.334, pruned_loss=0.08552, over 18570.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2885, pruned_loss=0.06246, over 3713367.39 frames. ], batch size: 100, lr: 1.16e-02, grad_scale: 8.0 2022-12-23 00:24:55,786 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2022-12-23 00:24:56,498 WARNING [train.py:1060] (3/4) Exclude cut with ID 8565-290391-0049-67394-0_sp0.9 from training. Duration: 21.3166875 2022-12-23 00:24:59,910 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33216.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:25:10,464 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1523, 2.4519, 2.4746, 1.0740, 2.4977, 1.8617, 0.7870, 1.7057], device='cuda:3'), covar=tensor([0.1936, 0.1242, 0.1516, 0.3673, 0.1248, 0.1094, 0.4167, 0.1684], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0118, 0.0145, 0.0120, 0.0122, 0.0103, 0.0141, 0.0107], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 00:25:11,643 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.064e+02 4.439e+02 5.632e+02 7.567e+02 2.415e+03, threshold=1.126e+03, percent-clipped=5.0 2022-12-23 00:25:25,818 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0029-104863-0_sp0.9 from training. Duration: 22.1055625 2022-12-23 00:25:36,013 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp1.1 from training. Duration: 21.77725 2022-12-23 00:25:58,802 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp0.9 from training. Duration: 27.8166875 2022-12-23 00:26:00,242 INFO [train.py:894] (3/4) Epoch 10, batch 1700, loss[loss=0.2284, simple_loss=0.3066, pruned_loss=0.07514, over 18725.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2888, pruned_loss=0.06387, over 3713229.27 frames. ], batch size: 52, lr: 1.16e-02, grad_scale: 8.0 2022-12-23 00:26:21,851 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp1.1 from training. Duration: 22.5090625 2022-12-23 00:26:30,229 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0 from training. Duration: 25.035 2022-12-23 00:26:47,912 WARNING [train.py:1060] (3/4) Exclude cut with ID 774-127930-0014-10412-0_sp1.1 from training. Duration: 0.95 2022-12-23 00:27:07,605 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp0.9 from training. Duration: 0.92225 2022-12-23 00:27:16,776 INFO [train.py:894] (3/4) Epoch 10, batch 1750, loss[loss=0.2365, simple_loss=0.321, pruned_loss=0.07598, over 18693.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2908, pruned_loss=0.06639, over 3714252.16 frames. ], batch size: 60, lr: 1.16e-02, grad_scale: 8.0 2022-12-23 00:27:34,236 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0 from training. Duration: 21.97 2022-12-23 00:27:42,355 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33322.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:27:45,020 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.376e+02 5.253e+02 6.174e+02 7.896e+02 1.469e+03, threshold=1.235e+03, percent-clipped=6.0 2022-12-23 00:27:53,800 WARNING [train.py:1060] (3/4) Exclude cut with ID 7492-105653-0055-62765-0_sp0.9 from training. Duration: 21.97225 2022-12-23 00:27:55,147 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp0.9 from training. Duration: 25.3333125 2022-12-23 00:28:07,183 WARNING [train.py:1060] (3/4) Exclude cut with ID 5172-29468-0015-19128-0_sp0.9 from training. Duration: 21.5055625 2022-12-23 00:28:15,093 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp1.1 from training. Duration: 20.72725 2022-12-23 00:28:28,625 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3333, 1.9532, 1.4119, 0.5866, 1.4323, 1.9410, 1.5899, 1.7392], device='cuda:3'), covar=tensor([0.0525, 0.0441, 0.0937, 0.1288, 0.0967, 0.1137, 0.1316, 0.0537], device='cuda:3'), in_proj_covar=tensor([0.0164, 0.0179, 0.0201, 0.0195, 0.0204, 0.0188, 0.0204, 0.0190], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 00:28:33,085 INFO [train.py:894] (3/4) Epoch 10, batch 1800, loss[loss=0.2099, simple_loss=0.2947, pruned_loss=0.06252, over 18386.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2922, pruned_loss=0.06865, over 3713844.72 frames. ], batch size: 53, lr: 1.16e-02, grad_scale: 8.0 2022-12-23 00:28:49,045 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp0.9 from training. Duration: 26.32775 2022-12-23 00:29:20,125 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0 from training. Duration: 20.025 2022-12-23 00:29:24,462 WARNING [train.py:1060] (3/4) Exclude cut with ID 6709-74022-0004-86860-0_sp1.1 from training. Duration: 0.9409375 2022-12-23 00:29:24,478 WARNING [train.py:1060] (3/4) Exclude cut with ID 4757-1811-0023-62229-0_sp0.9 from training. Duration: 21.37775 2022-12-23 00:29:48,019 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0004-25974-0_sp0.9 from training. Duration: 21.17225 2022-12-23 00:29:48,036 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp0.9 from training. Duration: 27.511125 2022-12-23 00:29:51,067 INFO [train.py:894] (3/4) Epoch 10, batch 1850, loss[loss=0.2034, simple_loss=0.282, pruned_loss=0.0624, over 18500.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.2943, pruned_loss=0.0716, over 3712956.10 frames. ], batch size: 52, lr: 1.16e-02, grad_scale: 4.0 2022-12-23 00:30:19,988 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.458e+02 5.783e+02 6.744e+02 8.262e+02 1.612e+03, threshold=1.349e+03, percent-clipped=7.0 2022-12-23 00:30:20,233 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33425.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:30:21,501 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0 from training. Duration: 22.8 2022-12-23 00:30:24,295 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0 from training. Duration: 22.585 2022-12-23 00:30:55,930 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0001-146967-0_sp0.9 from training. Duration: 22.0166875 2022-12-23 00:31:07,983 INFO [train.py:894] (3/4) Epoch 10, batch 1900, loss[loss=0.2235, simple_loss=0.2951, pruned_loss=0.07598, over 18699.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.2958, pruned_loss=0.0732, over 3713684.19 frames. ], batch size: 50, lr: 1.16e-02, grad_scale: 4.0 2022-12-23 00:31:11,628 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp1.1 from training. Duration: 24.395375 2022-12-23 00:31:19,008 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp0.9 from training. Duration: 27.47775 2022-12-23 00:31:23,445 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp0.9 from training. Duration: 24.8833125 2022-12-23 00:31:26,468 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0 from training. Duration: 23.39 2022-12-23 00:31:31,985 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp0.9 from training. Duration: 28.72225 2022-12-23 00:31:32,257 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33472.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:31:40,796 WARNING [train.py:1060] (3/4) Exclude cut with ID 585-294811-0110-133686-0_sp0.9 from training. Duration: 20.8944375 2022-12-23 00:31:58,322 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0_sp0.9 from training. Duration: 23.8444375 2022-12-23 00:32:24,040 INFO [train.py:894] (3/4) Epoch 10, batch 1950, loss[loss=0.1859, simple_loss=0.2578, pruned_loss=0.05704, over 18390.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.2951, pruned_loss=0.07364, over 3712666.52 frames. ], batch size: 46, lr: 1.16e-02, grad_scale: 4.0 2022-12-23 00:32:25,489 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0 from training. Duration: 25.85 2022-12-23 00:32:25,501 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0 from training. Duration: 21.39 2022-12-23 00:32:31,514 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33511.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:32:35,894 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0 from training. Duration: 27.92 2022-12-23 00:32:51,917 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.820e+02 5.421e+02 6.446e+02 7.879e+02 3.029e+03, threshold=1.289e+03, percent-clipped=5.0 2022-12-23 00:33:02,264 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0039-130165-0_sp0.9 from training. Duration: 20.661125 2022-12-23 00:33:04,215 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33533.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:33:19,505 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8182, 1.8651, 1.3390, 2.0351, 2.0197, 1.7182, 2.6533, 1.8950], device='cuda:3'), covar=tensor([0.0842, 0.1442, 0.2486, 0.1536, 0.1556, 0.0834, 0.0852, 0.1101], device='cuda:3'), in_proj_covar=tensor([0.0169, 0.0190, 0.0228, 0.0274, 0.0217, 0.0178, 0.0204, 0.0191], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 00:33:21,205 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.47 vs. limit=5.0 2022-12-23 00:33:29,384 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0043-15874-0_sp0.9 from training. Duration: 20.07225 2022-12-23 00:33:36,930 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0 from training. Duration: 21.01 2022-12-23 00:33:39,319 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6045, 1.2998, 0.9463, 1.6875, 1.6568, 2.9504, 1.3532, 1.4911], device='cuda:3'), covar=tensor([0.0827, 0.1750, 0.1249, 0.0883, 0.1404, 0.0270, 0.1332, 0.1521], device='cuda:3'), in_proj_covar=tensor([0.0075, 0.0083, 0.0077, 0.0077, 0.0093, 0.0073, 0.0087, 0.0079], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 00:33:40,398 INFO [train.py:894] (3/4) Epoch 10, batch 2000, loss[loss=0.204, simple_loss=0.2768, pruned_loss=0.06561, over 18713.00 frames. ], tot_loss[loss=0.2222, simple_loss=0.2955, pruned_loss=0.07443, over 3713357.17 frames. ], batch size: 50, lr: 1.16e-02, grad_scale: 8.0 2022-12-23 00:33:48,221 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7606, 1.5462, 1.2872, 2.0395, 1.8054, 3.5011, 1.3741, 1.4398], device='cuda:3'), covar=tensor([0.0880, 0.1649, 0.1246, 0.0917, 0.1407, 0.0245, 0.1443, 0.1743], device='cuda:3'), in_proj_covar=tensor([0.0075, 0.0083, 0.0077, 0.0077, 0.0093, 0.0073, 0.0087, 0.0079], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 00:34:51,273 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0 from training. Duration: 20.65 2022-12-23 00:34:56,515 INFO [train.py:894] (3/4) Epoch 10, batch 2050, loss[loss=0.2253, simple_loss=0.3052, pruned_loss=0.07265, over 18462.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.2946, pruned_loss=0.07384, over 3712649.57 frames. ], batch size: 54, lr: 1.15e-02, grad_scale: 8.0 2022-12-23 00:34:58,013 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0 from training. Duration: 21.46 2022-12-23 00:35:21,220 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33622.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:35:25,502 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.406e+02 5.626e+02 7.229e+02 8.971e+02 2.818e+03, threshold=1.446e+03, percent-clipped=6.0 2022-12-23 00:35:41,092 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0278, 2.0941, 1.3302, 2.3062, 2.1365, 1.8806, 2.8324, 2.0211], device='cuda:3'), covar=tensor([0.0871, 0.1392, 0.2605, 0.1708, 0.1675, 0.0876, 0.1007, 0.1158], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0192, 0.0231, 0.0277, 0.0219, 0.0179, 0.0205, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 00:35:42,041 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0 from training. Duration: 0.92 2022-12-23 00:35:48,283 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0_sp0.9 from training. Duration: 23.7666875 2022-12-23 00:36:12,694 INFO [train.py:894] (3/4) Epoch 10, batch 2100, loss[loss=0.2708, simple_loss=0.3234, pruned_loss=0.1092, over 18624.00 frames. ], tot_loss[loss=0.2231, simple_loss=0.2961, pruned_loss=0.07502, over 3713174.22 frames. ], batch size: 177, lr: 1.15e-02, grad_scale: 8.0 2022-12-23 00:36:25,855 WARNING [train.py:1060] (3/4) Exclude cut with ID 8544-281189-0060-101339-0_sp0.9 from training. Duration: 20.861125 2022-12-23 00:36:33,439 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=33670.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:36:37,501 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0_sp0.9 from training. Duration: 22.711125 2022-12-23 00:36:43,656 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33677.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:37:07,256 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7881, 1.5387, 1.2009, 0.2933, 1.2576, 1.6058, 1.3714, 1.5737], device='cuda:3'), covar=tensor([0.0609, 0.0438, 0.0817, 0.1371, 0.0955, 0.1336, 0.1419, 0.0545], device='cuda:3'), in_proj_covar=tensor([0.0166, 0.0180, 0.0202, 0.0197, 0.0207, 0.0189, 0.0205, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 00:37:22,028 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp1.1 from training. Duration: 22.986375 2022-12-23 00:37:27,738 INFO [train.py:894] (3/4) Epoch 10, batch 2150, loss[loss=0.2301, simple_loss=0.304, pruned_loss=0.07807, over 18525.00 frames. ], tot_loss[loss=0.2234, simple_loss=0.2958, pruned_loss=0.07549, over 3713126.43 frames. ], batch size: 55, lr: 1.15e-02, grad_scale: 8.0 2022-12-23 00:37:33,620 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6888, 3.9466, 3.8275, 1.8191, 3.9325, 2.9876, 0.9860, 2.7788], device='cuda:3'), covar=tensor([0.2011, 0.0844, 0.1212, 0.3450, 0.0889, 0.1015, 0.5155, 0.1470], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0121, 0.0150, 0.0121, 0.0125, 0.0106, 0.0143, 0.0109], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 00:37:37,931 WARNING [train.py:1060] (3/4) Exclude cut with ID 8040-260924-0003-80960-0_sp0.9 from training. Duration: 22.07225 2022-12-23 00:37:42,261 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0045-26330-0_sp0.9 from training. Duration: 20.3055625 2022-12-23 00:37:45,567 WARNING [train.py:1060] (3/4) Exclude cut with ID 6356-271890-0060-94317-0_sp0.9 from training. Duration: 20.72225 2022-12-23 00:37:55,476 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.380e+02 5.930e+02 7.248e+02 8.853e+02 1.852e+03, threshold=1.450e+03, percent-clipped=4.0 2022-12-23 00:37:55,953 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33725.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:38:02,989 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp1.1 from training. Duration: 22.4818125 2022-12-23 00:38:17,468 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33738.0, num_to_drop=1, layers_to_drop={3} 2022-12-23 00:38:29,741 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp0.9 from training. Duration: 25.0944375 2022-12-23 00:38:32,597 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0 from training. Duration: 21.515 2022-12-23 00:38:40,113 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp0.9 from training. Duration: 27.02225 2022-12-23 00:38:44,315 INFO [train.py:894] (3/4) Epoch 10, batch 2200, loss[loss=0.1961, simple_loss=0.2711, pruned_loss=0.06056, over 18658.00 frames. ], tot_loss[loss=0.2235, simple_loss=0.2954, pruned_loss=0.07583, over 3714598.92 frames. ], batch size: 48, lr: 1.15e-02, grad_scale: 8.0 2022-12-23 00:38:45,867 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0010-62480-0_sp0.9 from training. Duration: 22.22225 2022-12-23 00:38:53,491 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0085-44554-0_sp0.9 from training. Duration: 20.85 2022-12-23 00:38:55,552 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5385, 1.2375, 1.3226, 1.5333, 1.7996, 1.6119, 1.7093, 1.1508], device='cuda:3'), covar=tensor([0.0314, 0.0243, 0.0418, 0.0199, 0.0182, 0.0351, 0.0246, 0.0270], device='cuda:3'), in_proj_covar=tensor([0.0083, 0.0117, 0.0142, 0.0121, 0.0107, 0.0106, 0.0089, 0.0117], device='cuda:3'), out_proj_covar=tensor([7.0632e-05, 9.8422e-05, 1.2546e-04, 1.0300e-04, 9.4131e-05, 8.7130e-05, 7.4858e-05, 9.8413e-05], device='cuda:3') 2022-12-23 00:38:59,490 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7961, 3.9237, 3.9459, 1.5826, 3.8420, 2.9399, 0.6519, 2.6521], device='cuda:3'), covar=tensor([0.1729, 0.0956, 0.1207, 0.3592, 0.0942, 0.0967, 0.5200, 0.1422], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0122, 0.0150, 0.0121, 0.0126, 0.0106, 0.0144, 0.0109], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 00:39:09,567 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=33773.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:39:26,271 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0 from training. Duration: 21.54 2022-12-23 00:39:30,709 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp1.1 from training. Duration: 20.5318125 2022-12-23 00:39:41,712 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0012-134311-0_sp0.9 from training. Duration: 21.9333125 2022-12-23 00:40:00,939 INFO [train.py:894] (3/4) Epoch 10, batch 2250, loss[loss=0.2331, simple_loss=0.3088, pruned_loss=0.07874, over 18451.00 frames. ], tot_loss[loss=0.2239, simple_loss=0.2955, pruned_loss=0.07609, over 3713872.82 frames. ], batch size: 54, lr: 1.15e-02, grad_scale: 8.0 2022-12-23 00:40:08,722 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33811.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:40:29,048 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.109e+02 5.120e+02 6.230e+02 8.011e+02 1.544e+03, threshold=1.246e+03, percent-clipped=1.0 2022-12-23 00:40:29,144 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0025-130151-0_sp0.9 from training. Duration: 21.7944375 2022-12-23 00:40:34,266 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33828.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:40:36,699 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7722, 1.5725, 1.5323, 1.6644, 1.8094, 1.9102, 2.2334, 1.3400], device='cuda:3'), covar=tensor([0.0317, 0.0279, 0.0388, 0.0223, 0.0193, 0.0311, 0.0200, 0.0283], device='cuda:3'), in_proj_covar=tensor([0.0084, 0.0118, 0.0144, 0.0122, 0.0108, 0.0106, 0.0089, 0.0118], device='cuda:3'), out_proj_covar=tensor([7.1441e-05, 9.9327e-05, 1.2698e-04, 1.0352e-04, 9.4660e-05, 8.6856e-05, 7.5226e-05, 9.8975e-05], device='cuda:3') 2022-12-23 00:40:42,414 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0_sp0.9 from training. Duration: 22.4666875 2022-12-23 00:40:50,308 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0 from training. Duration: 21.635 2022-12-23 00:40:54,733 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0_sp0.9 from training. Duration: 24.038875 2022-12-23 00:41:12,537 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.5878, 3.9339, 3.9495, 4.4820, 4.1395, 4.1442, 4.6668, 1.8677], device='cuda:3'), covar=tensor([0.0566, 0.0518, 0.0541, 0.0669, 0.1152, 0.0823, 0.0475, 0.3924], device='cuda:3'), in_proj_covar=tensor([0.0300, 0.0200, 0.0211, 0.0215, 0.0287, 0.0241, 0.0243, 0.0260], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 00:41:16,727 INFO [train.py:894] (3/4) Epoch 10, batch 2300, loss[loss=0.2246, simple_loss=0.2963, pruned_loss=0.07643, over 18659.00 frames. ], tot_loss[loss=0.223, simple_loss=0.2947, pruned_loss=0.07567, over 3713674.11 frames. ], batch size: 60, lr: 1.15e-02, grad_scale: 8.0 2022-12-23 00:41:21,739 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=33859.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:41:37,419 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp1.1 from training. Duration: 21.786375 2022-12-23 00:41:50,722 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0 from training. Duration: 20.22 2022-12-23 00:42:33,585 INFO [train.py:894] (3/4) Epoch 10, batch 2350, loss[loss=0.2619, simple_loss=0.3248, pruned_loss=0.09956, over 18580.00 frames. ], tot_loss[loss=0.2235, simple_loss=0.2952, pruned_loss=0.07586, over 3713791.64 frames. ], batch size: 56, lr: 1.15e-02, grad_scale: 8.0 2022-12-23 00:42:44,952 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2022-12-23 00:43:02,608 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.812e+02 5.316e+02 6.848e+02 8.119e+02 2.151e+03, threshold=1.370e+03, percent-clipped=4.0 2022-12-23 00:43:13,873 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0572, 1.4755, 1.0301, 1.5440, 2.0179, 1.5595, 1.8708, 2.1149], device='cuda:3'), covar=tensor([0.1437, 0.1996, 0.2478, 0.1495, 0.1919, 0.1559, 0.1329, 0.1389], device='cuda:3'), in_proj_covar=tensor([0.0091, 0.0101, 0.0121, 0.0097, 0.0114, 0.0090, 0.0097, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 00:43:24,606 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33939.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:43:38,059 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.92 vs. limit=5.0 2022-12-23 00:43:48,629 INFO [train.py:894] (3/4) Epoch 10, batch 2400, loss[loss=0.2017, simple_loss=0.2814, pruned_loss=0.06101, over 18630.00 frames. ], tot_loss[loss=0.2237, simple_loss=0.2954, pruned_loss=0.07599, over 3714725.67 frames. ], batch size: 53, lr: 1.15e-02, grad_scale: 8.0 2022-12-23 00:43:51,546 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0 from training. Duration: 25.285 2022-12-23 00:44:58,809 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34000.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:44:59,809 WARNING [train.py:1060] (3/4) Exclude cut with ID 811-130148-0001-63453-0_sp0.9 from training. Duration: 20.861125 2022-12-23 00:45:07,225 INFO [train.py:894] (3/4) Epoch 10, batch 2450, loss[loss=0.2097, simple_loss=0.2819, pruned_loss=0.06878, over 18569.00 frames. ], tot_loss[loss=0.2237, simple_loss=0.2956, pruned_loss=0.07594, over 3715148.95 frames. ], batch size: 49, lr: 1.15e-02, grad_scale: 8.0 2022-12-23 00:45:21,400 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0 from training. Duration: 20.88 2022-12-23 00:45:38,424 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.233e+02 5.336e+02 6.652e+02 9.061e+02 1.953e+03, threshold=1.330e+03, percent-clipped=5.0 2022-12-23 00:45:41,120 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2022-12-23 00:45:50,117 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34033.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 00:45:54,365 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0_sp0.9 from training. Duration: 23.4166875 2022-12-23 00:46:07,117 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34044.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:46:24,997 INFO [train.py:894] (3/4) Epoch 10, batch 2500, loss[loss=0.2166, simple_loss=0.2863, pruned_loss=0.07345, over 18577.00 frames. ], tot_loss[loss=0.2228, simple_loss=0.2951, pruned_loss=0.07526, over 3715842.22 frames. ], batch size: 49, lr: 1.15e-02, grad_scale: 8.0 2022-12-23 00:46:30,349 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.55 vs. limit=5.0 2022-12-23 00:46:48,101 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1594, 2.2276, 1.3596, 2.6554, 2.4031, 1.9910, 3.1537, 2.1839], device='cuda:3'), covar=tensor([0.0816, 0.1577, 0.2648, 0.1681, 0.1547, 0.0817, 0.0974, 0.1035], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0196, 0.0236, 0.0283, 0.0224, 0.0182, 0.0210, 0.0196], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 00:47:10,390 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0 from training. Duration: 21.24 2022-12-23 00:47:10,411 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0_sp0.9 from training. Duration: 23.9055625 2022-12-23 00:47:29,079 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2022-12-23 00:47:40,802 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34105.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:47:41,790 INFO [train.py:894] (3/4) Epoch 10, batch 2550, loss[loss=0.246, simple_loss=0.3158, pruned_loss=0.08816, over 18643.00 frames. ], tot_loss[loss=0.222, simple_loss=0.2946, pruned_loss=0.07474, over 3715293.55 frames. ], batch size: 69, lr: 1.15e-02, grad_scale: 8.0 2022-12-23 00:47:43,608 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp0.9 from training. Duration: 25.988875 2022-12-23 00:47:53,012 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0001-134300-0_sp0.9 from training. Duration: 20.67225 2022-12-23 00:48:10,954 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.099e+02 5.662e+02 6.760e+02 8.493e+02 2.029e+03, threshold=1.352e+03, percent-clipped=2.0 2022-12-23 00:48:15,834 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34128.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:48:39,485 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0 from training. Duration: 20.34 2022-12-23 00:48:41,222 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34145.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:48:59,220 INFO [train.py:894] (3/4) Epoch 10, batch 2600, loss[loss=0.2261, simple_loss=0.2965, pruned_loss=0.07786, over 18592.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.2942, pruned_loss=0.07449, over 3715551.94 frames. ], batch size: 51, lr: 1.15e-02, grad_scale: 8.0 2022-12-23 00:49:03,600 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2022-12-23 00:49:30,100 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=34176.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:49:36,154 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6123, 1.0361, 1.6735, 3.0917, 2.2077, 2.4040, 0.9016, 2.1322], device='cuda:3'), covar=tensor([0.1905, 0.2159, 0.1843, 0.0635, 0.1249, 0.1418, 0.2517, 0.1213], device='cuda:3'), in_proj_covar=tensor([0.0105, 0.0116, 0.0132, 0.0127, 0.0105, 0.0134, 0.0132, 0.0111], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 00:49:54,849 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp0.9 from training. Duration: 25.061125 2022-12-23 00:50:06,349 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0 from training. Duration: 0.83 2022-12-23 00:50:17,348 INFO [train.py:894] (3/4) Epoch 10, batch 2650, loss[loss=0.2142, simple_loss=0.3008, pruned_loss=0.06377, over 18673.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.2948, pruned_loss=0.07444, over 3715780.22 frames. ], batch size: 62, lr: 1.14e-02, grad_scale: 8.0 2022-12-23 00:50:17,762 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34206.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:50:34,651 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0 from training. Duration: 24.73 2022-12-23 00:50:44,952 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.713e+02 5.212e+02 6.485e+02 7.863e+02 1.303e+03, threshold=1.297e+03, percent-clipped=0.0 2022-12-23 00:50:46,264 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0 from training. Duration: 23.965 2022-12-23 00:50:54,387 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0030-146996-0_sp0.9 from training. Duration: 22.088875 2022-12-23 00:51:10,971 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0_sp0.9 from training. Duration: 23.6 2022-12-23 00:51:33,678 INFO [train.py:894] (3/4) Epoch 10, batch 2700, loss[loss=0.2185, simple_loss=0.2841, pruned_loss=0.07639, over 18554.00 frames. ], tot_loss[loss=0.2215, simple_loss=0.2943, pruned_loss=0.07431, over 3715129.66 frames. ], batch size: 44, lr: 1.14e-02, grad_scale: 8.0 2022-12-23 00:51:37,803 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2022-12-23 00:51:41,960 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5515, 1.0929, 1.5672, 2.9888, 2.0999, 2.2935, 0.7078, 2.0145], device='cuda:3'), covar=tensor([0.1952, 0.1974, 0.1810, 0.0630, 0.1182, 0.1388, 0.2575, 0.1294], device='cuda:3'), in_proj_covar=tensor([0.0105, 0.0116, 0.0132, 0.0127, 0.0105, 0.0134, 0.0131, 0.0111], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 00:52:32,380 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34295.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:52:50,117 INFO [train.py:894] (3/4) Epoch 10, batch 2750, loss[loss=0.2436, simple_loss=0.3087, pruned_loss=0.08931, over 18657.00 frames. ], tot_loss[loss=0.222, simple_loss=0.2944, pruned_loss=0.07477, over 3715024.91 frames. ], batch size: 176, lr: 1.14e-02, grad_scale: 8.0 2022-12-23 00:52:54,383 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0 from training. Duration: 23.795 2022-12-23 00:53:11,517 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp1.1 from training. Duration: 21.5409375 2022-12-23 00:53:14,422 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp0.9 from training. Duration: 24.97775 2022-12-23 00:53:17,279 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.657e+02 5.849e+02 7.144e+02 9.388e+02 2.029e+03, threshold=1.429e+03, percent-clipped=11.0 2022-12-23 00:53:21,182 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.02 vs. limit=5.0 2022-12-23 00:53:26,037 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0_sp0.9 from training. Duration: 23.3444375 2022-12-23 00:53:29,318 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34333.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:53:54,442 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0_sp0.9 from training. Duration: 23.2 2022-12-23 00:53:59,415 WARNING [train.py:1060] (3/4) Exclude cut with ID 5653-46179-0060-117930-0_sp0.9 from training. Duration: 21.17225 2022-12-23 00:54:05,531 INFO [train.py:894] (3/4) Epoch 10, batch 2800, loss[loss=0.2349, simple_loss=0.3139, pruned_loss=0.07796, over 18670.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.2946, pruned_loss=0.07455, over 3715217.05 frames. ], batch size: 78, lr: 1.14e-02, grad_scale: 8.0 2022-12-23 00:54:20,262 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp0.9 from training. Duration: 24.6555625 2022-12-23 00:54:42,538 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=34381.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:55:01,224 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1959, 2.3302, 2.0169, 2.4817, 3.1617, 1.9517, 2.4506, 1.9515], device='cuda:3'), covar=tensor([0.1726, 0.1663, 0.1537, 0.1044, 0.1260, 0.1319, 0.1607, 0.1468], device='cuda:3'), in_proj_covar=tensor([0.0240, 0.0216, 0.0205, 0.0186, 0.0257, 0.0191, 0.0214, 0.0190], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 00:55:12,427 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34400.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:55:13,623 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0 from training. Duration: 20.44 2022-12-23 00:55:20,882 INFO [train.py:894] (3/4) Epoch 10, batch 2850, loss[loss=0.2247, simple_loss=0.2979, pruned_loss=0.07572, over 18692.00 frames. ], tot_loss[loss=0.2235, simple_loss=0.2955, pruned_loss=0.07569, over 3715888.91 frames. ], batch size: 97, lr: 1.14e-02, grad_scale: 8.0 2022-12-23 00:55:29,919 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0_sp0.9 from training. Duration: 23.45 2022-12-23 00:55:49,870 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.394e+02 5.367e+02 6.476e+02 7.793e+02 1.701e+03, threshold=1.295e+03, percent-clipped=3.0 2022-12-23 00:55:56,743 WARNING [train.py:1060] (3/4) Exclude cut with ID 6945-60535-0076-12784-0_sp0.9 from training. Duration: 20.52225 2022-12-23 00:56:04,359 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0 from training. Duration: 22.19 2022-12-23 00:56:16,143 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp1.1 from training. Duration: 25.3818125 2022-12-23 00:56:24,143 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8450, 1.8942, 1.3668, 1.8501, 2.0700, 1.6930, 2.6630, 1.9436], device='cuda:3'), covar=tensor([0.0881, 0.1467, 0.2536, 0.1723, 0.1545, 0.0868, 0.0874, 0.1065], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0194, 0.0234, 0.0279, 0.0220, 0.0179, 0.0208, 0.0192], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 00:56:32,892 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp0.9 from training. Duration: 28.0944375 2022-12-23 00:56:38,513 INFO [train.py:894] (3/4) Epoch 10, batch 2900, loss[loss=0.2322, simple_loss=0.311, pruned_loss=0.07669, over 18514.00 frames. ], tot_loss[loss=0.2234, simple_loss=0.2957, pruned_loss=0.07556, over 3715291.24 frames. ], batch size: 52, lr: 1.14e-02, grad_scale: 8.0 2022-12-23 00:56:38,618 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0_sp0.9 from training. Duration: 22.9444375 2022-12-23 00:56:45,772 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp1.1 from training. Duration: 21.6318125 2022-12-23 00:57:02,554 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0 from training. Duration: 23.695 2022-12-23 00:57:07,808 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5001, 2.0320, 1.3491, 2.2248, 2.4624, 1.3787, 1.6552, 1.2292], device='cuda:3'), covar=tensor([0.2278, 0.1752, 0.1891, 0.1062, 0.1557, 0.1504, 0.2125, 0.1771], device='cuda:3'), in_proj_covar=tensor([0.0238, 0.0212, 0.0203, 0.0184, 0.0255, 0.0189, 0.0212, 0.0188], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 00:57:29,881 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0 from training. Duration: 23.955 2022-12-23 00:57:49,537 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34501.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 00:57:56,470 INFO [train.py:894] (3/4) Epoch 10, batch 2950, loss[loss=0.2479, simple_loss=0.3142, pruned_loss=0.0908, over 18591.00 frames. ], tot_loss[loss=0.2222, simple_loss=0.2948, pruned_loss=0.07484, over 3715169.11 frames. ], batch size: 51, lr: 1.14e-02, grad_scale: 8.0 2022-12-23 00:57:59,737 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.8220, 3.9523, 3.9674, 1.8499, 4.0555, 2.9865, 0.9984, 2.9134], device='cuda:3'), covar=tensor([0.1878, 0.1019, 0.1235, 0.3559, 0.0906, 0.0968, 0.5271, 0.1481], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0125, 0.0151, 0.0121, 0.0127, 0.0106, 0.0144, 0.0110], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 00:58:02,557 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp0.9 from training. Duration: 26.438875 2022-12-23 00:58:25,800 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.157e+02 5.308e+02 6.490e+02 8.090e+02 1.978e+03, threshold=1.298e+03, percent-clipped=5.0 2022-12-23 00:58:43,341 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3046, 2.6390, 1.6320, 3.0379, 2.6766, 2.1525, 3.8032, 2.4712], device='cuda:3'), covar=tensor([0.0742, 0.1462, 0.2381, 0.1673, 0.1390, 0.0798, 0.0765, 0.0967], device='cuda:3'), in_proj_covar=tensor([0.0173, 0.0196, 0.0237, 0.0280, 0.0221, 0.0180, 0.0211, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 00:58:45,969 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0021-26306-0_sp0.9 from training. Duration: 21.2444375 2022-12-23 00:58:45,995 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp0.9 from training. Duration: 31.02225 2022-12-23 00:58:56,990 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0 from training. Duration: 22.395 2022-12-23 00:59:13,108 INFO [train.py:894] (3/4) Epoch 10, batch 3000, loss[loss=0.2257, simple_loss=0.2978, pruned_loss=0.07677, over 18570.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.294, pruned_loss=0.07424, over 3715825.16 frames. ], batch size: 57, lr: 1.14e-02, grad_scale: 8.0 2022-12-23 00:59:13,108 INFO [train.py:919] (3/4) Computing validation loss 2022-12-23 00:59:16,776 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9299, 1.5993, 1.8629, 2.6232, 2.0275, 4.2479, 1.6491, 1.9445], device='cuda:3'), covar=tensor([0.0951, 0.2024, 0.1112, 0.0868, 0.1556, 0.0143, 0.1430, 0.1633], device='cuda:3'), in_proj_covar=tensor([0.0075, 0.0082, 0.0075, 0.0075, 0.0093, 0.0073, 0.0086, 0.0078], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 00:59:24,295 INFO [train.py:928] (3/4) Epoch 10, validation: loss=0.175, simple_loss=0.2744, pruned_loss=0.03777, over 944034.00 frames. 2022-12-23 00:59:24,295 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24676MB 2022-12-23 00:59:27,399 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0 from training. Duration: 21.075 2022-12-23 00:59:29,350 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.6761, 4.0390, 4.0004, 4.5925, 4.1408, 4.1132, 4.7699, 1.3069], device='cuda:3'), covar=tensor([0.0642, 0.0597, 0.0604, 0.0618, 0.1502, 0.1080, 0.0525, 0.5015], device='cuda:3'), in_proj_covar=tensor([0.0300, 0.0198, 0.0207, 0.0215, 0.0285, 0.0238, 0.0242, 0.0257], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 00:59:32,121 WARNING [train.py:1060] (3/4) Exclude cut with ID 6482-98857-0025-147532-0_sp0.9 from training. Duration: 20.0055625 2022-12-23 00:59:32,128 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0037-132304-0_sp0.9 from training. Duration: 22.05 2022-12-23 00:59:32,139 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0 from training. Duration: 26.8349375 2022-12-23 00:59:35,209 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp1.1 from training. Duration: 22.1090625 2022-12-23 00:59:43,950 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp0.9 from training. Duration: 26.6166875 2022-12-23 01:00:01,140 WARNING [train.py:1060] (3/4) Exclude cut with ID 2046-178027-0000-53705-0_sp0.9 from training. Duration: 20.3055625 2022-12-23 01:00:21,761 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34593.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:00:24,356 WARNING [train.py:1060] (3/4) Exclude cut with ID 7205-50138-0008-5373-0_sp0.9 from training. Duration: 20.7 2022-12-23 01:00:24,652 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34595.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:00:41,531 INFO [train.py:894] (3/4) Epoch 10, batch 3050, loss[loss=0.1861, simple_loss=0.2531, pruned_loss=0.05954, over 18479.00 frames. ], tot_loss[loss=0.2215, simple_loss=0.2942, pruned_loss=0.07438, over 3714404.27 frames. ], batch size: 43, lr: 1.14e-02, grad_scale: 8.0 2022-12-23 01:01:06,526 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0 from training. Duration: 22.48 2022-12-23 01:01:06,777 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34622.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:01:11,058 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.837e+02 5.772e+02 6.600e+02 8.445e+02 2.116e+03, threshold=1.320e+03, percent-clipped=3.0 2022-12-23 01:01:12,886 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34626.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:01:21,243 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp0.9 from training. Duration: 29.816625 2022-12-23 01:01:29,702 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6430, 1.4099, 1.1992, 1.9395, 1.6066, 3.5753, 1.3586, 1.5684], device='cuda:3'), covar=tensor([0.0910, 0.1817, 0.1278, 0.0983, 0.1457, 0.0225, 0.1378, 0.1525], device='cuda:3'), in_proj_covar=tensor([0.0075, 0.0082, 0.0076, 0.0075, 0.0093, 0.0072, 0.0086, 0.0078], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 01:01:34,357 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.1137, 1.0025, 1.1680, 0.4979, 0.5658, 1.2560, 1.2003, 1.1614], device='cuda:3'), covar=tensor([0.0651, 0.0292, 0.0298, 0.0337, 0.0425, 0.0447, 0.0233, 0.0574], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0153, 0.0108, 0.0129, 0.0139, 0.0125, 0.0142, 0.0147], device='cuda:3'), out_proj_covar=tensor([1.1364e-04, 1.2862e-04, 8.9398e-05, 1.0493e-04, 1.1539e-04, 1.0558e-04, 1.2137e-04, 1.2279e-04], device='cuda:3') 2022-12-23 01:01:39,199 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=34643.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:01:42,102 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp1.1 from training. Duration: 22.7590625 2022-12-23 01:01:47,855 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0 from training. Duration: 22.555 2022-12-23 01:01:56,460 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34654.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:01:59,070 INFO [train.py:894] (3/4) Epoch 10, batch 3100, loss[loss=0.2187, simple_loss=0.2932, pruned_loss=0.07212, over 18591.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.2941, pruned_loss=0.07438, over 3713883.85 frames. ], batch size: 78, lr: 1.14e-02, grad_scale: 8.0 2022-12-23 01:02:07,576 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0005-25975-0_sp0.9 from training. Duration: 21.688875 2022-12-23 01:02:41,049 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34683.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 01:02:43,684 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0_sp0.9 from training. Duration: 22.6 2022-12-23 01:02:47,384 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34687.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:02:47,754 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.85 vs. limit=2.0 2022-12-23 01:03:08,244 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34700.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:03:17,606 INFO [train.py:894] (3/4) Epoch 10, batch 3150, loss[loss=0.2346, simple_loss=0.3072, pruned_loss=0.08097, over 18620.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.2943, pruned_loss=0.07427, over 3714184.38 frames. ], batch size: 53, lr: 1.14e-02, grad_scale: 8.0 2022-12-23 01:03:23,745 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9553, 1.2636, 2.5959, 4.1015, 3.0245, 2.5970, 0.7915, 2.8837], device='cuda:3'), covar=tensor([0.2002, 0.2123, 0.1538, 0.0506, 0.1078, 0.1385, 0.2690, 0.1158], device='cuda:3'), in_proj_covar=tensor([0.0103, 0.0116, 0.0130, 0.0128, 0.0104, 0.0133, 0.0131, 0.0111], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 01:03:23,800 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34710.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 01:03:24,936 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0 from training. Duration: 24.32 2022-12-23 01:03:46,377 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.556e+02 5.245e+02 6.711e+02 8.370e+02 2.399e+03, threshold=1.342e+03, percent-clipped=5.0 2022-12-23 01:04:03,077 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6160, 2.2263, 1.7322, 0.7669, 1.7552, 2.1298, 1.8578, 2.0081], device='cuda:3'), covar=tensor([0.0535, 0.0420, 0.1058, 0.1419, 0.1009, 0.1272, 0.1277, 0.0601], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0181, 0.0201, 0.0196, 0.0205, 0.0191, 0.0204, 0.0191], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 01:04:21,927 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-276745-0093-13116-0_sp0.9 from training. Duration: 21.061125 2022-12-23 01:04:22,032 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=34748.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:04:24,297 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2022-12-23 01:04:34,358 INFO [train.py:894] (3/4) Epoch 10, batch 3200, loss[loss=0.2234, simple_loss=0.2853, pruned_loss=0.08073, over 18693.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.2939, pruned_loss=0.07389, over 3714594.31 frames. ], batch size: 46, lr: 1.14e-02, grad_scale: 8.0 2022-12-23 01:04:35,841 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0024-15855-0_sp0.9 from training. Duration: 20.32225 2022-12-23 01:04:50,760 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp1.1 from training. Duration: 0.7545625 2022-12-23 01:04:53,919 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3713, 0.9370, 1.2833, 1.2635, 1.6141, 1.5025, 1.4966, 1.0804], device='cuda:3'), covar=tensor([0.0297, 0.0244, 0.0454, 0.0214, 0.0173, 0.0308, 0.0266, 0.0263], device='cuda:3'), in_proj_covar=tensor([0.0086, 0.0118, 0.0145, 0.0126, 0.0109, 0.0110, 0.0092, 0.0120], device='cuda:3'), out_proj_covar=tensor([7.3413e-05, 9.9027e-05, 1.2737e-04, 1.0681e-04, 9.5150e-05, 9.0046e-05, 7.7515e-05, 1.0053e-04], device='cuda:3') 2022-12-23 01:04:56,733 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34771.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 01:05:05,756 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0_sp0.9 from training. Duration: 23.9333125 2022-12-23 01:05:25,930 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5934, 3.6554, 3.5995, 1.5848, 3.7704, 2.7744, 0.7716, 2.5739], device='cuda:3'), covar=tensor([0.1919, 0.0988, 0.1354, 0.3532, 0.0824, 0.0979, 0.4914, 0.1465], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0124, 0.0152, 0.0120, 0.0127, 0.0106, 0.0141, 0.0111], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 01:05:36,268 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2022-12-23 01:05:40,024 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp1.1 from training. Duration: 20.17275 2022-12-23 01:05:43,450 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34801.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:05:46,269 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp1.1 from training. Duration: 20.436375 2022-12-23 01:05:50,480 INFO [train.py:894] (3/4) Epoch 10, batch 3250, loss[loss=0.2501, simple_loss=0.3126, pruned_loss=0.09378, over 18693.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.2934, pruned_loss=0.07357, over 3715814.41 frames. ], batch size: 50, lr: 1.14e-02, grad_scale: 8.0 2022-12-23 01:06:03,242 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.22 vs. limit=5.0 2022-12-23 01:06:19,412 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.549e+02 5.831e+02 6.841e+02 8.207e+02 1.292e+03, threshold=1.368e+03, percent-clipped=0.0 2022-12-23 01:06:56,430 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=34849.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:07:06,009 INFO [train.py:894] (3/4) Epoch 10, batch 3300, loss[loss=0.2253, simple_loss=0.3085, pruned_loss=0.07107, over 18694.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.2934, pruned_loss=0.07349, over 3715865.31 frames. ], batch size: 62, lr: 1.13e-02, grad_scale: 8.0 2022-12-23 01:07:07,652 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0_sp0.9 from training. Duration: 23.1055625 2022-12-23 01:07:09,475 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp1.1 from training. Duration: 23.5 2022-12-23 01:07:21,464 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp0.9 from training. Duration: 26.62775 2022-12-23 01:07:35,168 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0 from training. Duration: 21.105 2022-12-23 01:07:39,473 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0_sp0.9 from training. Duration: 24.411125 2022-12-23 01:07:54,277 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2861, 1.8922, 1.4748, 0.6153, 1.5568, 1.8997, 1.5416, 1.8864], device='cuda:3'), covar=tensor([0.0544, 0.0449, 0.0922, 0.1389, 0.0929, 0.1159, 0.1273, 0.0516], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0181, 0.0200, 0.0193, 0.0205, 0.0188, 0.0203, 0.0190], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 01:08:08,207 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp1.1 from training. Duration: 21.263625 2022-12-23 01:08:23,709 INFO [train.py:894] (3/4) Epoch 10, batch 3350, loss[loss=0.1975, simple_loss=0.262, pruned_loss=0.06647, over 18555.00 frames. ], tot_loss[loss=0.2198, simple_loss=0.2933, pruned_loss=0.07313, over 3715321.84 frames. ], batch size: 44, lr: 1.13e-02, grad_scale: 8.0 2022-12-23 01:08:35,903 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8577, 1.1577, 0.7386, 1.1821, 1.9487, 1.2216, 1.6301, 2.0083], device='cuda:3'), covar=tensor([0.1543, 0.2235, 0.2577, 0.1633, 0.1885, 0.1747, 0.1416, 0.1430], device='cuda:3'), in_proj_covar=tensor([0.0091, 0.0099, 0.0118, 0.0095, 0.0112, 0.0089, 0.0096, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 01:08:41,563 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0 from training. Duration: 20.795 2022-12-23 01:08:51,656 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.438e+02 5.551e+02 6.581e+02 7.829e+02 1.593e+03, threshold=1.316e+03, percent-clipped=1.0 2022-12-23 01:08:51,717 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0 from training. Duration: 24.76 2022-12-23 01:08:52,870 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0_sp0.9 from training. Duration: 22.25 2022-12-23 01:09:17,917 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp1.1 from training. Duration: 20.5045625 2022-12-23 01:09:28,721 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34949.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:09:38,885 INFO [train.py:894] (3/4) Epoch 10, batch 3400, loss[loss=0.208, simple_loss=0.2895, pruned_loss=0.06323, over 18562.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.2931, pruned_loss=0.07317, over 3714508.49 frames. ], batch size: 98, lr: 1.13e-02, grad_scale: 8.0 2022-12-23 01:09:57,436 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5436, 1.9394, 1.3931, 2.3037, 2.7417, 1.5776, 1.5795, 1.2387], device='cuda:3'), covar=tensor([0.2035, 0.1690, 0.1689, 0.0973, 0.1201, 0.1273, 0.1878, 0.1681], device='cuda:3'), in_proj_covar=tensor([0.0234, 0.0211, 0.0200, 0.0184, 0.0252, 0.0188, 0.0210, 0.0186], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 01:10:07,236 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3880, 2.0077, 1.3791, 2.2699, 2.7116, 1.4895, 1.5788, 1.1636], device='cuda:3'), covar=tensor([0.2103, 0.1651, 0.1724, 0.0982, 0.1244, 0.1315, 0.1876, 0.1716], device='cuda:3'), in_proj_covar=tensor([0.0234, 0.0211, 0.0200, 0.0185, 0.0252, 0.0188, 0.0210, 0.0186], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 01:10:11,158 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34978.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 01:10:16,851 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34982.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:10:36,326 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2716, 2.8269, 2.7192, 1.1838, 2.9062, 2.1363, 0.7614, 1.8439], device='cuda:3'), covar=tensor([0.2111, 0.1355, 0.1788, 0.4103, 0.1130, 0.1239, 0.4704, 0.1819], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0126, 0.0154, 0.0123, 0.0130, 0.0107, 0.0144, 0.0113], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 01:10:52,543 INFO [train.py:894] (3/4) Epoch 10, batch 3450, loss[loss=0.1964, simple_loss=0.2836, pruned_loss=0.0546, over 18719.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.2935, pruned_loss=0.07384, over 3714549.64 frames. ], batch size: 52, lr: 1.13e-02, grad_scale: 8.0 2022-12-23 01:11:20,658 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.418e+02 5.044e+02 6.339e+02 8.246e+02 2.038e+03, threshold=1.268e+03, percent-clipped=3.0 2022-12-23 01:11:29,399 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5889, 1.3993, 1.2841, 0.8279, 1.8275, 1.5843, 1.4994, 1.2167], device='cuda:3'), covar=tensor([0.0384, 0.0445, 0.0488, 0.0671, 0.0274, 0.0352, 0.0426, 0.0847], device='cuda:3'), in_proj_covar=tensor([0.0122, 0.0118, 0.0128, 0.0119, 0.0087, 0.0119, 0.0137, 0.0152], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 01:12:06,958 INFO [train.py:894] (3/4) Epoch 10, batch 3500, loss[loss=0.2551, simple_loss=0.3194, pruned_loss=0.09543, over 18549.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.2946, pruned_loss=0.07436, over 3714035.97 frames. ], batch size: 175, lr: 1.13e-02, grad_scale: 8.0 2022-12-23 01:12:31,010 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp1.1 from training. Duration: 22.2954375 2022-12-23 01:12:41,279 INFO [train.py:894] (3/4) Epoch 11, batch 0, loss[loss=0.2087, simple_loss=0.2952, pruned_loss=0.06108, over 18726.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2952, pruned_loss=0.06108, over 18726.00 frames. ], batch size: 52, lr: 1.08e-02, grad_scale: 8.0 2022-12-23 01:12:41,280 INFO [train.py:919] (3/4) Computing validation loss 2022-12-23 01:12:51,364 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5310, 3.4679, 3.2166, 1.3755, 3.3557, 2.4126, 0.6966, 2.3433], device='cuda:3'), covar=tensor([0.1829, 0.0763, 0.1432, 0.4469, 0.0976, 0.1243, 0.5203, 0.1697], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0125, 0.0152, 0.0121, 0.0129, 0.0105, 0.0142, 0.0111], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 01:12:52,252 INFO [train.py:928] (3/4) Epoch 11, validation: loss=0.1739, simple_loss=0.2739, pruned_loss=0.03697, over 944034.00 frames. 2022-12-23 01:12:52,252 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24676MB 2022-12-23 01:12:56,513 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35066.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 01:13:24,734 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2022-12-23 01:13:25,475 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7747, 1.7911, 1.7202, 1.7739, 1.4030, 4.0153, 1.9165, 2.3253], device='cuda:3'), covar=tensor([0.3264, 0.1948, 0.1972, 0.1955, 0.1390, 0.0132, 0.1392, 0.0810], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0119, 0.0129, 0.0120, 0.0103, 0.0102, 0.0099, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 01:13:35,566 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35092.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:13:37,006 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.8780, 3.5963, 2.9085, 1.3224, 2.5456, 2.7479, 2.5389, 2.5080], device='cuda:3'), covar=tensor([0.0484, 0.0376, 0.1207, 0.1676, 0.1594, 0.1071, 0.1141, 0.0948], device='cuda:3'), in_proj_covar=tensor([0.0164, 0.0181, 0.0200, 0.0192, 0.0205, 0.0188, 0.0201, 0.0190], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 01:13:44,301 WARNING [train.py:1060] (3/4) Exclude cut with ID 298-126791-0067-24026-0_sp0.9 from training. Duration: 21.438875 2022-12-23 01:13:48,866 WARNING [train.py:1060] (3/4) Exclude cut with ID 5652-39938-0025-23684-0_sp0.9 from training. Duration: 22.2055625 2022-12-23 01:14:06,705 INFO [train.py:894] (3/4) Epoch 11, batch 50, loss[loss=0.1834, simple_loss=0.2753, pruned_loss=0.04573, over 18632.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2866, pruned_loss=0.06155, over 839369.31 frames. ], batch size: 53, lr: 1.08e-02, grad_scale: 8.0 2022-12-23 01:14:17,362 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1486, 2.1129, 1.3639, 2.5018, 2.3740, 1.9167, 3.2805, 2.1067], device='cuda:3'), covar=tensor([0.0783, 0.1691, 0.2709, 0.1818, 0.1516, 0.0850, 0.0732, 0.1100], device='cuda:3'), in_proj_covar=tensor([0.0173, 0.0195, 0.0236, 0.0279, 0.0222, 0.0181, 0.0205, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 01:14:25,781 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.877e+02 4.401e+02 5.411e+02 7.454e+02 1.490e+03, threshold=1.082e+03, percent-clipped=5.0 2022-12-23 01:14:44,505 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5211, 1.2507, 1.1605, 1.8281, 1.4485, 3.1782, 1.1581, 1.1957], device='cuda:3'), covar=tensor([0.1114, 0.2517, 0.1405, 0.1114, 0.1875, 0.0263, 0.1942, 0.2195], device='cuda:3'), in_proj_covar=tensor([0.0074, 0.0083, 0.0075, 0.0075, 0.0092, 0.0072, 0.0086, 0.0078], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 01:14:44,867 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2022-12-23 01:14:59,158 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.71 vs. limit=5.0 2022-12-23 01:15:07,629 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35153.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:15:21,344 INFO [train.py:894] (3/4) Epoch 11, batch 100, loss[loss=0.2273, simple_loss=0.3101, pruned_loss=0.07224, over 18715.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2855, pruned_loss=0.06066, over 1477128.57 frames. ], batch size: 60, lr: 1.08e-02, grad_scale: 8.0 2022-12-23 01:16:30,918 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8201, 1.5788, 1.6290, 1.3144, 1.8319, 1.9934, 1.8849, 1.2803], device='cuda:3'), covar=tensor([0.0279, 0.0237, 0.0400, 0.0253, 0.0192, 0.0319, 0.0286, 0.0294], device='cuda:3'), in_proj_covar=tensor([0.0086, 0.0117, 0.0146, 0.0126, 0.0111, 0.0110, 0.0094, 0.0119], device='cuda:3'), out_proj_covar=tensor([7.3162e-05, 9.8195e-05, 1.2806e-04, 1.0645e-04, 9.6798e-05, 8.9827e-05, 7.8955e-05, 9.8938e-05], device='cuda:3') 2022-12-23 01:16:33,981 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3783, 1.4318, 1.3544, 1.5066, 1.1056, 3.2886, 1.5059, 1.9410], device='cuda:3'), covar=tensor([0.4793, 0.3027, 0.2591, 0.2632, 0.1575, 0.0285, 0.1622, 0.1008], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0117, 0.0127, 0.0119, 0.0102, 0.0100, 0.0097, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 01:16:38,020 INFO [train.py:894] (3/4) Epoch 11, batch 150, loss[loss=0.237, simple_loss=0.316, pruned_loss=0.07894, over 18471.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2818, pruned_loss=0.05833, over 1973779.08 frames. ], batch size: 54, lr: 1.08e-02, grad_scale: 8.0 2022-12-23 01:16:53,225 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0 from training. Duration: 24.525 2022-12-23 01:16:57,295 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.549e+02 4.088e+02 5.112e+02 6.380e+02 1.930e+03, threshold=1.022e+03, percent-clipped=3.0 2022-12-23 01:17:26,034 WARNING [train.py:1060] (3/4) Exclude cut with ID 3699-47246-0007-3408-0_sp0.9 from training. Duration: 20.26675 2022-12-23 01:17:33,892 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35249.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:17:39,298 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp0.9 from training. Duration: 27.25 2022-12-23 01:17:54,750 INFO [train.py:894] (3/4) Epoch 11, batch 200, loss[loss=0.1808, simple_loss=0.254, pruned_loss=0.05379, over 18582.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2801, pruned_loss=0.05735, over 2360140.82 frames. ], batch size: 45, lr: 1.08e-02, grad_scale: 8.0 2022-12-23 01:17:58,063 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4559, 3.0201, 2.9690, 1.5549, 2.8111, 2.7533, 1.9382, 3.8082], device='cuda:3'), covar=tensor([0.1235, 0.1385, 0.1247, 0.2099, 0.0805, 0.1269, 0.2142, 0.0449], device='cuda:3'), in_proj_covar=tensor([0.0199, 0.0198, 0.0203, 0.0192, 0.0180, 0.0211, 0.0210, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 01:18:09,693 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3105, 2.6676, 3.0918, 0.7612, 2.5359, 3.3060, 2.4044, 2.8584], device='cuda:3'), covar=tensor([0.0737, 0.0374, 0.0303, 0.0564, 0.0376, 0.0271, 0.0320, 0.0504], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0157, 0.0109, 0.0132, 0.0141, 0.0126, 0.0144, 0.0146], device='cuda:3'), out_proj_covar=tensor([1.1285e-04, 1.3203e-04, 8.9251e-05, 1.0686e-04, 1.1671e-04, 1.0561e-04, 1.2264e-04, 1.2158e-04], device='cuda:3') 2022-12-23 01:18:16,510 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35278.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 01:18:22,917 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35282.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:18:45,701 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=35297.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:18:53,671 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0 from training. Duration: 21.68 2022-12-23 01:19:04,897 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0 from training. Duration: 21.6300625 2022-12-23 01:19:10,608 INFO [train.py:894] (3/4) Epoch 11, batch 250, loss[loss=0.1783, simple_loss=0.2503, pruned_loss=0.05314, over 18524.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2812, pruned_loss=0.05789, over 2660883.28 frames. ], batch size: 44, lr: 1.08e-02, grad_scale: 8.0 2022-12-23 01:19:29,239 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.669e+02 4.290e+02 5.310e+02 7.018e+02 1.757e+03, threshold=1.062e+03, percent-clipped=3.0 2022-12-23 01:19:29,347 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0_sp0.9 from training. Duration: 24.033375 2022-12-23 01:19:30,835 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=35326.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:19:37,102 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=35330.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:19:59,268 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35345.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 01:20:05,745 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.8949, 2.4822, 2.1605, 0.9060, 2.0436, 2.2044, 1.8984, 2.1753], device='cuda:3'), covar=tensor([0.0575, 0.0470, 0.1084, 0.1696, 0.1272, 0.1234, 0.1371, 0.0798], device='cuda:3'), in_proj_covar=tensor([0.0160, 0.0177, 0.0195, 0.0190, 0.0202, 0.0186, 0.0199, 0.0187], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 01:20:26,272 INFO [train.py:894] (3/4) Epoch 11, batch 300, loss[loss=0.1719, simple_loss=0.2554, pruned_loss=0.04417, over 18535.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2793, pruned_loss=0.05688, over 2893089.94 frames. ], batch size: 44, lr: 1.08e-02, grad_scale: 8.0 2022-12-23 01:20:31,282 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0 from training. Duration: 22.905 2022-12-23 01:20:31,578 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35366.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 01:20:32,881 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp1.1 from training. Duration: 23.4318125 2022-12-23 01:20:43,071 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35374.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:21:30,435 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35406.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 01:21:41,817 INFO [train.py:894] (3/4) Epoch 11, batch 350, loss[loss=0.252, simple_loss=0.3292, pruned_loss=0.08743, over 18516.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2811, pruned_loss=0.05727, over 3075364.16 frames. ], batch size: 98, lr: 1.07e-02, grad_scale: 16.0 2022-12-23 01:21:42,236 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6121, 1.4821, 1.7172, 1.7759, 1.2525, 3.7917, 1.6789, 2.2613], device='cuda:3'), covar=tensor([0.3354, 0.2159, 0.1944, 0.1915, 0.1421, 0.0156, 0.1573, 0.0851], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0119, 0.0129, 0.0119, 0.0104, 0.0101, 0.0098, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 01:21:43,369 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=35414.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 01:22:00,005 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.294e+02 3.714e+02 4.522e+02 5.692e+02 1.135e+03, threshold=9.043e+02, percent-clipped=2.0 2022-12-23 01:22:15,608 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35435.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:22:26,091 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp1.1 from training. Duration: 20.82275 2022-12-23 01:22:26,135 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp0.9 from training. Duration: 25.45 2022-12-23 01:22:34,078 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35448.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:22:56,855 INFO [train.py:894] (3/4) Epoch 11, batch 400, loss[loss=0.1948, simple_loss=0.2776, pruned_loss=0.05606, over 18714.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2822, pruned_loss=0.05837, over 3216605.78 frames. ], batch size: 50, lr: 1.07e-02, grad_scale: 8.0 2022-12-23 01:23:25,337 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0 from training. Duration: 25.775 2022-12-23 01:23:27,450 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4838, 1.8812, 2.1589, 2.1088, 2.4619, 2.2735, 2.3757, 1.8233], device='cuda:3'), covar=tensor([0.1608, 0.2607, 0.1864, 0.2247, 0.1416, 0.0687, 0.2376, 0.0877], device='cuda:3'), in_proj_covar=tensor([0.0251, 0.0284, 0.0256, 0.0292, 0.0275, 0.0236, 0.0302, 0.0222], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 01:23:48,180 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0_sp0.9 from training. Duration: 22.25 2022-12-23 01:24:10,525 INFO [train.py:894] (3/4) Epoch 11, batch 450, loss[loss=0.196, simple_loss=0.2856, pruned_loss=0.05316, over 18697.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2826, pruned_loss=0.05872, over 3327640.51 frames. ], batch size: 50, lr: 1.07e-02, grad_scale: 8.0 2022-12-23 01:24:15,172 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0 from training. Duration: 26.205 2022-12-23 01:24:29,645 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.813e+02 4.201e+02 5.115e+02 6.796e+02 1.250e+03, threshold=1.023e+03, percent-clipped=10.0 2022-12-23 01:24:31,142 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp0.9 from training. Duration: 30.1555625 2022-12-23 01:24:36,002 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35530.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:24:37,070 WARNING [train.py:1060] (3/4) Exclude cut with ID 1265-135635-0050-6781-0_sp0.9 from training. Duration: 21.8333125 2022-12-23 01:24:41,681 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.4275, 3.7855, 3.7847, 4.3216, 4.0034, 4.0152, 4.5559, 1.4228], device='cuda:3'), covar=tensor([0.0622, 0.0568, 0.0537, 0.0644, 0.1182, 0.0889, 0.0468, 0.4435], device='cuda:3'), in_proj_covar=tensor([0.0293, 0.0193, 0.0199, 0.0208, 0.0274, 0.0230, 0.0238, 0.0250], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 01:24:47,336 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp1.1 from training. Duration: 20.6545625 2022-12-23 01:25:25,208 INFO [train.py:894] (3/4) Epoch 11, batch 500, loss[loss=0.2131, simple_loss=0.2999, pruned_loss=0.06318, over 18475.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2844, pruned_loss=0.06004, over 3413352.70 frames. ], batch size: 54, lr: 1.07e-02, grad_scale: 8.0 2022-12-23 01:25:28,175 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0045-39920-0_sp0.9 from training. Duration: 20.52225 2022-12-23 01:25:49,190 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp0.9 from training. Duration: 29.1166875 2022-12-23 01:26:07,871 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35591.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:26:40,523 INFO [train.py:894] (3/4) Epoch 11, batch 550, loss[loss=0.1637, simple_loss=0.2436, pruned_loss=0.04187, over 18608.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.285, pruned_loss=0.06039, over 3480834.14 frames. ], batch size: 45, lr: 1.07e-02, grad_scale: 8.0 2022-12-23 01:26:47,628 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133211-0007-59831-0_sp0.9 from training. Duration: 21.388875 2022-12-23 01:27:00,169 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.945e+02 4.347e+02 5.387e+02 6.927e+02 1.155e+03, threshold=1.077e+03, percent-clipped=5.0 2022-12-23 01:27:26,096 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0 from training. Duration: 22.72 2022-12-23 01:27:27,546 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0_sp0.9 from training. Duration: 22.7444375 2022-12-23 01:27:29,527 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6414, 1.6053, 1.2003, 1.5191, 1.7195, 1.5188, 2.1014, 1.6745], device='cuda:3'), covar=tensor([0.0889, 0.1539, 0.2611, 0.1742, 0.1608, 0.0873, 0.0952, 0.1222], device='cuda:3'), in_proj_covar=tensor([0.0169, 0.0192, 0.0233, 0.0275, 0.0216, 0.0178, 0.0201, 0.0191], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 01:27:56,864 INFO [train.py:894] (3/4) Epoch 11, batch 600, loss[loss=0.198, simple_loss=0.2832, pruned_loss=0.05644, over 18705.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2854, pruned_loss=0.06012, over 3532351.90 frames. ], batch size: 50, lr: 1.07e-02, grad_scale: 8.0 2022-12-23 01:27:58,777 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5475, 3.5513, 3.4666, 1.5048, 3.5051, 2.7103, 0.5776, 2.2992], device='cuda:3'), covar=tensor([0.2048, 0.1018, 0.1597, 0.3887, 0.1057, 0.1120, 0.5642, 0.1684], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0121, 0.0149, 0.0121, 0.0126, 0.0103, 0.0141, 0.0108], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 01:28:12,210 WARNING [train.py:1060] (3/4) Exclude cut with ID 4133-6541-0027-40495-0_sp1.1 from training. Duration: 0.9681875 2022-12-23 01:28:15,388 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0_sp0.9 from training. Duration: 22.3166875 2022-12-23 01:28:21,075 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133212-0015-59917-0_sp0.9 from training. Duration: 21.8166875 2022-12-23 01:28:49,998 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.2999, 2.3651, 1.8007, 3.2066, 2.3308, 2.2339, 2.4637, 3.6428], device='cuda:3'), covar=tensor([0.1489, 0.2633, 0.1582, 0.2424, 0.3119, 0.0930, 0.2682, 0.0514], device='cuda:3'), in_proj_covar=tensor([0.0272, 0.0263, 0.0223, 0.0337, 0.0247, 0.0208, 0.0263, 0.0186], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 01:28:51,107 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8487, 1.3278, 1.7789, 2.1966, 1.5084, 4.2004, 1.1477, 1.4497], device='cuda:3'), covar=tensor([0.1149, 0.2874, 0.1424, 0.1207, 0.2087, 0.0242, 0.2205, 0.2472], device='cuda:3'), in_proj_covar=tensor([0.0075, 0.0084, 0.0077, 0.0076, 0.0093, 0.0073, 0.0088, 0.0079], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 01:28:55,306 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35701.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 01:29:13,353 INFO [train.py:894] (3/4) Epoch 11, batch 650, loss[loss=0.1702, simple_loss=0.2475, pruned_loss=0.04647, over 18606.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2851, pruned_loss=0.05997, over 3572622.15 frames. ], batch size: 45, lr: 1.07e-02, grad_scale: 8.0 2022-12-23 01:29:19,640 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3989, 2.0165, 1.4572, 2.3137, 1.8043, 1.8452, 1.8528, 2.4294], device='cuda:3'), covar=tensor([0.1727, 0.2649, 0.1686, 0.2391, 0.2966, 0.0952, 0.2457, 0.0700], device='cuda:3'), in_proj_covar=tensor([0.0274, 0.0265, 0.0223, 0.0338, 0.0249, 0.0209, 0.0264, 0.0187], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 01:29:32,666 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.598e+02 4.065e+02 4.860e+02 5.934e+02 1.772e+03, threshold=9.721e+02, percent-clipped=1.0 2022-12-23 01:29:39,345 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35730.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:30:03,881 WARNING [train.py:1060] (3/4) Exclude cut with ID 4957-30119-0041-23990-0_sp0.9 from training. Duration: 20.22775 2022-12-23 01:30:06,533 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35748.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:30:28,940 INFO [train.py:894] (3/4) Epoch 11, batch 700, loss[loss=0.2073, simple_loss=0.2922, pruned_loss=0.0612, over 18625.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2855, pruned_loss=0.0601, over 3603965.55 frames. ], batch size: 53, lr: 1.07e-02, grad_scale: 8.0 2022-12-23 01:30:29,860 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2022-12-23 01:30:41,290 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35771.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:30:49,645 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp1.1 from training. Duration: 24.67275 2022-12-23 01:31:15,383 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1579, 2.0029, 2.2387, 1.2447, 2.4685, 2.3589, 1.4850, 2.7306], device='cuda:3'), covar=tensor([0.1094, 0.1517, 0.1124, 0.1877, 0.0603, 0.1097, 0.2122, 0.0420], device='cuda:3'), in_proj_covar=tensor([0.0196, 0.0197, 0.0197, 0.0189, 0.0175, 0.0208, 0.0206, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 01:31:17,946 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=35796.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:31:19,261 WARNING [train.py:1060] (3/4) Exclude cut with ID 3082-165428-0081-50734-0_sp0.9 from training. Duration: 21.8055625 2022-12-23 01:31:42,758 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35811.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:31:45,338 INFO [train.py:894] (3/4) Epoch 11, batch 750, loss[loss=0.2013, simple_loss=0.2823, pruned_loss=0.06017, over 18531.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2847, pruned_loss=0.05999, over 3627975.82 frames. ], batch size: 47, lr: 1.07e-02, grad_scale: 8.0 2022-12-23 01:31:47,225 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.3729, 1.7077, 1.7974, 0.7337, 1.0032, 1.9985, 1.7960, 1.6172], device='cuda:3'), covar=tensor([0.0684, 0.0264, 0.0289, 0.0406, 0.0415, 0.0374, 0.0235, 0.0519], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0157, 0.0109, 0.0132, 0.0141, 0.0127, 0.0145, 0.0146], device='cuda:3'), out_proj_covar=tensor([1.1397e-04, 1.3134e-04, 8.9004e-05, 1.0670e-04, 1.1629e-04, 1.0656e-04, 1.2255e-04, 1.2130e-04], device='cuda:3') 2022-12-23 01:31:54,638 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0_sp0.9 from training. Duration: 22.6666875 2022-12-23 01:32:05,505 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.908e+02 4.276e+02 5.128e+02 6.060e+02 1.074e+03, threshold=1.026e+03, percent-clipped=3.0 2022-12-23 01:32:14,951 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35832.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:32:30,607 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0289, 1.2865, 1.6602, 1.7810, 2.0642, 1.9068, 1.8750, 1.3983], device='cuda:3'), covar=tensor([0.1687, 0.2805, 0.2009, 0.2158, 0.1472, 0.0784, 0.2272, 0.1045], device='cuda:3'), in_proj_covar=tensor([0.0250, 0.0285, 0.0256, 0.0293, 0.0277, 0.0235, 0.0301, 0.0222], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 01:32:57,780 WARNING [train.py:1060] (3/4) Exclude cut with ID 2411-132532-0017-82279-0_sp1.1 from training. Duration: 0.9681875 2022-12-23 01:33:02,621 INFO [train.py:894] (3/4) Epoch 11, batch 800, loss[loss=0.2048, simple_loss=0.2932, pruned_loss=0.05821, over 18468.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2854, pruned_loss=0.06047, over 3647707.24 frames. ], batch size: 50, lr: 1.07e-02, grad_scale: 8.0 2022-12-23 01:33:16,446 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35872.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:33:23,182 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0 from training. Duration: 22.485 2022-12-23 01:33:37,226 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35886.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:34:01,374 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp1.1 from training. Duration: 23.82275 2022-12-23 01:34:03,108 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4799, 1.9506, 1.3223, 2.3458, 2.5094, 1.4208, 1.4449, 1.2175], device='cuda:3'), covar=tensor([0.1801, 0.1581, 0.1530, 0.0836, 0.1260, 0.1177, 0.1972, 0.1469], device='cuda:3'), in_proj_covar=tensor([0.0235, 0.0213, 0.0202, 0.0187, 0.0252, 0.0189, 0.0213, 0.0189], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 01:34:14,547 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0 from training. Duration: 20.77 2022-12-23 01:34:14,972 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.1302, 1.3968, 1.6552, 0.6618, 0.8713, 1.7583, 1.6723, 1.5421], device='cuda:3'), covar=tensor([0.0587, 0.0265, 0.0287, 0.0309, 0.0364, 0.0446, 0.0194, 0.0488], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0158, 0.0110, 0.0130, 0.0140, 0.0127, 0.0144, 0.0146], device='cuda:3'), out_proj_covar=tensor([1.1314e-04, 1.3161e-04, 8.9259e-05, 1.0547e-04, 1.1568e-04, 1.0610e-04, 1.2183e-04, 1.2074e-04], device='cuda:3') 2022-12-23 01:34:17,531 INFO [train.py:894] (3/4) Epoch 11, batch 850, loss[loss=0.2053, simple_loss=0.2941, pruned_loss=0.05824, over 18552.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2867, pruned_loss=0.06119, over 3663338.06 frames. ], batch size: 58, lr: 1.07e-02, grad_scale: 8.0 2022-12-23 01:34:23,664 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0_sp0.9 from training. Duration: 24.088875 2022-12-23 01:34:36,371 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.023e+02 4.512e+02 5.363e+02 6.717e+02 1.218e+03, threshold=1.073e+03, percent-clipped=5.0 2022-12-23 01:34:53,701 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp1.1 from training. Duration: 20.4409375 2022-12-23 01:35:27,089 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2022-12-23 01:35:34,149 INFO [train.py:894] (3/4) Epoch 11, batch 900, loss[loss=0.1805, simple_loss=0.2657, pruned_loss=0.04758, over 18529.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2856, pruned_loss=0.06032, over 3675242.47 frames. ], batch size: 47, lr: 1.07e-02, grad_scale: 8.0 2022-12-23 01:36:08,702 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0_sp0.9 from training. Duration: 22.511125 2022-12-23 01:36:10,120 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0 from training. Duration: 20.675 2022-12-23 01:36:16,294 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2022-12-23 01:36:36,755 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36001.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 01:36:54,000 INFO [train.py:894] (3/4) Epoch 11, batch 950, loss[loss=0.1885, simple_loss=0.2762, pruned_loss=0.05038, over 18525.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2865, pruned_loss=0.06045, over 3682886.74 frames. ], batch size: 47, lr: 1.07e-02, grad_scale: 8.0 2022-12-23 01:37:11,940 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.706e+02 4.318e+02 5.198e+02 6.420e+02 1.014e+03, threshold=1.040e+03, percent-clipped=0.0 2022-12-23 01:37:12,396 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0141, 1.5779, 1.9933, 1.1673, 2.0308, 1.9950, 1.3621, 2.4207], device='cuda:3'), covar=tensor([0.0860, 0.1796, 0.1200, 0.1736, 0.0661, 0.1161, 0.2168, 0.0443], device='cuda:3'), in_proj_covar=tensor([0.0197, 0.0199, 0.0200, 0.0192, 0.0178, 0.0212, 0.0208, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 01:37:18,486 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36030.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:37:47,171 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=36049.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 01:37:49,972 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp0.9 from training. Duration: 24.9833125 2022-12-23 01:38:08,627 INFO [train.py:894] (3/4) Epoch 11, batch 1000, loss[loss=0.2531, simple_loss=0.3258, pruned_loss=0.09016, over 18649.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2869, pruned_loss=0.0606, over 3688255.92 frames. ], batch size: 62, lr: 1.06e-02, grad_scale: 8.0 2022-12-23 01:38:21,458 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0 from training. Duration: 27.14 2022-12-23 01:38:31,037 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=36078.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:38:38,910 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0 from training. Duration: 22.44 2022-12-23 01:39:24,161 INFO [train.py:894] (3/4) Epoch 11, batch 1050, loss[loss=0.1835, simple_loss=0.2693, pruned_loss=0.04887, over 18680.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2862, pruned_loss=0.05982, over 3694710.08 frames. ], batch size: 48, lr: 1.06e-02, grad_scale: 8.0 2022-12-23 01:39:42,566 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.372e+02 4.570e+02 5.343e+02 6.692e+02 1.425e+03, threshold=1.069e+03, percent-clipped=5.0 2022-12-23 01:39:44,838 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36127.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:39:58,699 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0060-62364-0_sp0.9 from training. Duration: 21.361125 2022-12-23 01:40:05,617 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp1.1 from training. Duration: 27.0318125 2022-12-23 01:40:07,874 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.85 vs. limit=5.0 2022-12-23 01:40:14,419 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp0.9 from training. Duration: 28.638875 2022-12-23 01:40:21,305 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3970, 1.0074, 1.6882, 2.8459, 2.1019, 2.2756, 0.5976, 1.9022], device='cuda:3'), covar=tensor([0.2034, 0.1874, 0.1617, 0.0584, 0.1051, 0.1304, 0.2578, 0.1224], device='cuda:3'), in_proj_covar=tensor([0.0104, 0.0115, 0.0129, 0.0126, 0.0102, 0.0133, 0.0131, 0.0109], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 01:40:32,423 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0 from training. Duration: 20.4 2022-12-23 01:40:39,901 INFO [train.py:894] (3/4) Epoch 11, batch 1100, loss[loss=0.181, simple_loss=0.2633, pruned_loss=0.04937, over 18674.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2855, pruned_loss=0.05937, over 3698819.09 frames. ], batch size: 46, lr: 1.06e-02, grad_scale: 8.0 2022-12-23 01:40:45,879 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36167.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:40:58,926 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2022-12-23 01:41:04,472 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0 from training. Duration: 20.025 2022-12-23 01:41:04,482 WARNING [train.py:1060] (3/4) Exclude cut with ID 2364-131735-0112-64612-0_sp0.9 from training. Duration: 20.488875 2022-12-23 01:41:08,447 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36181.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:41:11,103 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0 from training. Duration: 29.735 2022-12-23 01:41:15,795 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36186.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:41:55,937 INFO [train.py:894] (3/4) Epoch 11, batch 1150, loss[loss=0.2064, simple_loss=0.2829, pruned_loss=0.06498, over 18606.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2858, pruned_loss=0.05966, over 3701544.29 frames. ], batch size: 51, lr: 1.06e-02, grad_scale: 8.0 2022-12-23 01:42:15,997 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.182e+02 4.452e+02 5.516e+02 6.913e+02 1.165e+03, threshold=1.103e+03, percent-clipped=1.0 2022-12-23 01:42:29,113 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=36234.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:42:34,052 WARNING [train.py:1060] (3/4) Exclude cut with ID 7276-92427-0014-12983-0_sp0.9 from training. Duration: 21.3055625 2022-12-23 01:42:35,456 WARNING [train.py:1060] (3/4) Exclude cut with ID 1025-75365-0008-79168-0_sp0.9 from training. Duration: 22.0666875 2022-12-23 01:42:41,744 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36242.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:43:13,122 INFO [train.py:894] (3/4) Epoch 11, batch 1200, loss[loss=0.1972, simple_loss=0.2846, pruned_loss=0.05493, over 18614.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.285, pruned_loss=0.05917, over 3704640.55 frames. ], batch size: 53, lr: 1.06e-02, grad_scale: 8.0 2022-12-23 01:43:57,559 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.7632, 3.2123, 3.1092, 3.6995, 3.3087, 3.3037, 3.9246, 1.1675], device='cuda:3'), covar=tensor([0.0884, 0.0708, 0.0816, 0.0812, 0.1673, 0.1194, 0.0668, 0.4987], device='cuda:3'), in_proj_covar=tensor([0.0298, 0.0195, 0.0202, 0.0209, 0.0278, 0.0230, 0.0240, 0.0255], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 01:44:07,184 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0608, 1.6951, 2.0275, 1.4790, 1.9866, 2.0095, 1.5426, 2.2251], device='cuda:3'), covar=tensor([0.0778, 0.1355, 0.1202, 0.1397, 0.0554, 0.0866, 0.1508, 0.0484], device='cuda:3'), in_proj_covar=tensor([0.0193, 0.0196, 0.0197, 0.0187, 0.0175, 0.0207, 0.0203, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 01:44:25,720 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0 from training. Duration: 20.26 2022-12-23 01:44:28,605 INFO [train.py:894] (3/4) Epoch 11, batch 1250, loss[loss=0.2244, simple_loss=0.305, pruned_loss=0.07188, over 18593.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2858, pruned_loss=0.05932, over 3706899.02 frames. ], batch size: 56, lr: 1.06e-02, grad_scale: 8.0 2022-12-23 01:44:38,836 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0030-9324-0_sp0.9 from training. Duration: 21.3444375 2022-12-23 01:44:48,463 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.493e+02 3.840e+02 4.722e+02 6.009e+02 1.141e+03, threshold=9.444e+02, percent-clipped=3.0 2022-12-23 01:45:34,605 WARNING [train.py:1060] (3/4) Exclude cut with ID 497-129325-0061-62254-0_sp1.1 from training. Duration: 0.97725 2022-12-23 01:45:44,712 INFO [train.py:894] (3/4) Epoch 11, batch 1300, loss[loss=0.1859, simple_loss=0.2709, pruned_loss=0.05048, over 18669.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2855, pruned_loss=0.05942, over 3709141.89 frames. ], batch size: 48, lr: 1.06e-02, grad_scale: 8.0 2022-12-23 01:46:08,958 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0988, 1.8389, 1.6386, 1.2010, 2.3607, 2.0673, 1.7228, 1.5592], device='cuda:3'), covar=tensor([0.0312, 0.0372, 0.0480, 0.0671, 0.0222, 0.0332, 0.0459, 0.0789], device='cuda:3'), in_proj_covar=tensor([0.0120, 0.0119, 0.0129, 0.0122, 0.0088, 0.0120, 0.0137, 0.0151], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 01:46:17,844 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0_sp0.9 from training. Duration: 22.97225 2022-12-23 01:46:43,508 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2022-12-23 01:46:50,848 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0047-39922-0_sp0.9 from training. Duration: 21.97775 2022-12-23 01:46:53,275 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.71 vs. limit=5.0 2022-12-23 01:46:58,423 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3369, 1.0815, 1.4194, 2.1721, 1.5507, 2.3877, 0.7560, 1.5739], device='cuda:3'), covar=tensor([0.1790, 0.1734, 0.1292, 0.0647, 0.1179, 0.0773, 0.1939, 0.1214], device='cuda:3'), in_proj_covar=tensor([0.0102, 0.0113, 0.0129, 0.0126, 0.0102, 0.0134, 0.0130, 0.0109], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 01:46:59,620 INFO [train.py:894] (3/4) Epoch 11, batch 1350, loss[loss=0.1883, simple_loss=0.2669, pruned_loss=0.05485, over 18681.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2856, pruned_loss=0.05914, over 3710338.17 frames. ], batch size: 46, lr: 1.06e-02, grad_scale: 8.0 2022-12-23 01:47:05,227 WARNING [train.py:1060] (3/4) Exclude cut with ID 1112-1043-0006-89194-0_sp0.9 from training. Duration: 21.8333125 2022-12-23 01:47:14,679 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0 from training. Duration: 20.47 2022-12-23 01:47:19,945 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.513e+02 4.062e+02 4.949e+02 6.479e+02 1.390e+03, threshold=9.899e+02, percent-clipped=5.0 2022-12-23 01:47:21,834 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36427.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:47:40,193 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.3553, 3.6761, 3.6151, 4.2590, 3.9326, 3.8529, 4.5293, 1.0994], device='cuda:3'), covar=tensor([0.0717, 0.0693, 0.0625, 0.0644, 0.1448, 0.1277, 0.0569, 0.5093], device='cuda:3'), in_proj_covar=tensor([0.0297, 0.0194, 0.0203, 0.0211, 0.0276, 0.0231, 0.0240, 0.0255], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 01:47:48,420 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9770, 1.2115, 1.6743, 1.6730, 1.9787, 1.9458, 1.8089, 1.4590], device='cuda:3'), covar=tensor([0.1683, 0.2639, 0.2036, 0.2185, 0.1476, 0.0740, 0.2294, 0.0977], device='cuda:3'), in_proj_covar=tensor([0.0249, 0.0286, 0.0258, 0.0296, 0.0278, 0.0235, 0.0302, 0.0222], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 01:48:15,441 INFO [train.py:894] (3/4) Epoch 11, batch 1400, loss[loss=0.1806, simple_loss=0.2623, pruned_loss=0.04948, over 18541.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2842, pruned_loss=0.05834, over 3711377.04 frames. ], batch size: 41, lr: 1.06e-02, grad_scale: 8.0 2022-12-23 01:48:21,481 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36467.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:48:23,925 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0037-39912-0_sp0.9 from training. Duration: 20.67225 2022-12-23 01:48:34,532 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=36475.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:48:43,055 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp0.9 from training. Duration: 25.2444375 2022-12-23 01:49:05,140 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0021-76797-0_sp0.9 from training. Duration: 21.1445 2022-12-23 01:49:33,106 INFO [train.py:894] (3/4) Epoch 11, batch 1450, loss[loss=0.2058, simple_loss=0.2832, pruned_loss=0.06423, over 18449.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2848, pruned_loss=0.05842, over 3711361.00 frames. ], batch size: 50, lr: 1.06e-02, grad_scale: 8.0 2022-12-23 01:49:36,317 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=36515.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:49:52,528 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.495e+02 4.266e+02 5.326e+02 7.004e+02 1.890e+03, threshold=1.065e+03, percent-clipped=3.0 2022-12-23 01:50:10,301 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36537.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:50:17,708 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp0.9 from training. Duration: 33.038875 2022-12-23 01:50:49,210 INFO [train.py:894] (3/4) Epoch 11, batch 1500, loss[loss=0.1834, simple_loss=0.2674, pruned_loss=0.04969, over 18444.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2851, pruned_loss=0.0586, over 3712493.94 frames. ], batch size: 48, lr: 1.06e-02, grad_scale: 8.0 2022-12-23 01:50:54,318 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64291-0000-16059-0_sp0.9 from training. Duration: 20.0944375 2022-12-23 01:51:09,650 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp1.1 from training. Duration: 20.4 2022-12-23 01:51:17,518 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0 from training. Duration: 20.085 2022-12-23 01:51:29,577 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0_sp0.9 from training. Duration: 23.07775 2022-12-23 01:51:52,501 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0715, 1.3757, 1.6814, 1.7481, 2.0418, 1.9281, 1.9107, 1.5089], device='cuda:3'), covar=tensor([0.1771, 0.2676, 0.2129, 0.2395, 0.1553, 0.0789, 0.2435, 0.1012], device='cuda:3'), in_proj_covar=tensor([0.0252, 0.0289, 0.0261, 0.0299, 0.0282, 0.0238, 0.0307, 0.0224], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 01:51:53,748 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4221, 1.3619, 1.0184, 1.6013, 1.6489, 2.9230, 1.2609, 1.5358], device='cuda:3'), covar=tensor([0.0903, 0.1699, 0.1239, 0.0911, 0.1346, 0.0249, 0.1350, 0.1465], device='cuda:3'), in_proj_covar=tensor([0.0074, 0.0083, 0.0076, 0.0076, 0.0092, 0.0072, 0.0086, 0.0078], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 01:51:53,802 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36605.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:51:58,281 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36608.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:52:04,750 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36612.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:52:05,724 INFO [train.py:894] (3/4) Epoch 11, batch 1550, loss[loss=0.1904, simple_loss=0.279, pruned_loss=0.05089, over 18566.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2849, pruned_loss=0.05839, over 3711694.99 frames. ], batch size: 57, lr: 1.06e-02, grad_scale: 8.0 2022-12-23 01:52:18,022 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp0.9 from training. Duration: 24.9333125 2022-12-23 01:52:25,466 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.783e+02 4.206e+02 5.136e+02 5.999e+02 2.048e+03, threshold=1.027e+03, percent-clipped=3.0 2022-12-23 01:53:01,350 WARNING [train.py:1060] (3/4) Exclude cut with ID 5118-111612-0016-124680-0_sp0.9 from training. Duration: 20.388875 2022-12-23 01:53:04,705 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.68 vs. limit=5.0 2022-12-23 01:53:07,040 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp1.1 from training. Duration: 20.3590625 2022-12-23 01:53:22,342 INFO [train.py:894] (3/4) Epoch 11, batch 1600, loss[loss=0.1933, simple_loss=0.2641, pruned_loss=0.06126, over 18588.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2851, pruned_loss=0.05878, over 3712553.85 frames. ], batch size: 45, lr: 1.06e-02, grad_scale: 8.0 2022-12-23 01:53:27,137 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36666.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:53:31,425 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36669.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:53:36,873 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36673.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:53:48,590 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2022-12-23 01:54:09,512 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9365, 1.7046, 2.2448, 1.2564, 2.2692, 2.2980, 1.3264, 2.6546], device='cuda:3'), covar=tensor([0.1166, 0.1731, 0.1249, 0.1928, 0.0749, 0.1119, 0.2273, 0.0447], device='cuda:3'), in_proj_covar=tensor([0.0201, 0.0203, 0.0204, 0.0195, 0.0182, 0.0215, 0.0210, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 01:54:15,017 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0_sp1.1 from training. Duration: 0.836375 2022-12-23 01:54:16,697 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36700.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 01:54:31,258 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5474, 2.1207, 1.4707, 2.6713, 1.9051, 1.9583, 2.0216, 2.7674], device='cuda:3'), covar=tensor([0.1678, 0.2710, 0.1625, 0.2442, 0.3094, 0.0944, 0.2490, 0.0632], device='cuda:3'), in_proj_covar=tensor([0.0276, 0.0270, 0.0225, 0.0339, 0.0252, 0.0211, 0.0265, 0.0189], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 01:54:36,541 INFO [train.py:894] (3/4) Epoch 11, batch 1650, loss[loss=0.2023, simple_loss=0.2862, pruned_loss=0.05918, over 18498.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.286, pruned_loss=0.05952, over 3711984.99 frames. ], batch size: 52, lr: 1.06e-02, grad_scale: 8.0 2022-12-23 01:54:55,465 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.822e+02 4.941e+02 6.129e+02 7.618e+02 1.418e+03, threshold=1.226e+03, percent-clipped=7.0 2022-12-23 01:54:59,912 WARNING [train.py:1060] (3/4) Exclude cut with ID 8565-290391-0049-67394-0_sp0.9 from training. Duration: 21.3166875 2022-12-23 01:55:29,921 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0029-104863-0_sp0.9 from training. Duration: 22.1055625 2022-12-23 01:55:42,774 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp1.1 from training. Duration: 21.77725 2022-12-23 01:55:49,088 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36761.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 01:55:51,584 INFO [train.py:894] (3/4) Epoch 11, batch 1700, loss[loss=0.2608, simple_loss=0.3248, pruned_loss=0.09836, over 18466.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2873, pruned_loss=0.06153, over 3712639.83 frames. ], batch size: 54, lr: 1.05e-02, grad_scale: 8.0 2022-12-23 01:55:59,820 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp0.9 from training. Duration: 27.8166875 2022-12-23 01:56:26,867 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp1.1 from training. Duration: 22.5090625 2022-12-23 01:56:32,920 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0 from training. Duration: 25.035 2022-12-23 01:56:50,776 WARNING [train.py:1060] (3/4) Exclude cut with ID 774-127930-0014-10412-0_sp1.1 from training. Duration: 0.95 2022-12-23 01:56:58,221 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36805.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 01:57:09,355 INFO [train.py:894] (3/4) Epoch 11, batch 1750, loss[loss=0.255, simple_loss=0.3231, pruned_loss=0.09342, over 18719.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2891, pruned_loss=0.06443, over 3712776.52 frames. ], batch size: 78, lr: 1.05e-02, grad_scale: 8.0 2022-12-23 01:57:09,394 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp0.9 from training. Duration: 0.92225 2022-12-23 01:57:30,109 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.720e+02 5.303e+02 6.412e+02 8.124e+02 1.465e+03, threshold=1.282e+03, percent-clipped=3.0 2022-12-23 01:57:37,414 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0 from training. Duration: 21.97 2022-12-23 01:57:46,829 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36837.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:57:57,092 WARNING [train.py:1060] (3/4) Exclude cut with ID 7492-105653-0055-62765-0_sp0.9 from training. Duration: 21.97225 2022-12-23 01:57:58,474 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp0.9 from training. Duration: 25.3333125 2022-12-23 01:58:10,038 WARNING [train.py:1060] (3/4) Exclude cut with ID 5172-29468-0015-19128-0_sp0.9 from training. Duration: 21.5055625 2022-12-23 01:58:21,090 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp1.1 from training. Duration: 20.72725 2022-12-23 01:58:25,594 INFO [train.py:894] (3/4) Epoch 11, batch 1800, loss[loss=0.2415, simple_loss=0.3074, pruned_loss=0.08783, over 18693.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2899, pruned_loss=0.06655, over 3713147.49 frames. ], batch size: 50, lr: 1.05e-02, grad_scale: 8.0 2022-12-23 01:58:30,318 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36866.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 01:58:55,792 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp0.9 from training. Duration: 26.32775 2022-12-23 01:58:58,749 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=36885.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:59:27,404 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0 from training. Duration: 20.025 2022-12-23 01:59:33,137 WARNING [train.py:1060] (3/4) Exclude cut with ID 6709-74022-0004-86860-0_sp1.1 from training. Duration: 0.9409375 2022-12-23 01:59:33,149 WARNING [train.py:1060] (3/4) Exclude cut with ID 4757-1811-0023-62229-0_sp0.9 from training. Duration: 21.37775 2022-12-23 01:59:39,371 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36911.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 01:59:42,056 INFO [train.py:894] (3/4) Epoch 11, batch 1850, loss[loss=0.1986, simple_loss=0.2801, pruned_loss=0.05858, over 18559.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2909, pruned_loss=0.06873, over 3713078.34 frames. ], batch size: 69, lr: 1.05e-02, grad_scale: 8.0 2022-12-23 01:59:55,870 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0004-25974-0_sp0.9 from training. Duration: 21.17225 2022-12-23 01:59:55,881 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp0.9 from training. Duration: 27.511125 2022-12-23 02:00:02,483 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.556e+02 5.200e+02 6.324e+02 7.983e+02 2.314e+03, threshold=1.265e+03, percent-clipped=3.0 2022-12-23 02:00:29,926 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0 from training. Duration: 22.8 2022-12-23 02:00:35,061 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0 from training. Duration: 22.585 2022-12-23 02:00:52,212 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36958.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:00:55,144 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.1667, 1.4644, 1.6289, 0.7477, 0.8918, 1.7373, 1.6725, 1.5256], device='cuda:3'), covar=tensor([0.0577, 0.0263, 0.0259, 0.0293, 0.0358, 0.0361, 0.0186, 0.0427], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0158, 0.0111, 0.0128, 0.0136, 0.0127, 0.0142, 0.0146], device='cuda:3'), out_proj_covar=tensor([1.1149e-04, 1.3134e-04, 9.0526e-05, 1.0289e-04, 1.1091e-04, 1.0635e-04, 1.1913e-04, 1.2069e-04], device='cuda:3') 2022-12-23 02:00:56,524 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36961.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:00:59,183 INFO [train.py:894] (3/4) Epoch 11, batch 1900, loss[loss=0.1593, simple_loss=0.2376, pruned_loss=0.04046, over 18496.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.2916, pruned_loss=0.07029, over 3713442.10 frames. ], batch size: 43, lr: 1.05e-02, grad_scale: 8.0 2022-12-23 02:01:00,836 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36964.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:01:06,885 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0001-146967-0_sp0.9 from training. Duration: 22.0166875 2022-12-23 02:01:07,048 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36968.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:01:13,697 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36972.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:01:24,096 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp1.1 from training. Duration: 24.395375 2022-12-23 02:01:31,348 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp0.9 from training. Duration: 27.47775 2022-12-23 02:01:35,843 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp0.9 from training. Duration: 24.8833125 2022-12-23 02:01:38,783 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0 from training. Duration: 23.39 2022-12-23 02:01:44,718 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp0.9 from training. Duration: 28.72225 2022-12-23 02:01:52,311 WARNING [train.py:1060] (3/4) Exclude cut with ID 585-294811-0110-133686-0_sp0.9 from training. Duration: 20.8944375 2022-12-23 02:02:02,458 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37004.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:02:09,328 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0_sp0.9 from training. Duration: 23.8444375 2022-12-23 02:02:15,056 INFO [train.py:894] (3/4) Epoch 11, batch 1950, loss[loss=0.2298, simple_loss=0.296, pruned_loss=0.08186, over 18693.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.2922, pruned_loss=0.07123, over 3713318.51 frames. ], batch size: 50, lr: 1.05e-02, grad_scale: 8.0 2022-12-23 02:02:24,859 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37019.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:02:29,995 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6145, 1.0994, 1.7468, 3.0843, 2.2736, 2.4630, 1.0967, 2.0845], device='cuda:3'), covar=tensor([0.1954, 0.1889, 0.1638, 0.0634, 0.1188, 0.1289, 0.2220, 0.1264], device='cuda:3'), in_proj_covar=tensor([0.0102, 0.0114, 0.0128, 0.0126, 0.0103, 0.0130, 0.0130, 0.0108], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 02:02:34,230 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0 from training. Duration: 25.85 2022-12-23 02:02:35,549 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.431e+02 5.428e+02 6.678e+02 8.110e+02 1.320e+03, threshold=1.336e+03, percent-clipped=2.0 2022-12-23 02:02:35,605 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0 from training. Duration: 21.39 2022-12-23 02:02:43,511 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0 from training. Duration: 27.92 2022-12-23 02:02:57,932 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2022-12-23 02:03:12,191 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0039-130165-0_sp0.9 from training. Duration: 20.661125 2022-12-23 02:03:21,137 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37056.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 02:03:31,610 INFO [train.py:894] (3/4) Epoch 11, batch 2000, loss[loss=0.2468, simple_loss=0.3165, pruned_loss=0.08854, over 18688.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.2937, pruned_loss=0.07301, over 3713266.70 frames. ], batch size: 99, lr: 1.05e-02, grad_scale: 8.0 2022-12-23 02:03:35,021 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37065.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:03:36,118 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0043-15874-0_sp0.9 from training. Duration: 20.07225 2022-12-23 02:03:42,577 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0 from training. Duration: 21.01 2022-12-23 02:04:47,947 INFO [train.py:894] (3/4) Epoch 11, batch 2050, loss[loss=0.2059, simple_loss=0.2728, pruned_loss=0.06947, over 18671.00 frames. ], tot_loss[loss=0.2198, simple_loss=0.2935, pruned_loss=0.07312, over 3714045.19 frames. ], batch size: 46, lr: 1.05e-02, grad_scale: 8.0 2022-12-23 02:04:50,192 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0 from training. Duration: 20.65 2022-12-23 02:04:55,718 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2022-12-23 02:04:56,558 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0 from training. Duration: 21.46 2022-12-23 02:04:59,873 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3384, 1.3434, 1.0845, 1.5512, 1.5911, 2.8953, 1.3103, 1.4392], device='cuda:3'), covar=tensor([0.1057, 0.1938, 0.1207, 0.0965, 0.1521, 0.0292, 0.1511, 0.1679], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0083, 0.0075, 0.0075, 0.0092, 0.0072, 0.0086, 0.0078], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 02:05:08,584 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.252e+02 5.274e+02 6.467e+02 7.854e+02 2.150e+03, threshold=1.293e+03, percent-clipped=5.0 2022-12-23 02:05:44,419 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0 from training. Duration: 0.92 2022-12-23 02:05:50,692 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0_sp0.9 from training. Duration: 23.7666875 2022-12-23 02:05:51,684 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2022-12-23 02:06:03,468 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37161.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 02:06:06,099 INFO [train.py:894] (3/4) Epoch 11, batch 2100, loss[loss=0.1906, simple_loss=0.2607, pruned_loss=0.06025, over 18537.00 frames. ], tot_loss[loss=0.2198, simple_loss=0.2934, pruned_loss=0.07315, over 3714014.84 frames. ], batch size: 47, lr: 1.05e-02, grad_scale: 8.0 2022-12-23 02:06:30,121 WARNING [train.py:1060] (3/4) Exclude cut with ID 8544-281189-0060-101339-0_sp0.9 from training. Duration: 20.861125 2022-12-23 02:06:40,479 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0_sp0.9 from training. Duration: 22.711125 2022-12-23 02:07:22,255 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp1.1 from training. Duration: 22.986375 2022-12-23 02:07:23,579 INFO [train.py:894] (3/4) Epoch 11, batch 2150, loss[loss=0.2325, simple_loss=0.2949, pruned_loss=0.08509, over 18628.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.2938, pruned_loss=0.07401, over 3713662.91 frames. ], batch size: 98, lr: 1.05e-02, grad_scale: 8.0 2022-12-23 02:07:38,284 WARNING [train.py:1060] (3/4) Exclude cut with ID 8040-260924-0003-80960-0_sp0.9 from training. Duration: 22.07225 2022-12-23 02:07:43,050 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.116e+02 5.173e+02 6.018e+02 7.074e+02 1.637e+03, threshold=1.204e+03, percent-clipped=3.0 2022-12-23 02:07:43,145 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0045-26330-0_sp0.9 from training. Duration: 20.3055625 2022-12-23 02:07:46,792 WARNING [train.py:1060] (3/4) Exclude cut with ID 6356-271890-0060-94317-0_sp0.9 from training. Duration: 20.72225 2022-12-23 02:08:05,139 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp1.1 from training. Duration: 22.4818125 2022-12-23 02:08:33,365 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp0.9 from training. Duration: 25.0944375 2022-12-23 02:08:36,380 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0 from training. Duration: 21.515 2022-12-23 02:08:36,593 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37261.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:08:39,102 INFO [train.py:894] (3/4) Epoch 11, batch 2200, loss[loss=0.2471, simple_loss=0.3157, pruned_loss=0.08928, over 18695.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.2918, pruned_loss=0.07295, over 3713129.19 frames. ], batch size: 62, lr: 1.05e-02, grad_scale: 8.0 2022-12-23 02:08:40,868 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37264.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:08:42,420 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp0.9 from training. Duration: 27.02225 2022-12-23 02:08:45,158 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37267.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:08:46,440 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0010-62480-0_sp0.9 from training. Duration: 22.22225 2022-12-23 02:08:46,774 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37268.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:08:46,849 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37268.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:08:52,592 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0085-44554-0_sp0.9 from training. Duration: 20.85 2022-12-23 02:08:54,469 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9420, 1.3495, 1.5175, 1.7646, 1.9523, 1.8927, 1.8769, 1.3537], device='cuda:3'), covar=tensor([0.2275, 0.3149, 0.2692, 0.2552, 0.2008, 0.1042, 0.2817, 0.1373], device='cuda:3'), in_proj_covar=tensor([0.0252, 0.0285, 0.0261, 0.0292, 0.0278, 0.0237, 0.0302, 0.0222], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 02:09:26,714 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0 from training. Duration: 21.54 2022-12-23 02:09:32,167 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp1.1 from training. Duration: 20.5318125 2022-12-23 02:09:41,236 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0012-134311-0_sp0.9 from training. Duration: 21.9333125 2022-12-23 02:09:48,547 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=37309.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:09:53,209 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=37312.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:09:54,469 INFO [train.py:894] (3/4) Epoch 11, batch 2250, loss[loss=0.1714, simple_loss=0.2472, pruned_loss=0.04779, over 18594.00 frames. ], tot_loss[loss=0.2184, simple_loss=0.2914, pruned_loss=0.07271, over 3713011.69 frames. ], batch size: 41, lr: 1.05e-02, grad_scale: 8.0 2022-12-23 02:09:56,456 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37314.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:09:57,988 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.8255, 5.1186, 4.6599, 1.9560, 5.0790, 3.8104, 0.7588, 3.2386], device='cuda:3'), covar=tensor([0.1932, 0.0726, 0.1180, 0.3610, 0.0651, 0.0854, 0.5771, 0.1461], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0124, 0.0154, 0.0121, 0.0129, 0.0105, 0.0145, 0.0112], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 02:09:59,319 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=37316.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:10:15,915 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.192e+02 5.513e+02 6.965e+02 8.801e+02 2.114e+03, threshold=1.393e+03, percent-clipped=10.0 2022-12-23 02:10:20,676 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37329.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:10:27,382 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0025-130151-0_sp0.9 from training. Duration: 21.7944375 2022-12-23 02:10:32,934 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.06 vs. limit=5.0 2022-12-23 02:10:38,308 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0_sp0.9 from training. Duration: 22.4666875 2022-12-23 02:10:45,102 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0 from training. Duration: 21.635 2022-12-23 02:10:51,324 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0_sp0.9 from training. Duration: 24.038875 2022-12-23 02:11:00,732 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37356.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 02:11:06,298 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37360.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:11:10,405 INFO [train.py:894] (3/4) Epoch 11, batch 2300, loss[loss=0.2523, simple_loss=0.3268, pruned_loss=0.08888, over 18578.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.2927, pruned_loss=0.07339, over 3713203.14 frames. ], batch size: 57, lr: 1.05e-02, grad_scale: 8.0 2022-12-23 02:11:12,145 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37364.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:11:35,150 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp1.1 from training. Duration: 21.786375 2022-12-23 02:11:48,674 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0 from training. Duration: 20.22 2022-12-23 02:11:59,499 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2022-12-23 02:12:12,853 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=37404.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 02:12:16,040 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.2834, 2.8250, 3.1345, 0.9688, 2.4893, 3.4575, 2.2996, 2.8808], device='cuda:3'), covar=tensor([0.0684, 0.0292, 0.0296, 0.0442, 0.0363, 0.0264, 0.0319, 0.0472], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0159, 0.0115, 0.0132, 0.0138, 0.0129, 0.0144, 0.0148], device='cuda:3'), out_proj_covar=tensor([1.1296e-04, 1.3158e-04, 9.4191e-05, 1.0595e-04, 1.1217e-04, 1.0804e-04, 1.2083e-04, 1.2147e-04], device='cuda:3') 2022-12-23 02:12:25,845 INFO [train.py:894] (3/4) Epoch 11, batch 2350, loss[loss=0.1795, simple_loss=0.2492, pruned_loss=0.05495, over 18620.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.2926, pruned_loss=0.07353, over 3714217.87 frames. ], batch size: 45, lr: 1.05e-02, grad_scale: 8.0 2022-12-23 02:12:45,502 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37425.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:12:46,710 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.259e+02 5.576e+02 6.263e+02 7.534e+02 2.266e+03, threshold=1.253e+03, percent-clipped=1.0 2022-12-23 02:13:35,250 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4855, 1.9248, 2.0707, 2.1717, 2.3944, 2.2049, 2.3297, 1.6651], device='cuda:3'), covar=tensor([0.1658, 0.2531, 0.1958, 0.2219, 0.1375, 0.0735, 0.2414, 0.1011], device='cuda:3'), in_proj_covar=tensor([0.0253, 0.0287, 0.0261, 0.0294, 0.0277, 0.0236, 0.0304, 0.0223], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 02:13:41,992 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37461.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 02:13:44,321 INFO [train.py:894] (3/4) Epoch 11, batch 2400, loss[loss=0.2392, simple_loss=0.3047, pruned_loss=0.08689, over 18576.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.292, pruned_loss=0.07312, over 3715075.76 frames. ], batch size: 57, lr: 1.05e-02, grad_scale: 16.0 2022-12-23 02:13:48,944 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0 from training. Duration: 25.285 2022-12-23 02:13:52,260 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7682, 1.3251, 1.2118, 1.1943, 1.7673, 1.8435, 1.9750, 1.1498], device='cuda:3'), covar=tensor([0.0296, 0.0278, 0.0561, 0.0270, 0.0228, 0.0322, 0.0286, 0.0336], device='cuda:3'), in_proj_covar=tensor([0.0087, 0.0115, 0.0144, 0.0122, 0.0113, 0.0112, 0.0093, 0.0121], device='cuda:3'), out_proj_covar=tensor([7.2929e-05, 9.5538e-05, 1.2551e-04, 1.0245e-04, 9.6765e-05, 9.0751e-05, 7.6399e-05, 1.0049e-04], device='cuda:3') 2022-12-23 02:14:52,569 WARNING [train.py:1060] (3/4) Exclude cut with ID 811-130148-0001-63453-0_sp0.9 from training. Duration: 20.861125 2022-12-23 02:14:55,317 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=37509.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 02:15:00,716 INFO [train.py:894] (3/4) Epoch 11, batch 2450, loss[loss=0.2049, simple_loss=0.2752, pruned_loss=0.06729, over 18430.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.2917, pruned_loss=0.07242, over 3713258.60 frames. ], batch size: 48, lr: 1.04e-02, grad_scale: 16.0 2022-12-23 02:15:13,260 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0 from training. Duration: 20.88 2022-12-23 02:15:20,403 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.193e+02 5.351e+02 6.733e+02 8.810e+02 1.600e+03, threshold=1.347e+03, percent-clipped=5.0 2022-12-23 02:15:44,389 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0_sp0.9 from training. Duration: 23.4166875 2022-12-23 02:16:03,849 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2022-12-23 02:16:16,708 INFO [train.py:894] (3/4) Epoch 11, batch 2500, loss[loss=0.2185, simple_loss=0.2957, pruned_loss=0.07066, over 18580.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.2916, pruned_loss=0.0725, over 3713044.47 frames. ], batch size: 57, lr: 1.04e-02, grad_scale: 16.0 2022-12-23 02:16:23,542 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37567.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:17:02,926 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0 from training. Duration: 21.24 2022-12-23 02:17:02,941 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0_sp0.9 from training. Duration: 23.9055625 2022-12-23 02:17:18,718 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.91 vs. limit=5.0 2022-12-23 02:17:21,289 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.5205, 1.6487, 2.0468, 0.8864, 1.3973, 2.3085, 1.8770, 1.7136], device='cuda:3'), covar=tensor([0.0630, 0.0394, 0.0366, 0.0394, 0.0385, 0.0327, 0.0263, 0.0559], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0158, 0.0114, 0.0132, 0.0139, 0.0129, 0.0145, 0.0148], device='cuda:3'), out_proj_covar=tensor([1.1207e-04, 1.3049e-04, 9.2997e-05, 1.0583e-04, 1.1220e-04, 1.0757e-04, 1.2107e-04, 1.2175e-04], device='cuda:3') 2022-12-23 02:17:34,329 INFO [train.py:894] (3/4) Epoch 11, batch 2550, loss[loss=0.2112, simple_loss=0.3014, pruned_loss=0.06054, over 18564.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.2921, pruned_loss=0.07214, over 3713481.22 frames. ], batch size: 56, lr: 1.04e-02, grad_scale: 16.0 2022-12-23 02:17:35,954 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37614.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:17:37,406 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=37615.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:17:38,911 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp0.9 from training. Duration: 25.988875 2022-12-23 02:17:48,138 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0001-134300-0_sp0.9 from training. Duration: 20.67225 2022-12-23 02:17:51,426 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37624.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:17:54,168 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.571e+02 5.424e+02 6.401e+02 7.705e+02 1.573e+03, threshold=1.280e+03, percent-clipped=3.0 2022-12-23 02:18:35,905 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0 from training. Duration: 20.34 2022-12-23 02:18:47,641 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37660.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:18:50,117 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=37662.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:18:51,376 INFO [train.py:894] (3/4) Epoch 11, batch 2600, loss[loss=0.2541, simple_loss=0.3254, pruned_loss=0.09137, over 18692.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.2921, pruned_loss=0.07262, over 3713198.15 frames. ], batch size: 78, lr: 1.04e-02, grad_scale: 16.0 2022-12-23 02:19:35,066 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2022-12-23 02:19:50,105 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp0.9 from training. Duration: 25.061125 2022-12-23 02:20:01,312 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0 from training. Duration: 0.83 2022-12-23 02:20:01,495 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=37708.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:20:08,859 INFO [train.py:894] (3/4) Epoch 11, batch 2650, loss[loss=0.1904, simple_loss=0.2773, pruned_loss=0.0517, over 18717.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.2903, pruned_loss=0.0716, over 3713073.10 frames. ], batch size: 52, lr: 1.04e-02, grad_scale: 8.0 2022-12-23 02:20:18,964 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37720.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:20:25,242 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0 from training. Duration: 24.73 2022-12-23 02:20:29,881 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.225e+02 5.554e+02 6.745e+02 9.322e+02 2.077e+03, threshold=1.349e+03, percent-clipped=8.0 2022-12-23 02:20:37,089 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0 from training. Duration: 23.965 2022-12-23 02:20:40,839 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0826, 1.5101, 2.4349, 4.3160, 3.1105, 2.6114, 0.7145, 2.8192], device='cuda:3'), covar=tensor([0.1808, 0.1912, 0.1501, 0.0505, 0.1071, 0.1416, 0.2703, 0.1124], device='cuda:3'), in_proj_covar=tensor([0.0105, 0.0117, 0.0128, 0.0131, 0.0105, 0.0135, 0.0131, 0.0110], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 02:20:46,980 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0030-146996-0_sp0.9 from training. Duration: 22.088875 2022-12-23 02:21:05,928 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0_sp0.9 from training. Duration: 23.6 2022-12-23 02:21:25,764 INFO [train.py:894] (3/4) Epoch 11, batch 2700, loss[loss=0.2191, simple_loss=0.2925, pruned_loss=0.07287, over 18576.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.2898, pruned_loss=0.0712, over 3713358.85 frames. ], batch size: 57, lr: 1.04e-02, grad_scale: 8.0 2022-12-23 02:21:31,479 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2022-12-23 02:21:56,182 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0364, 1.9431, 2.0116, 2.1621, 1.5246, 5.1048, 2.0598, 2.7452], device='cuda:3'), covar=tensor([0.3035, 0.1939, 0.1760, 0.1750, 0.1344, 0.0097, 0.1382, 0.0761], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0119, 0.0128, 0.0121, 0.0105, 0.0102, 0.0098, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 02:22:42,367 INFO [train.py:894] (3/4) Epoch 11, batch 2750, loss[loss=0.2808, simple_loss=0.3404, pruned_loss=0.1106, over 18523.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2891, pruned_loss=0.07102, over 3713539.35 frames. ], batch size: 58, lr: 1.04e-02, grad_scale: 8.0 2022-12-23 02:22:45,612 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0 from training. Duration: 23.795 2022-12-23 02:23:02,370 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp1.1 from training. Duration: 21.5409375 2022-12-23 02:23:03,845 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.734e+02 4.899e+02 5.772e+02 7.300e+02 1.732e+03, threshold=1.154e+03, percent-clipped=2.0 2022-12-23 02:23:06,108 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp0.9 from training. Duration: 24.97775 2022-12-23 02:23:18,543 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0_sp0.9 from training. Duration: 23.3444375 2022-12-23 02:23:44,573 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0_sp0.9 from training. Duration: 23.2 2022-12-23 02:23:50,672 WARNING [train.py:1060] (3/4) Exclude cut with ID 5653-46179-0060-117930-0_sp0.9 from training. Duration: 21.17225 2022-12-23 02:24:00,146 INFO [train.py:894] (3/4) Epoch 11, batch 2800, loss[loss=0.2134, simple_loss=0.2936, pruned_loss=0.06658, over 18501.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.2898, pruned_loss=0.07124, over 3713218.82 frames. ], batch size: 52, lr: 1.04e-02, grad_scale: 8.0 2022-12-23 02:24:09,091 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp0.9 from training. Duration: 24.6555625 2022-12-23 02:25:08,461 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0 from training. Duration: 20.44 2022-12-23 02:25:16,330 INFO [train.py:894] (3/4) Epoch 11, batch 2850, loss[loss=0.1959, simple_loss=0.2707, pruned_loss=0.06054, over 18706.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2889, pruned_loss=0.0707, over 3713089.12 frames. ], batch size: 50, lr: 1.04e-02, grad_scale: 8.0 2022-12-23 02:25:25,190 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0_sp0.9 from training. Duration: 23.45 2022-12-23 02:25:33,367 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37924.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:25:37,331 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.605e+02 5.185e+02 6.329e+02 8.068e+02 1.822e+03, threshold=1.266e+03, percent-clipped=5.0 2022-12-23 02:25:43,349 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2022-12-23 02:25:56,659 WARNING [train.py:1060] (3/4) Exclude cut with ID 6945-60535-0076-12784-0_sp0.9 from training. Duration: 20.52225 2022-12-23 02:26:00,744 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2022-12-23 02:26:03,978 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0 from training. Duration: 22.19 2022-12-23 02:26:14,384 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp1.1 from training. Duration: 25.3818125 2022-12-23 02:26:32,287 INFO [train.py:894] (3/4) Epoch 11, batch 2900, loss[loss=0.2409, simple_loss=0.31, pruned_loss=0.08589, over 18616.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2892, pruned_loss=0.07058, over 3713414.94 frames. ], batch size: 53, lr: 1.04e-02, grad_scale: 8.0 2022-12-23 02:26:32,340 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp0.9 from training. Duration: 28.0944375 2022-12-23 02:26:38,762 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0_sp0.9 from training. Duration: 22.9444375 2022-12-23 02:26:46,226 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp1.1 from training. Duration: 21.6318125 2022-12-23 02:26:46,367 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=37972.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:26:57,695 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37979.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:27:06,569 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0 from training. Duration: 23.695 2022-12-23 02:27:10,098 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6636, 2.2003, 1.6839, 2.5742, 2.0194, 2.0098, 2.2329, 2.6259], device='cuda:3'), covar=tensor([0.1804, 0.2863, 0.1683, 0.2686, 0.3024, 0.1003, 0.2733, 0.0690], device='cuda:3'), in_proj_covar=tensor([0.0284, 0.0273, 0.0232, 0.0349, 0.0255, 0.0217, 0.0270, 0.0194], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 02:27:20,184 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4277, 1.7880, 1.8171, 1.2856, 2.1793, 2.8519, 2.9820, 1.7851], device='cuda:3'), covar=tensor([0.0359, 0.0361, 0.0406, 0.0327, 0.0260, 0.0332, 0.0233, 0.0371], device='cuda:3'), in_proj_covar=tensor([0.0088, 0.0116, 0.0142, 0.0121, 0.0111, 0.0110, 0.0091, 0.0120], device='cuda:3'), out_proj_covar=tensor([7.3549e-05, 9.5855e-05, 1.2378e-04, 1.0184e-04, 9.4969e-05, 8.8456e-05, 7.4706e-05, 9.9941e-05], device='cuda:3') 2022-12-23 02:27:34,727 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0 from training. Duration: 23.955 2022-12-23 02:27:39,915 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3034, 2.1396, 1.7841, 1.1611, 2.9158, 2.4663, 1.9769, 1.6639], device='cuda:3'), covar=tensor([0.0395, 0.0405, 0.0567, 0.0804, 0.0181, 0.0346, 0.0540, 0.0895], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0121, 0.0129, 0.0123, 0.0091, 0.0121, 0.0138, 0.0155], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 02:27:50,584 INFO [train.py:894] (3/4) Epoch 11, batch 2950, loss[loss=0.1864, simple_loss=0.2627, pruned_loss=0.055, over 18528.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.2887, pruned_loss=0.07006, over 3713298.24 frames. ], batch size: 44, lr: 1.04e-02, grad_scale: 8.0 2022-12-23 02:28:01,723 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=38020.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:28:09,527 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp0.9 from training. Duration: 26.438875 2022-12-23 02:28:12,093 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.511e+02 5.339e+02 6.601e+02 8.126e+02 2.369e+03, threshold=1.320e+03, percent-clipped=4.0 2022-12-23 02:28:32,300 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38040.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:28:51,644 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0021-26306-0_sp0.9 from training. Duration: 21.2444375 2022-12-23 02:28:52,941 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp0.9 from training. Duration: 31.02225 2022-12-23 02:29:01,991 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0 from training. Duration: 22.395 2022-12-23 02:29:06,272 INFO [train.py:894] (3/4) Epoch 11, batch 3000, loss[loss=0.2087, simple_loss=0.2819, pruned_loss=0.06777, over 18400.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2888, pruned_loss=0.06975, over 3713517.50 frames. ], batch size: 46, lr: 1.04e-02, grad_scale: 8.0 2022-12-23 02:29:06,273 INFO [train.py:919] (3/4) Computing validation loss 2022-12-23 02:29:17,973 INFO [train.py:928] (3/4) Epoch 11, validation: loss=0.1727, simple_loss=0.2723, pruned_loss=0.03652, over 944034.00 frames. 2022-12-23 02:29:17,974 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24676MB 2022-12-23 02:29:25,799 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=38068.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:29:28,949 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0 from training. Duration: 21.075 2022-12-23 02:29:33,144 WARNING [train.py:1060] (3/4) Exclude cut with ID 6482-98857-0025-147532-0_sp0.9 from training. Duration: 20.0055625 2022-12-23 02:29:33,163 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0037-132304-0_sp0.9 from training. Duration: 22.05 2022-12-23 02:29:34,399 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0 from training. Duration: 26.8349375 2022-12-23 02:29:37,588 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp1.1 from training. Duration: 22.1090625 2022-12-23 02:29:43,548 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp0.9 from training. Duration: 26.6166875 2022-12-23 02:30:02,809 WARNING [train.py:1060] (3/4) Exclude cut with ID 2046-178027-0000-53705-0_sp0.9 from training. Duration: 20.3055625 2022-12-23 02:30:23,341 WARNING [train.py:1060] (3/4) Exclude cut with ID 7205-50138-0008-5373-0_sp0.9 from training. Duration: 20.7 2022-12-23 02:30:34,197 INFO [train.py:894] (3/4) Epoch 11, batch 3050, loss[loss=0.2257, simple_loss=0.2988, pruned_loss=0.07633, over 18596.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.2884, pruned_loss=0.06967, over 3713003.11 frames. ], batch size: 78, lr: 1.04e-02, grad_scale: 8.0 2022-12-23 02:30:49,938 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.88 vs. limit=2.0 2022-12-23 02:30:55,964 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.049e+02 5.489e+02 6.697e+02 9.055e+02 1.991e+03, threshold=1.339e+03, percent-clipped=3.0 2022-12-23 02:31:05,467 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0 from training. Duration: 22.48 2022-12-23 02:31:20,720 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp0.9 from training. Duration: 29.816625 2022-12-23 02:31:41,867 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp1.1 from training. Duration: 22.7590625 2022-12-23 02:31:42,491 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2022-12-23 02:31:46,235 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0 from training. Duration: 22.555 2022-12-23 02:31:50,329 INFO [train.py:894] (3/4) Epoch 11, batch 3100, loss[loss=0.227, simple_loss=0.2937, pruned_loss=0.08019, over 18427.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.289, pruned_loss=0.06999, over 3713260.55 frames. ], batch size: 48, lr: 1.04e-02, grad_scale: 8.0 2022-12-23 02:32:07,488 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0005-25975-0_sp0.9 from training. Duration: 21.688875 2022-12-23 02:32:41,684 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0_sp0.9 from training. Duration: 22.6 2022-12-23 02:33:07,470 INFO [train.py:894] (3/4) Epoch 11, batch 3150, loss[loss=0.2351, simple_loss=0.3098, pruned_loss=0.08017, over 18583.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.289, pruned_loss=0.07022, over 3713081.84 frames. ], batch size: 56, lr: 1.04e-02, grad_scale: 8.0 2022-12-23 02:33:19,409 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0 from training. Duration: 24.32 2022-12-23 02:33:25,890 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4226, 1.2410, 1.6111, 0.9226, 1.4508, 1.3993, 1.1790, 1.6111], device='cuda:3'), covar=tensor([0.0925, 0.1813, 0.0991, 0.1364, 0.0658, 0.1026, 0.2250, 0.0544], device='cuda:3'), in_proj_covar=tensor([0.0201, 0.0204, 0.0202, 0.0194, 0.0181, 0.0214, 0.0209, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 02:33:28,385 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.849e+02 5.249e+02 6.448e+02 8.290e+02 2.318e+03, threshold=1.290e+03, percent-clipped=4.0 2022-12-23 02:34:08,329 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2881, 2.1818, 2.5113, 1.2363, 2.8106, 2.6299, 1.4686, 3.1243], device='cuda:3'), covar=tensor([0.1336, 0.1711, 0.1268, 0.2066, 0.0853, 0.1296, 0.2250, 0.0607], device='cuda:3'), in_proj_covar=tensor([0.0200, 0.0202, 0.0201, 0.0192, 0.0180, 0.0212, 0.0207, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 02:34:16,606 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-276745-0093-13116-0_sp0.9 from training. Duration: 21.061125 2022-12-23 02:34:25,190 INFO [train.py:894] (3/4) Epoch 11, batch 3200, loss[loss=0.1957, simple_loss=0.2845, pruned_loss=0.05343, over 18616.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2893, pruned_loss=0.07046, over 3713830.16 frames. ], batch size: 53, lr: 1.03e-02, grad_scale: 8.0 2022-12-23 02:34:33,975 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0024-15855-0_sp0.9 from training. Duration: 20.32225 2022-12-23 02:34:44,794 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp1.1 from training. Duration: 0.7545625 2022-12-23 02:35:01,767 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0_sp0.9 from training. Duration: 23.9333125 2022-12-23 02:35:34,536 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp1.1 from training. Duration: 20.17275 2022-12-23 02:35:39,730 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp1.1 from training. Duration: 20.436375 2022-12-23 02:35:41,178 INFO [train.py:894] (3/4) Epoch 11, batch 3250, loss[loss=0.224, simple_loss=0.3022, pruned_loss=0.07287, over 18665.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2894, pruned_loss=0.07039, over 3714358.56 frames. ], batch size: 60, lr: 1.03e-02, grad_scale: 8.0 2022-12-23 02:36:02,550 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.401e+02 5.639e+02 6.485e+02 8.508e+02 1.298e+03, threshold=1.297e+03, percent-clipped=2.0 2022-12-23 02:36:16,099 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=38335.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:36:21,287 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2022-12-23 02:36:53,559 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=38360.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:36:57,182 INFO [train.py:894] (3/4) Epoch 11, batch 3300, loss[loss=0.2075, simple_loss=0.2752, pruned_loss=0.0699, over 18541.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2895, pruned_loss=0.07072, over 3714148.80 frames. ], batch size: 47, lr: 1.03e-02, grad_scale: 8.0 2022-12-23 02:36:59,832 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0_sp0.9 from training. Duration: 23.1055625 2022-12-23 02:37:02,552 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp1.1 from training. Duration: 23.5 2022-12-23 02:37:12,979 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp0.9 from training. Duration: 26.62775 2022-12-23 02:37:26,282 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0 from training. Duration: 21.105 2022-12-23 02:37:30,567 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0_sp0.9 from training. Duration: 24.411125 2022-12-23 02:37:55,828 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp1.1 from training. Duration: 21.263625 2022-12-23 02:38:13,631 INFO [train.py:894] (3/4) Epoch 11, batch 3350, loss[loss=0.2134, simple_loss=0.289, pruned_loss=0.0689, over 18510.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2884, pruned_loss=0.06987, over 3714821.63 frames. ], batch size: 52, lr: 1.03e-02, grad_scale: 8.0 2022-12-23 02:38:26,244 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38421.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:38:30,639 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0 from training. Duration: 20.795 2022-12-23 02:38:35,720 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.061e+02 5.055e+02 5.996e+02 7.502e+02 1.899e+03, threshold=1.199e+03, percent-clipped=4.0 2022-12-23 02:38:41,820 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0 from training. Duration: 24.76 2022-12-23 02:38:43,160 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0_sp0.9 from training. Duration: 22.25 2022-12-23 02:38:45,188 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([5.8211, 4.9313, 5.1145, 5.7723, 5.2573, 5.1425, 5.7994, 1.4813], device='cuda:3'), covar=tensor([0.0556, 0.0516, 0.0491, 0.0614, 0.1297, 0.0963, 0.0389, 0.5209], device='cuda:3'), in_proj_covar=tensor([0.0312, 0.0207, 0.0214, 0.0226, 0.0298, 0.0245, 0.0255, 0.0266], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 02:39:08,281 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp1.1 from training. Duration: 20.5045625 2022-12-23 02:39:11,704 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.0682, 0.8211, 0.9293, 1.2056, 1.2382, 1.1145, 1.0745, 0.9674], device='cuda:3'), covar=tensor([0.0251, 0.0250, 0.0502, 0.0185, 0.0214, 0.0345, 0.0252, 0.0304], device='cuda:3'), in_proj_covar=tensor([0.0088, 0.0118, 0.0145, 0.0123, 0.0112, 0.0112, 0.0092, 0.0122], device='cuda:3'), out_proj_covar=tensor([7.3684e-05, 9.7781e-05, 1.2587e-04, 1.0321e-04, 9.5662e-05, 8.9729e-05, 7.6063e-05, 1.0105e-04], device='cuda:3') 2022-12-23 02:39:18,772 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7926, 1.8450, 1.9536, 1.9836, 1.3997, 5.1223, 2.0679, 2.6647], device='cuda:3'), covar=tensor([0.3102, 0.1877, 0.1801, 0.1791, 0.1423, 0.0093, 0.1401, 0.0744], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0117, 0.0128, 0.0120, 0.0104, 0.0102, 0.0097, 0.0092], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 02:39:28,867 INFO [train.py:894] (3/4) Epoch 11, batch 3400, loss[loss=0.2382, simple_loss=0.3094, pruned_loss=0.08348, over 18398.00 frames. ], tot_loss[loss=0.2138, simple_loss=0.2882, pruned_loss=0.06969, over 3714055.68 frames. ], batch size: 53, lr: 1.03e-02, grad_scale: 8.0 2022-12-23 02:40:42,367 INFO [train.py:894] (3/4) Epoch 11, batch 3450, loss[loss=0.2284, simple_loss=0.3012, pruned_loss=0.07779, over 18504.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2895, pruned_loss=0.07034, over 3714781.51 frames. ], batch size: 52, lr: 1.03e-02, grad_scale: 8.0 2022-12-23 02:40:43,174 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2022-12-23 02:41:02,165 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.214e+02 4.989e+02 6.610e+02 8.130e+02 1.422e+03, threshold=1.322e+03, percent-clipped=2.0 2022-12-23 02:41:24,417 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.82 vs. limit=2.0 2022-12-23 02:41:55,279 INFO [train.py:894] (3/4) Epoch 11, batch 3500, loss[loss=0.2587, simple_loss=0.3245, pruned_loss=0.09648, over 18645.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2909, pruned_loss=0.07121, over 3714743.16 frames. ], batch size: 175, lr: 1.03e-02, grad_scale: 8.0 2022-12-23 02:42:17,265 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp1.1 from training. Duration: 22.2954375 2022-12-23 02:42:26,121 INFO [train.py:894] (3/4) Epoch 12, batch 0, loss[loss=0.1987, simple_loss=0.2678, pruned_loss=0.06485, over 18464.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2678, pruned_loss=0.06485, over 18464.00 frames. ], batch size: 43, lr: 9.87e-03, grad_scale: 8.0 2022-12-23 02:42:26,121 INFO [train.py:919] (3/4) Computing validation loss 2022-12-23 02:42:32,075 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7037, 1.6102, 1.7406, 1.7420, 1.2055, 3.2962, 1.7669, 2.2199], device='cuda:3'), covar=tensor([0.3641, 0.2305, 0.1984, 0.2000, 0.1603, 0.0232, 0.1715, 0.0805], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0118, 0.0127, 0.0120, 0.0104, 0.0101, 0.0097, 0.0092], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 02:42:36,023 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.5216, 4.0518, 3.8845, 4.4792, 3.8962, 3.9901, 4.5884, 1.6415], device='cuda:3'), covar=tensor([0.0648, 0.0662, 0.0803, 0.0607, 0.1699, 0.1034, 0.0460, 0.5075], device='cuda:3'), in_proj_covar=tensor([0.0305, 0.0202, 0.0209, 0.0221, 0.0291, 0.0242, 0.0251, 0.0260], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 02:42:36,864 INFO [train.py:928] (3/4) Epoch 12, validation: loss=0.1736, simple_loss=0.273, pruned_loss=0.03713, over 944034.00 frames. 2022-12-23 02:42:36,865 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24676MB 2022-12-23 02:43:26,958 WARNING [train.py:1060] (3/4) Exclude cut with ID 298-126791-0067-24026-0_sp0.9 from training. Duration: 21.438875 2022-12-23 02:43:31,234 WARNING [train.py:1060] (3/4) Exclude cut with ID 5652-39938-0025-23684-0_sp0.9 from training. Duration: 22.2055625 2022-12-23 02:43:54,089 INFO [train.py:894] (3/4) Epoch 12, batch 50, loss[loss=0.1757, simple_loss=0.2517, pruned_loss=0.0499, over 18523.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2852, pruned_loss=0.06063, over 837781.63 frames. ], batch size: 44, lr: 9.87e-03, grad_scale: 8.0 2022-12-23 02:44:05,888 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.87 vs. limit=5.0 2022-12-23 02:44:06,424 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.798e+02 4.356e+02 5.737e+02 7.037e+02 2.274e+03, threshold=1.147e+03, percent-clipped=3.0 2022-12-23 02:44:11,647 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2022-12-23 02:44:18,572 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=38635.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:44:23,456 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.4607, 2.9811, 3.0667, 3.3305, 3.1424, 3.1303, 3.5380, 1.7623], device='cuda:3'), covar=tensor([0.0801, 0.0662, 0.0636, 0.0878, 0.1388, 0.1055, 0.0945, 0.3693], device='cuda:3'), in_proj_covar=tensor([0.0303, 0.0202, 0.0208, 0.0221, 0.0289, 0.0241, 0.0250, 0.0261], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 02:45:09,827 INFO [train.py:894] (3/4) Epoch 12, batch 100, loss[loss=0.1899, simple_loss=0.2811, pruned_loss=0.04936, over 18720.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2801, pruned_loss=0.05737, over 1475559.66 frames. ], batch size: 54, lr: 9.86e-03, grad_scale: 8.0 2022-12-23 02:45:31,045 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=38683.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:46:20,189 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=38716.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:46:24,465 INFO [train.py:894] (3/4) Epoch 12, batch 150, loss[loss=0.192, simple_loss=0.2852, pruned_loss=0.04941, over 18650.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2788, pruned_loss=0.05642, over 1971031.24 frames. ], batch size: 78, lr: 9.85e-03, grad_scale: 8.0 2022-12-23 02:46:31,293 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9807, 1.8233, 1.8622, 2.3201, 2.0423, 4.5939, 1.7665, 1.7600], device='cuda:3'), covar=tensor([0.0828, 0.1777, 0.1115, 0.0970, 0.1471, 0.0147, 0.1371, 0.1665], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0081, 0.0074, 0.0074, 0.0090, 0.0071, 0.0085, 0.0077], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 02:46:32,279 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0 from training. Duration: 24.525 2022-12-23 02:46:36,584 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.508e+02 3.672e+02 4.513e+02 6.038e+02 1.379e+03, threshold=9.027e+02, percent-clipped=2.0 2022-12-23 02:47:03,546 WARNING [train.py:1060] (3/4) Exclude cut with ID 3699-47246-0007-3408-0_sp0.9 from training. Duration: 20.26675 2022-12-23 02:47:06,878 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6246, 1.4029, 1.3244, 1.8541, 1.6331, 3.3916, 1.4117, 1.4333], device='cuda:3'), covar=tensor([0.0871, 0.1788, 0.1138, 0.0936, 0.1490, 0.0202, 0.1368, 0.1658], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0081, 0.0074, 0.0074, 0.0090, 0.0071, 0.0085, 0.0076], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 02:47:18,630 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp0.9 from training. Duration: 27.25 2022-12-23 02:47:39,871 INFO [train.py:894] (3/4) Epoch 12, batch 200, loss[loss=0.197, simple_loss=0.2845, pruned_loss=0.05479, over 18492.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.2772, pruned_loss=0.05517, over 2356179.41 frames. ], batch size: 64, lr: 9.85e-03, grad_scale: 8.0 2022-12-23 02:48:34,273 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0 from training. Duration: 21.68 2022-12-23 02:48:36,394 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2022-12-23 02:48:44,357 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0 from training. Duration: 21.6300625 2022-12-23 02:48:45,169 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2022-12-23 02:48:55,432 INFO [train.py:894] (3/4) Epoch 12, batch 250, loss[loss=0.1678, simple_loss=0.247, pruned_loss=0.04433, over 18663.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.275, pruned_loss=0.05434, over 2656004.58 frames. ], batch size: 46, lr: 9.84e-03, grad_scale: 8.0 2022-12-23 02:49:02,119 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=38823.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:49:07,343 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.354e+02 3.798e+02 4.673e+02 6.319e+02 1.410e+03, threshold=9.347e+02, percent-clipped=6.0 2022-12-23 02:49:10,340 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0_sp0.9 from training. Duration: 24.033375 2022-12-23 02:49:15,394 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7019, 1.4223, 1.4280, 1.4600, 1.8700, 1.8035, 1.9435, 1.2171], device='cuda:3'), covar=tensor([0.0308, 0.0247, 0.0391, 0.0195, 0.0146, 0.0329, 0.0212, 0.0265], device='cuda:3'), in_proj_covar=tensor([0.0086, 0.0116, 0.0143, 0.0121, 0.0110, 0.0110, 0.0092, 0.0119], device='cuda:3'), out_proj_covar=tensor([7.1814e-05, 9.6304e-05, 1.2422e-04, 1.0110e-04, 9.3643e-05, 8.7717e-05, 7.5765e-05, 9.8289e-05], device='cuda:3') 2022-12-23 02:50:09,349 INFO [train.py:894] (3/4) Epoch 12, batch 300, loss[loss=0.1899, simple_loss=0.2812, pruned_loss=0.04925, over 18726.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2756, pruned_loss=0.05408, over 2889194.39 frames. ], batch size: 52, lr: 9.83e-03, grad_scale: 8.0 2022-12-23 02:50:09,407 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0 from training. Duration: 22.905 2022-12-23 02:50:09,460 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp1.1 from training. Duration: 23.4318125 2022-12-23 02:50:16,260 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5129, 2.6461, 1.7450, 1.1156, 3.1960, 2.8582, 2.0964, 1.7178], device='cuda:3'), covar=tensor([0.0304, 0.0252, 0.0502, 0.0746, 0.0139, 0.0238, 0.0444, 0.0813], device='cuda:3'), in_proj_covar=tensor([0.0118, 0.0117, 0.0125, 0.0118, 0.0089, 0.0115, 0.0132, 0.0149], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 02:50:32,572 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38884.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:50:58,284 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.5922, 2.8765, 2.7334, 1.0720, 2.5563, 2.3144, 2.1319, 2.2595], device='cuda:3'), covar=tensor([0.0551, 0.0588, 0.1298, 0.1770, 0.1575, 0.1477, 0.1429, 0.1060], device='cuda:3'), in_proj_covar=tensor([0.0164, 0.0177, 0.0197, 0.0192, 0.0202, 0.0192, 0.0202, 0.0192], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 02:51:24,900 INFO [train.py:894] (3/4) Epoch 12, batch 350, loss[loss=0.1806, simple_loss=0.2727, pruned_loss=0.0442, over 18588.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2765, pruned_loss=0.05433, over 3071582.26 frames. ], batch size: 51, lr: 9.83e-03, grad_scale: 8.0 2022-12-23 02:51:36,876 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.329e+02 4.042e+02 5.093e+02 6.152e+02 1.114e+03, threshold=1.019e+03, percent-clipped=2.0 2022-12-23 02:52:06,915 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp1.1 from training. Duration: 20.82275 2022-12-23 02:52:08,516 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp0.9 from training. Duration: 25.45 2022-12-23 02:52:32,858 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2022-12-23 02:52:39,574 INFO [train.py:894] (3/4) Epoch 12, batch 400, loss[loss=0.179, simple_loss=0.2659, pruned_loss=0.04603, over 18590.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.2784, pruned_loss=0.05506, over 3214451.30 frames. ], batch size: 51, lr: 9.82e-03, grad_scale: 8.0 2022-12-23 02:53:08,718 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0 from training. Duration: 25.775 2022-12-23 02:53:31,139 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0_sp0.9 from training. Duration: 22.25 2022-12-23 02:53:51,621 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39016.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:53:55,556 INFO [train.py:894] (3/4) Epoch 12, batch 450, loss[loss=0.2082, simple_loss=0.2958, pruned_loss=0.0603, over 18694.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2805, pruned_loss=0.05588, over 3325747.39 frames. ], batch size: 50, lr: 9.82e-03, grad_scale: 8.0 2022-12-23 02:54:00,470 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0 from training. Duration: 26.205 2022-12-23 02:54:02,368 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6391, 1.3330, 0.8933, 1.1628, 2.1894, 0.9456, 1.4468, 1.5885], device='cuda:3'), covar=tensor([0.1592, 0.2013, 0.2182, 0.1567, 0.1469, 0.1688, 0.1386, 0.1663], device='cuda:3'), in_proj_covar=tensor([0.0090, 0.0099, 0.0117, 0.0094, 0.0112, 0.0090, 0.0097, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 02:54:08,263 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.414e+02 3.857e+02 4.585e+02 5.760e+02 1.375e+03, threshold=9.170e+02, percent-clipped=1.0 2022-12-23 02:54:15,718 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp0.9 from training. Duration: 30.1555625 2022-12-23 02:54:21,310 WARNING [train.py:1060] (3/4) Exclude cut with ID 1265-135635-0050-6781-0_sp0.9 from training. Duration: 21.8333125 2022-12-23 02:54:26,385 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1822, 1.3737, 1.8766, 1.8334, 2.1306, 2.0345, 1.9494, 1.6308], device='cuda:3'), covar=tensor([0.1740, 0.2563, 0.2013, 0.2277, 0.1438, 0.0758, 0.2246, 0.1015], device='cuda:3'), in_proj_covar=tensor([0.0254, 0.0289, 0.0264, 0.0299, 0.0282, 0.0238, 0.0308, 0.0224], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 02:54:30,344 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp1.1 from training. Duration: 20.6545625 2022-12-23 02:55:03,963 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=39064.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:55:11,089 INFO [train.py:894] (3/4) Epoch 12, batch 500, loss[loss=0.2089, simple_loss=0.2842, pruned_loss=0.06683, over 18659.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2819, pruned_loss=0.05699, over 3412008.08 frames. ], batch size: 48, lr: 9.81e-03, grad_scale: 8.0 2022-12-23 02:55:14,591 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0045-39920-0_sp0.9 from training. Duration: 20.52225 2022-12-23 02:55:34,347 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp0.9 from training. Duration: 29.1166875 2022-12-23 02:56:26,653 INFO [train.py:894] (3/4) Epoch 12, batch 550, loss[loss=0.2099, simple_loss=0.2939, pruned_loss=0.063, over 18470.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2828, pruned_loss=0.05741, over 3478543.41 frames. ], batch size: 50, lr: 9.80e-03, grad_scale: 8.0 2022-12-23 02:56:32,896 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133211-0007-59831-0_sp0.9 from training. Duration: 21.388875 2022-12-23 02:56:39,181 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.465e+02 4.214e+02 4.900e+02 6.411e+02 1.265e+03, threshold=9.799e+02, percent-clipped=3.0 2022-12-23 02:56:49,878 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2022-12-23 02:56:56,515 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1781, 1.7719, 1.9074, 1.8589, 2.0403, 2.0271, 1.9911, 1.7192], device='cuda:3'), covar=tensor([0.1403, 0.2193, 0.1580, 0.2108, 0.1366, 0.0689, 0.2077, 0.0898], device='cuda:3'), in_proj_covar=tensor([0.0255, 0.0289, 0.0265, 0.0299, 0.0282, 0.0240, 0.0309, 0.0225], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 02:57:09,778 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0 from training. Duration: 22.72 2022-12-23 02:57:11,159 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0_sp0.9 from training. Duration: 22.7444375 2022-12-23 02:57:41,897 INFO [train.py:894] (3/4) Epoch 12, batch 600, loss[loss=0.2105, simple_loss=0.2966, pruned_loss=0.0622, over 18587.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.284, pruned_loss=0.05805, over 3530631.49 frames. ], batch size: 51, lr: 9.80e-03, grad_scale: 8.0 2022-12-23 02:57:53,064 WARNING [train.py:1060] (3/4) Exclude cut with ID 4133-6541-0027-40495-0_sp1.1 from training. Duration: 0.9681875 2022-12-23 02:57:56,116 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0_sp0.9 from training. Duration: 22.3166875 2022-12-23 02:57:57,769 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39179.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:58:02,030 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133212-0015-59917-0_sp0.9 from training. Duration: 21.8166875 2022-12-23 02:58:06,546 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39185.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:58:57,143 INFO [train.py:894] (3/4) Epoch 12, batch 650, loss[loss=0.2064, simple_loss=0.2986, pruned_loss=0.05709, over 18521.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2832, pruned_loss=0.05744, over 3570870.63 frames. ], batch size: 58, lr: 9.79e-03, grad_scale: 8.0 2022-12-23 02:59:07,015 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5179, 1.0347, 1.1834, 1.1310, 1.6802, 1.6093, 1.6476, 1.0769], device='cuda:3'), covar=tensor([0.0236, 0.0271, 0.0519, 0.0240, 0.0192, 0.0287, 0.0208, 0.0308], device='cuda:3'), in_proj_covar=tensor([0.0088, 0.0119, 0.0145, 0.0123, 0.0115, 0.0113, 0.0094, 0.0122], device='cuda:3'), out_proj_covar=tensor([7.2955e-05, 9.8092e-05, 1.2556e-04, 1.0221e-04, 9.7672e-05, 9.0681e-05, 7.6718e-05, 1.0090e-04], device='cuda:3') 2022-12-23 02:59:09,407 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.588e+02 3.961e+02 5.057e+02 6.119e+02 1.297e+03, threshold=1.011e+03, percent-clipped=1.0 2022-12-23 02:59:36,310 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.8971, 2.4379, 1.9660, 1.0634, 1.9382, 2.2930, 1.7963, 2.1867], device='cuda:3'), covar=tensor([0.0597, 0.0528, 0.1281, 0.1669, 0.1497, 0.1280, 0.1443, 0.0862], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0178, 0.0200, 0.0194, 0.0204, 0.0190, 0.0205, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 02:59:39,018 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39246.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 02:59:40,779 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.08 vs. limit=5.0 2022-12-23 02:59:44,317 WARNING [train.py:1060] (3/4) Exclude cut with ID 4957-30119-0041-23990-0_sp0.9 from training. Duration: 20.22775 2022-12-23 03:00:13,407 INFO [train.py:894] (3/4) Epoch 12, batch 700, loss[loss=0.1804, simple_loss=0.2623, pruned_loss=0.04927, over 18425.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.284, pruned_loss=0.05805, over 3603432.11 frames. ], batch size: 48, lr: 9.78e-03, grad_scale: 8.0 2022-12-23 03:00:28,220 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp1.1 from training. Duration: 24.67275 2022-12-23 03:00:36,247 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39284.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:00:54,837 WARNING [train.py:1060] (3/4) Exclude cut with ID 3082-165428-0081-50734-0_sp0.9 from training. Duration: 21.8055625 2022-12-23 03:01:27,293 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.2584, 1.5927, 1.8954, 1.1058, 1.0242, 2.0084, 1.7118, 1.6806], device='cuda:3'), covar=tensor([0.0709, 0.0344, 0.0291, 0.0331, 0.0389, 0.0381, 0.0251, 0.0528], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0157, 0.0113, 0.0129, 0.0137, 0.0129, 0.0146, 0.0148], device='cuda:3'), out_proj_covar=tensor([1.1047e-04, 1.2885e-04, 9.1057e-05, 1.0267e-04, 1.1031e-04, 1.0704e-04, 1.2133e-04, 1.2060e-04], device='cuda:3') 2022-12-23 03:01:28,248 INFO [train.py:894] (3/4) Epoch 12, batch 750, loss[loss=0.201, simple_loss=0.2907, pruned_loss=0.0557, over 18632.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2844, pruned_loss=0.05787, over 3628607.25 frames. ], batch size: 53, lr: 9.78e-03, grad_scale: 8.0 2022-12-23 03:01:32,564 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0_sp0.9 from training. Duration: 22.6666875 2022-12-23 03:01:39,380 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.862e+02 4.095e+02 4.892e+02 6.005e+02 1.372e+03, threshold=9.785e+02, percent-clipped=2.0 2022-12-23 03:02:07,109 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39345.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:02:33,295 WARNING [train.py:1060] (3/4) Exclude cut with ID 2411-132532-0017-82279-0_sp1.1 from training. Duration: 0.9681875 2022-12-23 03:02:43,260 INFO [train.py:894] (3/4) Epoch 12, batch 800, loss[loss=0.1789, simple_loss=0.2613, pruned_loss=0.0483, over 18412.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.284, pruned_loss=0.05773, over 3647674.76 frames. ], batch size: 48, lr: 9.77e-03, grad_scale: 8.0 2022-12-23 03:02:59,690 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0 from training. Duration: 22.485 2022-12-23 03:03:37,029 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp1.1 from training. Duration: 23.82275 2022-12-23 03:03:51,568 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0 from training. Duration: 20.77 2022-12-23 03:03:58,236 INFO [train.py:894] (3/4) Epoch 12, batch 850, loss[loss=0.1785, simple_loss=0.2573, pruned_loss=0.04988, over 18380.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2833, pruned_loss=0.0578, over 3662445.77 frames. ], batch size: 46, lr: 9.77e-03, grad_scale: 8.0 2022-12-23 03:03:59,641 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0_sp0.9 from training. Duration: 24.088875 2022-12-23 03:04:07,778 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2022-12-23 03:04:09,524 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.336e+02 4.122e+02 4.845e+02 6.172e+02 1.351e+03, threshold=9.691e+02, percent-clipped=1.0 2022-12-23 03:04:18,532 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2022-12-23 03:04:30,734 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp1.1 from training. Duration: 20.4409375 2022-12-23 03:05:04,225 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39463.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:05:12,887 INFO [train.py:894] (3/4) Epoch 12, batch 900, loss[loss=0.2173, simple_loss=0.2788, pruned_loss=0.07789, over 18415.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2832, pruned_loss=0.05794, over 3674664.34 frames. ], batch size: 42, lr: 9.76e-03, grad_scale: 8.0 2022-12-23 03:05:28,501 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39479.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:05:35,125 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=2.06 vs. limit=2.0 2022-12-23 03:05:47,534 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0_sp0.9 from training. Duration: 22.511125 2022-12-23 03:05:47,557 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0 from training. Duration: 20.675 2022-12-23 03:06:15,920 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4906, 2.3326, 1.7720, 1.2566, 2.8748, 2.6004, 2.1462, 1.7368], device='cuda:3'), covar=tensor([0.0297, 0.0325, 0.0511, 0.0736, 0.0163, 0.0277, 0.0400, 0.0735], device='cuda:3'), in_proj_covar=tensor([0.0121, 0.0119, 0.0127, 0.0120, 0.0090, 0.0117, 0.0134, 0.0151], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 03:06:30,622 INFO [train.py:894] (3/4) Epoch 12, batch 950, loss[loss=0.2219, simple_loss=0.3054, pruned_loss=0.06916, over 18562.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2832, pruned_loss=0.05772, over 3682890.00 frames. ], batch size: 69, lr: 9.75e-03, grad_scale: 8.0 2022-12-23 03:06:38,394 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39524.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:06:42,897 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.364e+02 4.631e+02 5.225e+02 6.813e+02 1.904e+03, threshold=1.045e+03, percent-clipped=5.0 2022-12-23 03:06:43,082 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=39527.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:06:54,771 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8389, 1.2139, 2.0561, 3.1809, 2.4437, 2.4811, 0.8456, 2.1427], device='cuda:3'), covar=tensor([0.1709, 0.1892, 0.1482, 0.0592, 0.1023, 0.1498, 0.2384, 0.1219], device='cuda:3'), in_proj_covar=tensor([0.0101, 0.0114, 0.0130, 0.0131, 0.0104, 0.0134, 0.0128, 0.0107], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 03:06:54,971 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1715, 2.2542, 1.4654, 2.5844, 2.3524, 2.0897, 2.9395, 2.1976], device='cuda:3'), covar=tensor([0.0812, 0.1581, 0.2693, 0.1854, 0.1604, 0.0832, 0.0961, 0.1160], device='cuda:3'), in_proj_covar=tensor([0.0173, 0.0195, 0.0236, 0.0281, 0.0224, 0.0182, 0.0206, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 03:07:03,408 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39541.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:07:28,500 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp0.9 from training. Duration: 24.9833125 2022-12-23 03:07:42,107 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3734, 1.7587, 0.3078, 1.7305, 2.7285, 1.7383, 2.4799, 2.3229], device='cuda:3'), covar=tensor([0.1512, 0.2100, 0.3233, 0.1620, 0.1526, 0.1600, 0.1457, 0.1663], device='cuda:3'), in_proj_covar=tensor([0.0091, 0.0099, 0.0119, 0.0095, 0.0114, 0.0091, 0.0098, 0.0095], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 03:07:46,163 INFO [train.py:894] (3/4) Epoch 12, batch 1000, loss[loss=0.1914, simple_loss=0.2708, pruned_loss=0.05602, over 18669.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2823, pruned_loss=0.05691, over 3690753.16 frames. ], batch size: 46, lr: 9.75e-03, grad_scale: 8.0 2022-12-23 03:07:58,945 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0 from training. Duration: 27.14 2022-12-23 03:08:14,338 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0 from training. Duration: 22.44 2022-12-23 03:08:49,034 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6464, 2.1576, 1.6573, 2.5287, 1.8707, 2.1136, 2.0121, 2.6764], device='cuda:3'), covar=tensor([0.1584, 0.2840, 0.1569, 0.2291, 0.3077, 0.0876, 0.2539, 0.0664], device='cuda:3'), in_proj_covar=tensor([0.0276, 0.0271, 0.0227, 0.0340, 0.0251, 0.0214, 0.0266, 0.0191], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 03:09:01,596 INFO [train.py:894] (3/4) Epoch 12, batch 1050, loss[loss=0.2155, simple_loss=0.3075, pruned_loss=0.06173, over 18717.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2822, pruned_loss=0.05699, over 3695934.87 frames. ], batch size: 60, lr: 9.74e-03, grad_scale: 8.0 2022-12-23 03:09:12,761 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.2587, 1.5825, 1.9223, 0.9284, 1.1991, 2.0438, 1.7652, 1.6898], device='cuda:3'), covar=tensor([0.0645, 0.0340, 0.0263, 0.0349, 0.0361, 0.0316, 0.0218, 0.0499], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0158, 0.0114, 0.0129, 0.0138, 0.0130, 0.0146, 0.0148], device='cuda:3'), out_proj_covar=tensor([1.1091e-04, 1.2928e-04, 9.2039e-05, 1.0297e-04, 1.1087e-04, 1.0801e-04, 1.2142e-04, 1.2059e-04], device='cuda:3') 2022-12-23 03:09:13,699 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.640e+02 3.885e+02 4.737e+02 5.602e+02 8.363e+02, threshold=9.474e+02, percent-clipped=0.0 2022-12-23 03:09:31,267 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0060-62364-0_sp0.9 from training. Duration: 21.361125 2022-12-23 03:09:31,577 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39639.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 03:09:32,779 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39640.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:09:37,422 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp1.1 from training. Duration: 27.0318125 2022-12-23 03:09:46,003 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp0.9 from training. Duration: 28.638875 2022-12-23 03:10:03,123 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0 from training. Duration: 20.4 2022-12-23 03:10:16,634 INFO [train.py:894] (3/4) Epoch 12, batch 1100, loss[loss=0.1691, simple_loss=0.2527, pruned_loss=0.04278, over 18476.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2827, pruned_loss=0.05721, over 3699393.35 frames. ], batch size: 43, lr: 9.74e-03, grad_scale: 8.0 2022-12-23 03:10:20,484 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2949, 1.3756, 2.4173, 4.3368, 3.2689, 2.6397, 0.9253, 3.0505], device='cuda:3'), covar=tensor([0.1556, 0.1889, 0.1550, 0.0492, 0.0921, 0.1279, 0.2311, 0.0865], device='cuda:3'), in_proj_covar=tensor([0.0103, 0.0114, 0.0130, 0.0131, 0.0105, 0.0134, 0.0129, 0.0108], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 03:10:36,376 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0 from training. Duration: 20.025 2022-12-23 03:10:36,387 WARNING [train.py:1060] (3/4) Exclude cut with ID 2364-131735-0112-64612-0_sp0.9 from training. Duration: 20.488875 2022-12-23 03:10:40,503 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0 from training. Duration: 29.735 2022-12-23 03:11:04,475 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39700.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 03:11:33,400 INFO [train.py:894] (3/4) Epoch 12, batch 1150, loss[loss=0.1941, simple_loss=0.2865, pruned_loss=0.05084, over 18580.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2825, pruned_loss=0.057, over 3703139.09 frames. ], batch size: 51, lr: 9.73e-03, grad_scale: 8.0 2022-12-23 03:11:46,696 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.615e+02 4.757e+02 5.682e+02 7.291e+02 2.564e+03, threshold=1.136e+03, percent-clipped=13.0 2022-12-23 03:11:59,930 WARNING [train.py:1060] (3/4) Exclude cut with ID 7276-92427-0014-12983-0_sp0.9 from training. Duration: 21.3055625 2022-12-23 03:12:01,350 WARNING [train.py:1060] (3/4) Exclude cut with ID 1025-75365-0008-79168-0_sp0.9 from training. Duration: 22.0666875 2022-12-23 03:12:16,678 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3191, 1.6591, 0.9951, 1.7787, 2.3898, 1.8240, 2.1552, 2.2042], device='cuda:3'), covar=tensor([0.1314, 0.1889, 0.2335, 0.1307, 0.1481, 0.1420, 0.1263, 0.1447], device='cuda:3'), in_proj_covar=tensor([0.0091, 0.0100, 0.0119, 0.0095, 0.0113, 0.0091, 0.0098, 0.0095], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 03:12:45,990 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39767.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:12:48,821 INFO [train.py:894] (3/4) Epoch 12, batch 1200, loss[loss=0.1956, simple_loss=0.288, pruned_loss=0.05156, over 18552.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2822, pruned_loss=0.05698, over 3705249.22 frames. ], batch size: 55, lr: 9.72e-03, grad_scale: 8.0 2022-12-23 03:12:57,521 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2022-12-23 03:13:46,132 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0 from training. Duration: 20.26 2022-12-23 03:13:58,625 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39815.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:14:01,029 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0030-9324-0_sp0.9 from training. Duration: 21.3444375 2022-12-23 03:14:03,876 INFO [train.py:894] (3/4) Epoch 12, batch 1250, loss[loss=0.22, simple_loss=0.3062, pruned_loss=0.0669, over 18689.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2828, pruned_loss=0.05659, over 3707852.65 frames. ], batch size: 60, lr: 9.72e-03, grad_scale: 8.0 2022-12-23 03:14:04,031 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39819.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:14:08,485 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6229, 1.4299, 1.3271, 0.7927, 1.8452, 1.5070, 1.4594, 1.2527], device='cuda:3'), covar=tensor([0.0350, 0.0472, 0.0491, 0.0742, 0.0295, 0.0365, 0.0447, 0.0865], device='cuda:3'), in_proj_covar=tensor([0.0121, 0.0117, 0.0126, 0.0118, 0.0088, 0.0116, 0.0133, 0.0149], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 03:14:13,517 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2022-12-23 03:14:16,863 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.560e+02 3.885e+02 4.860e+02 6.199e+02 1.184e+03, threshold=9.720e+02, percent-clipped=2.0 2022-12-23 03:14:17,270 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39828.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:14:36,694 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39841.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:14:56,803 WARNING [train.py:1060] (3/4) Exclude cut with ID 497-129325-0061-62254-0_sp1.1 from training. Duration: 0.97725 2022-12-23 03:14:58,590 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.0794, 0.7948, 0.5299, 0.8296, 1.3627, 0.5031, 0.8703, 0.9807], device='cuda:3'), covar=tensor([0.1243, 0.1748, 0.1811, 0.1215, 0.1412, 0.1551, 0.1184, 0.1341], device='cuda:3'), in_proj_covar=tensor([0.0091, 0.0100, 0.0119, 0.0095, 0.0113, 0.0091, 0.0098, 0.0095], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 03:15:19,543 INFO [train.py:894] (3/4) Epoch 12, batch 1300, loss[loss=0.1756, simple_loss=0.2595, pruned_loss=0.04582, over 18586.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2828, pruned_loss=0.05674, over 3709362.93 frames. ], batch size: 49, lr: 9.71e-03, grad_scale: 8.0 2022-12-23 03:15:29,815 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39876.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:15:40,577 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0_sp0.9 from training. Duration: 22.97225 2022-12-23 03:15:49,164 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=39889.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:15:58,552 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5414, 2.1335, 1.3904, 2.5843, 1.8633, 1.7137, 1.8816, 2.5858], device='cuda:3'), covar=tensor([0.1881, 0.2928, 0.1885, 0.2262, 0.3270, 0.1187, 0.3098, 0.0676], device='cuda:3'), in_proj_covar=tensor([0.0277, 0.0271, 0.0228, 0.0339, 0.0251, 0.0214, 0.0266, 0.0190], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 03:16:10,617 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2022-12-23 03:16:12,689 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0047-39922-0_sp0.9 from training. Duration: 21.97775 2022-12-23 03:16:13,023 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0296, 1.6884, 1.7757, 2.4183, 2.2050, 4.5517, 1.4632, 1.9001], device='cuda:3'), covar=tensor([0.0773, 0.1699, 0.0993, 0.0856, 0.1252, 0.0140, 0.1413, 0.1420], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0081, 0.0074, 0.0074, 0.0091, 0.0070, 0.0085, 0.0075], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 03:16:13,503 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2022-12-23 03:16:23,501 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.2661, 1.3194, 1.5360, 0.8765, 1.4140, 1.4069, 1.2134, 1.6445], device='cuda:3'), covar=tensor([0.0924, 0.1669, 0.1000, 0.1279, 0.0702, 0.0884, 0.2062, 0.0525], device='cuda:3'), in_proj_covar=tensor([0.0192, 0.0196, 0.0196, 0.0189, 0.0174, 0.0205, 0.0202, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 03:16:27,544 WARNING [train.py:1060] (3/4) Exclude cut with ID 1112-1043-0006-89194-0_sp0.9 from training. Duration: 21.8333125 2022-12-23 03:16:33,490 INFO [train.py:894] (3/4) Epoch 12, batch 1350, loss[loss=0.2069, simple_loss=0.2957, pruned_loss=0.05908, over 18623.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2819, pruned_loss=0.05612, over 3709971.78 frames. ], batch size: 53, lr: 9.71e-03, grad_scale: 8.0 2022-12-23 03:16:39,422 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0 from training. Duration: 20.47 2022-12-23 03:16:46,694 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.753e+02 3.746e+02 4.276e+02 5.136e+02 9.352e+02, threshold=8.553e+02, percent-clipped=0.0 2022-12-23 03:16:59,605 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0316, 1.8099, 1.5703, 0.9356, 2.3313, 1.9446, 1.7305, 1.4234], device='cuda:3'), covar=tensor([0.0329, 0.0400, 0.0483, 0.0726, 0.0205, 0.0348, 0.0484, 0.0818], device='cuda:3'), in_proj_covar=tensor([0.0122, 0.0119, 0.0127, 0.0119, 0.0087, 0.0117, 0.0134, 0.0150], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 03:17:04,984 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39940.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:17:44,763 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0037-39912-0_sp0.9 from training. Duration: 20.67225 2022-12-23 03:17:47,789 INFO [train.py:894] (3/4) Epoch 12, batch 1400, loss[loss=0.1918, simple_loss=0.2811, pruned_loss=0.05126, over 18701.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2822, pruned_loss=0.05663, over 3710869.48 frames. ], batch size: 62, lr: 9.70e-03, grad_scale: 8.0 2022-12-23 03:18:03,447 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp0.9 from training. Duration: 25.2444375 2022-12-23 03:18:08,538 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2022-12-23 03:18:16,682 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=39988.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:18:22,968 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2022-12-23 03:18:26,080 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0021-76797-0_sp0.9 from training. Duration: 21.1445 2022-12-23 03:18:28,275 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39995.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 03:19:07,132 INFO [train.py:894] (3/4) Epoch 12, batch 1450, loss[loss=0.2017, simple_loss=0.2864, pruned_loss=0.05851, over 18461.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2825, pruned_loss=0.05665, over 3711056.20 frames. ], batch size: 54, lr: 9.69e-03, grad_scale: 8.0 2022-12-23 03:19:20,694 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.732e+02 3.983e+02 4.941e+02 6.040e+02 1.228e+03, threshold=9.881e+02, percent-clipped=7.0 2022-12-23 03:19:22,711 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5670, 1.3925, 1.2446, 0.6586, 1.7911, 1.4406, 1.4094, 1.2536], device='cuda:3'), covar=tensor([0.0365, 0.0480, 0.0526, 0.0755, 0.0283, 0.0372, 0.0479, 0.0887], device='cuda:3'), in_proj_covar=tensor([0.0121, 0.0118, 0.0126, 0.0117, 0.0087, 0.0116, 0.0133, 0.0150], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 03:19:44,125 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp0.9 from training. Duration: 33.038875 2022-12-23 03:20:12,893 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2022-12-23 03:20:22,050 INFO [train.py:894] (3/4) Epoch 12, batch 1500, loss[loss=0.1877, simple_loss=0.2817, pruned_loss=0.04683, over 18625.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2809, pruned_loss=0.05593, over 3711549.87 frames. ], batch size: 98, lr: 9.69e-03, grad_scale: 8.0 2022-12-23 03:20:22,051 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64291-0000-16059-0_sp0.9 from training. Duration: 20.0944375 2022-12-23 03:20:37,028 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp1.1 from training. Duration: 20.4 2022-12-23 03:20:44,920 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0 from training. Duration: 20.085 2022-12-23 03:20:55,640 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0_sp0.9 from training. Duration: 23.07775 2022-12-23 03:21:17,010 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6445, 1.3884, 1.3508, 1.9353, 1.6432, 3.3996, 1.3171, 1.4399], device='cuda:3'), covar=tensor([0.0831, 0.1772, 0.1064, 0.0871, 0.1372, 0.0218, 0.1413, 0.1526], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0082, 0.0074, 0.0075, 0.0091, 0.0071, 0.0085, 0.0075], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 03:21:23,423 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2567, 2.0157, 1.7258, 1.0860, 2.5926, 2.2392, 2.0789, 1.6046], device='cuda:3'), covar=tensor([0.0313, 0.0382, 0.0498, 0.0736, 0.0181, 0.0322, 0.0393, 0.0760], device='cuda:3'), in_proj_covar=tensor([0.0122, 0.0118, 0.0127, 0.0117, 0.0087, 0.0117, 0.0133, 0.0150], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 03:21:37,396 INFO [train.py:894] (3/4) Epoch 12, batch 1550, loss[loss=0.186, simple_loss=0.2782, pruned_loss=0.04688, over 18592.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2808, pruned_loss=0.05589, over 3712561.33 frames. ], batch size: 51, lr: 9.68e-03, grad_scale: 8.0 2022-12-23 03:21:38,370 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40119.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:21:42,591 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp0.9 from training. Duration: 24.9333125 2022-12-23 03:21:44,054 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40123.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:21:51,315 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.523e+02 3.928e+02 5.193e+02 6.914e+02 1.537e+03, threshold=1.039e+03, percent-clipped=5.0 2022-12-23 03:22:20,574 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2022-12-23 03:22:25,770 WARNING [train.py:1060] (3/4) Exclude cut with ID 5118-111612-0016-124680-0_sp0.9 from training. Duration: 20.388875 2022-12-23 03:22:31,755 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp1.1 from training. Duration: 20.3590625 2022-12-23 03:22:51,394 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=40167.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:22:54,113 INFO [train.py:894] (3/4) Epoch 12, batch 1600, loss[loss=0.1892, simple_loss=0.2816, pruned_loss=0.04842, over 18724.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2792, pruned_loss=0.05495, over 3712710.67 frames. ], batch size: 52, lr: 9.68e-03, grad_scale: 8.0 2022-12-23 03:22:57,668 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40171.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:23:00,675 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8505, 1.4821, 2.2258, 1.3305, 2.3505, 2.2987, 1.3766, 2.5194], device='cuda:3'), covar=tensor([0.1072, 0.1929, 0.1108, 0.1673, 0.0613, 0.0899, 0.2241, 0.0463], device='cuda:3'), in_proj_covar=tensor([0.0194, 0.0200, 0.0198, 0.0192, 0.0175, 0.0209, 0.0206, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 03:23:23,301 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3755, 1.2774, 1.4352, 1.2570, 0.8385, 2.2609, 1.0972, 1.4833], device='cuda:3'), covar=tensor([0.3529, 0.2159, 0.2000, 0.2087, 0.1478, 0.0373, 0.1559, 0.0915], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0118, 0.0128, 0.0121, 0.0104, 0.0100, 0.0097, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 03:23:33,976 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40195.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:23:40,952 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0_sp1.1 from training. Duration: 0.836375 2022-12-23 03:24:03,936 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.2780, 1.5434, 1.9101, 0.7999, 1.0823, 2.0002, 1.8000, 1.6343], device='cuda:3'), covar=tensor([0.0779, 0.0374, 0.0271, 0.0398, 0.0397, 0.0415, 0.0221, 0.0588], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0160, 0.0114, 0.0132, 0.0140, 0.0131, 0.0145, 0.0150], device='cuda:3'), out_proj_covar=tensor([1.1220e-04, 1.3074e-04, 9.1794e-05, 1.0520e-04, 1.1220e-04, 1.0775e-04, 1.2037e-04, 1.2198e-04], device='cuda:3') 2022-12-23 03:24:08,540 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40218.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:24:09,532 INFO [train.py:894] (3/4) Epoch 12, batch 1650, loss[loss=0.2243, simple_loss=0.3053, pruned_loss=0.07162, over 18503.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2811, pruned_loss=0.05645, over 3713419.75 frames. ], batch size: 58, lr: 9.67e-03, grad_scale: 8.0 2022-12-23 03:24:23,015 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.557e+02 4.207e+02 4.862e+02 6.254e+02 1.082e+03, threshold=9.723e+02, percent-clipped=1.0 2022-12-23 03:24:23,080 WARNING [train.py:1060] (3/4) Exclude cut with ID 8565-290391-0049-67394-0_sp0.9 from training. Duration: 21.3166875 2022-12-23 03:24:55,313 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0029-104863-0_sp0.9 from training. Duration: 22.1055625 2022-12-23 03:25:01,194 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.5947, 1.8644, 2.2342, 1.0542, 1.3080, 2.3691, 1.9343, 1.7875], device='cuda:3'), covar=tensor([0.0688, 0.0311, 0.0235, 0.0361, 0.0348, 0.0353, 0.0231, 0.0542], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0158, 0.0113, 0.0130, 0.0138, 0.0129, 0.0144, 0.0148], device='cuda:3'), out_proj_covar=tensor([1.1080e-04, 1.2927e-04, 9.0810e-05, 1.0378e-04, 1.1093e-04, 1.0626e-04, 1.1948e-04, 1.2081e-04], device='cuda:3') 2022-12-23 03:25:05,137 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp1.1 from training. Duration: 21.77725 2022-12-23 03:25:05,490 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40256.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:25:24,598 INFO [train.py:894] (3/4) Epoch 12, batch 1700, loss[loss=0.2131, simple_loss=0.2918, pruned_loss=0.06721, over 18722.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2823, pruned_loss=0.0582, over 3712796.70 frames. ], batch size: 52, lr: 9.66e-03, grad_scale: 8.0 2022-12-23 03:25:27,497 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp0.9 from training. Duration: 27.8166875 2022-12-23 03:25:29,479 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40272.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:25:39,593 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40279.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:25:51,868 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp1.1 from training. Duration: 22.5090625 2022-12-23 03:25:59,956 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0 from training. Duration: 25.035 2022-12-23 03:26:04,301 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40295.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 03:26:17,742 WARNING [train.py:1060] (3/4) Exclude cut with ID 774-127930-0014-10412-0_sp1.1 from training. Duration: 0.95 2022-12-23 03:26:36,155 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp0.9 from training. Duration: 0.92225 2022-12-23 03:26:40,286 INFO [train.py:894] (3/4) Epoch 12, batch 1750, loss[loss=0.2121, simple_loss=0.2824, pruned_loss=0.07091, over 18537.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2834, pruned_loss=0.06085, over 3712694.36 frames. ], batch size: 44, lr: 9.66e-03, grad_scale: 8.0 2022-12-23 03:26:53,119 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.694e+02 5.133e+02 6.045e+02 7.323e+02 1.639e+03, threshold=1.209e+03, percent-clipped=7.0 2022-12-23 03:26:56,561 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40330.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:26:59,610 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0 from training. Duration: 21.97 2022-12-23 03:27:01,411 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40333.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:27:17,309 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=40343.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 03:27:20,142 WARNING [train.py:1060] (3/4) Exclude cut with ID 7492-105653-0055-62765-0_sp0.9 from training. Duration: 21.97225 2022-12-23 03:27:21,507 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp0.9 from training. Duration: 25.3333125 2022-12-23 03:27:32,378 WARNING [train.py:1060] (3/4) Exclude cut with ID 5172-29468-0015-19128-0_sp0.9 from training. Duration: 21.5055625 2022-12-23 03:27:42,285 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp1.1 from training. Duration: 20.72725 2022-12-23 03:27:57,162 INFO [train.py:894] (3/4) Epoch 12, batch 1800, loss[loss=0.2124, simple_loss=0.2941, pruned_loss=0.06532, over 18668.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2841, pruned_loss=0.06266, over 3712912.19 frames. ], batch size: 60, lr: 9.65e-03, grad_scale: 8.0 2022-12-23 03:28:16,447 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp0.9 from training. Duration: 26.32775 2022-12-23 03:28:30,915 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40391.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:28:37,347 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1504, 1.8798, 2.4429, 1.3258, 2.3815, 2.5027, 1.6440, 2.7051], device='cuda:3'), covar=tensor([0.1111, 0.1669, 0.1164, 0.1949, 0.0678, 0.1037, 0.2103, 0.0470], device='cuda:3'), in_proj_covar=tensor([0.0195, 0.0199, 0.0200, 0.0193, 0.0176, 0.0210, 0.0206, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 03:28:50,277 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0 from training. Duration: 20.025 2022-12-23 03:28:56,675 WARNING [train.py:1060] (3/4) Exclude cut with ID 6709-74022-0004-86860-0_sp1.1 from training. Duration: 0.9409375 2022-12-23 03:28:56,689 WARNING [train.py:1060] (3/4) Exclude cut with ID 4757-1811-0023-62229-0_sp0.9 from training. Duration: 21.37775 2022-12-23 03:29:12,702 INFO [train.py:894] (3/4) Epoch 12, batch 1850, loss[loss=0.2081, simple_loss=0.2907, pruned_loss=0.06274, over 18621.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2859, pruned_loss=0.06514, over 3711950.96 frames. ], batch size: 53, lr: 9.65e-03, grad_scale: 8.0 2022-12-23 03:29:17,362 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0004-25974-0_sp0.9 from training. Duration: 21.17225 2022-12-23 03:29:17,375 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp0.9 from training. Duration: 27.511125 2022-12-23 03:29:19,113 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40423.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:29:26,479 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.348e+02 4.947e+02 6.136e+02 7.915e+02 1.614e+03, threshold=1.227e+03, percent-clipped=1.0 2022-12-23 03:29:51,072 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0 from training. Duration: 22.8 2022-12-23 03:29:55,033 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0 from training. Duration: 22.585 2022-12-23 03:30:26,414 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0001-146967-0_sp0.9 from training. Duration: 22.0166875 2022-12-23 03:30:27,697 INFO [train.py:894] (3/4) Epoch 12, batch 1900, loss[loss=0.2669, simple_loss=0.3333, pruned_loss=0.1003, over 18674.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2864, pruned_loss=0.06632, over 3712124.62 frames. ], batch size: 78, lr: 9.64e-03, grad_scale: 8.0 2022-12-23 03:30:30,653 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=40471.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:30:30,877 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40471.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:30:42,836 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp1.1 from training. Duration: 24.395375 2022-12-23 03:30:50,558 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp0.9 from training. Duration: 27.47775 2022-12-23 03:30:54,894 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp0.9 from training. Duration: 24.8833125 2022-12-23 03:30:57,738 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0 from training. Duration: 23.39 2022-12-23 03:31:02,729 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40492.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:31:03,909 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp0.9 from training. Duration: 28.72225 2022-12-23 03:31:13,196 WARNING [train.py:1060] (3/4) Exclude cut with ID 585-294811-0110-133686-0_sp0.9 from training. Duration: 20.8944375 2022-12-23 03:31:23,308 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2022-12-23 03:31:26,782 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0_sp0.9 from training. Duration: 23.8444375 2022-12-23 03:31:42,939 INFO [train.py:894] (3/4) Epoch 12, batch 1950, loss[loss=0.1815, simple_loss=0.2597, pruned_loss=0.05164, over 18455.00 frames. ], tot_loss[loss=0.212, simple_loss=0.288, pruned_loss=0.06798, over 3712491.32 frames. ], batch size: 48, lr: 9.64e-03, grad_scale: 8.0 2022-12-23 03:31:43,119 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=40519.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:31:52,524 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0 from training. Duration: 25.85 2022-12-23 03:31:52,535 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0 from training. Duration: 21.39 2022-12-23 03:31:57,002 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.661e+02 5.030e+02 6.516e+02 8.267e+02 2.273e+03, threshold=1.303e+03, percent-clipped=7.0 2022-12-23 03:32:03,769 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0 from training. Duration: 27.92 2022-12-23 03:32:32,049 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0039-130165-0_sp0.9 from training. Duration: 20.661125 2022-12-23 03:32:32,208 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40551.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:32:35,255 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40553.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:32:55,402 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0043-15874-0_sp0.9 from training. Duration: 20.07225 2022-12-23 03:32:58,140 INFO [train.py:894] (3/4) Epoch 12, batch 2000, loss[loss=0.1964, simple_loss=0.2803, pruned_loss=0.05628, over 18458.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2883, pruned_loss=0.06856, over 3712954.07 frames. ], batch size: 54, lr: 9.63e-03, grad_scale: 8.0 2022-12-23 03:33:01,750 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0 from training. Duration: 21.01 2022-12-23 03:33:06,441 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40574.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:33:41,849 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2022-12-23 03:34:11,994 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0 from training. Duration: 20.65 2022-12-23 03:34:14,819 INFO [train.py:894] (3/4) Epoch 12, batch 2050, loss[loss=0.2356, simple_loss=0.3055, pruned_loss=0.0828, over 18525.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2877, pruned_loss=0.06876, over 3713062.30 frames. ], batch size: 97, lr: 9.62e-03, grad_scale: 8.0 2022-12-23 03:34:19,114 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0 from training. Duration: 21.46 2022-12-23 03:34:28,295 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.998e+02 4.881e+02 5.858e+02 6.986e+02 1.392e+03, threshold=1.172e+03, percent-clipped=2.0 2022-12-23 03:34:28,480 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40628.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:34:47,453 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.90 vs. limit=5.0 2022-12-23 03:35:00,390 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0 from training. Duration: 0.92 2022-12-23 03:35:04,184 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40651.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:35:06,954 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0_sp0.9 from training. Duration: 23.7666875 2022-12-23 03:35:32,190 INFO [train.py:894] (3/4) Epoch 12, batch 2100, loss[loss=0.1848, simple_loss=0.258, pruned_loss=0.05581, over 18680.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2881, pruned_loss=0.06924, over 3714455.02 frames. ], batch size: 46, lr: 9.62e-03, grad_scale: 8.0 2022-12-23 03:35:45,367 WARNING [train.py:1060] (3/4) Exclude cut with ID 8544-281189-0060-101339-0_sp0.9 from training. Duration: 20.861125 2022-12-23 03:35:57,733 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0_sp0.9 from training. Duration: 22.711125 2022-12-23 03:35:58,004 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40686.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:36:36,095 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40712.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:36:40,154 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp1.1 from training. Duration: 22.986375 2022-12-23 03:36:45,603 INFO [train.py:894] (3/4) Epoch 12, batch 2150, loss[loss=0.2299, simple_loss=0.296, pruned_loss=0.08188, over 18729.00 frames. ], tot_loss[loss=0.2138, simple_loss=0.2885, pruned_loss=0.0695, over 3715212.74 frames. ], batch size: 50, lr: 9.61e-03, grad_scale: 8.0 2022-12-23 03:36:55,132 WARNING [train.py:1060] (3/4) Exclude cut with ID 8040-260924-0003-80960-0_sp0.9 from training. Duration: 22.07225 2022-12-23 03:36:59,401 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.466e+02 5.161e+02 6.497e+02 7.799e+02 1.996e+03, threshold=1.299e+03, percent-clipped=5.0 2022-12-23 03:37:01,831 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0045-26330-0_sp0.9 from training. Duration: 20.3055625 2022-12-23 03:37:04,779 WARNING [train.py:1060] (3/4) Exclude cut with ID 6356-271890-0060-94317-0_sp0.9 from training. Duration: 20.72225 2022-12-23 03:37:19,997 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp1.1 from training. Duration: 22.4818125 2022-12-23 03:37:44,110 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp0.9 from training. Duration: 25.0944375 2022-12-23 03:37:48,609 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0 from training. Duration: 21.515 2022-12-23 03:37:52,990 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp0.9 from training. Duration: 27.02225 2022-12-23 03:37:59,477 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0010-62480-0_sp0.9 from training. Duration: 22.22225 2022-12-23 03:38:00,568 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2022-12-23 03:38:02,639 INFO [train.py:894] (3/4) Epoch 12, batch 2200, loss[loss=0.2033, simple_loss=0.2812, pruned_loss=0.06271, over 18579.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.2885, pruned_loss=0.07001, over 3715035.87 frames. ], batch size: 51, lr: 9.61e-03, grad_scale: 8.0 2022-12-23 03:38:04,198 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0085-44554-0_sp0.9 from training. Duration: 20.85 2022-12-23 03:38:23,924 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9631, 0.9625, 1.6227, 1.7185, 1.9058, 1.9410, 1.6489, 1.4836], device='cuda:3'), covar=tensor([0.1786, 0.2787, 0.2282, 0.2258, 0.1783, 0.0988, 0.2467, 0.1211], device='cuda:3'), in_proj_covar=tensor([0.0253, 0.0286, 0.0263, 0.0297, 0.0281, 0.0237, 0.0308, 0.0224], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 03:38:32,821 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.1620, 1.8924, 1.9318, 1.0434, 1.5518, 1.9996, 1.8982, 1.6622], device='cuda:3'), covar=tensor([0.0589, 0.0243, 0.0245, 0.0368, 0.0309, 0.0378, 0.0195, 0.0513], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0159, 0.0113, 0.0129, 0.0140, 0.0131, 0.0144, 0.0148], device='cuda:3'), out_proj_covar=tensor([1.1078e-04, 1.2920e-04, 9.1086e-05, 1.0238e-04, 1.1202e-04, 1.0777e-04, 1.1877e-04, 1.2018e-04], device='cuda:3') 2022-12-23 03:38:41,209 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0 from training. Duration: 21.54 2022-12-23 03:38:42,121 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2022-12-23 03:38:44,267 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp1.1 from training. Duration: 20.5318125 2022-12-23 03:38:49,386 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.3235, 1.6232, 1.9157, 0.7275, 1.0263, 2.0044, 1.8070, 1.5941], device='cuda:3'), covar=tensor([0.0613, 0.0244, 0.0256, 0.0317, 0.0375, 0.0380, 0.0190, 0.0501], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0159, 0.0114, 0.0130, 0.0141, 0.0132, 0.0144, 0.0149], device='cuda:3'), out_proj_covar=tensor([1.1133e-04, 1.2969e-04, 9.1384e-05, 1.0296e-04, 1.1274e-04, 1.0850e-04, 1.1935e-04, 1.2093e-04], device='cuda:3') 2022-12-23 03:38:53,641 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0012-134311-0_sp0.9 from training. Duration: 21.9333125 2022-12-23 03:39:19,322 INFO [train.py:894] (3/4) Epoch 12, batch 2250, loss[loss=0.2153, simple_loss=0.2871, pruned_loss=0.07177, over 18519.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2871, pruned_loss=0.06944, over 3714866.27 frames. ], batch size: 47, lr: 9.60e-03, grad_scale: 8.0 2022-12-23 03:39:33,368 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.699e+02 5.281e+02 6.293e+02 7.609e+02 1.404e+03, threshold=1.259e+03, percent-clipped=2.0 2022-12-23 03:39:44,657 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0025-130151-0_sp0.9 from training. Duration: 21.7944375 2022-12-23 03:39:53,763 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.8947, 2.5290, 2.0018, 0.8733, 2.1342, 2.1789, 1.9027, 2.3055], device='cuda:3'), covar=tensor([0.0500, 0.0466, 0.1100, 0.1522, 0.1147, 0.1168, 0.1176, 0.0663], device='cuda:3'), in_proj_covar=tensor([0.0166, 0.0181, 0.0202, 0.0196, 0.0207, 0.0192, 0.0205, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 03:39:56,178 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0_sp0.9 from training. Duration: 22.4666875 2022-12-23 03:40:02,791 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40848.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:40:03,974 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0 from training. Duration: 21.635 2022-12-23 03:40:07,019 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40851.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:40:09,749 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0_sp0.9 from training. Duration: 24.038875 2022-12-23 03:40:28,356 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40865.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:40:34,338 INFO [train.py:894] (3/4) Epoch 12, batch 2300, loss[loss=0.2295, simple_loss=0.311, pruned_loss=0.07406, over 18591.00 frames. ], tot_loss[loss=0.2138, simple_loss=0.2879, pruned_loss=0.06984, over 3715278.63 frames. ], batch size: 56, lr: 9.59e-03, grad_scale: 8.0 2022-12-23 03:40:42,643 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40874.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:40:55,518 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp1.1 from training. Duration: 21.786375 2022-12-23 03:41:03,861 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2022-12-23 03:41:06,425 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0 from training. Duration: 20.22 2022-12-23 03:41:20,692 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=40899.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:41:22,337 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40900.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:41:51,417 INFO [train.py:894] (3/4) Epoch 12, batch 2350, loss[loss=0.2042, simple_loss=0.2755, pruned_loss=0.06645, over 18557.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2891, pruned_loss=0.07059, over 3715310.98 frames. ], batch size: 41, lr: 9.59e-03, grad_scale: 8.0 2022-12-23 03:41:56,410 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=40922.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:42:02,499 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40926.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:42:04,881 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.147e+02 5.320e+02 6.167e+02 7.693e+02 1.883e+03, threshold=1.233e+03, percent-clipped=4.0 2022-12-23 03:42:05,166 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40928.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:42:17,949 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.50 vs. limit=5.0 2022-12-23 03:42:25,435 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8472, 1.8276, 1.3476, 1.9245, 1.9777, 1.7345, 2.5614, 1.9478], device='cuda:3'), covar=tensor([0.0870, 0.1483, 0.2573, 0.1667, 0.1725, 0.0870, 0.0944, 0.1060], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0197, 0.0240, 0.0285, 0.0228, 0.0184, 0.0208, 0.0196], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 03:42:55,887 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40961.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 03:43:05,005 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0 from training. Duration: 25.285 2022-12-23 03:43:07,762 INFO [train.py:894] (3/4) Epoch 12, batch 2400, loss[loss=0.2166, simple_loss=0.2838, pruned_loss=0.07474, over 18573.00 frames. ], tot_loss[loss=0.2138, simple_loss=0.2878, pruned_loss=0.0699, over 3714513.23 frames. ], batch size: 41, lr: 9.58e-03, grad_scale: 8.0 2022-12-23 03:43:17,936 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=40976.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:43:32,663 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40986.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:43:47,669 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40997.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:44:04,193 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41007.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:44:09,807 WARNING [train.py:1060] (3/4) Exclude cut with ID 811-130148-0001-63453-0_sp0.9 from training. Duration: 20.861125 2022-12-23 03:44:21,816 INFO [train.py:894] (3/4) Epoch 12, batch 2450, loss[loss=0.2055, simple_loss=0.2769, pruned_loss=0.06707, over 18562.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2889, pruned_loss=0.07035, over 3715465.11 frames. ], batch size: 49, lr: 9.58e-03, grad_scale: 8.0 2022-12-23 03:44:32,003 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0 from training. Duration: 20.88 2022-12-23 03:44:32,376 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9456, 1.7116, 2.0471, 1.1668, 2.2741, 2.1476, 1.5011, 2.5009], device='cuda:3'), covar=tensor([0.1002, 0.1685, 0.1189, 0.1842, 0.0611, 0.1149, 0.2118, 0.0475], device='cuda:3'), in_proj_covar=tensor([0.0195, 0.0199, 0.0202, 0.0193, 0.0176, 0.0211, 0.0206, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 03:44:34,731 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.360e+02 5.211e+02 6.060e+02 7.407e+02 1.596e+03, threshold=1.212e+03, percent-clipped=4.0 2022-12-23 03:44:44,637 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=41034.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:44:55,421 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3193, 2.6326, 1.5917, 3.0014, 2.8903, 2.2188, 3.5918, 2.3180], device='cuda:3'), covar=tensor([0.0758, 0.1550, 0.2542, 0.1659, 0.1422, 0.0866, 0.0802, 0.1064], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0199, 0.0240, 0.0286, 0.0229, 0.0186, 0.0210, 0.0196], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 03:45:01,422 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41045.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:45:02,385 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0_sp0.9 from training. Duration: 23.4166875 2022-12-23 03:45:21,712 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41058.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:45:38,343 INFO [train.py:894] (3/4) Epoch 12, batch 2500, loss[loss=0.2144, simple_loss=0.2901, pruned_loss=0.06932, over 18478.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2886, pruned_loss=0.07024, over 3714431.75 frames. ], batch size: 54, lr: 9.57e-03, grad_scale: 8.0 2022-12-23 03:46:14,149 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2022-12-23 03:46:19,969 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0 from training. Duration: 21.24 2022-12-23 03:46:19,981 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0_sp0.9 from training. Duration: 23.9055625 2022-12-23 03:46:30,283 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41102.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:46:36,690 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41106.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:46:54,155 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp0.9 from training. Duration: 25.988875 2022-12-23 03:46:55,585 INFO [train.py:894] (3/4) Epoch 12, batch 2550, loss[loss=0.1985, simple_loss=0.2715, pruned_loss=0.06276, over 18555.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2874, pruned_loss=0.06984, over 3714593.36 frames. ], batch size: 47, lr: 9.57e-03, grad_scale: 8.0 2022-12-23 03:47:03,502 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0001-134300-0_sp0.9 from training. Duration: 20.67225 2022-12-23 03:47:03,894 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5555, 1.4200, 1.2483, 0.7632, 1.8165, 1.4794, 1.4347, 1.2317], device='cuda:3'), covar=tensor([0.0353, 0.0457, 0.0516, 0.0701, 0.0278, 0.0348, 0.0458, 0.0850], device='cuda:3'), in_proj_covar=tensor([0.0119, 0.0119, 0.0126, 0.0119, 0.0088, 0.0115, 0.0133, 0.0150], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 03:47:07,148 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.9177, 2.4226, 1.9014, 0.8827, 2.1249, 2.2234, 1.5244, 2.0588], device='cuda:3'), covar=tensor([0.0568, 0.0650, 0.1311, 0.1799, 0.1350, 0.1395, 0.1852, 0.0880], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0182, 0.0203, 0.0197, 0.0208, 0.0193, 0.0209, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 03:47:09,335 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.299e+02 5.132e+02 6.345e+02 7.897e+02 1.932e+03, threshold=1.269e+03, percent-clipped=6.0 2022-12-23 03:47:39,659 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41148.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:47:50,024 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0 from training. Duration: 20.34 2022-12-23 03:48:01,664 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41163.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:48:10,560 INFO [train.py:894] (3/4) Epoch 12, batch 2600, loss[loss=0.1762, simple_loss=0.2581, pruned_loss=0.04717, over 18584.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.2883, pruned_loss=0.0702, over 3714443.47 frames. ], batch size: 51, lr: 9.56e-03, grad_scale: 8.0 2022-12-23 03:48:52,565 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=41196.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:49:02,388 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp0.9 from training. Duration: 25.061125 2022-12-23 03:49:12,624 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0 from training. Duration: 0.83 2022-12-23 03:49:26,844 INFO [train.py:894] (3/4) Epoch 12, batch 2650, loss[loss=0.2066, simple_loss=0.2838, pruned_loss=0.06476, over 18547.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2866, pruned_loss=0.06902, over 3714853.96 frames. ], batch size: 99, lr: 9.55e-03, grad_scale: 8.0 2022-12-23 03:49:30,186 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41221.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:49:38,834 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0 from training. Duration: 24.73 2022-12-23 03:49:40,119 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.597e+02 5.175e+02 6.585e+02 8.509e+02 1.944e+03, threshold=1.317e+03, percent-clipped=3.0 2022-12-23 03:49:52,605 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0 from training. Duration: 23.965 2022-12-23 03:50:00,559 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0030-146996-0_sp0.9 from training. Duration: 22.088875 2022-12-23 03:50:18,624 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0_sp0.9 from training. Duration: 23.6 2022-12-23 03:50:23,128 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41256.0, num_to_drop=1, layers_to_drop={3} 2022-12-23 03:50:42,834 INFO [train.py:894] (3/4) Epoch 12, batch 2700, loss[loss=0.206, simple_loss=0.29, pruned_loss=0.06102, over 18676.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2862, pruned_loss=0.06849, over 3714993.65 frames. ], batch size: 62, lr: 9.55e-03, grad_scale: 8.0 2022-12-23 03:51:41,055 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41307.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:51:59,146 INFO [train.py:894] (3/4) Epoch 12, batch 2750, loss[loss=0.2114, simple_loss=0.2892, pruned_loss=0.06677, over 18706.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2867, pruned_loss=0.06837, over 3716067.79 frames. ], batch size: 50, lr: 9.54e-03, grad_scale: 8.0 2022-12-23 03:52:00,688 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0 from training. Duration: 23.795 2022-12-23 03:52:12,939 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.607e+02 4.738e+02 5.981e+02 8.070e+02 1.875e+03, threshold=1.196e+03, percent-clipped=2.0 2022-12-23 03:52:18,620 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp1.1 from training. Duration: 21.5409375 2022-12-23 03:52:20,350 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp0.9 from training. Duration: 24.97775 2022-12-23 03:52:32,166 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0_sp0.9 from training. Duration: 23.3444375 2022-12-23 03:52:50,876 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41353.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:52:54,339 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=41355.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:53:00,195 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0_sp0.9 from training. Duration: 23.2 2022-12-23 03:53:05,889 WARNING [train.py:1060] (3/4) Exclude cut with ID 5653-46179-0060-117930-0_sp0.9 from training. Duration: 21.17225 2022-12-23 03:53:14,851 INFO [train.py:894] (3/4) Epoch 12, batch 2800, loss[loss=0.186, simple_loss=0.2611, pruned_loss=0.05541, over 18426.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2864, pruned_loss=0.0683, over 3717208.12 frames. ], batch size: 42, lr: 9.54e-03, grad_scale: 8.0 2022-12-23 03:53:25,418 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp0.9 from training. Duration: 24.6555625 2022-12-23 03:54:04,879 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41401.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:54:21,598 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0 from training. Duration: 20.44 2022-12-23 03:54:32,418 INFO [train.py:894] (3/4) Epoch 12, batch 2850, loss[loss=0.2109, simple_loss=0.285, pruned_loss=0.06845, over 18701.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2848, pruned_loss=0.06776, over 3716387.27 frames. ], batch size: 60, lr: 9.53e-03, grad_scale: 8.0 2022-12-23 03:54:36,911 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0_sp0.9 from training. Duration: 23.45 2022-12-23 03:54:46,857 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.602e+02 5.248e+02 6.316e+02 8.103e+02 1.885e+03, threshold=1.263e+03, percent-clipped=5.0 2022-12-23 03:55:04,792 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5025, 1.3521, 1.4456, 1.4003, 0.9366, 2.2417, 0.9735, 1.3867], device='cuda:3'), covar=tensor([0.3089, 0.2027, 0.1911, 0.1873, 0.1267, 0.0359, 0.1442, 0.0858], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0115, 0.0127, 0.0120, 0.0102, 0.0099, 0.0095, 0.0091], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 03:55:06,133 WARNING [train.py:1060] (3/4) Exclude cut with ID 6945-60535-0076-12784-0_sp0.9 from training. Duration: 20.52225 2022-12-23 03:55:14,440 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0 from training. Duration: 22.19 2022-12-23 03:55:24,799 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp1.1 from training. Duration: 25.3818125 2022-12-23 03:55:31,790 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41458.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:55:41,451 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp0.9 from training. Duration: 28.0944375 2022-12-23 03:55:47,921 INFO [train.py:894] (3/4) Epoch 12, batch 2900, loss[loss=0.2119, simple_loss=0.2887, pruned_loss=0.06751, over 18458.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2842, pruned_loss=0.06753, over 3715410.65 frames. ], batch size: 54, lr: 9.53e-03, grad_scale: 8.0 2022-12-23 03:55:47,960 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0_sp0.9 from training. Duration: 22.9444375 2022-12-23 03:55:56,787 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp1.1 from training. Duration: 21.6318125 2022-12-23 03:56:03,570 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2022-12-23 03:56:13,029 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0 from training. Duration: 23.695 2022-12-23 03:56:23,764 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8967, 1.8659, 1.8389, 1.8303, 1.4263, 5.0234, 2.1171, 2.5347], device='cuda:3'), covar=tensor([0.3728, 0.2201, 0.2070, 0.2071, 0.1380, 0.0117, 0.1393, 0.0877], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0116, 0.0127, 0.0120, 0.0102, 0.0099, 0.0095, 0.0091], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 03:56:38,850 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0 from training. Duration: 23.955 2022-12-23 03:57:03,507 INFO [train.py:894] (3/4) Epoch 12, batch 2950, loss[loss=0.2316, simple_loss=0.3077, pruned_loss=0.07773, over 18707.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2854, pruned_loss=0.06797, over 3714652.16 frames. ], batch size: 60, lr: 9.52e-03, grad_scale: 8.0 2022-12-23 03:57:07,484 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41521.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:57:12,781 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp0.9 from training. Duration: 26.438875 2022-12-23 03:57:17,825 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.973e+02 4.599e+02 6.243e+02 8.139e+02 2.376e+03, threshold=1.249e+03, percent-clipped=3.0 2022-12-23 03:57:57,841 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0021-26306-0_sp0.9 from training. Duration: 21.2444375 2022-12-23 03:57:59,360 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp0.9 from training. Duration: 31.02225 2022-12-23 03:57:59,547 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41556.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 03:58:07,182 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0 from training. Duration: 22.395 2022-12-23 03:58:20,143 INFO [train.py:894] (3/4) Epoch 12, batch 3000, loss[loss=0.2241, simple_loss=0.2994, pruned_loss=0.07438, over 18463.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2853, pruned_loss=0.06782, over 3714733.13 frames. ], batch size: 54, lr: 9.51e-03, grad_scale: 8.0 2022-12-23 03:58:20,143 INFO [train.py:919] (3/4) Computing validation loss 2022-12-23 03:58:31,130 INFO [train.py:928] (3/4) Epoch 12, validation: loss=0.1703, simple_loss=0.2698, pruned_loss=0.03541, over 944034.00 frames. 2022-12-23 03:58:31,131 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24676MB 2022-12-23 03:58:31,281 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=41569.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:58:35,615 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0 from training. Duration: 21.075 2022-12-23 03:58:36,058 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3671, 2.6969, 3.0029, 0.8816, 2.5536, 3.5720, 2.4048, 2.8108], device='cuda:3'), covar=tensor([0.0782, 0.0423, 0.0326, 0.0492, 0.0425, 0.0257, 0.0355, 0.0567], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0159, 0.0114, 0.0131, 0.0140, 0.0132, 0.0148, 0.0150], device='cuda:3'), out_proj_covar=tensor([1.1152e-04, 1.2966e-04, 9.1631e-05, 1.0311e-04, 1.1146e-04, 1.0834e-04, 1.2195e-04, 1.2127e-04], device='cuda:3') 2022-12-23 03:58:40,111 WARNING [train.py:1060] (3/4) Exclude cut with ID 6482-98857-0025-147532-0_sp0.9 from training. Duration: 20.0055625 2022-12-23 03:58:41,474 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0037-132304-0_sp0.9 from training. Duration: 22.05 2022-12-23 03:58:41,487 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0 from training. Duration: 26.8349375 2022-12-23 03:58:44,308 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp1.1 from training. Duration: 22.1090625 2022-12-23 03:58:51,894 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp0.9 from training. Duration: 26.6166875 2022-12-23 03:59:08,972 WARNING [train.py:1060] (3/4) Exclude cut with ID 2046-178027-0000-53705-0_sp0.9 from training. Duration: 20.3055625 2022-12-23 03:59:22,304 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.9872, 5.2433, 4.7809, 2.4639, 5.1206, 4.0198, 0.6793, 3.3982], device='cuda:3'), covar=tensor([0.1813, 0.0827, 0.1159, 0.3144, 0.0719, 0.0732, 0.5320, 0.1263], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0128, 0.0150, 0.0121, 0.0129, 0.0107, 0.0142, 0.0110], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 03:59:25,864 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=41604.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 03:59:32,135 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1068, 1.8556, 2.0348, 2.4688, 2.3001, 4.1103, 1.7140, 2.0037], device='cuda:3'), covar=tensor([0.0824, 0.1579, 0.1038, 0.0806, 0.1171, 0.0210, 0.1316, 0.1329], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0082, 0.0073, 0.0074, 0.0090, 0.0072, 0.0085, 0.0076], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 03:59:33,362 WARNING [train.py:1060] (3/4) Exclude cut with ID 7205-50138-0008-5373-0_sp0.9 from training. Duration: 20.7 2022-12-23 03:59:48,388 INFO [train.py:894] (3/4) Epoch 12, batch 3050, loss[loss=0.239, simple_loss=0.3136, pruned_loss=0.08217, over 18710.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2858, pruned_loss=0.06783, over 3715046.86 frames. ], batch size: 65, lr: 9.51e-03, grad_scale: 8.0 2022-12-23 04:00:01,463 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.194e+02 4.690e+02 5.641e+02 7.048e+02 1.458e+03, threshold=1.128e+03, percent-clipped=1.0 2022-12-23 04:00:16,407 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0 from training. Duration: 22.48 2022-12-23 04:00:33,408 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp0.9 from training. Duration: 29.816625 2022-12-23 04:00:40,209 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41653.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:00:41,843 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41654.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:00:52,026 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7613, 1.4757, 1.5478, 2.1359, 1.7397, 3.6196, 1.2957, 1.6242], device='cuda:3'), covar=tensor([0.0839, 0.1796, 0.1077, 0.0849, 0.1422, 0.0217, 0.1520, 0.1517], device='cuda:3'), in_proj_covar=tensor([0.0072, 0.0081, 0.0072, 0.0073, 0.0090, 0.0072, 0.0085, 0.0076], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 04:00:53,242 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp1.1 from training. Duration: 22.7590625 2022-12-23 04:00:55,389 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4113, 1.6729, 1.3392, 2.0189, 2.1226, 1.4877, 1.1946, 1.1918], device='cuda:3'), covar=tensor([0.2163, 0.1816, 0.1790, 0.1058, 0.1282, 0.1271, 0.2220, 0.1677], device='cuda:3'), in_proj_covar=tensor([0.0239, 0.0218, 0.0203, 0.0188, 0.0257, 0.0190, 0.0212, 0.0189], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 04:00:59,635 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0 from training. Duration: 22.555 2022-12-23 04:01:03,868 INFO [train.py:894] (3/4) Epoch 12, batch 3100, loss[loss=0.175, simple_loss=0.2579, pruned_loss=0.04608, over 18677.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2849, pruned_loss=0.06691, over 3714631.01 frames. ], batch size: 50, lr: 9.50e-03, grad_scale: 8.0 2022-12-23 04:01:19,173 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.2288, 1.0421, 0.8126, 0.2739, 0.8840, 1.0646, 0.8941, 1.0597], device='cuda:3'), covar=tensor([0.0437, 0.0372, 0.0689, 0.1073, 0.0762, 0.0976, 0.1062, 0.0442], device='cuda:3'), in_proj_covar=tensor([0.0165, 0.0182, 0.0203, 0.0196, 0.0206, 0.0194, 0.0208, 0.0194], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 04:01:20,266 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0005-25975-0_sp0.9 from training. Duration: 21.688875 2022-12-23 04:01:52,846 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=41701.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:01:52,965 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41701.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:01:55,655 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0_sp0.9 from training. Duration: 22.6 2022-12-23 04:02:13,989 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41715.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 04:02:19,244 INFO [train.py:894] (3/4) Epoch 12, batch 3150, loss[loss=0.228, simple_loss=0.2997, pruned_loss=0.07816, over 18453.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2849, pruned_loss=0.06676, over 3714457.07 frames. ], batch size: 54, lr: 9.50e-03, grad_scale: 16.0 2022-12-23 04:02:32,305 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.643e+02 4.988e+02 6.154e+02 7.740e+02 1.353e+03, threshold=1.231e+03, percent-clipped=2.0 2022-12-23 04:02:33,810 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0 from training. Duration: 24.32 2022-12-23 04:03:05,274 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=41749.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:03:19,906 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41758.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:03:29,812 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-276745-0093-13116-0_sp0.9 from training. Duration: 21.061125 2022-12-23 04:03:34,818 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7114, 1.6488, 1.2844, 1.6869, 1.7924, 1.5585, 2.0803, 1.7932], device='cuda:3'), covar=tensor([0.0922, 0.1501, 0.2547, 0.1510, 0.1777, 0.0893, 0.0962, 0.1129], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0197, 0.0240, 0.0285, 0.0228, 0.0185, 0.0208, 0.0194], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 04:03:35,748 INFO [train.py:894] (3/4) Epoch 12, batch 3200, loss[loss=0.2206, simple_loss=0.2877, pruned_loss=0.07677, over 18575.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2855, pruned_loss=0.06708, over 3713699.79 frames. ], batch size: 49, lr: 9.49e-03, grad_scale: 16.0 2022-12-23 04:03:43,574 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0024-15855-0_sp0.9 from training. Duration: 20.32225 2022-12-23 04:03:55,554 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp1.1 from training. Duration: 0.7545625 2022-12-23 04:04:11,622 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0_sp0.9 from training. Duration: 23.9333125 2022-12-23 04:04:34,193 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=41806.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:04:39,394 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2022-12-23 04:04:44,452 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp1.1 from training. Duration: 20.17275 2022-12-23 04:04:48,908 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp1.1 from training. Duration: 20.436375 2022-12-23 04:04:53,940 INFO [train.py:894] (3/4) Epoch 12, batch 3250, loss[loss=0.1946, simple_loss=0.2793, pruned_loss=0.05493, over 18519.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2858, pruned_loss=0.06685, over 3714057.03 frames. ], batch size: 52, lr: 9.49e-03, grad_scale: 16.0 2022-12-23 04:05:07,766 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.427e+02 5.153e+02 6.270e+02 8.432e+02 2.144e+03, threshold=1.254e+03, percent-clipped=7.0 2022-12-23 04:06:08,335 INFO [train.py:894] (3/4) Epoch 12, batch 3300, loss[loss=0.2216, simple_loss=0.301, pruned_loss=0.0711, over 18578.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2862, pruned_loss=0.06731, over 3714321.47 frames. ], batch size: 56, lr: 9.48e-03, grad_scale: 16.0 2022-12-23 04:06:08,396 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0_sp0.9 from training. Duration: 23.1055625 2022-12-23 04:06:09,873 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp1.1 from training. Duration: 23.5 2022-12-23 04:06:20,383 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp0.9 from training. Duration: 26.62775 2022-12-23 04:06:34,516 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0 from training. Duration: 21.105 2022-12-23 04:06:37,493 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0_sp0.9 from training. Duration: 24.411125 2022-12-23 04:07:04,329 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp1.1 from training. Duration: 21.263625 2022-12-23 04:07:04,693 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41906.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:07:24,852 INFO [train.py:894] (3/4) Epoch 12, batch 3350, loss[loss=0.1473, simple_loss=0.2269, pruned_loss=0.03383, over 18485.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2864, pruned_loss=0.06762, over 3713814.08 frames. ], batch size: 43, lr: 9.48e-03, grad_scale: 16.0 2022-12-23 04:07:35,278 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5975, 3.8851, 3.8083, 1.7184, 3.8632, 2.7772, 0.6800, 2.5510], device='cuda:3'), covar=tensor([0.2007, 0.1117, 0.1410, 0.3667, 0.0935, 0.1075, 0.5152, 0.1672], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0129, 0.0151, 0.0120, 0.0130, 0.0106, 0.0141, 0.0111], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 04:07:37,876 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.521e+02 5.179e+02 6.368e+02 8.514e+02 1.589e+03, threshold=1.274e+03, percent-clipped=3.0 2022-12-23 04:07:37,972 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0 from training. Duration: 20.795 2022-12-23 04:07:49,282 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0 from training. Duration: 24.76 2022-12-23 04:07:49,297 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0_sp0.9 from training. Duration: 22.25 2022-12-23 04:08:08,795 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41948.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 04:08:14,601 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp1.1 from training. Duration: 20.5045625 2022-12-23 04:08:37,806 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41967.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:08:40,235 INFO [train.py:894] (3/4) Epoch 12, batch 3400, loss[loss=0.2065, simple_loss=0.2898, pruned_loss=0.0616, over 18622.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2862, pruned_loss=0.06764, over 3714199.62 frames. ], batch size: 53, lr: 9.47e-03, grad_scale: 16.0 2022-12-23 04:08:48,655 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2022-12-23 04:09:31,184 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2022-12-23 04:09:42,051 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42009.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 04:09:43,231 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42010.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 04:09:56,441 INFO [train.py:894] (3/4) Epoch 12, batch 3450, loss[loss=0.1996, simple_loss=0.2728, pruned_loss=0.06323, over 18538.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2862, pruned_loss=0.06747, over 3714473.39 frames. ], batch size: 47, lr: 9.46e-03, grad_scale: 16.0 2022-12-23 04:10:09,836 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.367e+02 5.696e+02 6.525e+02 7.684e+02 1.344e+03, threshold=1.305e+03, percent-clipped=1.0 2022-12-23 04:11:10,035 INFO [train.py:894] (3/4) Epoch 12, batch 3500, loss[loss=0.2192, simple_loss=0.2982, pruned_loss=0.07013, over 18661.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2861, pruned_loss=0.06732, over 3714436.26 frames. ], batch size: 99, lr: 9.46e-03, grad_scale: 16.0 2022-12-23 04:11:31,517 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp1.1 from training. Duration: 22.2954375 2022-12-23 04:11:42,001 INFO [train.py:894] (3/4) Epoch 13, batch 0, loss[loss=0.2114, simple_loss=0.2955, pruned_loss=0.06368, over 18588.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.2955, pruned_loss=0.06368, over 18588.00 frames. ], batch size: 51, lr: 9.09e-03, grad_scale: 16.0 2022-12-23 04:11:42,002 INFO [train.py:919] (3/4) Computing validation loss 2022-12-23 04:11:52,899 INFO [train.py:928] (3/4) Epoch 13, validation: loss=0.1708, simple_loss=0.27, pruned_loss=0.03586, over 944034.00 frames. 2022-12-23 04:11:52,900 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24676MB 2022-12-23 04:12:44,163 WARNING [train.py:1060] (3/4) Exclude cut with ID 298-126791-0067-24026-0_sp0.9 from training. Duration: 21.438875 2022-12-23 04:12:50,515 WARNING [train.py:1060] (3/4) Exclude cut with ID 5652-39938-0025-23684-0_sp0.9 from training. Duration: 22.2055625 2022-12-23 04:13:10,458 INFO [train.py:894] (3/4) Epoch 13, batch 50, loss[loss=0.1986, simple_loss=0.2829, pruned_loss=0.05712, over 18469.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2815, pruned_loss=0.05614, over 837253.23 frames. ], batch size: 54, lr: 9.08e-03, grad_scale: 16.0 2022-12-23 04:13:14,626 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.917e+02 4.264e+02 5.110e+02 6.496e+02 1.876e+03, threshold=1.022e+03, percent-clipped=1.0 2022-12-23 04:14:19,148 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.05 vs. limit=5.0 2022-12-23 04:14:20,413 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2022-12-23 04:14:26,126 INFO [train.py:894] (3/4) Epoch 13, batch 100, loss[loss=0.2455, simple_loss=0.3225, pruned_loss=0.08425, over 18712.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2785, pruned_loss=0.05477, over 1475100.62 frames. ], batch size: 62, lr: 9.08e-03, grad_scale: 16.0 2022-12-23 04:15:26,335 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5519, 3.7170, 3.5910, 1.5769, 3.7880, 2.6242, 0.8964, 2.4641], device='cuda:3'), covar=tensor([0.1947, 0.1015, 0.1417, 0.3720, 0.0792, 0.1037, 0.4713, 0.1620], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0127, 0.0147, 0.0119, 0.0128, 0.0106, 0.0137, 0.0110], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 04:15:42,616 INFO [train.py:894] (3/4) Epoch 13, batch 150, loss[loss=0.1708, simple_loss=0.2522, pruned_loss=0.04471, over 18524.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2746, pruned_loss=0.05297, over 1970583.91 frames. ], batch size: 44, lr: 9.07e-03, grad_scale: 16.0 2022-12-23 04:15:47,722 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.470e+02 3.667e+02 4.662e+02 5.480e+02 1.007e+03, threshold=9.324e+02, percent-clipped=0.0 2022-12-23 04:15:52,214 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0 from training. Duration: 24.525 2022-12-23 04:16:27,789 WARNING [train.py:1060] (3/4) Exclude cut with ID 3699-47246-0007-3408-0_sp0.9 from training. Duration: 20.26675 2022-12-23 04:16:37,776 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42262.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:16:41,720 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp0.9 from training. Duration: 27.25 2022-12-23 04:16:56,934 INFO [train.py:894] (3/4) Epoch 13, batch 200, loss[loss=0.1872, simple_loss=0.2759, pruned_loss=0.04924, over 18454.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2746, pruned_loss=0.05263, over 2356778.23 frames. ], batch size: 50, lr: 9.07e-03, grad_scale: 16.0 2022-12-23 04:17:41,283 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42304.0, num_to_drop=1, layers_to_drop={3} 2022-12-23 04:17:50,546 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42310.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 04:17:55,012 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0 from training. Duration: 21.68 2022-12-23 04:18:07,834 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0 from training. Duration: 21.6300625 2022-12-23 04:18:13,377 INFO [train.py:894] (3/4) Epoch 13, batch 250, loss[loss=0.1548, simple_loss=0.2408, pruned_loss=0.03435, over 18386.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2734, pruned_loss=0.05208, over 2657044.54 frames. ], batch size: 46, lr: 9.06e-03, grad_scale: 16.0 2022-12-23 04:18:18,286 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.488e+02 3.659e+02 4.624e+02 5.942e+02 1.816e+03, threshold=9.248e+02, percent-clipped=5.0 2022-12-23 04:18:32,856 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0_sp0.9 from training. Duration: 24.033375 2022-12-23 04:18:55,499 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=5.09 vs. limit=5.0 2022-12-23 04:19:02,817 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=42358.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:19:29,153 INFO [train.py:894] (3/4) Epoch 13, batch 300, loss[loss=0.2046, simple_loss=0.2946, pruned_loss=0.05728, over 18532.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2733, pruned_loss=0.05168, over 2891280.27 frames. ], batch size: 55, lr: 9.06e-03, grad_scale: 8.0 2022-12-23 04:19:29,209 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0 from training. Duration: 22.905 2022-12-23 04:19:30,637 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp1.1 from training. Duration: 23.4318125 2022-12-23 04:19:36,361 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42379.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:19:57,588 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2022-12-23 04:20:46,373 INFO [train.py:894] (3/4) Epoch 13, batch 350, loss[loss=0.2034, simple_loss=0.291, pruned_loss=0.05784, over 18586.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2754, pruned_loss=0.05264, over 3075213.58 frames. ], batch size: 57, lr: 9.05e-03, grad_scale: 8.0 2022-12-23 04:20:52,408 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.737e+02 3.882e+02 4.906e+02 5.915e+02 1.901e+03, threshold=9.812e+02, percent-clipped=4.0 2022-12-23 04:21:09,394 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42440.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:21:27,903 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp1.1 from training. Duration: 20.82275 2022-12-23 04:21:29,287 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp0.9 from training. Duration: 25.45 2022-12-23 04:21:33,827 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.8292, 1.9497, 0.9823, 2.4389, 2.9835, 1.8110, 2.5194, 2.7368], device='cuda:3'), covar=tensor([0.1287, 0.1855, 0.2441, 0.1183, 0.1228, 0.1436, 0.1196, 0.1326], device='cuda:3'), in_proj_covar=tensor([0.0092, 0.0100, 0.0119, 0.0096, 0.0114, 0.0090, 0.0098, 0.0095], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 04:22:02,672 INFO [train.py:894] (3/4) Epoch 13, batch 400, loss[loss=0.191, simple_loss=0.2697, pruned_loss=0.05618, over 18704.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2759, pruned_loss=0.05312, over 3216810.64 frames. ], batch size: 46, lr: 9.05e-03, grad_scale: 8.0 2022-12-23 04:22:28,963 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3348, 1.1103, 1.6790, 2.6122, 1.9319, 2.1418, 0.7007, 1.7407], device='cuda:3'), covar=tensor([0.1915, 0.1721, 0.1437, 0.0589, 0.1097, 0.1253, 0.2229, 0.1303], device='cuda:3'), in_proj_covar=tensor([0.0102, 0.0115, 0.0132, 0.0132, 0.0104, 0.0133, 0.0129, 0.0109], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 04:22:30,153 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0 from training. Duration: 25.775 2022-12-23 04:22:48,582 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42506.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:22:52,803 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0_sp0.9 from training. Duration: 22.25 2022-12-23 04:23:17,537 INFO [train.py:894] (3/4) Epoch 13, batch 450, loss[loss=0.2002, simple_loss=0.2975, pruned_loss=0.05144, over 18465.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2761, pruned_loss=0.05329, over 3326569.53 frames. ], batch size: 54, lr: 9.04e-03, grad_scale: 8.0 2022-12-23 04:23:21,600 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0 from training. Duration: 26.205 2022-12-23 04:23:22,903 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.137e+02 3.894e+02 4.578e+02 5.727e+02 1.078e+03, threshold=9.156e+02, percent-clipped=2.0 2022-12-23 04:23:39,310 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp0.9 from training. Duration: 30.1555625 2022-12-23 04:23:44,940 WARNING [train.py:1060] (3/4) Exclude cut with ID 1265-135635-0050-6781-0_sp0.9 from training. Duration: 21.8333125 2022-12-23 04:23:54,432 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp1.1 from training. Duration: 20.6545625 2022-12-23 04:24:13,051 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42562.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:24:21,386 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42567.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:24:34,375 INFO [train.py:894] (3/4) Epoch 13, batch 500, loss[loss=0.1893, simple_loss=0.269, pruned_loss=0.0548, over 18681.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2759, pruned_loss=0.05358, over 3412547.77 frames. ], batch size: 48, lr: 9.04e-03, grad_scale: 8.0 2022-12-23 04:24:34,476 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0045-39920-0_sp0.9 from training. Duration: 20.52225 2022-12-23 04:24:56,190 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp0.9 from training. Duration: 29.1166875 2022-12-23 04:25:17,490 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42604.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 04:25:26,268 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=42610.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:25:49,351 INFO [train.py:894] (3/4) Epoch 13, batch 550, loss[loss=0.183, simple_loss=0.2711, pruned_loss=0.04745, over 18718.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2771, pruned_loss=0.05393, over 3478464.36 frames. ], batch size: 52, lr: 9.03e-03, grad_scale: 8.0 2022-12-23 04:25:55,203 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.626e+02 4.162e+02 5.238e+02 6.247e+02 1.388e+03, threshold=1.048e+03, percent-clipped=3.0 2022-12-23 04:25:55,247 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133211-0007-59831-0_sp0.9 from training. Duration: 21.388875 2022-12-23 04:26:24,821 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42649.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:26:29,072 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=42652.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 04:26:30,290 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0 from training. Duration: 22.72 2022-12-23 04:26:32,051 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0_sp0.9 from training. Duration: 22.7444375 2022-12-23 04:26:52,074 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1866, 1.9441, 1.4668, 2.0013, 1.7161, 1.9032, 1.7821, 2.1829], device='cuda:3'), covar=tensor([0.1983, 0.2937, 0.1717, 0.2654, 0.2971, 0.0955, 0.2754, 0.0813], device='cuda:3'), in_proj_covar=tensor([0.0278, 0.0274, 0.0231, 0.0343, 0.0256, 0.0217, 0.0269, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 04:26:52,265 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2022-12-23 04:27:04,013 INFO [train.py:894] (3/4) Epoch 13, batch 600, loss[loss=0.2088, simple_loss=0.2909, pruned_loss=0.0634, over 18447.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2787, pruned_loss=0.05419, over 3531994.80 frames. ], batch size: 50, lr: 9.03e-03, grad_scale: 8.0 2022-12-23 04:27:13,843 WARNING [train.py:1060] (3/4) Exclude cut with ID 4133-6541-0027-40495-0_sp1.1 from training. Duration: 0.9681875 2022-12-23 04:27:18,142 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0_sp0.9 from training. Duration: 22.3166875 2022-12-23 04:27:22,745 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4936, 1.5020, 1.0668, 1.5831, 1.7979, 2.9577, 1.4188, 1.5197], device='cuda:3'), covar=tensor([0.0858, 0.1649, 0.1189, 0.0896, 0.1254, 0.0228, 0.1284, 0.1419], device='cuda:3'), in_proj_covar=tensor([0.0072, 0.0082, 0.0074, 0.0073, 0.0091, 0.0072, 0.0084, 0.0075], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 04:27:24,085 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133212-0015-59917-0_sp0.9 from training. Duration: 21.8166875 2022-12-23 04:27:47,655 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2022-12-23 04:27:56,182 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42710.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:28:19,776 INFO [train.py:894] (3/4) Epoch 13, batch 650, loss[loss=0.1905, simple_loss=0.2808, pruned_loss=0.0501, over 18529.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2796, pruned_loss=0.05486, over 3573201.78 frames. ], batch size: 55, lr: 9.02e-03, grad_scale: 8.0 2022-12-23 04:28:25,612 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.822e+02 4.139e+02 5.026e+02 6.152e+02 1.523e+03, threshold=1.005e+03, percent-clipped=4.0 2022-12-23 04:28:34,838 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42735.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:29:07,434 WARNING [train.py:1060] (3/4) Exclude cut with ID 4957-30119-0041-23990-0_sp0.9 from training. Duration: 20.22775 2022-12-23 04:29:35,750 INFO [train.py:894] (3/4) Epoch 13, batch 700, loss[loss=0.1769, simple_loss=0.2529, pruned_loss=0.05046, over 18418.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.28, pruned_loss=0.05514, over 3603811.16 frames. ], batch size: 42, lr: 9.01e-03, grad_scale: 8.0 2022-12-23 04:29:53,022 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp1.1 from training. Duration: 24.67275 2022-12-23 04:30:22,235 WARNING [train.py:1060] (3/4) Exclude cut with ID 3082-165428-0081-50734-0_sp0.9 from training. Duration: 21.8055625 2022-12-23 04:30:23,070 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2022-12-23 04:30:51,743 INFO [train.py:894] (3/4) Epoch 13, batch 750, loss[loss=0.2023, simple_loss=0.2942, pruned_loss=0.05523, over 18713.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.2804, pruned_loss=0.05488, over 3628369.88 frames. ], batch size: 65, lr: 9.01e-03, grad_scale: 8.0 2022-12-23 04:30:57,839 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.780e+02 4.039e+02 4.835e+02 6.109e+02 2.370e+03, threshold=9.671e+02, percent-clipped=2.0 2022-12-23 04:31:00,797 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0_sp0.9 from training. Duration: 22.6666875 2022-12-23 04:31:12,439 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2022-12-23 04:31:43,098 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42859.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:31:47,247 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42862.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:31:59,961 WARNING [train.py:1060] (3/4) Exclude cut with ID 2411-132532-0017-82279-0_sp1.1 from training. Duration: 0.9681875 2022-12-23 04:32:07,703 INFO [train.py:894] (3/4) Epoch 13, batch 800, loss[loss=0.2165, simple_loss=0.3163, pruned_loss=0.0583, over 18560.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2806, pruned_loss=0.0553, over 3646904.14 frames. ], batch size: 56, lr: 9.00e-03, grad_scale: 8.0 2022-12-23 04:32:14,063 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42879.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:32:20,296 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2884, 2.0277, 1.5623, 0.6073, 1.4334, 1.9315, 1.6627, 1.8186], device='cuda:3'), covar=tensor([0.0555, 0.0417, 0.0922, 0.1413, 0.1069, 0.1258, 0.1345, 0.0547], device='cuda:3'), in_proj_covar=tensor([0.0165, 0.0179, 0.0199, 0.0193, 0.0206, 0.0191, 0.0204, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 04:32:24,559 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0 from training. Duration: 22.485 2022-12-23 04:33:02,839 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp1.1 from training. Duration: 23.82275 2022-12-23 04:33:15,016 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42920.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:33:16,091 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0 from training. Duration: 20.77 2022-12-23 04:33:21,780 INFO [train.py:894] (3/4) Epoch 13, batch 850, loss[loss=0.2046, simple_loss=0.291, pruned_loss=0.05907, over 18590.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.2809, pruned_loss=0.05543, over 3661247.45 frames. ], batch size: 57, lr: 9.00e-03, grad_scale: 8.0 2022-12-23 04:33:23,450 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0_sp0.9 from training. Duration: 24.088875 2022-12-23 04:33:27,849 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.652e+02 4.085e+02 4.833e+02 5.843e+02 1.315e+03, threshold=9.667e+02, percent-clipped=6.0 2022-12-23 04:33:44,882 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42940.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:33:54,526 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp1.1 from training. Duration: 20.4409375 2022-12-23 04:34:03,335 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9837, 2.0826, 1.3477, 2.5788, 2.2878, 1.8770, 3.0345, 2.0005], device='cuda:3'), covar=tensor([0.0931, 0.1705, 0.2860, 0.1671, 0.1667, 0.0956, 0.0895, 0.1195], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0196, 0.0238, 0.0279, 0.0226, 0.0184, 0.0207, 0.0194], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 04:34:26,861 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9106, 1.8536, 1.4753, 0.8920, 2.3545, 1.9465, 1.6484, 1.4129], device='cuda:3'), covar=tensor([0.0345, 0.0347, 0.0480, 0.0746, 0.0234, 0.0336, 0.0458, 0.0830], device='cuda:3'), in_proj_covar=tensor([0.0118, 0.0117, 0.0123, 0.0116, 0.0087, 0.0112, 0.0129, 0.0149], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 04:34:36,972 INFO [train.py:894] (3/4) Epoch 13, batch 900, loss[loss=0.2002, simple_loss=0.2862, pruned_loss=0.05715, over 18676.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2809, pruned_loss=0.05536, over 3673399.22 frames. ], batch size: 62, lr: 8.99e-03, grad_scale: 8.0 2022-12-23 04:35:09,736 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0_sp0.9 from training. Duration: 22.511125 2022-12-23 04:35:11,146 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0 from training. Duration: 20.675 2022-12-23 04:35:22,828 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43005.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:35:53,122 INFO [train.py:894] (3/4) Epoch 13, batch 950, loss[loss=0.197, simple_loss=0.2835, pruned_loss=0.05525, over 18701.00 frames. ], tot_loss[loss=0.195, simple_loss=0.2803, pruned_loss=0.05488, over 3682296.12 frames. ], batch size: 62, lr: 8.99e-03, grad_scale: 8.0 2022-12-23 04:35:59,525 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.591e+02 3.779e+02 4.609e+02 5.645e+02 1.005e+03, threshold=9.218e+02, percent-clipped=1.0 2022-12-23 04:36:08,741 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43035.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:36:26,356 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9253, 2.0508, 1.4834, 2.3027, 2.1049, 1.8207, 2.7763, 2.0474], device='cuda:3'), covar=tensor([0.0928, 0.1569, 0.2540, 0.1630, 0.1587, 0.0894, 0.0926, 0.1179], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0197, 0.0239, 0.0281, 0.0227, 0.0184, 0.0209, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 04:36:50,802 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp0.9 from training. Duration: 24.9833125 2022-12-23 04:37:00,823 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2022-12-23 04:37:08,455 INFO [train.py:894] (3/4) Epoch 13, batch 1000, loss[loss=0.1878, simple_loss=0.2656, pruned_loss=0.055, over 18433.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.2795, pruned_loss=0.05505, over 3689702.25 frames. ], batch size: 42, lr: 8.98e-03, grad_scale: 8.0 2022-12-23 04:37:20,510 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=43083.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:37:23,243 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0 from training. Duration: 27.14 2022-12-23 04:37:37,451 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0 from training. Duration: 22.44 2022-12-23 04:38:23,697 INFO [train.py:894] (3/4) Epoch 13, batch 1050, loss[loss=0.2342, simple_loss=0.3114, pruned_loss=0.07847, over 18713.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2794, pruned_loss=0.05496, over 3696232.53 frames. ], batch size: 69, lr: 8.98e-03, grad_scale: 8.0 2022-12-23 04:38:29,575 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.711e+02 3.874e+02 4.915e+02 6.506e+02 1.252e+03, threshold=9.831e+02, percent-clipped=4.0 2022-12-23 04:38:57,197 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0060-62364-0_sp0.9 from training. Duration: 21.361125 2022-12-23 04:39:04,043 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp1.1 from training. Duration: 27.0318125 2022-12-23 04:39:13,001 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp0.9 from training. Duration: 28.638875 2022-12-23 04:39:19,065 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43162.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:39:26,311 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0 from training. Duration: 20.4 2022-12-23 04:39:38,927 INFO [train.py:894] (3/4) Epoch 13, batch 1100, loss[loss=0.1907, simple_loss=0.2608, pruned_loss=0.06032, over 18585.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2793, pruned_loss=0.05493, over 3700463.21 frames. ], batch size: 45, lr: 8.97e-03, grad_scale: 8.0 2022-12-23 04:39:57,022 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0 from training. Duration: 20.025 2022-12-23 04:39:57,030 WARNING [train.py:1060] (3/4) Exclude cut with ID 2364-131735-0112-64612-0_sp0.9 from training. Duration: 20.488875 2022-12-23 04:39:59,741 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2022-12-23 04:40:01,642 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0 from training. Duration: 29.735 2022-12-23 04:40:30,956 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3527, 1.9139, 2.5659, 3.4462, 2.8130, 3.0119, 1.7389, 2.4714], device='cuda:3'), covar=tensor([0.1434, 0.1519, 0.1217, 0.0526, 0.0808, 0.1954, 0.1851, 0.1022], device='cuda:3'), in_proj_covar=tensor([0.0103, 0.0115, 0.0133, 0.0132, 0.0104, 0.0135, 0.0130, 0.0109], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 04:40:33,952 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=43210.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:40:41,295 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43215.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:40:56,301 INFO [train.py:894] (3/4) Epoch 13, batch 1150, loss[loss=0.2172, simple_loss=0.3007, pruned_loss=0.06688, over 18710.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2779, pruned_loss=0.05436, over 3702637.24 frames. ], batch size: 78, lr: 8.97e-03, grad_scale: 8.0 2022-12-23 04:41:02,039 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.407e+02 3.889e+02 4.613e+02 6.003e+02 1.099e+03, threshold=9.226e+02, percent-clipped=3.0 2022-12-23 04:41:03,159 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2022-12-23 04:41:11,031 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43235.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:41:26,076 WARNING [train.py:1060] (3/4) Exclude cut with ID 7276-92427-0014-12983-0_sp0.9 from training. Duration: 21.3055625 2022-12-23 04:41:28,163 WARNING [train.py:1060] (3/4) Exclude cut with ID 1025-75365-0008-79168-0_sp0.9 from training. Duration: 22.0666875 2022-12-23 04:42:11,536 INFO [train.py:894] (3/4) Epoch 13, batch 1200, loss[loss=0.1647, simple_loss=0.2536, pruned_loss=0.03789, over 18432.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2771, pruned_loss=0.05371, over 3704418.58 frames. ], batch size: 48, lr: 8.96e-03, grad_scale: 8.0 2022-12-23 04:42:51,121 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3119, 3.4020, 3.1994, 1.1185, 3.3096, 2.4242, 0.6579, 2.2148], device='cuda:3'), covar=tensor([0.2277, 0.1156, 0.1541, 0.4113, 0.0959, 0.1088, 0.4882, 0.1592], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0128, 0.0148, 0.0119, 0.0128, 0.0104, 0.0139, 0.0109], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 04:42:57,231 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43305.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:43:16,829 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0 from training. Duration: 20.26 2022-12-23 04:43:26,884 INFO [train.py:894] (3/4) Epoch 13, batch 1250, loss[loss=0.2051, simple_loss=0.2906, pruned_loss=0.05981, over 18700.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2771, pruned_loss=0.05375, over 3707055.66 frames. ], batch size: 60, lr: 8.96e-03, grad_scale: 8.0 2022-12-23 04:43:28,627 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8305, 1.2349, 0.7201, 1.3594, 1.9752, 1.1822, 1.6529, 1.7654], device='cuda:3'), covar=tensor([0.1710, 0.2344, 0.2678, 0.1775, 0.1903, 0.1962, 0.1464, 0.1771], device='cuda:3'), in_proj_covar=tensor([0.0092, 0.0100, 0.0119, 0.0098, 0.0115, 0.0092, 0.0098, 0.0095], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 04:43:29,620 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0030-9324-0_sp0.9 from training. Duration: 21.3444375 2022-12-23 04:43:32,668 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.493e+02 3.857e+02 4.657e+02 6.544e+02 3.664e+03, threshold=9.315e+02, percent-clipped=7.0 2022-12-23 04:44:09,760 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=43353.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:44:27,113 WARNING [train.py:1060] (3/4) Exclude cut with ID 497-129325-0061-62254-0_sp1.1 from training. Duration: 0.97725 2022-12-23 04:44:42,515 INFO [train.py:894] (3/4) Epoch 13, batch 1300, loss[loss=0.1731, simple_loss=0.2608, pruned_loss=0.04265, over 18685.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2772, pruned_loss=0.05365, over 3707439.14 frames. ], batch size: 48, lr: 8.95e-03, grad_scale: 8.0 2022-12-23 04:45:09,790 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0_sp0.9 from training. Duration: 22.97225 2022-12-23 04:45:41,073 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0047-39922-0_sp0.9 from training. Duration: 21.97775 2022-12-23 04:45:55,096 WARNING [train.py:1060] (3/4) Exclude cut with ID 1112-1043-0006-89194-0_sp0.9 from training. Duration: 21.8333125 2022-12-23 04:45:57,947 INFO [train.py:894] (3/4) Epoch 13, batch 1350, loss[loss=0.1993, simple_loss=0.2828, pruned_loss=0.05787, over 18590.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2771, pruned_loss=0.05367, over 3709061.90 frames. ], batch size: 57, lr: 8.95e-03, grad_scale: 8.0 2022-12-23 04:46:03,898 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.083e+02 3.554e+02 4.386e+02 5.691e+02 1.592e+03, threshold=8.773e+02, percent-clipped=6.0 2022-12-23 04:46:06,835 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0 from training. Duration: 20.47 2022-12-23 04:47:12,577 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0037-39912-0_sp0.9 from training. Duration: 20.67225 2022-12-23 04:47:14,272 INFO [train.py:894] (3/4) Epoch 13, batch 1400, loss[loss=0.2333, simple_loss=0.3042, pruned_loss=0.08123, over 18562.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2772, pruned_loss=0.05352, over 3711287.77 frames. ], batch size: 182, lr: 8.94e-03, grad_scale: 8.0 2022-12-23 04:47:16,269 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3179, 2.8246, 2.8180, 1.4222, 3.0597, 2.8384, 1.9699, 3.2374], device='cuda:3'), covar=tensor([0.1317, 0.1454, 0.1443, 0.2535, 0.0740, 0.1324, 0.2154, 0.0570], device='cuda:3'), in_proj_covar=tensor([0.0196, 0.0199, 0.0204, 0.0193, 0.0174, 0.0209, 0.0205, 0.0181], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 04:47:32,059 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp0.9 from training. Duration: 25.2444375 2022-12-23 04:47:33,028 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4839, 1.0835, 0.6727, 1.0653, 1.8697, 0.6246, 1.1396, 1.4484], device='cuda:3'), covar=tensor([0.1704, 0.2278, 0.2273, 0.1756, 0.1869, 0.1862, 0.1575, 0.1691], device='cuda:3'), in_proj_covar=tensor([0.0093, 0.0100, 0.0121, 0.0099, 0.0115, 0.0092, 0.0098, 0.0096], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 04:47:55,096 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0021-76797-0_sp0.9 from training. Duration: 21.1445 2022-12-23 04:48:13,601 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.5321, 3.8532, 3.8520, 4.5281, 4.1146, 4.0269, 4.6866, 1.3593], device='cuda:3'), covar=tensor([0.0635, 0.0683, 0.0603, 0.0578, 0.1241, 0.1065, 0.0515, 0.4796], device='cuda:3'), in_proj_covar=tensor([0.0312, 0.0204, 0.0209, 0.0231, 0.0289, 0.0242, 0.0253, 0.0264], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 04:48:13,700 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43514.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:48:15,101 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43515.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:48:28,975 INFO [train.py:894] (3/4) Epoch 13, batch 1450, loss[loss=0.1878, simple_loss=0.2783, pruned_loss=0.04865, over 18597.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2782, pruned_loss=0.05375, over 3711982.29 frames. ], batch size: 51, lr: 8.94e-03, grad_scale: 8.0 2022-12-23 04:48:29,681 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2022-12-23 04:48:34,934 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.266e+02 3.498e+02 4.440e+02 5.498e+02 1.255e+03, threshold=8.879e+02, percent-clipped=4.0 2022-12-23 04:48:45,710 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43535.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:49:09,063 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp0.9 from training. Duration: 33.038875 2022-12-23 04:49:27,629 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=43563.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:49:27,956 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.3387, 1.6079, 1.9653, 0.8639, 1.0208, 2.0643, 1.8397, 1.6979], device='cuda:3'), covar=tensor([0.0706, 0.0327, 0.0295, 0.0354, 0.0394, 0.0376, 0.0211, 0.0603], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0158, 0.0114, 0.0130, 0.0141, 0.0131, 0.0148, 0.0151], device='cuda:3'), out_proj_covar=tensor([1.1206e-04, 1.2768e-04, 9.1003e-05, 1.0143e-04, 1.1190e-04, 1.0670e-04, 1.2113e-04, 1.2068e-04], device='cuda:3') 2022-12-23 04:49:30,783 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5903, 1.7739, 1.6892, 1.6262, 1.1535, 3.8787, 1.7266, 2.3728], device='cuda:3'), covar=tensor([0.3303, 0.1808, 0.1887, 0.1901, 0.1497, 0.0151, 0.1406, 0.0767], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0116, 0.0125, 0.0119, 0.0103, 0.0097, 0.0093, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 04:49:44,996 INFO [train.py:894] (3/4) Epoch 13, batch 1500, loss[loss=0.1743, simple_loss=0.2513, pruned_loss=0.04868, over 18661.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2783, pruned_loss=0.05407, over 3713580.78 frames. ], batch size: 41, lr: 8.93e-03, grad_scale: 8.0 2022-12-23 04:49:45,497 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43575.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:49:47,984 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64291-0000-16059-0_sp0.9 from training. Duration: 20.0944375 2022-12-23 04:49:57,679 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=43583.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:49:59,337 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4933, 0.9543, 0.6347, 1.0231, 1.9091, 0.5925, 1.0866, 1.4015], device='cuda:3'), covar=tensor([0.1811, 0.2559, 0.2319, 0.1889, 0.1910, 0.1945, 0.1706, 0.1855], device='cuda:3'), in_proj_covar=tensor([0.0091, 0.0099, 0.0119, 0.0096, 0.0113, 0.0091, 0.0097, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 04:50:04,026 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp1.1 from training. Duration: 20.4 2022-12-23 04:50:12,847 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0 from training. Duration: 20.085 2022-12-23 04:50:17,825 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.04 vs. limit=5.0 2022-12-23 04:50:22,988 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0_sp0.9 from training. Duration: 23.07775 2022-12-23 04:50:45,421 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6644, 1.6481, 1.3556, 1.7611, 1.7909, 1.5421, 2.1305, 1.8124], device='cuda:3'), covar=tensor([0.0934, 0.1584, 0.2502, 0.1423, 0.1716, 0.0915, 0.1006, 0.1162], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0195, 0.0236, 0.0277, 0.0224, 0.0181, 0.0205, 0.0192], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 04:51:00,798 INFO [train.py:894] (3/4) Epoch 13, batch 1550, loss[loss=0.2002, simple_loss=0.2867, pruned_loss=0.05679, over 18586.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2781, pruned_loss=0.05405, over 3713545.72 frames. ], batch size: 51, lr: 8.93e-03, grad_scale: 8.0 2022-12-23 04:51:07,027 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.501e+02 4.191e+02 4.765e+02 6.382e+02 1.278e+03, threshold=9.530e+02, percent-clipped=5.0 2022-12-23 04:51:09,056 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp0.9 from training. Duration: 24.9333125 2022-12-23 04:51:22,113 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.67 vs. limit=5.0 2022-12-23 04:51:52,609 WARNING [train.py:1060] (3/4) Exclude cut with ID 5118-111612-0016-124680-0_sp0.9 from training. Duration: 20.388875 2022-12-23 04:51:58,436 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp1.1 from training. Duration: 20.3590625 2022-12-23 04:52:10,312 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1815, 1.4789, 0.7911, 1.7798, 2.3714, 1.6619, 1.9342, 2.1920], device='cuda:3'), covar=tensor([0.1609, 0.2267, 0.2596, 0.1606, 0.1659, 0.1779, 0.1544, 0.1725], device='cuda:3'), in_proj_covar=tensor([0.0091, 0.0098, 0.0118, 0.0096, 0.0113, 0.0090, 0.0096, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 04:52:15,951 INFO [train.py:894] (3/4) Epoch 13, batch 1600, loss[loss=0.193, simple_loss=0.2812, pruned_loss=0.05245, over 18559.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2788, pruned_loss=0.05424, over 3713437.43 frames. ], batch size: 56, lr: 8.92e-03, grad_scale: 8.0 2022-12-23 04:52:48,971 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2022-12-23 04:53:07,413 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0_sp1.1 from training. Duration: 0.836375 2022-12-23 04:53:15,443 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2022-12-23 04:53:31,655 INFO [train.py:894] (3/4) Epoch 13, batch 1650, loss[loss=0.2299, simple_loss=0.3082, pruned_loss=0.07583, over 18523.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2793, pruned_loss=0.055, over 3713978.67 frames. ], batch size: 58, lr: 8.92e-03, grad_scale: 8.0 2022-12-23 04:53:38,416 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.973e+02 4.370e+02 5.353e+02 6.332e+02 1.180e+03, threshold=1.071e+03, percent-clipped=2.0 2022-12-23 04:53:49,759 WARNING [train.py:1060] (3/4) Exclude cut with ID 8565-290391-0049-67394-0_sp0.9 from training. Duration: 21.3166875 2022-12-23 04:53:51,735 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8685, 1.8537, 1.3581, 1.8029, 1.6575, 1.6884, 1.5976, 1.8769], device='cuda:3'), covar=tensor([0.1892, 0.2433, 0.1642, 0.1881, 0.2513, 0.0922, 0.2229, 0.0741], device='cuda:3'), in_proj_covar=tensor([0.0280, 0.0273, 0.0229, 0.0340, 0.0254, 0.0213, 0.0268, 0.0191], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 04:54:15,175 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5130, 2.1904, 1.5646, 2.4740, 1.9263, 1.8791, 1.9890, 2.4686], device='cuda:3'), covar=tensor([0.1842, 0.2758, 0.1770, 0.2296, 0.3075, 0.0970, 0.2655, 0.0714], device='cuda:3'), in_proj_covar=tensor([0.0283, 0.0276, 0.0232, 0.0344, 0.0257, 0.0215, 0.0270, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 04:54:19,400 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0029-104863-0_sp0.9 from training. Duration: 22.1055625 2022-12-23 04:54:29,637 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp1.1 from training. Duration: 21.77725 2022-12-23 04:54:47,435 INFO [train.py:894] (3/4) Epoch 13, batch 1700, loss[loss=0.1798, simple_loss=0.2668, pruned_loss=0.04644, over 18613.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2808, pruned_loss=0.05689, over 3714498.39 frames. ], batch size: 51, lr: 8.91e-03, grad_scale: 8.0 2022-12-23 04:54:50,298 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp0.9 from training. Duration: 27.8166875 2022-12-23 04:55:06,650 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43787.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:55:15,061 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp1.1 from training. Duration: 22.5090625 2022-12-23 04:55:23,043 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0 from training. Duration: 25.035 2022-12-23 04:55:26,667 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.2141, 1.0018, 1.4710, 2.5498, 1.7816, 2.0605, 0.7920, 1.7556], device='cuda:3'), covar=tensor([0.1992, 0.1986, 0.1543, 0.0670, 0.1249, 0.1263, 0.2209, 0.1240], device='cuda:3'), in_proj_covar=tensor([0.0101, 0.0114, 0.0130, 0.0132, 0.0104, 0.0132, 0.0129, 0.0108], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 04:55:39,649 WARNING [train.py:1060] (3/4) Exclude cut with ID 774-127930-0014-10412-0_sp1.1 from training. Duration: 0.95 2022-12-23 04:55:57,266 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp0.9 from training. Duration: 0.92225 2022-12-23 04:56:04,339 INFO [train.py:894] (3/4) Epoch 13, batch 1750, loss[loss=0.2675, simple_loss=0.3268, pruned_loss=0.1041, over 18648.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2831, pruned_loss=0.05999, over 3714663.38 frames. ], batch size: 181, lr: 8.91e-03, grad_scale: 8.0 2022-12-23 04:56:10,722 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.858e+02 4.539e+02 5.407e+02 6.945e+02 1.348e+03, threshold=1.081e+03, percent-clipped=3.0 2022-12-23 04:56:25,811 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0 from training. Duration: 21.97 2022-12-23 04:56:32,422 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6455, 1.3722, 1.4157, 0.7573, 1.7401, 1.5613, 1.3815, 1.1159], device='cuda:3'), covar=tensor([0.0353, 0.0475, 0.0419, 0.0660, 0.0312, 0.0325, 0.0427, 0.1033], device='cuda:3'), in_proj_covar=tensor([0.0118, 0.0115, 0.0124, 0.0114, 0.0086, 0.0113, 0.0127, 0.0148], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 04:56:40,222 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43848.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:56:45,399 WARNING [train.py:1060] (3/4) Exclude cut with ID 7492-105653-0055-62765-0_sp0.9 from training. Duration: 21.97225 2022-12-23 04:56:47,276 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp0.9 from training. Duration: 25.3333125 2022-12-23 04:56:57,378 WARNING [train.py:1060] (3/4) Exclude cut with ID 5172-29468-0015-19128-0_sp0.9 from training. Duration: 21.5055625 2022-12-23 04:57:06,946 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp1.1 from training. Duration: 20.72725 2022-12-23 04:57:12,916 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43870.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:57:20,807 INFO [train.py:894] (3/4) Epoch 13, batch 1800, loss[loss=0.2011, simple_loss=0.2853, pruned_loss=0.05843, over 18634.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2831, pruned_loss=0.06104, over 3714472.74 frames. ], batch size: 53, lr: 8.90e-03, grad_scale: 8.0 2022-12-23 04:57:36,076 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7004, 1.4327, 1.3188, 1.4707, 1.8812, 1.8048, 1.9600, 1.2218], device='cuda:3'), covar=tensor([0.0315, 0.0268, 0.0434, 0.0208, 0.0162, 0.0275, 0.0172, 0.0283], device='cuda:3'), in_proj_covar=tensor([0.0089, 0.0122, 0.0146, 0.0123, 0.0111, 0.0112, 0.0095, 0.0123], device='cuda:3'), out_proj_covar=tensor([7.2913e-05, 9.9445e-05, 1.2469e-04, 1.0198e-04, 9.3345e-05, 8.8350e-05, 7.6444e-05, 1.0029e-04], device='cuda:3') 2022-12-23 04:57:41,334 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp0.9 from training. Duration: 26.32775 2022-12-23 04:57:58,564 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1987, 2.5241, 1.5065, 3.0484, 2.6070, 2.2449, 3.7331, 2.4295], device='cuda:3'), covar=tensor([0.0866, 0.1723, 0.2749, 0.1744, 0.1629, 0.0871, 0.0776, 0.1125], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0195, 0.0236, 0.0281, 0.0226, 0.0182, 0.0206, 0.0194], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 04:58:05,527 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7283, 3.8347, 3.7050, 1.5183, 3.7728, 2.7486, 0.7350, 2.5599], device='cuda:3'), covar=tensor([0.2073, 0.1055, 0.1397, 0.3974, 0.1050, 0.1197, 0.5395, 0.1591], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0130, 0.0152, 0.0121, 0.0131, 0.0108, 0.0143, 0.0111], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 04:58:12,253 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43909.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:58:13,315 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0 from training. Duration: 20.025 2022-12-23 04:58:18,612 WARNING [train.py:1060] (3/4) Exclude cut with ID 6709-74022-0004-86860-0_sp1.1 from training. Duration: 0.9409375 2022-12-23 04:58:18,619 WARNING [train.py:1060] (3/4) Exclude cut with ID 4757-1811-0023-62229-0_sp0.9 from training. Duration: 21.37775 2022-12-23 04:58:21,835 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.4583, 3.8981, 3.8540, 4.2999, 4.0375, 3.9071, 4.5416, 1.5188], device='cuda:3'), covar=tensor([0.0622, 0.0558, 0.0618, 0.0698, 0.1266, 0.0981, 0.0513, 0.4521], device='cuda:3'), in_proj_covar=tensor([0.0308, 0.0201, 0.0207, 0.0228, 0.0288, 0.0241, 0.0250, 0.0260], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 04:58:25,151 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2022-12-23 04:58:27,245 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7173, 1.3002, 0.9677, 1.4608, 2.0062, 1.1714, 1.3903, 1.7529], device='cuda:3'), covar=tensor([0.1538, 0.2021, 0.2143, 0.1395, 0.1740, 0.1670, 0.1467, 0.1441], device='cuda:3'), in_proj_covar=tensor([0.0091, 0.0099, 0.0118, 0.0097, 0.0113, 0.0091, 0.0096, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 04:58:36,324 INFO [train.py:894] (3/4) Epoch 13, batch 1850, loss[loss=0.2075, simple_loss=0.2755, pruned_loss=0.06974, over 18397.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2838, pruned_loss=0.06256, over 3714011.57 frames. ], batch size: 46, lr: 8.90e-03, grad_scale: 8.0 2022-12-23 04:58:41,372 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0004-25974-0_sp0.9 from training. Duration: 21.17225 2022-12-23 04:58:41,383 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp0.9 from training. Duration: 27.511125 2022-12-23 04:58:42,660 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.374e+02 4.790e+02 5.650e+02 6.943e+02 1.599e+03, threshold=1.130e+03, percent-clipped=7.0 2022-12-23 04:58:53,472 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6092, 2.0294, 1.8816, 1.2134, 3.0089, 2.7285, 2.2434, 1.2312], device='cuda:3'), covar=tensor([0.0370, 0.0482, 0.0528, 0.0795, 0.0189, 0.0326, 0.0476, 0.1372], device='cuda:3'), in_proj_covar=tensor([0.0118, 0.0115, 0.0123, 0.0114, 0.0087, 0.0113, 0.0126, 0.0147], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 04:59:14,332 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.1574, 3.6050, 3.5258, 4.0291, 3.7559, 3.6876, 4.3114, 1.2281], device='cuda:3'), covar=tensor([0.0820, 0.0676, 0.0693, 0.0860, 0.1632, 0.1378, 0.0613, 0.4993], device='cuda:3'), in_proj_covar=tensor([0.0312, 0.0204, 0.0210, 0.0231, 0.0293, 0.0244, 0.0252, 0.0263], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 04:59:18,516 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0 from training. Duration: 22.8 2022-12-23 04:59:22,944 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0 from training. Duration: 22.585 2022-12-23 04:59:27,712 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43959.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:59:44,753 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43970.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 04:59:51,439 INFO [train.py:894] (3/4) Epoch 13, batch 1900, loss[loss=0.2463, simple_loss=0.3177, pruned_loss=0.08747, over 18658.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.283, pruned_loss=0.06289, over 3713681.62 frames. ], batch size: 60, lr: 8.89e-03, grad_scale: 8.0 2022-12-23 04:59:56,601 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0001-146967-0_sp0.9 from training. Duration: 22.0166875 2022-12-23 05:00:14,059 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp1.1 from training. Duration: 24.395375 2022-12-23 05:00:20,230 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp0.9 from training. Duration: 27.47775 2022-12-23 05:00:24,665 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp0.9 from training. Duration: 24.8833125 2022-12-23 05:00:27,599 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0 from training. Duration: 23.39 2022-12-23 05:00:37,079 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp0.9 from training. Duration: 28.72225 2022-12-23 05:00:45,740 WARNING [train.py:1060] (3/4) Exclude cut with ID 585-294811-0110-133686-0_sp0.9 from training. Duration: 20.8944375 2022-12-23 05:01:00,971 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0_sp0.9 from training. Duration: 23.8444375 2022-12-23 05:01:04,336 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44020.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:01:11,924 INFO [train.py:894] (3/4) Epoch 13, batch 1950, loss[loss=0.2062, simple_loss=0.2884, pruned_loss=0.062, over 18552.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2829, pruned_loss=0.06348, over 3714621.84 frames. ], batch size: 55, lr: 8.89e-03, grad_scale: 8.0 2022-12-23 05:01:18,255 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.985e+02 4.904e+02 5.983e+02 6.955e+02 1.317e+03, threshold=1.197e+03, percent-clipped=1.0 2022-12-23 05:01:26,847 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0 from training. Duration: 25.85 2022-12-23 05:01:26,858 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0 from training. Duration: 21.39 2022-12-23 05:01:38,033 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0 from training. Duration: 27.92 2022-12-23 05:02:03,838 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0039-130165-0_sp0.9 from training. Duration: 20.661125 2022-12-23 05:02:27,774 INFO [train.py:894] (3/4) Epoch 13, batch 2000, loss[loss=0.226, simple_loss=0.2983, pruned_loss=0.07688, over 18535.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2833, pruned_loss=0.06435, over 3714011.38 frames. ], batch size: 58, lr: 8.88e-03, grad_scale: 8.0 2022-12-23 05:02:27,849 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0043-15874-0_sp0.9 from training. Duration: 20.07225 2022-12-23 05:02:36,666 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0 from training. Duration: 21.01 2022-12-23 05:03:41,975 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0 from training. Duration: 20.65 2022-12-23 05:03:44,898 INFO [train.py:894] (3/4) Epoch 13, batch 2050, loss[loss=0.2109, simple_loss=0.2887, pruned_loss=0.06656, over 18556.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2839, pruned_loss=0.06499, over 3714558.22 frames. ], batch size: 77, lr: 8.88e-03, grad_scale: 8.0 2022-12-23 05:03:51,452 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.848e+02 4.873e+02 5.730e+02 6.896e+02 1.341e+03, threshold=1.146e+03, percent-clipped=1.0 2022-12-23 05:03:51,473 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0 from training. Duration: 21.46 2022-12-23 05:04:12,930 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44143.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:04:35,166 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0 from training. Duration: 0.92 2022-12-23 05:04:39,213 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.79 vs. limit=5.0 2022-12-23 05:04:42,100 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0_sp0.9 from training. Duration: 23.7666875 2022-12-23 05:04:54,368 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44170.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:05:01,739 INFO [train.py:894] (3/4) Epoch 13, batch 2100, loss[loss=0.2058, simple_loss=0.2859, pruned_loss=0.06287, over 18386.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.284, pruned_loss=0.06518, over 3714830.55 frames. ], batch size: 53, lr: 8.87e-03, grad_scale: 8.0 2022-12-23 05:05:12,016 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.8275, 5.2165, 4.7504, 2.8192, 5.3320, 4.1079, 0.6617, 3.7256], device='cuda:3'), covar=tensor([0.2210, 0.0836, 0.1340, 0.3200, 0.0761, 0.0896, 0.5976, 0.1314], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0130, 0.0152, 0.0121, 0.0132, 0.0108, 0.0143, 0.0111], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 05:05:19,051 WARNING [train.py:1060] (3/4) Exclude cut with ID 8544-281189-0060-101339-0_sp0.9 from training. Duration: 20.861125 2022-12-23 05:05:28,315 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0_sp0.9 from training. Duration: 22.711125 2022-12-23 05:05:56,776 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.8477, 5.1184, 4.6849, 2.7799, 5.1460, 3.7601, 0.7726, 3.5015], device='cuda:3'), covar=tensor([0.1977, 0.0844, 0.1173, 0.2924, 0.0774, 0.0981, 0.5534, 0.1303], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0129, 0.0151, 0.0120, 0.0131, 0.0107, 0.0141, 0.0110], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 05:06:06,803 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=44218.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:06:08,221 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp1.1 from training. Duration: 22.986375 2022-12-23 05:06:17,674 INFO [train.py:894] (3/4) Epoch 13, batch 2150, loss[loss=0.1964, simple_loss=0.2746, pruned_loss=0.05913, over 18539.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2852, pruned_loss=0.06637, over 3715501.66 frames. ], batch size: 55, lr: 8.87e-03, grad_scale: 8.0 2022-12-23 05:06:23,827 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.378e+02 5.026e+02 6.284e+02 7.553e+02 1.602e+03, threshold=1.257e+03, percent-clipped=4.0 2022-12-23 05:06:26,702 WARNING [train.py:1060] (3/4) Exclude cut with ID 8040-260924-0003-80960-0_sp0.9 from training. Duration: 22.07225 2022-12-23 05:06:31,056 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0045-26330-0_sp0.9 from training. Duration: 20.3055625 2022-12-23 05:06:32,461 WARNING [train.py:1060] (3/4) Exclude cut with ID 6356-271890-0060-94317-0_sp0.9 from training. Duration: 20.72225 2022-12-23 05:06:51,058 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp1.1 from training. Duration: 22.4818125 2022-12-23 05:07:18,066 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp0.9 from training. Duration: 25.0944375 2022-12-23 05:07:18,218 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44265.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:07:20,984 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0 from training. Duration: 21.515 2022-12-23 05:07:26,007 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp0.9 from training. Duration: 27.02225 2022-12-23 05:07:31,750 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0010-62480-0_sp0.9 from training. Duration: 22.22225 2022-12-23 05:07:33,323 INFO [train.py:894] (3/4) Epoch 13, batch 2200, loss[loss=0.2214, simple_loss=0.297, pruned_loss=0.07287, over 18627.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2849, pruned_loss=0.06622, over 3714861.18 frames. ], batch size: 99, lr: 8.86e-03, grad_scale: 8.0 2022-12-23 05:07:39,403 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0085-44554-0_sp0.9 from training. Duration: 20.85 2022-12-23 05:07:44,888 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2022-12-23 05:08:14,766 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0 from training. Duration: 21.54 2022-12-23 05:08:19,510 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp1.1 from training. Duration: 20.5318125 2022-12-23 05:08:24,187 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2022-12-23 05:08:28,229 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0012-134311-0_sp0.9 from training. Duration: 21.9333125 2022-12-23 05:08:31,987 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2022-12-23 05:08:32,778 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44315.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:08:48,075 INFO [train.py:894] (3/4) Epoch 13, batch 2250, loss[loss=0.2205, simple_loss=0.2983, pruned_loss=0.07139, over 18583.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2838, pruned_loss=0.06576, over 3713469.29 frames. ], batch size: 51, lr: 8.86e-03, grad_scale: 8.0 2022-12-23 05:08:54,075 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.356e+02 5.266e+02 6.249e+02 7.363e+02 1.689e+03, threshold=1.250e+03, percent-clipped=3.0 2022-12-23 05:09:15,288 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0025-130151-0_sp0.9 from training. Duration: 21.7944375 2022-12-23 05:09:28,103 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0_sp0.9 from training. Duration: 22.4666875 2022-12-23 05:09:34,071 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0 from training. Duration: 21.635 2022-12-23 05:09:40,159 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0_sp0.9 from training. Duration: 24.038875 2022-12-23 05:10:03,930 INFO [train.py:894] (3/4) Epoch 13, batch 2300, loss[loss=0.2167, simple_loss=0.2786, pruned_loss=0.07742, over 18615.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2837, pruned_loss=0.06562, over 3713599.01 frames. ], batch size: 41, lr: 8.85e-03, grad_scale: 16.0 2022-12-23 05:10:12,191 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2798, 2.2360, 1.7527, 1.1173, 2.8731, 2.5271, 2.1770, 1.6337], device='cuda:3'), covar=tensor([0.0361, 0.0329, 0.0481, 0.0716, 0.0177, 0.0294, 0.0385, 0.0780], device='cuda:3'), in_proj_covar=tensor([0.0118, 0.0117, 0.0125, 0.0116, 0.0087, 0.0114, 0.0128, 0.0148], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 05:10:23,922 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp1.1 from training. Duration: 21.786375 2022-12-23 05:10:33,469 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.35 vs. limit=5.0 2022-12-23 05:10:36,272 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0 from training. Duration: 20.22 2022-12-23 05:10:46,698 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6417, 1.4676, 1.4942, 1.8333, 1.6438, 2.5578, 1.4171, 1.4315], device='cuda:3'), covar=tensor([0.0826, 0.1410, 0.1173, 0.0796, 0.1244, 0.0375, 0.1198, 0.1306], device='cuda:3'), in_proj_covar=tensor([0.0072, 0.0082, 0.0074, 0.0074, 0.0091, 0.0072, 0.0084, 0.0075], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 05:11:19,833 INFO [train.py:894] (3/4) Epoch 13, batch 2350, loss[loss=0.2004, simple_loss=0.2804, pruned_loss=0.06015, over 18542.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2842, pruned_loss=0.06582, over 3713726.42 frames. ], batch size: 98, lr: 8.85e-03, grad_scale: 16.0 2022-12-23 05:11:26,062 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.295e+02 5.245e+02 6.691e+02 8.586e+02 1.798e+03, threshold=1.338e+03, percent-clipped=9.0 2022-12-23 05:11:30,639 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4131, 1.4505, 1.2857, 1.7514, 1.6738, 3.1618, 1.3671, 1.5152], device='cuda:3'), covar=tensor([0.0929, 0.1811, 0.1198, 0.0908, 0.1366, 0.0251, 0.1403, 0.1532], device='cuda:3'), in_proj_covar=tensor([0.0071, 0.0081, 0.0073, 0.0073, 0.0090, 0.0072, 0.0083, 0.0075], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 05:11:47,457 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44443.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:12:13,955 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.73 vs. limit=5.0 2022-12-23 05:12:33,966 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0 from training. Duration: 25.285 2022-12-23 05:12:35,374 INFO [train.py:894] (3/4) Epoch 13, batch 2400, loss[loss=0.1901, simple_loss=0.265, pruned_loss=0.05763, over 18618.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2846, pruned_loss=0.06618, over 3713250.51 frames. ], batch size: 45, lr: 8.84e-03, grad_scale: 16.0 2022-12-23 05:13:00,058 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=44491.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:13:38,766 WARNING [train.py:1060] (3/4) Exclude cut with ID 811-130148-0001-63453-0_sp0.9 from training. Duration: 20.861125 2022-12-23 05:13:49,408 INFO [train.py:894] (3/4) Epoch 13, batch 2450, loss[loss=0.2351, simple_loss=0.3146, pruned_loss=0.07779, over 18713.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2852, pruned_loss=0.06705, over 3713823.18 frames. ], batch size: 52, lr: 8.84e-03, grad_scale: 8.0 2022-12-23 05:13:53,052 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=44527.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:13:56,731 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.765e+02 5.091e+02 6.317e+02 7.743e+02 2.009e+03, threshold=1.263e+03, percent-clipped=3.0 2022-12-23 05:14:01,036 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0 from training. Duration: 20.88 2022-12-23 05:14:27,785 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=44550.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:14:33,419 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0_sp0.9 from training. Duration: 23.4166875 2022-12-23 05:14:50,363 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44565.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:15:05,134 INFO [train.py:894] (3/4) Epoch 13, batch 2500, loss[loss=0.2354, simple_loss=0.3144, pruned_loss=0.07816, over 18453.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2838, pruned_loss=0.06659, over 3712562.42 frames. ], batch size: 54, lr: 8.83e-03, grad_scale: 8.0 2022-12-23 05:15:26,570 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44588.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:15:50,549 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2022-12-23 05:15:51,033 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0 from training. Duration: 21.24 2022-12-23 05:15:51,049 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0_sp0.9 from training. Duration: 23.9055625 2022-12-23 05:16:01,738 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44611.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:16:05,155 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=44613.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:16:08,082 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44615.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:16:22,382 INFO [train.py:894] (3/4) Epoch 13, batch 2550, loss[loss=0.2109, simple_loss=0.2921, pruned_loss=0.06482, over 18692.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2825, pruned_loss=0.0657, over 3712790.74 frames. ], batch size: 96, lr: 8.83e-03, grad_scale: 8.0 2022-12-23 05:16:25,828 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp0.9 from training. Duration: 25.988875 2022-12-23 05:16:30,042 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.522e+02 4.934e+02 5.713e+02 7.366e+02 1.644e+03, threshold=1.143e+03, percent-clipped=2.0 2022-12-23 05:16:34,500 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0001-134300-0_sp0.9 from training. Duration: 20.67225 2022-12-23 05:17:20,721 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0 from training. Duration: 20.34 2022-12-23 05:17:20,806 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=44663.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:17:38,666 INFO [train.py:894] (3/4) Epoch 13, batch 2600, loss[loss=0.1953, simple_loss=0.2824, pruned_loss=0.0541, over 18537.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2826, pruned_loss=0.06538, over 3713090.45 frames. ], batch size: 58, lr: 8.82e-03, grad_scale: 8.0 2022-12-23 05:18:33,347 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp0.9 from training. Duration: 25.061125 2022-12-23 05:18:46,936 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0 from training. Duration: 0.83 2022-12-23 05:18:54,359 INFO [train.py:894] (3/4) Epoch 13, batch 2650, loss[loss=0.2131, simple_loss=0.2949, pruned_loss=0.06566, over 18679.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2831, pruned_loss=0.06559, over 3713474.61 frames. ], batch size: 99, lr: 8.82e-03, grad_scale: 8.0 2022-12-23 05:19:02,485 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.920e+02 4.661e+02 5.769e+02 6.751e+02 1.660e+03, threshold=1.154e+03, percent-clipped=4.0 2022-12-23 05:19:11,978 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0 from training. Duration: 24.73 2022-12-23 05:19:22,468 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0 from training. Duration: 23.965 2022-12-23 05:19:31,310 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0030-146996-0_sp0.9 from training. Duration: 22.088875 2022-12-23 05:19:48,785 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0_sp0.9 from training. Duration: 23.6 2022-12-23 05:20:10,132 INFO [train.py:894] (3/4) Epoch 13, batch 2700, loss[loss=0.1936, simple_loss=0.2814, pruned_loss=0.05287, over 18441.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2823, pruned_loss=0.06513, over 3712678.21 frames. ], batch size: 50, lr: 8.81e-03, grad_scale: 8.0 2022-12-23 05:20:24,511 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4706, 1.7661, 0.6217, 2.1248, 2.5438, 1.6167, 2.3090, 2.6871], device='cuda:3'), covar=tensor([0.1392, 0.1941, 0.2624, 0.1331, 0.1640, 0.1645, 0.1259, 0.1356], device='cuda:3'), in_proj_covar=tensor([0.0091, 0.0098, 0.0118, 0.0096, 0.0114, 0.0090, 0.0097, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 05:20:37,376 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4252, 2.1134, 2.0165, 1.7160, 2.3490, 2.9389, 2.9024, 1.9960], device='cuda:3'), covar=tensor([0.0313, 0.0306, 0.0374, 0.0263, 0.0233, 0.0253, 0.0209, 0.0279], device='cuda:3'), in_proj_covar=tensor([0.0087, 0.0122, 0.0144, 0.0122, 0.0111, 0.0111, 0.0093, 0.0122], device='cuda:3'), out_proj_covar=tensor([7.1585e-05, 9.9398e-05, 1.2319e-04, 1.0044e-04, 9.3111e-05, 8.7525e-05, 7.4954e-05, 9.9648e-05], device='cuda:3') 2022-12-23 05:20:54,978 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4410, 3.7530, 3.7702, 1.5262, 3.7979, 2.7906, 0.5950, 2.4829], device='cuda:3'), covar=tensor([0.2215, 0.1225, 0.1324, 0.3847, 0.0970, 0.1174, 0.5467, 0.1713], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0132, 0.0152, 0.0122, 0.0133, 0.0108, 0.0142, 0.0112], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 05:21:25,757 INFO [train.py:894] (3/4) Epoch 13, batch 2750, loss[loss=0.1661, simple_loss=0.2465, pruned_loss=0.04287, over 18391.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2831, pruned_loss=0.06541, over 3714021.95 frames. ], batch size: 46, lr: 8.81e-03, grad_scale: 8.0 2022-12-23 05:21:32,091 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0 from training. Duration: 23.795 2022-12-23 05:21:33,372 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.534e+02 5.060e+02 5.932e+02 7.826e+02 1.329e+03, threshold=1.186e+03, percent-clipped=3.0 2022-12-23 05:21:47,333 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp1.1 from training. Duration: 21.5409375 2022-12-23 05:21:50,110 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp0.9 from training. Duration: 24.97775 2022-12-23 05:22:02,453 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0_sp0.9 from training. Duration: 23.3444375 2022-12-23 05:22:15,346 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7941, 1.6211, 1.8241, 1.6899, 1.2866, 4.7480, 2.1353, 2.5296], device='cuda:3'), covar=tensor([0.3337, 0.2052, 0.1952, 0.2065, 0.1398, 0.0119, 0.1396, 0.0828], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0116, 0.0126, 0.0120, 0.0103, 0.0099, 0.0093, 0.0091], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 05:22:29,070 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0_sp0.9 from training. Duration: 23.2 2022-12-23 05:22:31,480 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5497, 1.4510, 1.3428, 0.7618, 1.7108, 1.4657, 1.3944, 1.2861], device='cuda:3'), covar=tensor([0.0386, 0.0447, 0.0464, 0.0690, 0.0287, 0.0367, 0.0466, 0.0833], device='cuda:3'), in_proj_covar=tensor([0.0119, 0.0119, 0.0127, 0.0116, 0.0088, 0.0115, 0.0130, 0.0149], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 05:22:35,362 WARNING [train.py:1060] (3/4) Exclude cut with ID 5653-46179-0060-117930-0_sp0.9 from training. Duration: 21.17225 2022-12-23 05:22:41,452 INFO [train.py:894] (3/4) Epoch 13, batch 2800, loss[loss=0.2184, simple_loss=0.3005, pruned_loss=0.06818, over 18605.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.285, pruned_loss=0.06663, over 3713844.28 frames. ], batch size: 97, lr: 8.80e-03, grad_scale: 8.0 2022-12-23 05:22:53,794 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44883.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:22:55,096 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp0.9 from training. Duration: 24.6555625 2022-12-23 05:23:28,754 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44906.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:23:47,920 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0 from training. Duration: 20.44 2022-12-23 05:23:57,007 INFO [train.py:894] (3/4) Epoch 13, batch 2850, loss[loss=0.2265, simple_loss=0.306, pruned_loss=0.07346, over 18631.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2858, pruned_loss=0.06668, over 3714582.17 frames. ], batch size: 53, lr: 8.80e-03, grad_scale: 8.0 2022-12-23 05:24:03,961 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.255e+02 5.409e+02 6.227e+02 8.445e+02 1.567e+03, threshold=1.245e+03, percent-clipped=8.0 2022-12-23 05:24:04,068 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0_sp0.9 from training. Duration: 23.45 2022-12-23 05:24:19,681 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.6716, 3.2308, 2.6945, 1.2521, 2.5110, 2.6458, 2.5160, 2.5048], device='cuda:3'), covar=tensor([0.0510, 0.0506, 0.1375, 0.1712, 0.1427, 0.1157, 0.1168, 0.1001], device='cuda:3'), in_proj_covar=tensor([0.0166, 0.0180, 0.0202, 0.0192, 0.0205, 0.0194, 0.0207, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 05:24:25,600 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1485, 1.5846, 2.6074, 4.5018, 3.3064, 2.7277, 0.9166, 3.3202], device='cuda:3'), covar=tensor([0.1751, 0.1731, 0.1431, 0.0448, 0.0921, 0.1210, 0.2420, 0.0795], device='cuda:3'), in_proj_covar=tensor([0.0102, 0.0114, 0.0130, 0.0134, 0.0105, 0.0133, 0.0129, 0.0109], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 05:24:33,617 WARNING [train.py:1060] (3/4) Exclude cut with ID 6945-60535-0076-12784-0_sp0.9 from training. Duration: 20.52225 2022-12-23 05:24:42,391 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0 from training. Duration: 22.19 2022-12-23 05:24:51,279 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp1.1 from training. Duration: 25.3818125 2022-12-23 05:25:10,158 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp0.9 from training. Duration: 28.0944375 2022-12-23 05:25:12,980 INFO [train.py:894] (3/4) Epoch 13, batch 2900, loss[loss=0.173, simple_loss=0.2499, pruned_loss=0.04808, over 18578.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2847, pruned_loss=0.06635, over 3714868.22 frames. ], batch size: 41, lr: 8.79e-03, grad_scale: 8.0 2022-12-23 05:25:14,339 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0_sp0.9 from training. Duration: 22.9444375 2022-12-23 05:25:23,988 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp1.1 from training. Duration: 21.6318125 2022-12-23 05:25:38,870 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0 from training. Duration: 23.695 2022-12-23 05:26:04,029 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0 from training. Duration: 23.955 2022-12-23 05:26:29,265 INFO [train.py:894] (3/4) Epoch 13, batch 2950, loss[loss=0.241, simple_loss=0.3103, pruned_loss=0.0859, over 18679.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2845, pruned_loss=0.06624, over 3714303.60 frames. ], batch size: 99, lr: 8.79e-03, grad_scale: 8.0 2022-12-23 05:26:37,232 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.731e+02 4.907e+02 6.071e+02 7.847e+02 1.453e+03, threshold=1.214e+03, percent-clipped=2.0 2022-12-23 05:26:38,583 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp0.9 from training. Duration: 26.438875 2022-12-23 05:27:16,140 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45056.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:27:21,625 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0021-26306-0_sp0.9 from training. Duration: 21.2444375 2022-12-23 05:27:21,659 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp0.9 from training. Duration: 31.02225 2022-12-23 05:27:32,347 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0 from training. Duration: 22.395 2022-12-23 05:27:42,505 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8696, 1.7765, 1.3223, 1.7542, 1.6102, 1.6362, 1.6262, 1.8130], device='cuda:3'), covar=tensor([0.2121, 0.2649, 0.1890, 0.2301, 0.2770, 0.1105, 0.2398, 0.0894], device='cuda:3'), in_proj_covar=tensor([0.0282, 0.0274, 0.0236, 0.0348, 0.0256, 0.0218, 0.0271, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 05:27:44,694 INFO [train.py:894] (3/4) Epoch 13, batch 3000, loss[loss=0.2029, simple_loss=0.2762, pruned_loss=0.0648, over 18394.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.285, pruned_loss=0.0659, over 3714867.84 frames. ], batch size: 42, lr: 8.78e-03, grad_scale: 8.0 2022-12-23 05:27:44,694 INFO [train.py:919] (3/4) Computing validation loss 2022-12-23 05:27:55,635 INFO [train.py:928] (3/4) Epoch 13, validation: loss=0.1686, simple_loss=0.268, pruned_loss=0.03461, over 944034.00 frames. 2022-12-23 05:27:55,636 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-23 05:28:01,387 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0 from training. Duration: 21.075 2022-12-23 05:28:04,779 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45081.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:28:05,902 WARNING [train.py:1060] (3/4) Exclude cut with ID 6482-98857-0025-147532-0_sp0.9 from training. Duration: 20.0055625 2022-12-23 05:28:05,909 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0037-132304-0_sp0.9 from training. Duration: 22.05 2022-12-23 05:28:05,921 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0 from training. Duration: 26.8349375 2022-12-23 05:28:10,241 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp1.1 from training. Duration: 22.1090625 2022-12-23 05:28:17,097 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp0.9 from training. Duration: 26.6166875 2022-12-23 05:28:22,051 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2759, 2.0798, 1.8273, 1.7362, 2.2528, 2.9466, 2.6932, 1.9809], device='cuda:3'), covar=tensor([0.0361, 0.0307, 0.0405, 0.0277, 0.0214, 0.0268, 0.0316, 0.0300], device='cuda:3'), in_proj_covar=tensor([0.0087, 0.0121, 0.0143, 0.0122, 0.0110, 0.0110, 0.0093, 0.0122], device='cuda:3'), out_proj_covar=tensor([7.1178e-05, 9.8130e-05, 1.2261e-04, 1.0009e-04, 9.2174e-05, 8.6858e-05, 7.5021e-05, 9.9151e-05], device='cuda:3') 2022-12-23 05:28:36,781 WARNING [train.py:1060] (3/4) Exclude cut with ID 2046-178027-0000-53705-0_sp0.9 from training. Duration: 20.3055625 2022-12-23 05:28:50,181 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.1554, 0.9583, 1.1508, 0.3435, 0.5093, 1.2415, 1.2432, 1.1303], device='cuda:3'), covar=tensor([0.0657, 0.0294, 0.0360, 0.0400, 0.0438, 0.0520, 0.0241, 0.0573], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0161, 0.0117, 0.0134, 0.0143, 0.0135, 0.0152, 0.0156], device='cuda:3'), out_proj_covar=tensor([1.1411e-04, 1.2964e-04, 9.2736e-05, 1.0384e-04, 1.1350e-04, 1.0937e-04, 1.2340e-04, 1.2465e-04], device='cuda:3') 2022-12-23 05:28:53,441 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2631, 2.1359, 1.7110, 1.0613, 2.6890, 2.3995, 2.0979, 1.5925], device='cuda:3'), covar=tensor([0.0314, 0.0356, 0.0465, 0.0629, 0.0186, 0.0268, 0.0368, 0.0785], device='cuda:3'), in_proj_covar=tensor([0.0120, 0.0119, 0.0127, 0.0116, 0.0087, 0.0115, 0.0130, 0.0149], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 05:28:57,476 WARNING [train.py:1060] (3/4) Exclude cut with ID 7205-50138-0008-5373-0_sp0.9 from training. Duration: 20.7 2022-12-23 05:28:59,544 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45117.0, num_to_drop=1, layers_to_drop={3} 2022-12-23 05:29:11,124 INFO [train.py:894] (3/4) Epoch 13, batch 3050, loss[loss=0.2408, simple_loss=0.3065, pruned_loss=0.08751, over 18651.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2848, pruned_loss=0.06611, over 3715929.90 frames. ], batch size: 175, lr: 8.78e-03, grad_scale: 8.0 2022-12-23 05:29:18,124 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.268e+02 5.093e+02 6.049e+02 7.784e+02 1.480e+03, threshold=1.210e+03, percent-clipped=4.0 2022-12-23 05:29:30,053 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2022-12-23 05:29:32,799 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.91 vs. limit=5.0 2022-12-23 05:29:36,806 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45142.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:29:42,998 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0 from training. Duration: 22.48 2022-12-23 05:29:58,124 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp0.9 from training. Duration: 29.816625 2022-12-23 05:30:17,710 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp1.1 from training. Duration: 22.7590625 2022-12-23 05:30:23,388 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0 from training. Duration: 22.555 2022-12-23 05:30:26,298 INFO [train.py:894] (3/4) Epoch 13, batch 3100, loss[loss=0.2099, simple_loss=0.2926, pruned_loss=0.06361, over 18684.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2843, pruned_loss=0.06541, over 3715367.98 frames. ], batch size: 62, lr: 8.78e-03, grad_scale: 8.0 2022-12-23 05:30:39,075 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45183.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:30:43,685 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0005-25975-0_sp0.9 from training. Duration: 21.688875 2022-12-23 05:31:04,497 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6800, 1.1761, 1.2323, 1.3213, 1.8177, 1.7535, 1.8500, 1.1403], device='cuda:3'), covar=tensor([0.0261, 0.0274, 0.0515, 0.0230, 0.0170, 0.0304, 0.0211, 0.0319], device='cuda:3'), in_proj_covar=tensor([0.0087, 0.0120, 0.0143, 0.0122, 0.0110, 0.0110, 0.0093, 0.0122], device='cuda:3'), out_proj_covar=tensor([7.1206e-05, 9.7701e-05, 1.2251e-04, 1.0010e-04, 9.1799e-05, 8.7201e-05, 7.4629e-05, 9.8961e-05], device='cuda:3') 2022-12-23 05:31:14,843 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45206.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:31:16,246 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0_sp0.9 from training. Duration: 22.6 2022-12-23 05:31:42,791 INFO [train.py:894] (3/4) Epoch 13, batch 3150, loss[loss=0.2083, simple_loss=0.2947, pruned_loss=0.06096, over 18633.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2838, pruned_loss=0.06535, over 3715758.38 frames. ], batch size: 53, lr: 8.77e-03, grad_scale: 8.0 2022-12-23 05:31:43,958 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9766, 2.1087, 1.4144, 2.1949, 2.1619, 1.9041, 2.8208, 2.1363], device='cuda:3'), covar=tensor([0.0874, 0.1411, 0.2613, 0.1730, 0.1691, 0.0842, 0.0907, 0.1056], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0198, 0.0241, 0.0287, 0.0229, 0.0186, 0.0211, 0.0199], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 05:31:50,731 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.862e+02 4.724e+02 5.980e+02 7.666e+02 1.502e+03, threshold=1.196e+03, percent-clipped=2.0 2022-12-23 05:31:52,595 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0 from training. Duration: 24.32 2022-12-23 05:31:52,714 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=45231.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:32:21,078 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.0268, 0.7709, 0.9158, 1.1467, 1.2694, 1.0992, 1.0586, 0.9059], device='cuda:3'), covar=tensor([0.0259, 0.0253, 0.0516, 0.0226, 0.0218, 0.0336, 0.0248, 0.0295], device='cuda:3'), in_proj_covar=tensor([0.0087, 0.0120, 0.0144, 0.0122, 0.0110, 0.0109, 0.0093, 0.0122], device='cuda:3'), out_proj_covar=tensor([7.0910e-05, 9.7510e-05, 1.2262e-04, 1.0006e-04, 9.1615e-05, 8.6567e-05, 7.4403e-05, 9.8715e-05], device='cuda:3') 2022-12-23 05:32:24,032 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5998, 1.9092, 1.3529, 2.1495, 2.4963, 1.5719, 1.4785, 1.2997], device='cuda:3'), covar=tensor([0.1922, 0.1708, 0.1654, 0.0960, 0.1245, 0.1176, 0.2073, 0.1474], device='cuda:3'), in_proj_covar=tensor([0.0241, 0.0218, 0.0207, 0.0193, 0.0257, 0.0193, 0.0217, 0.0192], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 05:32:29,054 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=45254.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:32:53,409 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-276745-0093-13116-0_sp0.9 from training. Duration: 21.061125 2022-12-23 05:32:59,693 INFO [train.py:894] (3/4) Epoch 13, batch 3200, loss[loss=0.225, simple_loss=0.2979, pruned_loss=0.07599, over 18528.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2823, pruned_loss=0.06439, over 3715063.11 frames. ], batch size: 58, lr: 8.77e-03, grad_scale: 8.0 2022-12-23 05:33:07,090 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0024-15855-0_sp0.9 from training. Duration: 20.32225 2022-12-23 05:33:22,376 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp1.1 from training. Duration: 0.7545625 2022-12-23 05:33:37,195 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0_sp0.9 from training. Duration: 23.9333125 2022-12-23 05:34:09,854 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp1.1 from training. Duration: 20.17275 2022-12-23 05:34:16,010 INFO [train.py:894] (3/4) Epoch 13, batch 3250, loss[loss=0.2451, simple_loss=0.3016, pruned_loss=0.09432, over 18681.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2823, pruned_loss=0.06427, over 3715322.75 frames. ], batch size: 48, lr: 8.76e-03, grad_scale: 8.0 2022-12-23 05:34:16,033 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp1.1 from training. Duration: 20.436375 2022-12-23 05:34:24,014 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.138e+02 5.303e+02 6.543e+02 7.796e+02 1.452e+03, threshold=1.309e+03, percent-clipped=4.0 2022-12-23 05:34:52,058 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45348.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 05:35:32,627 INFO [train.py:894] (3/4) Epoch 13, batch 3300, loss[loss=0.2038, simple_loss=0.2916, pruned_loss=0.05802, over 18560.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2823, pruned_loss=0.06406, over 3715044.45 frames. ], batch size: 55, lr: 8.76e-03, grad_scale: 8.0 2022-12-23 05:35:38,570 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0_sp0.9 from training. Duration: 23.1055625 2022-12-23 05:35:40,141 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp1.1 from training. Duration: 23.5 2022-12-23 05:35:52,651 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp0.9 from training. Duration: 26.62775 2022-12-23 05:36:04,497 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0 from training. Duration: 21.105 2022-12-23 05:36:09,155 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0_sp0.9 from training. Duration: 24.411125 2022-12-23 05:36:24,773 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45409.0, num_to_drop=1, layers_to_drop={3} 2022-12-23 05:36:29,117 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45412.0, num_to_drop=1, layers_to_drop={3} 2022-12-23 05:36:35,310 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp1.1 from training. Duration: 21.263625 2022-12-23 05:36:49,535 INFO [train.py:894] (3/4) Epoch 13, batch 3350, loss[loss=0.22, simple_loss=0.2971, pruned_loss=0.07143, over 18614.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2822, pruned_loss=0.06427, over 3715381.14 frames. ], batch size: 53, lr: 8.75e-03, grad_scale: 8.0 2022-12-23 05:36:57,612 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.516e+02 4.658e+02 5.777e+02 7.131e+02 1.099e+03, threshold=1.155e+03, percent-clipped=0.0 2022-12-23 05:37:05,363 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0 from training. Duration: 20.795 2022-12-23 05:37:08,427 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45437.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:37:16,371 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0 from training. Duration: 24.76 2022-12-23 05:37:17,973 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0_sp0.9 from training. Duration: 22.25 2022-12-23 05:37:30,147 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0491, 1.5123, 0.9292, 1.7365, 2.2228, 1.6367, 1.9004, 2.2154], device='cuda:3'), covar=tensor([0.1471, 0.2072, 0.2452, 0.1459, 0.1816, 0.1691, 0.1334, 0.1325], device='cuda:3'), in_proj_covar=tensor([0.0091, 0.0099, 0.0118, 0.0096, 0.0115, 0.0091, 0.0098, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 05:37:42,598 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp1.1 from training. Duration: 20.5045625 2022-12-23 05:38:05,791 INFO [train.py:894] (3/4) Epoch 13, batch 3400, loss[loss=0.2152, simple_loss=0.2867, pruned_loss=0.07189, over 18585.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2815, pruned_loss=0.06447, over 3715193.22 frames. ], batch size: 57, lr: 8.75e-03, grad_scale: 8.0 2022-12-23 05:39:16,508 INFO [train.py:894] (3/4) Epoch 13, batch 3450, loss[loss=0.2469, simple_loss=0.3065, pruned_loss=0.09362, over 18436.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2819, pruned_loss=0.06478, over 3715095.44 frames. ], batch size: 48, lr: 8.74e-03, grad_scale: 8.0 2022-12-23 05:39:23,734 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.221e+02 5.288e+02 6.220e+02 8.428e+02 1.484e+03, threshold=1.244e+03, percent-clipped=4.0 2022-12-23 05:39:38,110 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6085, 1.3485, 1.2261, 1.8821, 1.5803, 3.3859, 1.3055, 1.4534], device='cuda:3'), covar=tensor([0.0906, 0.2041, 0.1336, 0.0999, 0.1641, 0.0284, 0.1574, 0.1839], device='cuda:3'), in_proj_covar=tensor([0.0072, 0.0081, 0.0074, 0.0074, 0.0091, 0.0073, 0.0084, 0.0077], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 05:40:29,607 INFO [train.py:894] (3/4) Epoch 13, batch 3500, loss[loss=0.2455, simple_loss=0.3129, pruned_loss=0.08903, over 18651.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2825, pruned_loss=0.06508, over 3715211.61 frames. ], batch size: 180, lr: 8.74e-03, grad_scale: 8.0 2022-12-23 05:40:50,933 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp1.1 from training. Duration: 22.2954375 2022-12-23 05:41:02,151 INFO [train.py:894] (3/4) Epoch 14, batch 0, loss[loss=0.1835, simple_loss=0.2589, pruned_loss=0.05405, over 18709.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2589, pruned_loss=0.05405, over 18709.00 frames. ], batch size: 41, lr: 8.42e-03, grad_scale: 8.0 2022-12-23 05:41:02,152 INFO [train.py:919] (3/4) Computing validation loss 2022-12-23 05:41:12,887 INFO [train.py:928] (3/4) Epoch 14, validation: loss=0.1689, simple_loss=0.2684, pruned_loss=0.03472, over 944034.00 frames. 2022-12-23 05:41:12,888 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-23 05:41:38,230 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4964, 2.0494, 1.5815, 1.4199, 2.1564, 3.0451, 3.0437, 1.8971], device='cuda:3'), covar=tensor([0.0311, 0.0343, 0.0574, 0.0339, 0.0271, 0.0263, 0.0201, 0.0352], device='cuda:3'), in_proj_covar=tensor([0.0088, 0.0122, 0.0144, 0.0123, 0.0112, 0.0111, 0.0093, 0.0124], device='cuda:3'), out_proj_covar=tensor([7.2431e-05, 9.8929e-05, 1.2311e-04, 1.0093e-04, 9.3779e-05, 8.7287e-05, 7.4922e-05, 1.0009e-04], device='cuda:3') 2022-12-23 05:42:09,138 WARNING [train.py:1060] (3/4) Exclude cut with ID 298-126791-0067-24026-0_sp0.9 from training. Duration: 21.438875 2022-12-23 05:42:13,477 WARNING [train.py:1060] (3/4) Exclude cut with ID 5652-39938-0025-23684-0_sp0.9 from training. Duration: 22.2055625 2022-12-23 05:42:26,609 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.512e+02 4.529e+02 5.962e+02 7.073e+02 1.880e+03, threshold=1.192e+03, percent-clipped=5.0 2022-12-23 05:42:28,160 INFO [train.py:894] (3/4) Epoch 14, batch 50, loss[loss=0.1984, simple_loss=0.2817, pruned_loss=0.05757, over 18696.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2774, pruned_loss=0.05505, over 837275.86 frames. ], batch size: 50, lr: 8.41e-03, grad_scale: 8.0 2022-12-23 05:43:02,942 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4179, 2.0650, 1.5380, 2.2445, 1.8145, 1.9215, 1.8763, 2.4364], device='cuda:3'), covar=tensor([0.1968, 0.2932, 0.1761, 0.2749, 0.3456, 0.1011, 0.2752, 0.0783], device='cuda:3'), in_proj_covar=tensor([0.0285, 0.0276, 0.0234, 0.0348, 0.0258, 0.0219, 0.0272, 0.0197], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 05:43:43,697 INFO [train.py:894] (3/4) Epoch 14, batch 100, loss[loss=0.1972, simple_loss=0.2777, pruned_loss=0.0584, over 18642.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2761, pruned_loss=0.05348, over 1476111.70 frames. ], batch size: 174, lr: 8.41e-03, grad_scale: 8.0 2022-12-23 05:44:17,228 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45702.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:44:19,927 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45704.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 05:44:32,614 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45712.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 05:44:36,245 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2022-12-23 05:44:59,363 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.734e+02 3.533e+02 4.272e+02 5.188e+02 1.286e+03, threshold=8.545e+02, percent-clipped=1.0 2022-12-23 05:45:00,665 INFO [train.py:894] (3/4) Epoch 14, batch 150, loss[loss=0.1822, simple_loss=0.2657, pruned_loss=0.04942, over 18456.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2746, pruned_loss=0.05216, over 1971979.76 frames. ], batch size: 50, lr: 8.40e-03, grad_scale: 8.0 2022-12-23 05:45:09,950 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0 from training. Duration: 24.525 2022-12-23 05:45:10,180 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45737.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:45:43,366 WARNING [train.py:1060] (3/4) Exclude cut with ID 3699-47246-0007-3408-0_sp0.9 from training. Duration: 20.26675 2022-12-23 05:45:45,261 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=45760.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:45:49,780 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45763.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:45:56,271 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp0.9 from training. Duration: 27.25 2022-12-23 05:46:05,070 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6314, 3.7246, 3.6242, 1.4148, 3.7803, 2.6598, 0.6496, 2.4658], device='cuda:3'), covar=tensor([0.1930, 0.0866, 0.1385, 0.3910, 0.0822, 0.0975, 0.5126, 0.1589], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0132, 0.0151, 0.0122, 0.0132, 0.0108, 0.0142, 0.0111], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 05:46:05,224 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45774.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:46:15,003 INFO [train.py:894] (3/4) Epoch 14, batch 200, loss[loss=0.1682, simple_loss=0.2522, pruned_loss=0.04209, over 18558.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2735, pruned_loss=0.05192, over 2358705.61 frames. ], batch size: 44, lr: 8.40e-03, grad_scale: 8.0 2022-12-23 05:46:15,384 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.2613, 3.1605, 2.2652, 1.5272, 3.7251, 3.5580, 2.8871, 2.1718], device='cuda:3'), covar=tensor([0.0333, 0.0328, 0.0527, 0.0737, 0.0151, 0.0290, 0.0451, 0.0783], device='cuda:3'), in_proj_covar=tensor([0.0121, 0.0121, 0.0128, 0.0118, 0.0089, 0.0118, 0.0133, 0.0152], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 05:46:21,453 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7141, 1.5093, 1.1832, 0.2094, 1.1681, 1.5508, 1.3709, 1.4736], device='cuda:3'), covar=tensor([0.0633, 0.0521, 0.0956, 0.1627, 0.1033, 0.1553, 0.1579, 0.0630], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0182, 0.0202, 0.0193, 0.0209, 0.0195, 0.0211, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 05:46:21,562 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9075, 1.8275, 1.3639, 1.6673, 1.6258, 1.7429, 1.5891, 1.8197], device='cuda:3'), covar=tensor([0.2074, 0.2914, 0.1909, 0.2517, 0.3265, 0.1076, 0.2708, 0.0908], device='cuda:3'), in_proj_covar=tensor([0.0288, 0.0282, 0.0238, 0.0352, 0.0262, 0.0224, 0.0277, 0.0200], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 05:46:25,675 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=45785.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:47:08,595 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2965, 1.5537, 2.5208, 4.5260, 3.1886, 2.8064, 0.8765, 3.2366], device='cuda:3'), covar=tensor([0.1769, 0.1779, 0.1418, 0.0417, 0.0977, 0.1148, 0.2361, 0.0801], device='cuda:3'), in_proj_covar=tensor([0.0101, 0.0115, 0.0130, 0.0135, 0.0104, 0.0134, 0.0128, 0.0108], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 05:47:15,500 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0 from training. Duration: 21.68 2022-12-23 05:47:25,787 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0 from training. Duration: 21.6300625 2022-12-23 05:47:28,251 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2022-12-23 05:47:34,693 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.785e+02 3.863e+02 4.748e+02 5.932e+02 2.201e+03, threshold=9.496e+02, percent-clipped=7.0 2022-12-23 05:47:36,859 INFO [train.py:894] (3/4) Epoch 14, batch 250, loss[loss=0.1987, simple_loss=0.2861, pruned_loss=0.05563, over 18459.00 frames. ], tot_loss[loss=0.1869, simple_loss=0.2718, pruned_loss=0.05095, over 2658766.86 frames. ], batch size: 54, lr: 8.40e-03, grad_scale: 8.0 2022-12-23 05:47:44,289 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45835.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:47:51,504 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0_sp0.9 from training. Duration: 24.033375 2022-12-23 05:48:19,318 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45859.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:48:46,429 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0 from training. Duration: 22.905 2022-12-23 05:48:47,826 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp1.1 from training. Duration: 23.4318125 2022-12-23 05:48:51,806 INFO [train.py:894] (3/4) Epoch 14, batch 300, loss[loss=0.1624, simple_loss=0.258, pruned_loss=0.03337, over 18473.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2719, pruned_loss=0.05067, over 2892762.45 frames. ], batch size: 54, lr: 8.39e-03, grad_scale: 8.0 2022-12-23 05:49:09,871 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7335, 1.2753, 0.6064, 1.3817, 2.0153, 0.9441, 1.5484, 1.6092], device='cuda:3'), covar=tensor([0.1546, 0.2105, 0.2352, 0.1549, 0.1618, 0.1766, 0.1321, 0.1583], device='cuda:3'), in_proj_covar=tensor([0.0091, 0.0099, 0.0118, 0.0095, 0.0113, 0.0090, 0.0098, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 05:49:50,479 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45920.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:49:52,661 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2022-12-23 05:50:05,349 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.191e+02 3.620e+02 4.473e+02 5.511e+02 1.268e+03, threshold=8.946e+02, percent-clipped=1.0 2022-12-23 05:50:06,952 INFO [train.py:894] (3/4) Epoch 14, batch 350, loss[loss=0.1995, simple_loss=0.2796, pruned_loss=0.05972, over 18545.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2729, pruned_loss=0.05103, over 3076554.95 frames. ], batch size: 49, lr: 8.39e-03, grad_scale: 8.0 2022-12-23 05:50:46,125 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp1.1 from training. Duration: 20.82275 2022-12-23 05:50:47,661 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp0.9 from training. Duration: 25.45 2022-12-23 05:50:56,054 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2022-12-23 05:51:11,746 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2022-12-23 05:51:24,538 INFO [train.py:894] (3/4) Epoch 14, batch 400, loss[loss=0.2209, simple_loss=0.2988, pruned_loss=0.07151, over 18458.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2747, pruned_loss=0.0522, over 3217582.11 frames. ], batch size: 64, lr: 8.38e-03, grad_scale: 8.0 2022-12-23 05:51:48,666 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0 from training. Duration: 25.775 2022-12-23 05:52:01,480 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46004.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 05:52:03,661 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2022-12-23 05:52:13,154 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0_sp0.9 from training. Duration: 22.25 2022-12-23 05:52:40,421 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.282e+02 3.982e+02 5.008e+02 5.983e+02 8.979e+02, threshold=1.002e+03, percent-clipped=1.0 2022-12-23 05:52:40,477 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0 from training. Duration: 26.205 2022-12-23 05:52:41,921 INFO [train.py:894] (3/4) Epoch 14, batch 450, loss[loss=0.1802, simple_loss=0.2676, pruned_loss=0.04638, over 18701.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2759, pruned_loss=0.05256, over 3327351.90 frames. ], batch size: 50, lr: 8.38e-03, grad_scale: 8.0 2022-12-23 05:52:43,805 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4537, 2.0252, 1.7257, 0.6714, 1.8018, 1.8552, 1.3386, 2.0840], device='cuda:3'), covar=tensor([0.0681, 0.0785, 0.1498, 0.2055, 0.1401, 0.1571, 0.1997, 0.0836], device='cuda:3'), in_proj_covar=tensor([0.0165, 0.0181, 0.0201, 0.0192, 0.0206, 0.0193, 0.0208, 0.0192], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 05:52:49,630 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5963, 2.3273, 1.7064, 1.1891, 3.0129, 2.6111, 2.0783, 1.8961], device='cuda:3'), covar=tensor([0.0372, 0.0384, 0.0589, 0.0775, 0.0174, 0.0315, 0.0536, 0.0733], device='cuda:3'), in_proj_covar=tensor([0.0121, 0.0120, 0.0127, 0.0118, 0.0089, 0.0118, 0.0133, 0.0150], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 05:52:56,705 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp0.9 from training. Duration: 30.1555625 2022-12-23 05:53:01,227 WARNING [train.py:1060] (3/4) Exclude cut with ID 1265-135635-0050-6781-0_sp0.9 from training. Duration: 21.8333125 2022-12-23 05:53:10,779 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp1.1 from training. Duration: 20.6545625 2022-12-23 05:53:13,747 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=46052.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 05:53:22,315 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46058.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:53:49,344 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4466, 2.0444, 1.9811, 1.2826, 3.0923, 2.5821, 2.1484, 1.3280], device='cuda:3'), covar=tensor([0.0438, 0.0518, 0.0560, 0.0858, 0.0174, 0.0387, 0.0634, 0.1477], device='cuda:3'), in_proj_covar=tensor([0.0122, 0.0121, 0.0128, 0.0119, 0.0089, 0.0118, 0.0133, 0.0151], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 05:53:53,850 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0045-39920-0_sp0.9 from training. Duration: 20.52225 2022-12-23 05:53:56,434 INFO [train.py:894] (3/4) Epoch 14, batch 500, loss[loss=0.2108, simple_loss=0.294, pruned_loss=0.06383, over 18472.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2772, pruned_loss=0.05345, over 3413419.49 frames. ], batch size: 54, lr: 8.37e-03, grad_scale: 8.0 2022-12-23 05:54:12,943 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp0.9 from training. Duration: 29.1166875 2022-12-23 05:54:28,840 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6406, 1.5129, 1.6520, 1.6335, 1.1022, 3.3774, 1.5770, 2.1356], device='cuda:3'), covar=tensor([0.3177, 0.2028, 0.1840, 0.1981, 0.1461, 0.0179, 0.1416, 0.0766], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0116, 0.0128, 0.0120, 0.0103, 0.0098, 0.0092, 0.0091], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 05:55:10,416 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.469e+02 4.016e+02 4.935e+02 6.120e+02 1.386e+03, threshold=9.871e+02, percent-clipped=2.0 2022-12-23 05:55:10,604 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46130.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:55:12,151 INFO [train.py:894] (3/4) Epoch 14, batch 550, loss[loss=0.1947, simple_loss=0.2842, pruned_loss=0.05258, over 18500.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2771, pruned_loss=0.05336, over 3478288.18 frames. ], batch size: 64, lr: 8.37e-03, grad_scale: 8.0 2022-12-23 05:55:15,281 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133211-0007-59831-0_sp0.9 from training. Duration: 21.388875 2022-12-23 05:55:20,926 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2022-12-23 05:55:50,854 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0 from training. Duration: 22.72 2022-12-23 05:55:52,280 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0_sp0.9 from training. Duration: 22.7444375 2022-12-23 05:56:27,831 INFO [train.py:894] (3/4) Epoch 14, batch 600, loss[loss=0.1968, simple_loss=0.2896, pruned_loss=0.05199, over 18720.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2774, pruned_loss=0.05338, over 3530518.08 frames. ], batch size: 54, lr: 8.36e-03, grad_scale: 8.0 2022-12-23 05:56:32,144 WARNING [train.py:1060] (3/4) Exclude cut with ID 4133-6541-0027-40495-0_sp1.1 from training. Duration: 0.9681875 2022-12-23 05:56:35,082 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0_sp0.9 from training. Duration: 22.3166875 2022-12-23 05:56:41,127 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133212-0015-59917-0_sp0.9 from training. Duration: 21.8166875 2022-12-23 05:57:18,745 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46215.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:57:35,006 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6395, 1.2358, 1.8255, 3.1199, 2.2777, 2.3361, 0.4955, 2.2284], device='cuda:3'), covar=tensor([0.1846, 0.1809, 0.1565, 0.0575, 0.1077, 0.1179, 0.2545, 0.1087], device='cuda:3'), in_proj_covar=tensor([0.0101, 0.0115, 0.0129, 0.0134, 0.0104, 0.0134, 0.0128, 0.0107], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 05:57:41,977 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.383e+02 4.078e+02 4.914e+02 6.557e+02 2.060e+03, threshold=9.827e+02, percent-clipped=9.0 2022-12-23 05:57:43,540 INFO [train.py:894] (3/4) Epoch 14, batch 650, loss[loss=0.197, simple_loss=0.288, pruned_loss=0.053, over 18539.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.2774, pruned_loss=0.05368, over 3571020.49 frames. ], batch size: 98, lr: 8.36e-03, grad_scale: 8.0 2022-12-23 05:58:28,353 WARNING [train.py:1060] (3/4) Exclude cut with ID 4957-30119-0041-23990-0_sp0.9 from training. Duration: 20.22775 2022-12-23 05:58:52,471 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46276.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 05:58:59,235 INFO [train.py:894] (3/4) Epoch 14, batch 700, loss[loss=0.1944, simple_loss=0.2824, pruned_loss=0.05322, over 18429.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2772, pruned_loss=0.05368, over 3601572.86 frames. ], batch size: 48, lr: 8.36e-03, grad_scale: 8.0 2022-12-23 05:59:13,271 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2022-12-23 05:59:13,910 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp1.1 from training. Duration: 24.67275 2022-12-23 05:59:34,376 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2022-12-23 05:59:43,694 WARNING [train.py:1060] (3/4) Exclude cut with ID 3082-165428-0081-50734-0_sp0.9 from training. Duration: 21.8055625 2022-12-23 06:00:13,188 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.178e+02 4.025e+02 4.936e+02 5.531e+02 1.060e+03, threshold=9.873e+02, percent-clipped=1.0 2022-12-23 06:00:14,545 INFO [train.py:894] (3/4) Epoch 14, batch 750, loss[loss=0.2267, simple_loss=0.3058, pruned_loss=0.07377, over 18656.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2781, pruned_loss=0.05383, over 3626429.73 frames. ], batch size: 175, lr: 8.35e-03, grad_scale: 8.0 2022-12-23 06:00:21,785 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0_sp0.9 from training. Duration: 22.6666875 2022-12-23 06:00:23,621 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46337.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:00:54,962 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46358.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:00:55,361 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2022-12-23 06:01:23,560 WARNING [train.py:1060] (3/4) Exclude cut with ID 2411-132532-0017-82279-0_sp1.1 from training. Duration: 0.9681875 2022-12-23 06:01:29,304 INFO [train.py:894] (3/4) Epoch 14, batch 800, loss[loss=0.2066, simple_loss=0.298, pruned_loss=0.05763, over 18593.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2778, pruned_loss=0.05383, over 3646527.04 frames. ], batch size: 51, lr: 8.35e-03, grad_scale: 8.0 2022-12-23 06:01:50,105 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0 from training. Duration: 22.485 2022-12-23 06:02:06,130 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=46406.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:02:27,798 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp1.1 from training. Duration: 23.82275 2022-12-23 06:02:40,428 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0 from training. Duration: 20.77 2022-12-23 06:02:42,833 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2022-12-23 06:02:43,274 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.502e+02 3.825e+02 4.476e+02 5.573e+02 1.333e+03, threshold=8.951e+02, percent-clipped=4.0 2022-12-23 06:02:43,513 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46430.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:02:44,557 INFO [train.py:894] (3/4) Epoch 14, batch 850, loss[loss=0.1982, simple_loss=0.2858, pruned_loss=0.05525, over 18463.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2781, pruned_loss=0.05396, over 3661723.36 frames. ], batch size: 54, lr: 8.34e-03, grad_scale: 8.0 2022-12-23 06:02:47,786 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0_sp0.9 from training. Duration: 24.088875 2022-12-23 06:03:15,534 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp1.1 from training. Duration: 20.4409375 2022-12-23 06:03:55,977 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=46478.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:04:00,227 INFO [train.py:894] (3/4) Epoch 14, batch 900, loss[loss=0.1893, simple_loss=0.2818, pruned_loss=0.04841, over 18593.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2782, pruned_loss=0.05393, over 3672549.50 frames. ], batch size: 97, lr: 8.34e-03, grad_scale: 8.0 2022-12-23 06:04:31,600 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0_sp0.9 from training. Duration: 22.511125 2022-12-23 06:04:31,624 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0 from training. Duration: 20.675 2022-12-23 06:04:45,660 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3502, 1.3525, 1.5142, 0.8889, 1.5546, 1.4979, 1.2327, 1.6774], device='cuda:3'), covar=tensor([0.0997, 0.1744, 0.1093, 0.1401, 0.0669, 0.0966, 0.2159, 0.0539], device='cuda:3'), in_proj_covar=tensor([0.0200, 0.0201, 0.0206, 0.0194, 0.0176, 0.0214, 0.0208, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 06:04:52,123 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46515.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:05:17,323 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.454e+02 3.881e+02 4.626e+02 5.706e+02 1.079e+03, threshold=9.251e+02, percent-clipped=7.0 2022-12-23 06:05:17,340 INFO [train.py:894] (3/4) Epoch 14, batch 950, loss[loss=0.1677, simple_loss=0.2549, pruned_loss=0.04029, over 18451.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2773, pruned_loss=0.05362, over 3681464.07 frames. ], batch size: 50, lr: 8.33e-03, grad_scale: 8.0 2022-12-23 06:05:50,874 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.4788, 2.4938, 1.7150, 3.0698, 2.2231, 2.4319, 2.4306, 3.7425], device='cuda:3'), covar=tensor([0.1581, 0.3071, 0.1819, 0.2824, 0.3487, 0.0911, 0.2935, 0.0562], device='cuda:3'), in_proj_covar=tensor([0.0282, 0.0278, 0.0233, 0.0342, 0.0257, 0.0218, 0.0272, 0.0198], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 06:06:07,186 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=46563.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:06:11,672 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp0.9 from training. Duration: 24.9833125 2022-12-23 06:06:33,821 INFO [train.py:894] (3/4) Epoch 14, batch 1000, loss[loss=0.2084, simple_loss=0.2957, pruned_loss=0.06057, over 18588.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2783, pruned_loss=0.05392, over 3688834.68 frames. ], batch size: 98, lr: 8.33e-03, grad_scale: 8.0 2022-12-23 06:06:42,969 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0 from training. Duration: 27.14 2022-12-23 06:06:55,885 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46595.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:06:58,653 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0 from training. Duration: 22.44 2022-12-23 06:07:06,740 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46602.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:07:50,415 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.396e+02 3.981e+02 4.528e+02 6.046e+02 1.105e+03, threshold=9.056e+02, percent-clipped=1.0 2022-12-23 06:07:50,431 INFO [train.py:894] (3/4) Epoch 14, batch 1050, loss[loss=0.2098, simple_loss=0.301, pruned_loss=0.05933, over 18685.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2788, pruned_loss=0.05386, over 3694522.36 frames. ], batch size: 60, lr: 8.32e-03, grad_scale: 8.0 2022-12-23 06:07:51,962 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46632.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:08:12,659 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0060-62364-0_sp0.9 from training. Duration: 21.361125 2022-12-23 06:08:18,510 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp1.1 from training. Duration: 27.0318125 2022-12-23 06:08:28,926 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46656.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 06:08:30,071 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp0.9 from training. Duration: 28.638875 2022-12-23 06:08:39,724 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46663.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:08:46,661 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0 from training. Duration: 20.4 2022-12-23 06:09:06,401 INFO [train.py:894] (3/4) Epoch 14, batch 1100, loss[loss=0.2028, simple_loss=0.2883, pruned_loss=0.05864, over 18708.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2779, pruned_loss=0.05322, over 3700184.79 frames. ], batch size: 99, lr: 8.32e-03, grad_scale: 8.0 2022-12-23 06:09:19,492 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0 from training. Duration: 20.025 2022-12-23 06:09:19,506 WARNING [train.py:1060] (3/4) Exclude cut with ID 2364-131735-0112-64612-0_sp0.9 from training. Duration: 20.488875 2022-12-23 06:09:23,641 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0 from training. Duration: 29.735 2022-12-23 06:09:33,807 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6162, 2.7179, 2.1176, 1.8279, 2.5378, 3.1079, 2.9949, 2.4316], device='cuda:3'), covar=tensor([0.0369, 0.0218, 0.0362, 0.0232, 0.0194, 0.0236, 0.0255, 0.0224], device='cuda:3'), in_proj_covar=tensor([0.0086, 0.0119, 0.0144, 0.0122, 0.0110, 0.0109, 0.0093, 0.0121], device='cuda:3'), out_proj_covar=tensor([7.0666e-05, 9.6729e-05, 1.2239e-04, 9.9542e-05, 9.1870e-05, 8.6120e-05, 7.4255e-05, 9.7401e-05], device='cuda:3') 2022-12-23 06:09:42,665 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.68 vs. limit=5.0 2022-12-23 06:10:10,849 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2022-12-23 06:10:21,528 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.330e+02 4.103e+02 4.927e+02 5.923e+02 1.016e+03, threshold=9.854e+02, percent-clipped=2.0 2022-12-23 06:10:21,544 INFO [train.py:894] (3/4) Epoch 14, batch 1150, loss[loss=0.1938, simple_loss=0.2707, pruned_loss=0.0584, over 18573.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2777, pruned_loss=0.05286, over 3702648.82 frames. ], batch size: 49, lr: 8.32e-03, grad_scale: 8.0 2022-12-23 06:10:46,306 WARNING [train.py:1060] (3/4) Exclude cut with ID 7276-92427-0014-12983-0_sp0.9 from training. Duration: 21.3055625 2022-12-23 06:10:48,292 WARNING [train.py:1060] (3/4) Exclude cut with ID 1025-75365-0008-79168-0_sp0.9 from training. Duration: 22.0666875 2022-12-23 06:11:36,876 INFO [train.py:894] (3/4) Epoch 14, batch 1200, loss[loss=0.1985, simple_loss=0.2812, pruned_loss=0.05794, over 18459.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2768, pruned_loss=0.05257, over 3704204.94 frames. ], batch size: 54, lr: 8.31e-03, grad_scale: 8.0 2022-12-23 06:12:05,688 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.80 vs. limit=5.0 2022-12-23 06:12:16,209 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1258, 1.4390, 1.7223, 1.7952, 2.0835, 2.0992, 1.9368, 1.5982], device='cuda:3'), covar=tensor([0.1889, 0.2985, 0.2291, 0.2494, 0.1741, 0.0856, 0.2757, 0.1130], device='cuda:3'), in_proj_covar=tensor([0.0257, 0.0294, 0.0268, 0.0304, 0.0291, 0.0241, 0.0319, 0.0228], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 06:12:21,735 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46810.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:12:39,989 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0 from training. Duration: 20.26 2022-12-23 06:12:52,911 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.710e+02 4.042e+02 4.765e+02 5.744e+02 1.398e+03, threshold=9.530e+02, percent-clipped=5.0 2022-12-23 06:12:52,927 INFO [train.py:894] (3/4) Epoch 14, batch 1250, loss[loss=0.2326, simple_loss=0.3046, pruned_loss=0.08033, over 18600.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.2763, pruned_loss=0.05231, over 3706505.08 frames. ], batch size: 180, lr: 8.31e-03, grad_scale: 8.0 2022-12-23 06:12:52,978 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0030-9324-0_sp0.9 from training. Duration: 21.3444375 2022-12-23 06:13:53,574 WARNING [train.py:1060] (3/4) Exclude cut with ID 497-129325-0061-62254-0_sp1.1 from training. Duration: 0.97725 2022-12-23 06:13:54,005 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46871.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:14:08,372 INFO [train.py:894] (3/4) Epoch 14, batch 1300, loss[loss=0.1968, simple_loss=0.2862, pruned_loss=0.0537, over 18671.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2761, pruned_loss=0.05228, over 3708416.40 frames. ], batch size: 60, lr: 8.30e-03, grad_scale: 8.0 2022-12-23 06:14:16,181 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6541, 1.5413, 1.6942, 1.6430, 1.0947, 3.0190, 1.2870, 1.8921], device='cuda:3'), covar=tensor([0.2997, 0.1957, 0.1778, 0.1839, 0.1338, 0.0218, 0.1560, 0.0759], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0117, 0.0127, 0.0121, 0.0104, 0.0098, 0.0094, 0.0091], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 06:14:35,021 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0_sp0.9 from training. Duration: 22.97225 2022-12-23 06:15:04,764 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0047-39922-0_sp0.9 from training. Duration: 21.97775 2022-12-23 06:15:18,420 WARNING [train.py:1060] (3/4) Exclude cut with ID 1112-1043-0006-89194-0_sp0.9 from training. Duration: 21.8333125 2022-12-23 06:15:24,186 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.454e+02 4.083e+02 4.755e+02 5.873e+02 8.808e+02, threshold=9.511e+02, percent-clipped=0.0 2022-12-23 06:15:24,202 INFO [train.py:894] (3/4) Epoch 14, batch 1350, loss[loss=0.1946, simple_loss=0.2719, pruned_loss=0.05862, over 18671.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2766, pruned_loss=0.05274, over 3708566.76 frames. ], batch size: 48, lr: 8.30e-03, grad_scale: 8.0 2022-12-23 06:15:25,930 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46932.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:15:28,593 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0 from training. Duration: 20.47 2022-12-23 06:15:32,912 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.60 vs. limit=2.0 2022-12-23 06:15:54,701 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46951.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 06:16:05,828 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46958.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:16:33,584 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0037-39912-0_sp0.9 from training. Duration: 20.67225 2022-12-23 06:16:38,078 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=46980.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:16:39,249 INFO [train.py:894] (3/4) Epoch 14, batch 1400, loss[loss=0.1742, simple_loss=0.2589, pruned_loss=0.04481, over 18672.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2761, pruned_loss=0.05251, over 3709071.33 frames. ], batch size: 48, lr: 8.29e-03, grad_scale: 8.0 2022-12-23 06:16:50,106 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp0.9 from training. Duration: 25.2444375 2022-12-23 06:17:14,819 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0021-76797-0_sp0.9 from training. Duration: 21.1445 2022-12-23 06:17:56,990 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.138e+02 3.877e+02 4.597e+02 5.782e+02 1.688e+03, threshold=9.194e+02, percent-clipped=3.0 2022-12-23 06:17:57,006 INFO [train.py:894] (3/4) Epoch 14, batch 1450, loss[loss=0.1724, simple_loss=0.2634, pruned_loss=0.04074, over 18459.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2767, pruned_loss=0.05249, over 3710031.62 frames. ], batch size: 50, lr: 8.29e-03, grad_scale: 8.0 2022-12-23 06:18:29,687 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp0.9 from training. Duration: 33.038875 2022-12-23 06:19:06,813 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64291-0000-16059-0_sp0.9 from training. Duration: 20.0944375 2022-12-23 06:19:12,450 INFO [train.py:894] (3/4) Epoch 14, batch 1500, loss[loss=0.1932, simple_loss=0.2708, pruned_loss=0.05779, over 18697.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2774, pruned_loss=0.05248, over 3711328.48 frames. ], batch size: 46, lr: 8.28e-03, grad_scale: 8.0 2022-12-23 06:19:19,909 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp1.1 from training. Duration: 20.4 2022-12-23 06:19:29,869 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0 from training. Duration: 20.085 2022-12-23 06:19:40,525 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0_sp0.9 from training. Duration: 23.07775 2022-12-23 06:20:25,402 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp0.9 from training. Duration: 24.9333125 2022-12-23 06:20:28,369 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.178e+02 3.794e+02 4.635e+02 5.636e+02 1.005e+03, threshold=9.269e+02, percent-clipped=3.0 2022-12-23 06:20:28,385 INFO [train.py:894] (3/4) Epoch 14, batch 1550, loss[loss=0.1941, simple_loss=0.2802, pruned_loss=0.05399, over 18524.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.2762, pruned_loss=0.05227, over 3712555.36 frames. ], batch size: 58, lr: 8.28e-03, grad_scale: 8.0 2022-12-23 06:21:10,299 WARNING [train.py:1060] (3/4) Exclude cut with ID 5118-111612-0016-124680-0_sp0.9 from training. Duration: 20.388875 2022-12-23 06:21:17,534 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp1.1 from training. Duration: 20.3590625 2022-12-23 06:21:21,995 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47166.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:21:44,087 INFO [train.py:894] (3/4) Epoch 14, batch 1600, loss[loss=0.2105, simple_loss=0.296, pruned_loss=0.06247, over 18590.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2768, pruned_loss=0.05246, over 3713658.01 frames. ], batch size: 51, lr: 8.28e-03, grad_scale: 8.0 2022-12-23 06:22:25,302 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0_sp1.1 from training. Duration: 0.836375 2022-12-23 06:22:58,396 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.803e+02 4.204e+02 4.894e+02 6.203e+02 1.673e+03, threshold=9.789e+02, percent-clipped=2.0 2022-12-23 06:22:58,412 INFO [train.py:894] (3/4) Epoch 14, batch 1650, loss[loss=0.1988, simple_loss=0.285, pruned_loss=0.05625, over 18584.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2768, pruned_loss=0.05303, over 3714278.09 frames. ], batch size: 51, lr: 8.27e-03, grad_scale: 8.0 2022-12-23 06:22:59,452 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1518, 2.0541, 1.6771, 1.0271, 2.7325, 2.2893, 2.0357, 1.4655], device='cuda:3'), covar=tensor([0.0379, 0.0384, 0.0485, 0.0762, 0.0203, 0.0366, 0.0437, 0.0940], device='cuda:3'), in_proj_covar=tensor([0.0121, 0.0120, 0.0127, 0.0117, 0.0090, 0.0119, 0.0134, 0.0153], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 06:23:09,777 WARNING [train.py:1060] (3/4) Exclude cut with ID 8565-290391-0049-67394-0_sp0.9 from training. Duration: 21.3166875 2022-12-23 06:23:14,533 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47241.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:23:29,873 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47251.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:23:39,894 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47258.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:23:42,509 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0029-104863-0_sp0.9 from training. Duration: 22.1055625 2022-12-23 06:23:53,896 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp1.1 from training. Duration: 21.77725 2022-12-23 06:24:12,270 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp0.9 from training. Duration: 27.8166875 2022-12-23 06:24:13,402 INFO [train.py:894] (3/4) Epoch 14, batch 1700, loss[loss=0.2476, simple_loss=0.3142, pruned_loss=0.09051, over 18576.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2773, pruned_loss=0.05457, over 3713734.22 frames. ], batch size: 57, lr: 8.27e-03, grad_scale: 8.0 2022-12-23 06:24:31,353 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2022-12-23 06:24:36,226 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.80 vs. limit=2.0 2022-12-23 06:24:38,742 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp1.1 from training. Duration: 22.5090625 2022-12-23 06:24:41,900 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=47299.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:24:44,483 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0 from training. Duration: 25.035 2022-12-23 06:24:46,204 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47302.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:24:51,941 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=47306.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:25:03,549 WARNING [train.py:1060] (3/4) Exclude cut with ID 774-127930-0014-10412-0_sp1.1 from training. Duration: 0.95 2022-12-23 06:25:14,345 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2022-12-23 06:25:22,550 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp0.9 from training. Duration: 0.92225 2022-12-23 06:25:29,000 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.980e+02 4.300e+02 5.386e+02 7.051e+02 1.470e+03, threshold=1.077e+03, percent-clipped=6.0 2022-12-23 06:25:29,016 INFO [train.py:894] (3/4) Epoch 14, batch 1750, loss[loss=0.1811, simple_loss=0.2678, pruned_loss=0.04718, over 18595.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2774, pruned_loss=0.05605, over 3713220.78 frames. ], batch size: 51, lr: 8.26e-03, grad_scale: 8.0 2022-12-23 06:25:38,122 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2022-12-23 06:25:48,277 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0 from training. Duration: 21.97 2022-12-23 06:26:01,135 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7854, 2.1618, 1.7587, 2.5806, 1.9827, 2.1034, 2.0462, 2.6218], device='cuda:3'), covar=tensor([0.1672, 0.2964, 0.1657, 0.2440, 0.3380, 0.0986, 0.2806, 0.0785], device='cuda:3'), in_proj_covar=tensor([0.0284, 0.0277, 0.0233, 0.0343, 0.0258, 0.0218, 0.0272, 0.0198], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 06:26:05,031 WARNING [train.py:1060] (3/4) Exclude cut with ID 7492-105653-0055-62765-0_sp0.9 from training. Duration: 21.97225 2022-12-23 06:26:06,499 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp0.9 from training. Duration: 25.3333125 2022-12-23 06:26:16,837 WARNING [train.py:1060] (3/4) Exclude cut with ID 5172-29468-0015-19128-0_sp0.9 from training. Duration: 21.5055625 2022-12-23 06:26:24,068 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp1.1 from training. Duration: 20.72725 2022-12-23 06:26:43,261 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47379.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:26:46,096 INFO [train.py:894] (3/4) Epoch 14, batch 1800, loss[loss=0.2147, simple_loss=0.2964, pruned_loss=0.06647, over 18662.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.28, pruned_loss=0.05871, over 3714120.55 frames. ], batch size: 53, lr: 8.26e-03, grad_scale: 8.0 2022-12-23 06:26:58,862 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp0.9 from training. Duration: 26.32775 2022-12-23 06:27:28,800 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0 from training. Duration: 20.025 2022-12-23 06:27:32,988 WARNING [train.py:1060] (3/4) Exclude cut with ID 6709-74022-0004-86860-0_sp1.1 from training. Duration: 0.9409375 2022-12-23 06:27:32,998 WARNING [train.py:1060] (3/4) Exclude cut with ID 4757-1811-0023-62229-0_sp0.9 from training. Duration: 21.37775 2022-12-23 06:27:54,559 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0004-25974-0_sp0.9 from training. Duration: 21.17225 2022-12-23 06:27:54,567 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp0.9 from training. Duration: 27.511125 2022-12-23 06:28:04,825 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.184e+02 5.129e+02 6.200e+02 7.319e+02 2.210e+03, threshold=1.240e+03, percent-clipped=6.0 2022-12-23 06:28:04,841 INFO [train.py:894] (3/4) Epoch 14, batch 1850, loss[loss=0.2559, simple_loss=0.3199, pruned_loss=0.09594, over 18611.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2823, pruned_loss=0.06123, over 3714260.12 frames. ], batch size: 184, lr: 8.25e-03, grad_scale: 8.0 2022-12-23 06:28:19,107 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47440.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:28:28,768 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0 from training. Duration: 22.8 2022-12-23 06:28:31,632 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0 from training. Duration: 22.585 2022-12-23 06:28:56,810 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47466.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:29:03,630 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0001-146967-0_sp0.9 from training. Duration: 22.0166875 2022-12-23 06:29:19,651 INFO [train.py:894] (3/4) Epoch 14, batch 1900, loss[loss=0.2354, simple_loss=0.3066, pruned_loss=0.08214, over 18502.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.284, pruned_loss=0.0629, over 3714369.28 frames. ], batch size: 77, lr: 8.25e-03, grad_scale: 8.0 2022-12-23 06:29:19,697 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp1.1 from training. Duration: 24.395375 2022-12-23 06:29:27,105 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp0.9 from training. Duration: 27.47775 2022-12-23 06:29:30,067 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp0.9 from training. Duration: 24.8833125 2022-12-23 06:29:32,937 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0 from training. Duration: 23.39 2022-12-23 06:29:38,533 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp0.9 from training. Duration: 28.72225 2022-12-23 06:29:48,390 WARNING [train.py:1060] (3/4) Exclude cut with ID 585-294811-0110-133686-0_sp0.9 from training. Duration: 20.8944375 2022-12-23 06:30:02,799 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0_sp0.9 from training. Duration: 23.8444375 2022-12-23 06:30:07,858 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=47514.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:30:28,609 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0 from training. Duration: 25.85 2022-12-23 06:30:28,620 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0 from training. Duration: 21.39 2022-12-23 06:30:34,534 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.979e+02 5.048e+02 5.917e+02 7.311e+02 1.442e+03, threshold=1.183e+03, percent-clipped=2.0 2022-12-23 06:30:34,550 INFO [train.py:894] (3/4) Epoch 14, batch 1950, loss[loss=0.2114, simple_loss=0.2934, pruned_loss=0.06475, over 18462.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2835, pruned_loss=0.06333, over 3713252.31 frames. ], batch size: 54, lr: 8.25e-03, grad_scale: 8.0 2022-12-23 06:30:39,012 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0 from training. Duration: 27.92 2022-12-23 06:31:05,739 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0039-130165-0_sp0.9 from training. Duration: 20.661125 2022-12-23 06:31:30,321 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0043-15874-0_sp0.9 from training. Duration: 20.07225 2022-12-23 06:31:34,811 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.0614, 5.3216, 4.8821, 2.8525, 5.2578, 4.1072, 0.8510, 3.8269], device='cuda:3'), covar=tensor([0.1818, 0.0945, 0.1281, 0.3005, 0.0784, 0.0799, 0.5527, 0.1166], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0130, 0.0150, 0.0122, 0.0133, 0.0107, 0.0142, 0.0110], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 06:31:39,743 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0 from training. Duration: 21.01 2022-12-23 06:31:48,671 INFO [train.py:894] (3/4) Epoch 14, batch 2000, loss[loss=0.2299, simple_loss=0.3074, pruned_loss=0.07621, over 18594.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2845, pruned_loss=0.06438, over 3713492.67 frames. ], batch size: 56, lr: 8.24e-03, grad_scale: 8.0 2022-12-23 06:32:10,398 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5059, 2.0959, 1.4640, 2.3713, 2.4261, 1.4466, 1.6260, 1.2410], device='cuda:3'), covar=tensor([0.1990, 0.1647, 0.1629, 0.0923, 0.1426, 0.1326, 0.2022, 0.1623], device='cuda:3'), in_proj_covar=tensor([0.0241, 0.0217, 0.0206, 0.0191, 0.0256, 0.0193, 0.0217, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 06:32:13,316 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47597.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:32:44,335 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0 from training. Duration: 20.65 2022-12-23 06:32:51,866 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0 from training. Duration: 21.46 2022-12-23 06:33:05,408 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.860e+02 5.137e+02 6.273e+02 7.724e+02 1.894e+03, threshold=1.255e+03, percent-clipped=3.0 2022-12-23 06:33:05,424 INFO [train.py:894] (3/4) Epoch 14, batch 2050, loss[loss=0.2569, simple_loss=0.3121, pruned_loss=0.1008, over 18637.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2852, pruned_loss=0.06519, over 3714524.49 frames. ], batch size: 178, lr: 8.24e-03, grad_scale: 8.0 2022-12-23 06:33:20,611 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2022-12-23 06:33:30,703 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2022-12-23 06:33:36,890 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0 from training. Duration: 0.92 2022-12-23 06:33:42,984 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0_sp0.9 from training. Duration: 23.7666875 2022-12-23 06:34:20,887 INFO [train.py:894] (3/4) Epoch 14, batch 2100, loss[loss=0.2146, simple_loss=0.2949, pruned_loss=0.06716, over 18553.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.285, pruned_loss=0.0658, over 3714576.25 frames. ], batch size: 77, lr: 8.23e-03, grad_scale: 8.0 2022-12-23 06:34:20,938 WARNING [train.py:1060] (3/4) Exclude cut with ID 8544-281189-0060-101339-0_sp0.9 from training. Duration: 20.861125 2022-12-23 06:34:31,850 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0_sp0.9 from training. Duration: 22.711125 2022-12-23 06:35:16,041 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp1.1 from training. Duration: 22.986375 2022-12-23 06:35:19,887 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2022-12-23 06:35:32,272 WARNING [train.py:1060] (3/4) Exclude cut with ID 8040-260924-0003-80960-0_sp0.9 from training. Duration: 22.07225 2022-12-23 06:35:38,043 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.231e+02 4.876e+02 6.072e+02 7.489e+02 2.027e+03, threshold=1.214e+03, percent-clipped=4.0 2022-12-23 06:35:38,060 INFO [train.py:894] (3/4) Epoch 14, batch 2150, loss[loss=0.2319, simple_loss=0.3052, pruned_loss=0.07931, over 18715.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.284, pruned_loss=0.0656, over 3714561.49 frames. ], batch size: 79, lr: 8.23e-03, grad_scale: 8.0 2022-12-23 06:35:38,114 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0045-26330-0_sp0.9 from training. Duration: 20.3055625 2022-12-23 06:35:41,169 WARNING [train.py:1060] (3/4) Exclude cut with ID 6356-271890-0060-94317-0_sp0.9 from training. Duration: 20.72225 2022-12-23 06:35:44,072 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47735.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:35:48,505 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47738.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:36:00,237 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp1.1 from training. Duration: 22.4818125 2022-12-23 06:36:28,886 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp0.9 from training. Duration: 25.0944375 2022-12-23 06:36:33,972 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0 from training. Duration: 21.515 2022-12-23 06:36:40,270 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp0.9 from training. Duration: 27.02225 2022-12-23 06:36:46,025 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0010-62480-0_sp0.9 from training. Duration: 22.22225 2022-12-23 06:36:53,123 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0085-44554-0_sp0.9 from training. Duration: 20.85 2022-12-23 06:36:54,592 INFO [train.py:894] (3/4) Epoch 14, batch 2200, loss[loss=0.1865, simple_loss=0.2734, pruned_loss=0.04984, over 18545.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2828, pruned_loss=0.06483, over 3714818.64 frames. ], batch size: 55, lr: 8.22e-03, grad_scale: 8.0 2022-12-23 06:37:05,230 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47788.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:37:21,332 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47799.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:37:26,004 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0 from training. Duration: 21.54 2022-12-23 06:37:32,058 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp1.1 from training. Duration: 20.5318125 2022-12-23 06:37:37,432 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47809.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 06:37:40,251 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0012-134311-0_sp0.9 from training. Duration: 21.9333125 2022-12-23 06:38:09,846 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.708e+02 5.175e+02 6.084e+02 7.017e+02 2.042e+03, threshold=1.217e+03, percent-clipped=4.0 2022-12-23 06:38:09,862 INFO [train.py:894] (3/4) Epoch 14, batch 2250, loss[loss=0.2449, simple_loss=0.3133, pruned_loss=0.08826, over 18598.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2816, pruned_loss=0.06461, over 3714658.33 frames. ], batch size: 69, lr: 8.22e-03, grad_scale: 8.0 2022-12-23 06:38:27,215 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0025-130151-0_sp0.9 from training. Duration: 21.7944375 2022-12-23 06:38:36,862 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47849.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:38:41,612 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0_sp0.9 from training. Duration: 22.4666875 2022-12-23 06:38:47,504 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0 from training. Duration: 21.635 2022-12-23 06:38:53,199 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0_sp0.9 from training. Duration: 24.038875 2022-12-23 06:39:08,974 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47870.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 06:39:25,657 INFO [train.py:894] (3/4) Epoch 14, batch 2300, loss[loss=0.2166, simple_loss=0.2746, pruned_loss=0.07929, over 18685.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2807, pruned_loss=0.06406, over 3714042.50 frames. ], batch size: 46, lr: 8.22e-03, grad_scale: 8.0 2022-12-23 06:39:35,760 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp1.1 from training. Duration: 21.786375 2022-12-23 06:39:47,973 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0 from training. Duration: 20.22 2022-12-23 06:39:50,342 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47897.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:40:41,726 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.478e+02 5.323e+02 6.580e+02 8.724e+02 2.507e+03, threshold=1.316e+03, percent-clipped=6.0 2022-12-23 06:40:41,742 INFO [train.py:894] (3/4) Epoch 14, batch 2350, loss[loss=0.1724, simple_loss=0.2442, pruned_loss=0.0503, over 18405.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2811, pruned_loss=0.06387, over 3713917.35 frames. ], batch size: 42, lr: 8.21e-03, grad_scale: 8.0 2022-12-23 06:41:01,190 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4212, 1.9072, 1.3873, 2.1515, 2.1514, 1.4619, 1.4773, 1.1960], device='cuda:3'), covar=tensor([0.2123, 0.1786, 0.1745, 0.1066, 0.1430, 0.1323, 0.2209, 0.1718], device='cuda:3'), in_proj_covar=tensor([0.0238, 0.0215, 0.0204, 0.0190, 0.0255, 0.0191, 0.0215, 0.0192], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 06:41:02,305 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=47945.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:41:52,603 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0 from training. Duration: 25.285 2022-12-23 06:41:57,058 INFO [train.py:894] (3/4) Epoch 14, batch 2400, loss[loss=0.2251, simple_loss=0.3032, pruned_loss=0.07346, over 18511.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.281, pruned_loss=0.0637, over 3713870.55 frames. ], batch size: 77, lr: 8.21e-03, grad_scale: 8.0 2022-12-23 06:43:00,371 WARNING [train.py:1060] (3/4) Exclude cut with ID 811-130148-0001-63453-0_sp0.9 from training. Duration: 20.861125 2022-12-23 06:43:16,636 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.720e+02 4.531e+02 5.657e+02 7.183e+02 1.580e+03, threshold=1.131e+03, percent-clipped=2.0 2022-12-23 06:43:16,653 INFO [train.py:894] (3/4) Epoch 14, batch 2450, loss[loss=0.1918, simple_loss=0.276, pruned_loss=0.05383, over 18714.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2805, pruned_loss=0.06357, over 3714497.68 frames. ], batch size: 52, lr: 8.20e-03, grad_scale: 8.0 2022-12-23 06:43:20,804 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0 from training. Duration: 20.88 2022-12-23 06:43:23,067 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48035.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:43:57,556 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0_sp0.9 from training. Duration: 23.4166875 2022-12-23 06:44:18,211 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48072.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:44:30,892 INFO [train.py:894] (3/4) Epoch 14, batch 2500, loss[loss=0.1903, simple_loss=0.2749, pruned_loss=0.05291, over 18648.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2799, pruned_loss=0.0629, over 3713720.14 frames. ], batch size: 62, lr: 8.20e-03, grad_scale: 8.0 2022-12-23 06:44:34,715 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=48083.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:44:52,058 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48094.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:45:16,283 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0 from training. Duration: 21.24 2022-12-23 06:45:16,300 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0_sp0.9 from training. Duration: 23.9055625 2022-12-23 06:45:48,283 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.059e+02 4.733e+02 5.770e+02 7.680e+02 1.345e+03, threshold=1.154e+03, percent-clipped=2.0 2022-12-23 06:45:48,303 INFO [train.py:894] (3/4) Epoch 14, batch 2550, loss[loss=0.2509, simple_loss=0.3127, pruned_loss=0.09457, over 18672.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2788, pruned_loss=0.06233, over 3713700.35 frames. ], batch size: 181, lr: 8.19e-03, grad_scale: 8.0 2022-12-23 06:45:48,366 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp0.9 from training. Duration: 25.988875 2022-12-23 06:45:52,023 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48133.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:45:58,500 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0001-134300-0_sp0.9 from training. Duration: 20.67225 2022-12-23 06:46:09,349 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48144.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:46:40,966 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48165.0, num_to_drop=1, layers_to_drop={3} 2022-12-23 06:46:45,183 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0 from training. Duration: 20.34 2022-12-23 06:47:05,103 INFO [train.py:894] (3/4) Epoch 14, batch 2600, loss[loss=0.1798, simple_loss=0.2645, pruned_loss=0.04755, over 18663.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2783, pruned_loss=0.06237, over 3713931.45 frames. ], batch size: 48, lr: 8.19e-03, grad_scale: 8.0 2022-12-23 06:47:15,882 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6717, 2.3205, 1.9407, 0.7449, 1.9173, 2.1373, 1.6712, 1.8777], device='cuda:3'), covar=tensor([0.0694, 0.0648, 0.1338, 0.1812, 0.1305, 0.1508, 0.1782, 0.0940], device='cuda:3'), in_proj_covar=tensor([0.0168, 0.0181, 0.0204, 0.0192, 0.0209, 0.0195, 0.0209, 0.0196], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 06:47:52,372 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48212.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 06:47:55,010 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp0.9 from training. Duration: 25.061125 2022-12-23 06:48:07,557 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0 from training. Duration: 0.83 2022-12-23 06:48:21,175 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5867, 1.4029, 1.4792, 0.7961, 1.6415, 1.5395, 1.4658, 1.2328], device='cuda:3'), covar=tensor([0.0370, 0.0497, 0.0429, 0.0720, 0.0386, 0.0395, 0.0475, 0.0964], device='cuda:3'), in_proj_covar=tensor([0.0121, 0.0121, 0.0127, 0.0119, 0.0092, 0.0119, 0.0134, 0.0154], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 06:48:22,222 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.675e+02 5.046e+02 6.298e+02 8.011e+02 1.904e+03, threshold=1.260e+03, percent-clipped=4.0 2022-12-23 06:48:22,238 INFO [train.py:894] (3/4) Epoch 14, batch 2650, loss[loss=0.2095, simple_loss=0.2787, pruned_loss=0.07013, over 18533.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2786, pruned_loss=0.0626, over 3714141.01 frames. ], batch size: 47, lr: 8.19e-03, grad_scale: 8.0 2022-12-23 06:48:34,081 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0 from training. Duration: 24.73 2022-12-23 06:48:45,856 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0 from training. Duration: 23.965 2022-12-23 06:48:55,961 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0030-146996-0_sp0.9 from training. Duration: 22.088875 2022-12-23 06:49:10,113 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0_sp0.9 from training. Duration: 23.6 2022-12-23 06:49:26,296 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48273.0, num_to_drop=1, layers_to_drop={3} 2022-12-23 06:49:38,606 INFO [train.py:894] (3/4) Epoch 14, batch 2700, loss[loss=0.2192, simple_loss=0.2979, pruned_loss=0.07029, over 18510.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.281, pruned_loss=0.06338, over 3713338.60 frames. ], batch size: 52, lr: 8.18e-03, grad_scale: 8.0 2022-12-23 06:50:37,587 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7166, 2.2605, 1.6481, 2.6296, 1.9808, 2.0301, 2.0362, 2.7700], device='cuda:3'), covar=tensor([0.1751, 0.2836, 0.1744, 0.2421, 0.3261, 0.0964, 0.2896, 0.0738], device='cuda:3'), in_proj_covar=tensor([0.0288, 0.0280, 0.0235, 0.0350, 0.0261, 0.0219, 0.0275, 0.0202], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 06:50:44,310 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48324.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:50:54,773 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.448e+02 4.967e+02 5.745e+02 6.581e+02 2.319e+03, threshold=1.149e+03, percent-clipped=2.0 2022-12-23 06:50:54,789 INFO [train.py:894] (3/4) Epoch 14, batch 2750, loss[loss=0.2015, simple_loss=0.2848, pruned_loss=0.05906, over 18379.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2798, pruned_loss=0.06262, over 3712983.83 frames. ], batch size: 51, lr: 8.18e-03, grad_scale: 8.0 2022-12-23 06:50:54,822 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0 from training. Duration: 23.795 2022-12-23 06:51:10,205 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp1.1 from training. Duration: 21.5409375 2022-12-23 06:51:14,665 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp0.9 from training. Duration: 24.97775 2022-12-23 06:51:20,703 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.6025, 1.3031, 1.4683, 0.8579, 1.0788, 1.5185, 1.4229, 1.2728], device='cuda:3'), covar=tensor([0.0600, 0.0309, 0.0246, 0.0333, 0.0320, 0.0409, 0.0202, 0.0538], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0162, 0.0117, 0.0135, 0.0145, 0.0137, 0.0154, 0.0161], device='cuda:3'), out_proj_covar=tensor([1.1510e-04, 1.2925e-04, 9.1685e-05, 1.0481e-04, 1.1448e-04, 1.0975e-04, 1.2360e-04, 1.2771e-04], device='cuda:3') 2022-12-23 06:51:24,782 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0_sp0.9 from training. Duration: 23.3444375 2022-12-23 06:51:41,029 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4816, 2.0027, 2.1188, 2.1607, 2.4286, 2.3021, 2.3307, 1.8563], device='cuda:3'), covar=tensor([0.1861, 0.2720, 0.2008, 0.2586, 0.1493, 0.0803, 0.2656, 0.1063], device='cuda:3'), in_proj_covar=tensor([0.0261, 0.0294, 0.0268, 0.0303, 0.0293, 0.0244, 0.0322, 0.0231], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 06:51:53,345 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0_sp0.9 from training. Duration: 23.2 2022-12-23 06:51:54,992 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2022-12-23 06:51:59,882 WARNING [train.py:1060] (3/4) Exclude cut with ID 5653-46179-0060-117930-0_sp0.9 from training. Duration: 21.17225 2022-12-23 06:52:11,695 INFO [train.py:894] (3/4) Epoch 14, batch 2800, loss[loss=0.1867, simple_loss=0.269, pruned_loss=0.05222, over 18664.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2797, pruned_loss=0.0628, over 3712685.44 frames. ], batch size: 60, lr: 8.17e-03, grad_scale: 8.0 2022-12-23 06:52:18,229 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48385.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:52:19,450 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp0.9 from training. Duration: 24.6555625 2022-12-23 06:52:31,357 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48394.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:52:35,560 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48397.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:53:15,109 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0 from training. Duration: 20.44 2022-12-23 06:53:17,281 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48423.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:53:24,386 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48428.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:53:28,734 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.239e+02 4.838e+02 5.602e+02 6.913e+02 2.084e+03, threshold=1.120e+03, percent-clipped=3.0 2022-12-23 06:53:28,750 INFO [train.py:894] (3/4) Epoch 14, batch 2850, loss[loss=0.2079, simple_loss=0.2982, pruned_loss=0.0588, over 18684.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2795, pruned_loss=0.06267, over 3713072.87 frames. ], batch size: 60, lr: 8.17e-03, grad_scale: 8.0 2022-12-23 06:53:30,314 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0_sp0.9 from training. Duration: 23.45 2022-12-23 06:53:44,637 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=48442.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:53:47,599 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48444.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:54:01,277 WARNING [train.py:1060] (3/4) Exclude cut with ID 6945-60535-0076-12784-0_sp0.9 from training. Duration: 20.52225 2022-12-23 06:54:08,757 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0 from training. Duration: 22.19 2022-12-23 06:54:09,148 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48458.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:54:20,300 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp1.1 from training. Duration: 25.3818125 2022-12-23 06:54:20,520 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48465.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 06:54:25,116 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0186, 2.0502, 2.1606, 1.2218, 2.3585, 2.4077, 1.7004, 2.7318], device='cuda:3'), covar=tensor([0.1071, 0.1474, 0.1149, 0.1793, 0.0644, 0.0950, 0.1996, 0.0418], device='cuda:3'), in_proj_covar=tensor([0.0197, 0.0202, 0.0205, 0.0192, 0.0174, 0.0211, 0.0208, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 06:54:38,096 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp0.9 from training. Duration: 28.0944375 2022-12-23 06:54:42,567 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0_sp0.9 from training. Duration: 22.9444375 2022-12-23 06:54:44,028 INFO [train.py:894] (3/4) Epoch 14, batch 2900, loss[loss=0.1976, simple_loss=0.2678, pruned_loss=0.06368, over 18686.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2799, pruned_loss=0.06317, over 3712721.24 frames. ], batch size: 46, lr: 8.17e-03, grad_scale: 8.0 2022-12-23 06:54:48,967 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48484.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:54:50,309 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp1.1 from training. Duration: 21.6318125 2022-12-23 06:55:00,608 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=48492.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 06:55:08,080 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0 from training. Duration: 23.695 2022-12-23 06:55:33,869 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=48513.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 06:55:37,883 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0 from training. Duration: 23.955 2022-12-23 06:56:00,356 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.111e+02 4.661e+02 5.748e+02 7.181e+02 1.318e+03, threshold=1.150e+03, percent-clipped=2.0 2022-12-23 06:56:00,373 INFO [train.py:894] (3/4) Epoch 14, batch 2950, loss[loss=0.1722, simple_loss=0.2443, pruned_loss=0.05001, over 18545.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2793, pruned_loss=0.06278, over 3712294.58 frames. ], batch size: 44, lr: 8.16e-03, grad_scale: 16.0 2022-12-23 06:56:01,300 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2022-12-23 06:56:09,294 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp0.9 from training. Duration: 26.438875 2022-12-23 06:56:55,631 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0021-26306-0_sp0.9 from training. Duration: 21.2444375 2022-12-23 06:56:57,157 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp0.9 from training. Duration: 31.02225 2022-12-23 06:56:57,296 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48568.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 06:57:07,164 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0 from training. Duration: 22.395 2022-12-23 06:57:16,283 INFO [train.py:894] (3/4) Epoch 14, batch 3000, loss[loss=0.1817, simple_loss=0.2641, pruned_loss=0.04967, over 18676.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2792, pruned_loss=0.06265, over 3713207.11 frames. ], batch size: 48, lr: 8.16e-03, grad_scale: 16.0 2022-12-23 06:57:16,283 INFO [train.py:919] (3/4) Computing validation loss 2022-12-23 06:57:26,225 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.5223, 1.6316, 1.7820, 1.0534, 1.1862, 1.8386, 1.7436, 1.5808], device='cuda:3'), covar=tensor([0.0757, 0.0320, 0.0272, 0.0386, 0.0450, 0.0389, 0.0194, 0.0524], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0162, 0.0117, 0.0135, 0.0145, 0.0137, 0.0153, 0.0161], device='cuda:3'), out_proj_covar=tensor([1.1529e-04, 1.2909e-04, 9.1129e-05, 1.0467e-04, 1.1370e-04, 1.0988e-04, 1.2325e-04, 1.2713e-04], device='cuda:3') 2022-12-23 06:57:27,212 INFO [train.py:928] (3/4) Epoch 14, validation: loss=0.1688, simple_loss=0.2675, pruned_loss=0.03502, over 944034.00 frames. 2022-12-23 06:57:27,213 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-23 06:57:33,021 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0 from training. Duration: 21.075 2022-12-23 06:57:39,746 WARNING [train.py:1060] (3/4) Exclude cut with ID 6482-98857-0025-147532-0_sp0.9 from training. Duration: 20.0055625 2022-12-23 06:57:39,759 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0037-132304-0_sp0.9 from training. Duration: 22.05 2022-12-23 06:57:39,769 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0 from training. Duration: 26.8349375 2022-12-23 06:57:43,258 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp1.1 from training. Duration: 22.1090625 2022-12-23 06:57:48,000 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7497, 1.5977, 1.3874, 1.6393, 1.9369, 1.8553, 1.9836, 1.3102], device='cuda:3'), covar=tensor([0.0285, 0.0244, 0.0472, 0.0172, 0.0164, 0.0308, 0.0209, 0.0296], device='cuda:3'), in_proj_covar=tensor([0.0088, 0.0122, 0.0149, 0.0124, 0.0111, 0.0114, 0.0095, 0.0123], device='cuda:3'), out_proj_covar=tensor([7.1832e-05, 9.8417e-05, 1.2649e-04, 1.0045e-04, 9.2571e-05, 8.9447e-05, 7.6052e-05, 9.9273e-05], device='cuda:3') 2022-12-23 06:57:50,627 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp0.9 from training. Duration: 26.6166875 2022-12-23 06:58:07,558 WARNING [train.py:1060] (3/4) Exclude cut with ID 2046-178027-0000-53705-0_sp0.9 from training. Duration: 20.3055625 2022-12-23 06:58:28,557 WARNING [train.py:1060] (3/4) Exclude cut with ID 7205-50138-0008-5373-0_sp0.9 from training. Duration: 20.7 2022-12-23 06:58:43,454 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.119e+02 4.638e+02 5.784e+02 7.167e+02 1.598e+03, threshold=1.157e+03, percent-clipped=4.0 2022-12-23 06:58:43,470 INFO [train.py:894] (3/4) Epoch 14, batch 3050, loss[loss=0.1845, simple_loss=0.2592, pruned_loss=0.05492, over 18526.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2793, pruned_loss=0.06263, over 3714898.98 frames. ], batch size: 47, lr: 8.15e-03, grad_scale: 16.0 2022-12-23 06:59:14,999 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0 from training. Duration: 22.48 2022-12-23 06:59:30,163 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp0.9 from training. Duration: 29.816625 2022-12-23 06:59:50,851 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp1.1 from training. Duration: 22.7590625 2022-12-23 06:59:55,552 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0 from training. Duration: 22.555 2022-12-23 06:59:58,799 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48680.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:00:00,079 INFO [train.py:894] (3/4) Epoch 14, batch 3100, loss[loss=0.188, simple_loss=0.2757, pruned_loss=0.0501, over 18466.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2788, pruned_loss=0.0624, over 3714708.56 frames. ], batch size: 54, lr: 8.15e-03, grad_scale: 16.0 2022-12-23 07:00:17,804 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0005-25975-0_sp0.9 from training. Duration: 21.688875 2022-12-23 07:00:18,218 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.5482, 1.8950, 2.1839, 1.0462, 1.2924, 2.3482, 2.0194, 1.7613], device='cuda:3'), covar=tensor([0.0777, 0.0313, 0.0314, 0.0392, 0.0396, 0.0411, 0.0223, 0.0675], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0161, 0.0117, 0.0136, 0.0145, 0.0137, 0.0154, 0.0161], device='cuda:3'), out_proj_covar=tensor([1.1534e-04, 1.2863e-04, 9.1454e-05, 1.0531e-04, 1.1410e-04, 1.1002e-04, 1.2366e-04, 1.2706e-04], device='cuda:3') 2022-12-23 07:00:33,065 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9092, 1.7130, 1.6548, 1.6235, 1.9100, 2.0618, 2.1911, 1.5518], device='cuda:3'), covar=tensor([0.0327, 0.0267, 0.0434, 0.0213, 0.0187, 0.0350, 0.0275, 0.0290], device='cuda:3'), in_proj_covar=tensor([0.0088, 0.0121, 0.0149, 0.0124, 0.0112, 0.0114, 0.0095, 0.0123], device='cuda:3'), out_proj_covar=tensor([7.2188e-05, 9.8299e-05, 1.2642e-04, 1.0070e-04, 9.2782e-05, 8.9545e-05, 7.5611e-05, 9.9068e-05], device='cuda:3') 2022-12-23 07:00:50,886 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0_sp0.9 from training. Duration: 22.6 2022-12-23 07:01:14,225 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4800, 0.9763, 0.7088, 1.1850, 1.9903, 0.5996, 1.2420, 1.3974], device='cuda:3'), covar=tensor([0.1712, 0.2262, 0.2089, 0.1567, 0.1733, 0.1789, 0.1542, 0.1659], device='cuda:3'), in_proj_covar=tensor([0.0092, 0.0097, 0.0117, 0.0095, 0.0113, 0.0090, 0.0097, 0.0092], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 07:01:14,239 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48728.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:01:18,803 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.420e+02 4.890e+02 5.907e+02 7.278e+02 1.415e+03, threshold=1.181e+03, percent-clipped=1.0 2022-12-23 07:01:18,819 INFO [train.py:894] (3/4) Epoch 14, batch 3150, loss[loss=0.2439, simple_loss=0.3075, pruned_loss=0.09016, over 18530.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2791, pruned_loss=0.06212, over 3714689.68 frames. ], batch size: 58, lr: 8.14e-03, grad_scale: 16.0 2022-12-23 07:01:29,646 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0 from training. Duration: 24.32 2022-12-23 07:01:43,263 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1258, 1.2227, 1.8328, 1.7180, 2.1481, 2.0216, 1.8518, 1.7236], device='cuda:3'), covar=tensor([0.1878, 0.2755, 0.2164, 0.2312, 0.1669, 0.0884, 0.2476, 0.1097], device='cuda:3'), in_proj_covar=tensor([0.0260, 0.0295, 0.0269, 0.0304, 0.0293, 0.0244, 0.0322, 0.0232], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 07:01:51,694 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48753.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:02:26,873 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-276745-0093-13116-0_sp0.9 from training. Duration: 21.061125 2022-12-23 07:02:28,924 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=48776.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:02:33,328 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48779.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:02:36,151 INFO [train.py:894] (3/4) Epoch 14, batch 3200, loss[loss=0.1787, simple_loss=0.2684, pruned_loss=0.04453, over 18592.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2794, pruned_loss=0.0622, over 3714911.62 frames. ], batch size: 69, lr: 8.14e-03, grad_scale: 16.0 2022-12-23 07:02:42,171 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0024-15855-0_sp0.9 from training. Duration: 20.32225 2022-12-23 07:02:56,289 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp1.1 from training. Duration: 0.7545625 2022-12-23 07:03:10,405 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2022-12-23 07:03:12,310 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0_sp0.9 from training. Duration: 23.9333125 2022-12-23 07:03:35,155 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48821.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:03:45,358 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp1.1 from training. Duration: 20.17275 2022-12-23 07:03:51,308 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.097e+02 4.705e+02 5.372e+02 6.667e+02 1.296e+03, threshold=1.074e+03, percent-clipped=3.0 2022-12-23 07:03:51,324 INFO [train.py:894] (3/4) Epoch 14, batch 3250, loss[loss=0.1833, simple_loss=0.2641, pruned_loss=0.05124, over 18563.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2778, pruned_loss=0.06141, over 3714606.22 frames. ], batch size: 49, lr: 8.14e-03, grad_scale: 16.0 2022-12-23 07:03:52,791 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp1.1 from training. Duration: 20.436375 2022-12-23 07:04:02,127 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48838.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:04:52,022 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48868.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 07:05:12,150 INFO [train.py:894] (3/4) Epoch 14, batch 3300, loss[loss=0.2145, simple_loss=0.2975, pruned_loss=0.06572, over 18487.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2786, pruned_loss=0.06161, over 3715164.42 frames. ], batch size: 52, lr: 8.13e-03, grad_scale: 16.0 2022-12-23 07:05:14,091 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48882.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:05:15,221 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0_sp0.9 from training. Duration: 23.1055625 2022-12-23 07:05:16,846 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp1.1 from training. Duration: 23.5 2022-12-23 07:05:20,241 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6914, 3.0413, 2.7600, 1.3372, 3.1398, 3.2959, 2.5768, 3.8848], device='cuda:3'), covar=tensor([0.1187, 0.1350, 0.1502, 0.2300, 0.0747, 0.1250, 0.1677, 0.0455], device='cuda:3'), in_proj_covar=tensor([0.0200, 0.0206, 0.0206, 0.0196, 0.0177, 0.0216, 0.0213, 0.0188], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 07:05:27,139 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp0.9 from training. Duration: 26.62775 2022-12-23 07:05:38,870 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48899.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:05:39,970 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0 from training. Duration: 21.105 2022-12-23 07:05:44,462 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0_sp0.9 from training. Duration: 24.411125 2022-12-23 07:05:49,670 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.1429, 1.4905, 1.7003, 0.7765, 0.8446, 1.8576, 1.6671, 1.5348], device='cuda:3'), covar=tensor([0.0638, 0.0268, 0.0295, 0.0306, 0.0387, 0.0384, 0.0197, 0.0535], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0162, 0.0118, 0.0137, 0.0146, 0.0138, 0.0154, 0.0160], device='cuda:3'), out_proj_covar=tensor([1.1555e-04, 1.2922e-04, 9.1862e-05, 1.0586e-04, 1.1428e-04, 1.1088e-04, 1.2399e-04, 1.2662e-04], device='cuda:3') 2022-12-23 07:05:54,164 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5248, 1.3632, 1.4236, 1.9298, 1.7147, 3.3042, 1.3358, 1.5487], device='cuda:3'), covar=tensor([0.0877, 0.1808, 0.1045, 0.0806, 0.1377, 0.0231, 0.1399, 0.1485], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0082, 0.0073, 0.0074, 0.0090, 0.0074, 0.0084, 0.0077], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 07:06:04,729 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=48916.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 07:06:11,819 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp1.1 from training. Duration: 21.263625 2022-12-23 07:06:26,996 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.866e+02 5.170e+02 6.260e+02 9.545e+02 2.280e+03, threshold=1.252e+03, percent-clipped=16.0 2022-12-23 07:06:27,014 INFO [train.py:894] (3/4) Epoch 14, batch 3350, loss[loss=0.2054, simple_loss=0.2904, pruned_loss=0.06019, over 18542.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2794, pruned_loss=0.06196, over 3715114.88 frames. ], batch size: 55, lr: 8.13e-03, grad_scale: 16.0 2022-12-23 07:06:44,238 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0 from training. Duration: 20.795 2022-12-23 07:06:54,603 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0 from training. Duration: 24.76 2022-12-23 07:06:54,619 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0_sp0.9 from training. Duration: 22.25 2022-12-23 07:07:21,935 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp1.1 from training. Duration: 20.5045625 2022-12-23 07:07:35,489 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2022-12-23 07:07:41,876 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48980.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:07:43,308 INFO [train.py:894] (3/4) Epoch 14, batch 3400, loss[loss=0.2418, simple_loss=0.3097, pruned_loss=0.08697, over 18645.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2804, pruned_loss=0.06266, over 3716259.03 frames. ], batch size: 183, lr: 8.12e-03, grad_scale: 16.0 2022-12-23 07:08:51,318 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=49028.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:08:55,427 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.249e+02 4.856e+02 6.127e+02 7.528e+02 2.019e+03, threshold=1.225e+03, percent-clipped=5.0 2022-12-23 07:08:55,443 INFO [train.py:894] (3/4) Epoch 14, batch 3450, loss[loss=0.1882, simple_loss=0.2643, pruned_loss=0.05608, over 18522.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2798, pruned_loss=0.0621, over 3717527.28 frames. ], batch size: 55, lr: 8.12e-03, grad_scale: 16.0 2022-12-23 07:09:27,263 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49053.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:10:08,326 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49079.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:10:11,319 INFO [train.py:894] (3/4) Epoch 14, batch 3500, loss[loss=0.239, simple_loss=0.3027, pruned_loss=0.08764, over 18670.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2807, pruned_loss=0.06295, over 3717219.36 frames. ], batch size: 182, lr: 8.12e-03, grad_scale: 16.0 2022-12-23 07:10:37,266 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp1.1 from training. Duration: 22.2954375 2022-12-23 07:10:41,358 INFO [train.py:894] (3/4) Epoch 15, batch 0, loss[loss=0.2126, simple_loss=0.2863, pruned_loss=0.06939, over 18691.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2863, pruned_loss=0.06939, over 18691.00 frames. ], batch size: 46, lr: 7.84e-03, grad_scale: 16.0 2022-12-23 07:10:41,359 INFO [train.py:919] (3/4) Computing validation loss 2022-12-23 07:10:52,028 INFO [train.py:928] (3/4) Epoch 15, validation: loss=0.1674, simple_loss=0.2666, pruned_loss=0.03411, over 944034.00 frames. 2022-12-23 07:10:52,029 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-23 07:10:59,302 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7282, 1.5160, 1.6236, 1.6830, 1.4029, 3.7693, 1.8716, 2.2437], device='cuda:3'), covar=tensor([0.3175, 0.2170, 0.1948, 0.2023, 0.1431, 0.0165, 0.1325, 0.0797], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0117, 0.0127, 0.0121, 0.0104, 0.0099, 0.0094, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 07:11:13,618 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=49101.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:11:45,203 WARNING [train.py:1060] (3/4) Exclude cut with ID 298-126791-0067-24026-0_sp0.9 from training. Duration: 21.438875 2022-12-23 07:11:48,189 WARNING [train.py:1060] (3/4) Exclude cut with ID 5652-39938-0025-23684-0_sp0.9 from training. Duration: 22.2055625 2022-12-23 07:11:50,596 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.74 vs. limit=5.0 2022-12-23 07:11:52,699 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=49127.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:11:58,179 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.381e+02 4.087e+02 5.188e+02 6.525e+02 1.846e+03, threshold=1.038e+03, percent-clipped=3.0 2022-12-23 07:12:06,978 INFO [train.py:894] (3/4) Epoch 15, batch 50, loss[loss=0.1686, simple_loss=0.2552, pruned_loss=0.04099, over 18674.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2788, pruned_loss=0.05375, over 837842.92 frames. ], batch size: 48, lr: 7.83e-03, grad_scale: 16.0 2022-12-23 07:12:38,948 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.2312, 5.4764, 4.8725, 2.7071, 5.4626, 4.1506, 0.7123, 3.9452], device='cuda:3'), covar=tensor([0.1751, 0.0639, 0.1253, 0.3085, 0.0601, 0.0734, 0.5411, 0.1097], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0133, 0.0152, 0.0122, 0.0135, 0.0109, 0.0143, 0.0110], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 07:13:09,562 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49177.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:13:24,204 INFO [train.py:894] (3/4) Epoch 15, batch 100, loss[loss=0.1891, simple_loss=0.2601, pruned_loss=0.0591, over 18585.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2735, pruned_loss=0.05189, over 1476008.78 frames. ], batch size: 45, lr: 7.83e-03, grad_scale: 16.0 2022-12-23 07:13:35,685 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49194.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:14:30,645 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.987e+02 3.186e+02 4.193e+02 4.956e+02 1.857e+03, threshold=8.385e+02, percent-clipped=1.0 2022-12-23 07:14:38,206 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49236.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:14:39,181 INFO [train.py:894] (3/4) Epoch 15, batch 150, loss[loss=0.2157, simple_loss=0.3012, pruned_loss=0.06505, over 18712.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2722, pruned_loss=0.05071, over 1972502.86 frames. ], batch size: 97, lr: 7.83e-03, grad_scale: 16.0 2022-12-23 07:14:50,224 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0 from training. Duration: 24.525 2022-12-23 07:14:54,194 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.87 vs. limit=5.0 2022-12-23 07:15:22,274 WARNING [train.py:1060] (3/4) Exclude cut with ID 3699-47246-0007-3408-0_sp0.9 from training. Duration: 20.26675 2022-12-23 07:15:37,433 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp0.9 from training. Duration: 27.25 2022-12-23 07:15:55,581 INFO [train.py:894] (3/4) Epoch 15, batch 200, loss[loss=0.1615, simple_loss=0.2466, pruned_loss=0.03818, over 18585.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2713, pruned_loss=0.05005, over 2357778.06 frames. ], batch size: 41, lr: 7.82e-03, grad_scale: 16.0 2022-12-23 07:16:13,341 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49297.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:16:48,437 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0 from training. Duration: 21.68 2022-12-23 07:17:01,112 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0 from training. Duration: 21.6300625 2022-12-23 07:17:02,601 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.256e+02 3.589e+02 4.467e+02 5.180e+02 1.013e+03, threshold=8.934e+02, percent-clipped=1.0 2022-12-23 07:17:11,972 INFO [train.py:894] (3/4) Epoch 15, batch 250, loss[loss=0.1496, simple_loss=0.2298, pruned_loss=0.03474, over 18409.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2702, pruned_loss=0.04935, over 2658331.28 frames. ], batch size: 42, lr: 7.82e-03, grad_scale: 16.0 2022-12-23 07:17:26,519 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0_sp0.9 from training. Duration: 24.033375 2022-12-23 07:18:22,645 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0 from training. Duration: 22.905 2022-12-23 07:18:24,133 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp1.1 from training. Duration: 23.4318125 2022-12-23 07:18:26,179 INFO [train.py:894] (3/4) Epoch 15, batch 300, loss[loss=0.1773, simple_loss=0.2649, pruned_loss=0.04482, over 18667.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2695, pruned_loss=0.04877, over 2893353.62 frames. ], batch size: 48, lr: 7.82e-03, grad_scale: 16.0 2022-12-23 07:19:18,995 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49422.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:19:29,590 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7834, 1.1900, 0.8398, 1.3637, 2.0103, 1.2769, 1.7152, 1.9336], device='cuda:3'), covar=tensor([0.1626, 0.2137, 0.2383, 0.1635, 0.1808, 0.1793, 0.1424, 0.1573], device='cuda:3'), in_proj_covar=tensor([0.0092, 0.0097, 0.0116, 0.0094, 0.0112, 0.0090, 0.0097, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 07:19:32,364 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.246e+02 3.698e+02 4.337e+02 5.424e+02 1.306e+03, threshold=8.675e+02, percent-clipped=3.0 2022-12-23 07:19:34,178 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49432.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 07:19:41,985 INFO [train.py:894] (3/4) Epoch 15, batch 350, loss[loss=0.1663, simple_loss=0.2466, pruned_loss=0.04302, over 18600.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2701, pruned_loss=0.04938, over 3074892.93 frames. ], batch size: 45, lr: 7.81e-03, grad_scale: 16.0 2022-12-23 07:20:20,423 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp1.1 from training. Duration: 20.82275 2022-12-23 07:20:20,477 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp0.9 from training. Duration: 25.45 2022-12-23 07:20:41,219 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49477.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:20:50,740 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49483.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:20:56,138 INFO [train.py:894] (3/4) Epoch 15, batch 400, loss[loss=0.1593, simple_loss=0.2493, pruned_loss=0.03468, over 18692.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2711, pruned_loss=0.05026, over 3216819.22 frames. ], batch size: 48, lr: 7.81e-03, grad_scale: 16.0 2022-12-23 07:21:07,194 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49493.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 07:21:08,295 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49494.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:21:18,590 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7271, 2.1614, 1.5755, 2.6401, 1.9484, 2.0947, 2.1030, 2.6144], device='cuda:3'), covar=tensor([0.1701, 0.3016, 0.1824, 0.2572, 0.3217, 0.0962, 0.2779, 0.0798], device='cuda:3'), in_proj_covar=tensor([0.0284, 0.0279, 0.0235, 0.0348, 0.0259, 0.0217, 0.0275, 0.0201], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 07:21:23,822 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0 from training. Duration: 25.775 2022-12-23 07:21:27,140 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.0824, 1.0036, 0.9928, 1.1027, 1.2914, 1.1599, 1.0968, 0.9476], device='cuda:3'), covar=tensor([0.0258, 0.0254, 0.0517, 0.0198, 0.0217, 0.0373, 0.0252, 0.0280], device='cuda:3'), in_proj_covar=tensor([0.0090, 0.0124, 0.0150, 0.0124, 0.0115, 0.0116, 0.0096, 0.0123], device='cuda:3'), out_proj_covar=tensor([7.3474e-05, 1.0038e-04, 1.2752e-04, 1.0080e-04, 9.5513e-05, 9.0910e-05, 7.6931e-05, 9.9204e-05], device='cuda:3') 2022-12-23 07:21:44,147 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.5418, 1.9592, 2.1649, 1.2064, 1.3865, 2.3122, 2.1161, 1.7648], device='cuda:3'), covar=tensor([0.0799, 0.0301, 0.0285, 0.0341, 0.0344, 0.0422, 0.0213, 0.0646], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0163, 0.0118, 0.0136, 0.0146, 0.0138, 0.0153, 0.0159], device='cuda:3'), out_proj_covar=tensor([1.1581e-04, 1.2950e-04, 9.1720e-05, 1.0490e-04, 1.1440e-04, 1.1037e-04, 1.2248e-04, 1.2485e-04], device='cuda:3') 2022-12-23 07:21:46,635 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0_sp0.9 from training. Duration: 22.25 2022-12-23 07:21:52,754 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=49525.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:22:01,189 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.732e+02 3.654e+02 4.284e+02 4.993e+02 1.074e+03, threshold=8.568e+02, percent-clipped=3.0 2022-12-23 07:22:10,266 INFO [train.py:894] (3/4) Epoch 15, batch 450, loss[loss=0.1722, simple_loss=0.2495, pruned_loss=0.0475, over 18511.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.273, pruned_loss=0.05113, over 3327231.42 frames. ], batch size: 44, lr: 7.80e-03, grad_scale: 16.0 2022-12-23 07:22:14,161 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0 from training. Duration: 26.205 2022-12-23 07:22:18,516 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=49542.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:22:29,961 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp0.9 from training. Duration: 30.1555625 2022-12-23 07:22:34,944 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.0998, 1.0089, 1.1872, 0.4756, 0.5555, 1.2684, 1.2503, 1.1555], device='cuda:3'), covar=tensor([0.0690, 0.0292, 0.0325, 0.0358, 0.0435, 0.0460, 0.0250, 0.0543], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0163, 0.0117, 0.0136, 0.0145, 0.0138, 0.0152, 0.0158], device='cuda:3'), out_proj_covar=tensor([1.1565e-04, 1.2963e-04, 9.1721e-05, 1.0508e-04, 1.1417e-04, 1.1022e-04, 1.2191e-04, 1.2477e-04], device='cuda:3') 2022-12-23 07:22:35,767 WARNING [train.py:1060] (3/4) Exclude cut with ID 1265-135635-0050-6781-0_sp0.9 from training. Duration: 21.8333125 2022-12-23 07:22:45,956 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp1.1 from training. Duration: 20.6545625 2022-12-23 07:23:26,573 INFO [train.py:894] (3/4) Epoch 15, batch 500, loss[loss=0.2007, simple_loss=0.2815, pruned_loss=0.05996, over 18579.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2736, pruned_loss=0.05147, over 3412847.14 frames. ], batch size: 51, lr: 7.80e-03, grad_scale: 16.0 2022-12-23 07:23:27,272 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0045-39920-0_sp0.9 from training. Duration: 20.52225 2022-12-23 07:23:34,292 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49592.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:23:44,725 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp0.9 from training. Duration: 29.1166875 2022-12-23 07:23:57,650 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 2022-12-23 07:24:32,841 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.334e+02 4.035e+02 5.056e+02 6.562e+02 1.635e+03, threshold=1.011e+03, percent-clipped=6.0 2022-12-23 07:24:43,082 INFO [train.py:894] (3/4) Epoch 15, batch 550, loss[loss=0.1697, simple_loss=0.2429, pruned_loss=0.04824, over 18540.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2731, pruned_loss=0.05173, over 3480193.08 frames. ], batch size: 41, lr: 7.80e-03, grad_scale: 16.0 2022-12-23 07:24:47,972 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133211-0007-59831-0_sp0.9 from training. Duration: 21.388875 2022-12-23 07:25:21,178 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0 from training. Duration: 22.72 2022-12-23 07:25:21,220 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0_sp0.9 from training. Duration: 22.7444375 2022-12-23 07:25:57,315 INFO [train.py:894] (3/4) Epoch 15, batch 600, loss[loss=0.1809, simple_loss=0.2603, pruned_loss=0.05071, over 18596.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2738, pruned_loss=0.05165, over 3531633.56 frames. ], batch size: 45, lr: 7.79e-03, grad_scale: 16.0 2022-12-23 07:26:04,658 WARNING [train.py:1060] (3/4) Exclude cut with ID 4133-6541-0027-40495-0_sp1.1 from training. Duration: 0.9681875 2022-12-23 07:26:07,520 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0_sp0.9 from training. Duration: 22.3166875 2022-12-23 07:26:07,994 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3576, 2.5765, 1.6611, 3.1956, 2.7750, 2.3739, 3.7813, 2.3238], device='cuda:3'), covar=tensor([0.0793, 0.1673, 0.2737, 0.1693, 0.1589, 0.0859, 0.0859, 0.1174], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0200, 0.0243, 0.0285, 0.0229, 0.0185, 0.0206, 0.0200], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 07:26:12,122 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133212-0015-59917-0_sp0.9 from training. Duration: 21.8166875 2022-12-23 07:27:03,252 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.339e+02 3.610e+02 4.307e+02 5.287e+02 1.018e+03, threshold=8.614e+02, percent-clipped=1.0 2022-12-23 07:27:13,047 INFO [train.py:894] (3/4) Epoch 15, batch 650, loss[loss=0.2009, simple_loss=0.29, pruned_loss=0.05594, over 18619.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2741, pruned_loss=0.05192, over 3572052.74 frames. ], batch size: 69, lr: 7.79e-03, grad_scale: 16.0 2022-12-23 07:27:25,147 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6671, 2.3206, 1.6683, 2.8017, 3.2736, 1.7406, 2.0400, 1.5227], device='cuda:3'), covar=tensor([0.1854, 0.1539, 0.1440, 0.0803, 0.1160, 0.1016, 0.1676, 0.1406], device='cuda:3'), in_proj_covar=tensor([0.0239, 0.0215, 0.0204, 0.0190, 0.0254, 0.0189, 0.0214, 0.0192], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 07:27:56,411 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49766.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:27:57,474 WARNING [train.py:1060] (3/4) Exclude cut with ID 4957-30119-0041-23990-0_sp0.9 from training. Duration: 20.22775 2022-12-23 07:28:14,612 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49778.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:28:28,630 INFO [train.py:894] (3/4) Epoch 15, batch 700, loss[loss=0.1687, simple_loss=0.2554, pruned_loss=0.04107, over 18392.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2749, pruned_loss=0.0521, over 3604643.54 frames. ], batch size: 46, lr: 7.78e-03, grad_scale: 16.0 2022-12-23 07:28:30,300 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49788.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 07:28:41,853 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp1.1 from training. Duration: 24.67275 2022-12-23 07:29:08,697 WARNING [train.py:1060] (3/4) Exclude cut with ID 3082-165428-0081-50734-0_sp0.9 from training. Duration: 21.8055625 2022-12-23 07:29:28,981 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49827.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:29:34,936 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.205e+02 3.616e+02 4.297e+02 5.395e+02 9.947e+02, threshold=8.594e+02, percent-clipped=4.0 2022-12-23 07:29:44,196 INFO [train.py:894] (3/4) Epoch 15, batch 750, loss[loss=0.1756, simple_loss=0.2724, pruned_loss=0.03935, over 18479.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.275, pruned_loss=0.05178, over 3628828.47 frames. ], batch size: 54, lr: 7.78e-03, grad_scale: 16.0 2022-12-23 07:29:49,009 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0_sp0.9 from training. Duration: 22.6666875 2022-12-23 07:30:06,543 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4092, 1.0080, 1.4208, 2.2746, 1.6459, 2.2444, 0.8307, 1.6153], device='cuda:3'), covar=tensor([0.1837, 0.1918, 0.1354, 0.0642, 0.1144, 0.0892, 0.1861, 0.1233], device='cuda:3'), in_proj_covar=tensor([0.0101, 0.0116, 0.0131, 0.0137, 0.0104, 0.0135, 0.0129, 0.0110], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 07:30:49,893 WARNING [train.py:1060] (3/4) Exclude cut with ID 2411-132532-0017-82279-0_sp1.1 from training. Duration: 0.9681875 2022-12-23 07:30:58,950 INFO [train.py:894] (3/4) Epoch 15, batch 800, loss[loss=0.2091, simple_loss=0.2971, pruned_loss=0.06049, over 18555.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2751, pruned_loss=0.05179, over 3648156.87 frames. ], batch size: 97, lr: 7.78e-03, grad_scale: 8.0 2022-12-23 07:31:06,215 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49892.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:31:16,091 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0 from training. Duration: 22.485 2022-12-23 07:31:51,845 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp1.1 from training. Duration: 23.82275 2022-12-23 07:32:06,627 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.483e+02 3.966e+02 4.783e+02 5.574e+02 1.191e+03, threshold=9.565e+02, percent-clipped=3.0 2022-12-23 07:32:07,855 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0 from training. Duration: 20.77 2022-12-23 07:32:13,607 INFO [train.py:894] (3/4) Epoch 15, batch 850, loss[loss=0.1674, simple_loss=0.2517, pruned_loss=0.04151, over 18582.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2742, pruned_loss=0.05129, over 3661721.51 frames. ], batch size: 49, lr: 7.77e-03, grad_scale: 8.0 2022-12-23 07:32:13,614 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0_sp0.9 from training. Duration: 24.088875 2022-12-23 07:32:18,231 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=49940.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:32:22,979 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3987, 1.7073, 1.3675, 1.9572, 2.2558, 1.4331, 1.3774, 1.2867], device='cuda:3'), covar=tensor([0.1814, 0.1706, 0.1506, 0.0998, 0.1143, 0.1074, 0.1928, 0.1434], device='cuda:3'), in_proj_covar=tensor([0.0238, 0.0215, 0.0205, 0.0190, 0.0254, 0.0189, 0.0214, 0.0192], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 07:32:41,681 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp1.1 from training. Duration: 20.4409375 2022-12-23 07:32:43,381 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.8181, 4.0233, 3.9176, 1.8177, 4.2590, 3.1314, 0.6627, 2.6453], device='cuda:3'), covar=tensor([0.1952, 0.1065, 0.1362, 0.3499, 0.0767, 0.0937, 0.5386, 0.1591], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0129, 0.0152, 0.0121, 0.0132, 0.0108, 0.0141, 0.0110], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 07:32:51,933 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2022-12-23 07:33:29,290 INFO [train.py:894] (3/4) Epoch 15, batch 900, loss[loss=0.1524, simple_loss=0.2332, pruned_loss=0.03578, over 18640.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2753, pruned_loss=0.05174, over 3673645.30 frames. ], batch size: 45, lr: 7.77e-03, grad_scale: 8.0 2022-12-23 07:34:00,998 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0_sp0.9 from training. Duration: 22.511125 2022-12-23 07:34:02,403 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0 from training. Duration: 20.675 2022-12-23 07:34:07,252 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2022-12-23 07:34:40,525 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.188e+02 3.776e+02 4.591e+02 6.047e+02 1.033e+03, threshold=9.181e+02, percent-clipped=2.0 2022-12-23 07:34:42,635 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.3413, 2.5132, 2.1440, 3.0659, 2.3515, 2.7187, 2.5901, 3.7664], device='cuda:3'), covar=tensor([0.1677, 0.3295, 0.1661, 0.2893, 0.3606, 0.0872, 0.3041, 0.0590], device='cuda:3'), in_proj_covar=tensor([0.0285, 0.0279, 0.0235, 0.0346, 0.0259, 0.0216, 0.0276, 0.0200], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 07:34:47,834 INFO [train.py:894] (3/4) Epoch 15, batch 950, loss[loss=0.1909, simple_loss=0.2596, pruned_loss=0.06106, over 18498.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2747, pruned_loss=0.05152, over 3682184.41 frames. ], batch size: 43, lr: 7.76e-03, grad_scale: 8.0 2022-12-23 07:34:59,634 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5638, 1.0332, 1.5567, 2.6657, 1.9613, 2.2255, 0.8174, 1.7883], device='cuda:3'), covar=tensor([0.1826, 0.1930, 0.1487, 0.0603, 0.1068, 0.1103, 0.2122, 0.1271], device='cuda:3'), in_proj_covar=tensor([0.0103, 0.0119, 0.0134, 0.0140, 0.0106, 0.0137, 0.0132, 0.0113], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 07:35:39,002 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp0.9 from training. Duration: 24.9833125 2022-12-23 07:35:51,235 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50078.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:35:57,063 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8066, 1.3326, 0.8368, 1.1999, 2.1923, 0.9196, 1.5666, 1.6048], device='cuda:3'), covar=tensor([0.1474, 0.1955, 0.2229, 0.1671, 0.1535, 0.1755, 0.1367, 0.1671], device='cuda:3'), in_proj_covar=tensor([0.0093, 0.0098, 0.0118, 0.0096, 0.0112, 0.0092, 0.0098, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 07:36:03,970 INFO [train.py:894] (3/4) Epoch 15, batch 1000, loss[loss=0.1711, simple_loss=0.2585, pruned_loss=0.04186, over 18543.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2748, pruned_loss=0.05168, over 3688283.35 frames. ], batch size: 49, lr: 7.76e-03, grad_scale: 8.0 2022-12-23 07:36:05,709 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50088.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 07:36:09,999 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0 from training. Duration: 27.14 2022-12-23 07:36:24,725 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0 from training. Duration: 22.44 2022-12-23 07:36:43,430 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2022-12-23 07:36:54,531 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7225, 3.6691, 3.5883, 1.5779, 3.7960, 2.7116, 0.5294, 2.6598], device='cuda:3'), covar=tensor([0.1935, 0.0970, 0.1325, 0.3774, 0.0817, 0.1057, 0.5368, 0.1458], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0130, 0.0152, 0.0122, 0.0133, 0.0108, 0.0142, 0.0110], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 07:36:57,418 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50122.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:37:03,198 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=50126.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:37:06,257 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50128.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:37:11,411 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.398e+02 3.589e+02 4.498e+02 6.076e+02 1.358e+03, threshold=8.996e+02, percent-clipped=1.0 2022-12-23 07:37:17,322 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=50136.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 07:37:18,478 INFO [train.py:894] (3/4) Epoch 15, batch 1050, loss[loss=0.2214, simple_loss=0.302, pruned_loss=0.07046, over 18610.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2751, pruned_loss=0.0514, over 3693666.12 frames. ], batch size: 78, lr: 7.76e-03, grad_scale: 8.0 2022-12-23 07:37:46,419 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0060-62364-0_sp0.9 from training. Duration: 21.361125 2022-12-23 07:37:51,894 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp1.1 from training. Duration: 27.0318125 2022-12-23 07:38:03,171 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp0.9 from training. Duration: 28.638875 2022-12-23 07:38:05,040 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8229, 1.7045, 1.7299, 2.2977, 1.9465, 4.5387, 1.5921, 1.8422], device='cuda:3'), covar=tensor([0.0771, 0.1655, 0.1029, 0.0862, 0.1353, 0.0162, 0.1323, 0.1482], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0082, 0.0073, 0.0074, 0.0090, 0.0074, 0.0084, 0.0077], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 07:38:19,541 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0 from training. Duration: 20.4 2022-12-23 07:38:32,487 INFO [train.py:894] (3/4) Epoch 15, batch 1100, loss[loss=0.1964, simple_loss=0.2886, pruned_loss=0.05206, over 18505.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2738, pruned_loss=0.05045, over 3697973.73 frames. ], batch size: 52, lr: 7.75e-03, grad_scale: 8.0 2022-12-23 07:38:35,619 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50189.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:38:46,641 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50197.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 07:38:50,641 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0 from training. Duration: 20.025 2022-12-23 07:38:50,661 WARNING [train.py:1060] (3/4) Exclude cut with ID 2364-131735-0112-64612-0_sp0.9 from training. Duration: 20.488875 2022-12-23 07:38:57,868 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0 from training. Duration: 29.735 2022-12-23 07:39:39,227 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.273e+02 3.557e+02 4.399e+02 5.755e+02 2.138e+03, threshold=8.797e+02, percent-clipped=6.0 2022-12-23 07:39:46,189 INFO [train.py:894] (3/4) Epoch 15, batch 1150, loss[loss=0.1734, simple_loss=0.2459, pruned_loss=0.05045, over 18678.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2724, pruned_loss=0.04991, over 3701410.81 frames. ], batch size: 46, lr: 7.75e-03, grad_scale: 8.0 2022-12-23 07:40:17,429 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50258.0, num_to_drop=1, layers_to_drop={3} 2022-12-23 07:40:18,533 WARNING [train.py:1060] (3/4) Exclude cut with ID 7276-92427-0014-12983-0_sp0.9 from training. Duration: 21.3055625 2022-12-23 07:40:19,938 WARNING [train.py:1060] (3/4) Exclude cut with ID 1025-75365-0008-79168-0_sp0.9 from training. Duration: 22.0666875 2022-12-23 07:40:27,474 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2022-12-23 07:41:01,623 INFO [train.py:894] (3/4) Epoch 15, batch 1200, loss[loss=0.2095, simple_loss=0.2942, pruned_loss=0.06237, over 18506.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2726, pruned_loss=0.0499, over 3704233.85 frames. ], batch size: 52, lr: 7.75e-03, grad_scale: 8.0 2022-12-23 07:42:09,331 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.658e+02 3.741e+02 4.483e+02 5.877e+02 1.567e+03, threshold=8.966e+02, percent-clipped=2.0 2022-12-23 07:42:09,401 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0 from training. Duration: 20.26 2022-12-23 07:42:16,737 INFO [train.py:894] (3/4) Epoch 15, batch 1250, loss[loss=0.1656, simple_loss=0.2494, pruned_loss=0.04091, over 18529.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.272, pruned_loss=0.04984, over 3706497.38 frames. ], batch size: 47, lr: 7.74e-03, grad_scale: 8.0 2022-12-23 07:42:23,647 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0030-9324-0_sp0.9 from training. Duration: 21.3444375 2022-12-23 07:43:18,916 WARNING [train.py:1060] (3/4) Exclude cut with ID 497-129325-0061-62254-0_sp1.1 from training. Duration: 0.97725 2022-12-23 07:43:30,718 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5739, 2.0911, 1.6708, 2.3708, 1.8683, 2.0688, 1.9155, 2.4513], device='cuda:3'), covar=tensor([0.1875, 0.3281, 0.1851, 0.2546, 0.3440, 0.1010, 0.2920, 0.0850], device='cuda:3'), in_proj_covar=tensor([0.0287, 0.0282, 0.0238, 0.0349, 0.0262, 0.0219, 0.0278, 0.0203], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 07:43:31,604 INFO [train.py:894] (3/4) Epoch 15, batch 1300, loss[loss=0.1513, simple_loss=0.2296, pruned_loss=0.03646, over 18469.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2729, pruned_loss=0.04994, over 3707731.98 frames. ], batch size: 43, lr: 7.74e-03, grad_scale: 8.0 2022-12-23 07:44:02,820 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0_sp0.9 from training. Duration: 22.97225 2022-12-23 07:44:24,752 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50422.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:44:29,421 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.7389, 3.2192, 2.9330, 1.2054, 2.5179, 2.5606, 2.2227, 2.4165], device='cuda:3'), covar=tensor([0.0582, 0.0550, 0.1215, 0.1695, 0.1582, 0.1298, 0.1447, 0.0990], device='cuda:3'), in_proj_covar=tensor([0.0164, 0.0179, 0.0197, 0.0186, 0.0203, 0.0191, 0.0206, 0.0192], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 07:44:33,429 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0047-39922-0_sp0.9 from training. Duration: 21.97775 2022-12-23 07:44:39,441 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.804e+02 4.009e+02 4.601e+02 5.649e+02 1.183e+03, threshold=9.201e+02, percent-clipped=3.0 2022-12-23 07:44:46,985 INFO [train.py:894] (3/4) Epoch 15, batch 1350, loss[loss=0.1784, simple_loss=0.2637, pruned_loss=0.04657, over 18717.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2726, pruned_loss=0.04997, over 3710050.42 frames. ], batch size: 46, lr: 7.73e-03, grad_scale: 8.0 2022-12-23 07:44:47,042 WARNING [train.py:1060] (3/4) Exclude cut with ID 1112-1043-0006-89194-0_sp0.9 from training. Duration: 21.8333125 2022-12-23 07:44:57,090 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0 from training. Duration: 20.47 2022-12-23 07:45:17,278 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1098, 1.3830, 1.7845, 1.7641, 2.1412, 2.1040, 1.9363, 1.6386], device='cuda:3'), covar=tensor([0.1899, 0.2791, 0.2163, 0.2619, 0.1634, 0.0801, 0.2638, 0.1062], device='cuda:3'), in_proj_covar=tensor([0.0258, 0.0291, 0.0266, 0.0303, 0.0289, 0.0242, 0.0322, 0.0227], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 07:45:36,656 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=50470.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:45:56,920 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50484.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:46:00,939 INFO [train.py:894] (3/4) Epoch 15, batch 1400, loss[loss=0.209, simple_loss=0.298, pruned_loss=0.05997, over 18591.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2726, pruned_loss=0.0497, over 3710484.83 frames. ], batch size: 56, lr: 7.73e-03, grad_scale: 8.0 2022-12-23 07:46:02,443 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0037-39912-0_sp0.9 from training. Duration: 20.67225 2022-12-23 07:46:20,712 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp0.9 from training. Duration: 25.2444375 2022-12-23 07:46:42,480 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0021-76797-0_sp0.9 from training. Duration: 21.1445 2022-12-23 07:47:08,331 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.114e+02 3.582e+02 4.419e+02 5.521e+02 1.088e+03, threshold=8.838e+02, percent-clipped=3.0 2022-12-23 07:47:15,707 INFO [train.py:894] (3/4) Epoch 15, batch 1450, loss[loss=0.2013, simple_loss=0.2895, pruned_loss=0.0566, over 18596.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2735, pruned_loss=0.04999, over 3710623.18 frames. ], batch size: 69, lr: 7.73e-03, grad_scale: 8.0 2022-12-23 07:47:29,684 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50546.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:47:40,092 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50553.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 07:47:48,744 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2022-12-23 07:47:57,229 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp0.9 from training. Duration: 33.038875 2022-12-23 07:48:31,371 INFO [train.py:894] (3/4) Epoch 15, batch 1500, loss[loss=0.1991, simple_loss=0.2891, pruned_loss=0.05455, over 18687.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2731, pruned_loss=0.04988, over 3710590.25 frames. ], batch size: 96, lr: 7.72e-03, grad_scale: 8.0 2022-12-23 07:48:34,380 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64291-0000-16059-0_sp0.9 from training. Duration: 20.0944375 2022-12-23 07:48:51,371 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp1.1 from training. Duration: 20.4 2022-12-23 07:48:58,506 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0 from training. Duration: 20.085 2022-12-23 07:49:02,152 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50607.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:49:11,605 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0_sp0.9 from training. Duration: 23.07775 2022-12-23 07:49:40,186 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.951e+02 3.779e+02 4.506e+02 5.486e+02 9.567e+02, threshold=9.012e+02, percent-clipped=3.0 2022-12-23 07:49:47,215 INFO [train.py:894] (3/4) Epoch 15, batch 1550, loss[loss=0.1966, simple_loss=0.2851, pruned_loss=0.05407, over 18473.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2728, pruned_loss=0.04958, over 3711407.62 frames. ], batch size: 64, lr: 7.72e-03, grad_scale: 8.0 2022-12-23 07:49:53,895 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4600, 1.3423, 1.1104, 1.5985, 1.5787, 2.9278, 1.4295, 1.5905], device='cuda:3'), covar=tensor([0.0886, 0.1940, 0.1195, 0.0947, 0.1494, 0.0261, 0.1377, 0.1529], device='cuda:3'), in_proj_covar=tensor([0.0074, 0.0083, 0.0074, 0.0075, 0.0092, 0.0076, 0.0086, 0.0079], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 07:49:56,276 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp0.9 from training. Duration: 24.9333125 2022-12-23 07:50:33,212 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5216, 2.8002, 2.8595, 1.5072, 2.8918, 2.8382, 2.0023, 3.6094], device='cuda:3'), covar=tensor([0.1188, 0.1491, 0.1523, 0.2357, 0.0805, 0.1225, 0.2094, 0.0466], device='cuda:3'), in_proj_covar=tensor([0.0198, 0.0204, 0.0205, 0.0193, 0.0176, 0.0213, 0.0208, 0.0185], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 07:50:41,838 WARNING [train.py:1060] (3/4) Exclude cut with ID 5118-111612-0016-124680-0_sp0.9 from training. Duration: 20.388875 2022-12-23 07:50:49,368 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp1.1 from training. Duration: 20.3590625 2022-12-23 07:51:02,332 INFO [train.py:894] (3/4) Epoch 15, batch 1600, loss[loss=0.1828, simple_loss=0.2736, pruned_loss=0.04598, over 18486.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2724, pruned_loss=0.04912, over 3712777.84 frames. ], batch size: 52, lr: 7.72e-03, grad_scale: 8.0 2022-12-23 07:51:56,643 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0_sp1.1 from training. Duration: 0.836375 2022-12-23 07:52:09,726 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.709e+02 4.126e+02 5.150e+02 6.276e+02 1.539e+03, threshold=1.030e+03, percent-clipped=5.0 2022-12-23 07:52:17,182 INFO [train.py:894] (3/4) Epoch 15, batch 1650, loss[loss=0.1815, simple_loss=0.2575, pruned_loss=0.05273, over 18556.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2737, pruned_loss=0.05066, over 3713418.88 frames. ], batch size: 44, lr: 7.71e-03, grad_scale: 8.0 2022-12-23 07:52:37,957 WARNING [train.py:1060] (3/4) Exclude cut with ID 8565-290391-0049-67394-0_sp0.9 from training. Duration: 21.3166875 2022-12-23 07:53:09,255 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0029-104863-0_sp0.9 from training. Duration: 22.1055625 2022-12-23 07:53:15,431 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6610, 1.0396, 1.8245, 3.1420, 2.1708, 2.4637, 0.8512, 2.0576], device='cuda:3'), covar=tensor([0.1856, 0.2034, 0.1648, 0.0576, 0.1178, 0.1240, 0.2409, 0.1223], device='cuda:3'), in_proj_covar=tensor([0.0103, 0.0117, 0.0134, 0.0140, 0.0106, 0.0136, 0.0131, 0.0112], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 07:53:18,573 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5966, 2.3844, 1.8785, 1.2790, 2.9415, 2.6947, 2.3453, 1.7161], device='cuda:3'), covar=tensor([0.0342, 0.0349, 0.0512, 0.0757, 0.0198, 0.0326, 0.0419, 0.0893], device='cuda:3'), in_proj_covar=tensor([0.0122, 0.0122, 0.0129, 0.0121, 0.0091, 0.0121, 0.0135, 0.0154], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 07:53:20,916 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp1.1 from training. Duration: 21.77725 2022-12-23 07:53:24,760 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50782.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:53:25,438 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2022-12-23 07:53:27,449 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50784.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:53:29,228 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5950, 2.1509, 1.7904, 2.3620, 1.8881, 2.0874, 1.9671, 2.5396], device='cuda:3'), covar=tensor([0.1584, 0.2834, 0.1650, 0.2364, 0.3180, 0.0894, 0.2589, 0.0741], device='cuda:3'), in_proj_covar=tensor([0.0285, 0.0281, 0.0235, 0.0348, 0.0260, 0.0218, 0.0277, 0.0202], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 07:53:31,808 INFO [train.py:894] (3/4) Epoch 15, batch 1700, loss[loss=0.183, simple_loss=0.2667, pruned_loss=0.04963, over 18662.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2747, pruned_loss=0.05268, over 3712904.52 frames. ], batch size: 48, lr: 7.71e-03, grad_scale: 8.0 2022-12-23 07:53:40,181 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp0.9 from training. Duration: 27.8166875 2022-12-23 07:54:00,087 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.0144, 2.3602, 1.8724, 2.9443, 2.1868, 2.2220, 2.4018, 3.3738], device='cuda:3'), covar=tensor([0.1812, 0.3051, 0.1676, 0.2699, 0.3664, 0.0961, 0.2873, 0.0656], device='cuda:3'), in_proj_covar=tensor([0.0288, 0.0283, 0.0237, 0.0351, 0.0262, 0.0220, 0.0278, 0.0204], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 07:54:04,755 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp1.1 from training. Duration: 22.5090625 2022-12-23 07:54:05,014 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50808.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 07:54:10,685 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0 from training. Duration: 25.035 2022-12-23 07:54:27,983 WARNING [train.py:1060] (3/4) Exclude cut with ID 774-127930-0014-10412-0_sp1.1 from training. Duration: 0.95 2022-12-23 07:54:39,921 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.931e+02 4.643e+02 5.788e+02 7.435e+02 4.288e+03, threshold=1.158e+03, percent-clipped=5.0 2022-12-23 07:54:40,107 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=50832.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:54:45,942 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp0.9 from training. Duration: 0.92225 2022-12-23 07:54:47,261 INFO [train.py:894] (3/4) Epoch 15, batch 1750, loss[loss=0.1977, simple_loss=0.2823, pruned_loss=0.05657, over 18501.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2759, pruned_loss=0.05465, over 3713075.59 frames. ], batch size: 69, lr: 7.70e-03, grad_scale: 8.0 2022-12-23 07:54:57,662 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50843.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:55:12,304 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0 from training. Duration: 21.97 2022-12-23 07:55:13,263 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50853.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 07:55:32,010 WARNING [train.py:1060] (3/4) Exclude cut with ID 7492-105653-0055-62765-0_sp0.9 from training. Duration: 21.97225 2022-12-23 07:55:32,048 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp0.9 from training. Duration: 25.3333125 2022-12-23 07:55:36,758 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50869.0, num_to_drop=1, layers_to_drop={3} 2022-12-23 07:55:42,144 WARNING [train.py:1060] (3/4) Exclude cut with ID 5172-29468-0015-19128-0_sp0.9 from training. Duration: 21.5055625 2022-12-23 07:55:52,972 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp1.1 from training. Duration: 20.72725 2022-12-23 07:56:03,527 INFO [train.py:894] (3/4) Epoch 15, batch 1800, loss[loss=0.1845, simple_loss=0.2612, pruned_loss=0.05392, over 18659.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2774, pruned_loss=0.05651, over 3714621.53 frames. ], batch size: 46, lr: 7.70e-03, grad_scale: 8.0 2022-12-23 07:56:25,096 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp0.9 from training. Duration: 26.32775 2022-12-23 07:56:25,229 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=50901.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 07:56:27,014 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50902.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:56:55,989 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0 from training. Duration: 20.025 2022-12-23 07:57:01,806 WARNING [train.py:1060] (3/4) Exclude cut with ID 6709-74022-0004-86860-0_sp1.1 from training. Duration: 0.9409375 2022-12-23 07:57:01,814 WARNING [train.py:1060] (3/4) Exclude cut with ID 4757-1811-0023-62229-0_sp0.9 from training. Duration: 21.37775 2022-12-23 07:57:10,949 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.207e+02 5.141e+02 6.153e+02 7.821e+02 1.885e+03, threshold=1.231e+03, percent-clipped=4.0 2022-12-23 07:57:18,335 INFO [train.py:894] (3/4) Epoch 15, batch 1850, loss[loss=0.2243, simple_loss=0.2968, pruned_loss=0.07589, over 18576.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2781, pruned_loss=0.05823, over 3712988.73 frames. ], batch size: 56, lr: 7.70e-03, grad_scale: 8.0 2022-12-23 07:57:22,689 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0004-25974-0_sp0.9 from training. Duration: 21.17225 2022-12-23 07:57:22,701 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp0.9 from training. Duration: 27.511125 2022-12-23 07:57:32,276 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50946.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:57:57,406 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0 from training. Duration: 22.8 2022-12-23 07:57:58,545 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.58 vs. limit=5.0 2022-12-23 07:58:00,159 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0 from training. Duration: 22.585 2022-12-23 07:58:31,865 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0001-146967-0_sp0.9 from training. Duration: 22.0166875 2022-12-23 07:58:33,316 INFO [train.py:894] (3/4) Epoch 15, batch 1900, loss[loss=0.2125, simple_loss=0.2928, pruned_loss=0.06615, over 18540.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.278, pruned_loss=0.05887, over 3713126.02 frames. ], batch size: 55, lr: 7.69e-03, grad_scale: 8.0 2022-12-23 07:58:49,259 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp1.1 from training. Duration: 24.395375 2022-12-23 07:58:56,641 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp0.9 from training. Duration: 27.47775 2022-12-23 07:58:59,593 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp0.9 from training. Duration: 24.8833125 2022-12-23 07:59:02,533 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0 from training. Duration: 23.39 2022-12-23 07:59:04,340 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51007.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 07:59:08,698 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp0.9 from training. Duration: 28.72225 2022-12-23 07:59:19,837 WARNING [train.py:1060] (3/4) Exclude cut with ID 585-294811-0110-133686-0_sp0.9 from training. Duration: 20.8944375 2022-12-23 07:59:34,380 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0_sp0.9 from training. Duration: 23.8444375 2022-12-23 07:59:41,798 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.113e+02 4.916e+02 5.880e+02 7.888e+02 2.150e+03, threshold=1.176e+03, percent-clipped=4.0 2022-12-23 07:59:49,404 INFO [train.py:894] (3/4) Epoch 15, batch 1950, loss[loss=0.1942, simple_loss=0.2669, pruned_loss=0.06072, over 18677.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.279, pruned_loss=0.05967, over 3712103.77 frames. ], batch size: 46, lr: 7.69e-03, grad_scale: 8.0 2022-12-23 08:00:00,154 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0 from training. Duration: 25.85 2022-12-23 08:00:00,161 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0 from training. Duration: 21.39 2022-12-23 08:00:12,121 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0 from training. Duration: 27.92 2022-12-23 08:00:27,336 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51062.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:00:40,145 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0039-130165-0_sp0.9 from training. Duration: 20.661125 2022-12-23 08:01:03,630 INFO [train.py:894] (3/4) Epoch 15, batch 2000, loss[loss=0.2299, simple_loss=0.3026, pruned_loss=0.07858, over 18469.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2797, pruned_loss=0.0604, over 3712474.18 frames. ], batch size: 54, lr: 7.69e-03, grad_scale: 8.0 2022-12-23 08:01:03,699 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0043-15874-0_sp0.9 from training. Duration: 20.07225 2022-12-23 08:01:11,439 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0 from training. Duration: 21.01 2022-12-23 08:01:58,761 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51123.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:02:11,370 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.251e+02 4.739e+02 5.620e+02 7.194e+02 1.175e+03, threshold=1.124e+03, percent-clipped=0.0 2022-12-23 08:02:17,153 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0 from training. Duration: 20.65 2022-12-23 08:02:19,369 INFO [train.py:894] (3/4) Epoch 15, batch 2050, loss[loss=0.2263, simple_loss=0.3008, pruned_loss=0.07589, over 18580.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2801, pruned_loss=0.06111, over 3713212.61 frames. ], batch size: 78, lr: 7.68e-03, grad_scale: 8.0 2022-12-23 08:02:21,445 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51138.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:02:25,621 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0 from training. Duration: 21.46 2022-12-23 08:02:59,901 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51164.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 08:03:10,467 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0 from training. Duration: 0.92 2022-12-23 08:03:12,431 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51172.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 08:03:14,753 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0_sp0.9 from training. Duration: 23.7666875 2022-12-23 08:03:35,201 INFO [train.py:894] (3/4) Epoch 15, batch 2100, loss[loss=0.196, simple_loss=0.2709, pruned_loss=0.06054, over 18562.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2815, pruned_loss=0.0622, over 3713728.19 frames. ], batch size: 49, lr: 7.68e-03, grad_scale: 8.0 2022-12-23 08:03:54,301 WARNING [train.py:1060] (3/4) Exclude cut with ID 8544-281189-0060-101339-0_sp0.9 from training. Duration: 20.861125 2022-12-23 08:03:57,678 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51202.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:04:03,264 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0_sp0.9 from training. Duration: 22.711125 2022-12-23 08:04:44,103 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.951e+02 4.953e+02 6.008e+02 8.155e+02 2.142e+03, threshold=1.202e+03, percent-clipped=7.0 2022-12-23 08:04:45,716 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp1.1 from training. Duration: 22.986375 2022-12-23 08:04:46,079 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51233.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 08:04:51,360 INFO [train.py:894] (3/4) Epoch 15, batch 2150, loss[loss=0.2346, simple_loss=0.3026, pruned_loss=0.08329, over 18554.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2803, pruned_loss=0.06224, over 3713418.35 frames. ], batch size: 57, lr: 7.67e-03, grad_scale: 8.0 2022-12-23 08:05:01,825 WARNING [train.py:1060] (3/4) Exclude cut with ID 8040-260924-0003-80960-0_sp0.9 from training. Duration: 22.07225 2022-12-23 08:05:07,092 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0045-26330-0_sp0.9 from training. Duration: 20.3055625 2022-12-23 08:05:08,645 WARNING [train.py:1060] (3/4) Exclude cut with ID 6356-271890-0060-94317-0_sp0.9 from training. Duration: 20.72225 2022-12-23 08:05:09,145 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0245, 1.0831, 1.8055, 1.6613, 2.0179, 1.9655, 1.7630, 1.6247], device='cuda:3'), covar=tensor([0.1767, 0.2653, 0.2079, 0.2258, 0.1562, 0.0817, 0.2314, 0.1052], device='cuda:3'), in_proj_covar=tensor([0.0261, 0.0294, 0.0269, 0.0305, 0.0291, 0.0244, 0.0322, 0.0232], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 08:05:10,312 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=51250.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:05:28,917 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp1.1 from training. Duration: 22.4818125 2022-12-23 08:05:56,345 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp0.9 from training. Duration: 25.0944375 2022-12-23 08:05:59,318 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0 from training. Duration: 21.515 2022-12-23 08:06:06,790 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp0.9 from training. Duration: 27.02225 2022-12-23 08:06:08,202 INFO [train.py:894] (3/4) Epoch 15, batch 2200, loss[loss=0.1845, simple_loss=0.2737, pruned_loss=0.0477, over 18396.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2783, pruned_loss=0.06132, over 3713416.04 frames. ], batch size: 53, lr: 7.67e-03, grad_scale: 8.0 2022-12-23 08:06:11,071 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0010-62480-0_sp0.9 from training. Duration: 22.22225 2022-12-23 08:06:18,494 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0085-44554-0_sp0.9 from training. Duration: 20.85 2022-12-23 08:06:30,928 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51302.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:06:52,758 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0 from training. Duration: 21.54 2022-12-23 08:06:54,609 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6062, 2.3817, 2.0355, 1.6960, 2.2302, 2.9976, 2.9990, 2.2242], device='cuda:3'), covar=tensor([0.0341, 0.0285, 0.0376, 0.0250, 0.0233, 0.0284, 0.0216, 0.0241], device='cuda:3'), in_proj_covar=tensor([0.0091, 0.0123, 0.0149, 0.0124, 0.0113, 0.0115, 0.0094, 0.0124], device='cuda:3'), out_proj_covar=tensor([7.3967e-05, 9.9156e-05, 1.2585e-04, 1.0099e-04, 9.3926e-05, 9.0517e-05, 7.4466e-05, 9.9130e-05], device='cuda:3') 2022-12-23 08:06:57,155 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp1.1 from training. Duration: 20.5318125 2022-12-23 08:07:06,012 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0012-134311-0_sp0.9 from training. Duration: 21.9333125 2022-12-23 08:07:16,954 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.356e+02 4.912e+02 6.307e+02 7.418e+02 1.261e+03, threshold=1.261e+03, percent-clipped=3.0 2022-12-23 08:07:24,361 INFO [train.py:894] (3/4) Epoch 15, batch 2250, loss[loss=0.1776, simple_loss=0.2553, pruned_loss=0.04996, over 18397.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2788, pruned_loss=0.06151, over 3713633.78 frames. ], batch size: 46, lr: 7.67e-03, grad_scale: 8.0 2022-12-23 08:07:36,269 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.5934, 3.9820, 3.9477, 4.5138, 4.1315, 4.0512, 4.6638, 1.3094], device='cuda:3'), covar=tensor([0.0657, 0.0762, 0.0612, 0.0709, 0.1475, 0.1202, 0.0699, 0.5265], device='cuda:3'), in_proj_covar=tensor([0.0324, 0.0217, 0.0224, 0.0248, 0.0307, 0.0259, 0.0272, 0.0273], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 08:07:52,727 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0025-130151-0_sp0.9 from training. Duration: 21.7944375 2022-12-23 08:08:07,495 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0_sp0.9 from training. Duration: 22.4666875 2022-12-23 08:08:13,761 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0 from training. Duration: 21.635 2022-12-23 08:08:20,416 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0_sp0.9 from training. Duration: 24.038875 2022-12-23 08:08:32,082 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51382.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:08:39,394 INFO [train.py:894] (3/4) Epoch 15, batch 2300, loss[loss=0.1877, simple_loss=0.2742, pruned_loss=0.05064, over 18490.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2794, pruned_loss=0.06212, over 3713775.48 frames. ], batch size: 52, lr: 7.66e-03, grad_scale: 8.0 2022-12-23 08:08:47,368 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51392.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:08:59,328 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp1.1 from training. Duration: 21.786375 2022-12-23 08:09:09,776 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7810, 1.6848, 1.3309, 1.5967, 1.8288, 1.6022, 2.0843, 1.8748], device='cuda:3'), covar=tensor([0.0915, 0.1613, 0.2841, 0.1763, 0.1842, 0.0946, 0.1073, 0.1126], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0200, 0.0241, 0.0285, 0.0228, 0.0185, 0.0206, 0.0199], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 08:09:12,220 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0 from training. Duration: 20.22 2022-12-23 08:09:27,845 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51418.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:09:47,811 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.635e+02 5.135e+02 6.604e+02 7.825e+02 1.376e+03, threshold=1.321e+03, percent-clipped=3.0 2022-12-23 08:09:55,129 INFO [train.py:894] (3/4) Epoch 15, batch 2350, loss[loss=0.215, simple_loss=0.296, pruned_loss=0.06704, over 18603.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2789, pruned_loss=0.06207, over 3713642.23 frames. ], batch size: 51, lr: 7.66e-03, grad_scale: 8.0 2022-12-23 08:09:57,016 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51438.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:10:04,908 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51443.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:10:17,999 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6376, 2.2710, 1.9727, 1.5996, 2.2179, 2.9976, 2.8211, 2.2346], device='cuda:3'), covar=tensor([0.0261, 0.0281, 0.0387, 0.0270, 0.0232, 0.0253, 0.0239, 0.0247], device='cuda:3'), in_proj_covar=tensor([0.0091, 0.0123, 0.0150, 0.0126, 0.0114, 0.0115, 0.0094, 0.0124], device='cuda:3'), out_proj_covar=tensor([7.4020e-05, 9.9495e-05, 1.2696e-04, 1.0215e-04, 9.4751e-05, 9.0321e-05, 7.4802e-05, 9.9657e-05], device='cuda:3') 2022-12-23 08:10:19,480 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51453.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:10:36,452 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51464.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 08:11:08,171 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=51486.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:11:09,475 INFO [train.py:894] (3/4) Epoch 15, batch 2400, loss[loss=0.1903, simple_loss=0.2801, pruned_loss=0.05021, over 18689.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2795, pruned_loss=0.06225, over 3713285.24 frames. ], batch size: 60, lr: 7.66e-03, grad_scale: 8.0 2022-12-23 08:11:09,486 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0 from training. Duration: 25.285 2022-12-23 08:11:22,412 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51495.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:11:48,463 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=51512.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 08:12:11,664 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51528.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 08:12:13,254 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.8192, 4.2169, 4.3744, 4.7119, 4.4954, 4.3406, 4.8557, 2.5539], device='cuda:3'), covar=tensor([0.0592, 0.0537, 0.0477, 0.0746, 0.1149, 0.1052, 0.0647, 0.3745], device='cuda:3'), in_proj_covar=tensor([0.0328, 0.0219, 0.0227, 0.0250, 0.0310, 0.0263, 0.0276, 0.0274], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 08:12:13,348 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5325, 2.1376, 1.3212, 2.1456, 2.6940, 2.3555, 2.4637, 2.4476], device='cuda:3'), covar=tensor([0.1174, 0.1594, 0.2262, 0.1158, 0.1478, 0.1295, 0.1119, 0.1205], device='cuda:3'), in_proj_covar=tensor([0.0092, 0.0097, 0.0118, 0.0096, 0.0113, 0.0091, 0.0099, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 08:12:14,822 WARNING [train.py:1060] (3/4) Exclude cut with ID 811-130148-0001-63453-0_sp0.9 from training. Duration: 20.861125 2022-12-23 08:12:15,529 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 2022-12-23 08:12:17,683 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.080e+02 5.000e+02 6.055e+02 7.391e+02 1.821e+03, threshold=1.211e+03, percent-clipped=2.0 2022-12-23 08:12:25,896 INFO [train.py:894] (3/4) Epoch 15, batch 2450, loss[loss=0.2414, simple_loss=0.3158, pruned_loss=0.08347, over 18450.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2792, pruned_loss=0.06219, over 3714145.47 frames. ], batch size: 64, lr: 7.65e-03, grad_scale: 8.0 2022-12-23 08:12:38,274 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0 from training. Duration: 20.88 2022-12-23 08:12:55,445 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51556.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:13:09,673 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0_sp0.9 from training. Duration: 23.4166875 2022-12-23 08:13:33,051 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2022-12-23 08:13:40,943 INFO [train.py:894] (3/4) Epoch 15, batch 2500, loss[loss=0.2111, simple_loss=0.2835, pruned_loss=0.06938, over 18577.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2798, pruned_loss=0.06242, over 3714699.96 frames. ], batch size: 49, lr: 7.65e-03, grad_scale: 8.0 2022-12-23 08:13:49,500 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2022-12-23 08:14:04,539 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51602.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:14:28,539 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0 from training. Duration: 21.24 2022-12-23 08:14:28,562 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0_sp0.9 from training. Duration: 23.9055625 2022-12-23 08:14:49,858 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.827e+02 4.732e+02 5.703e+02 7.714e+02 1.744e+03, threshold=1.141e+03, percent-clipped=4.0 2022-12-23 08:14:58,036 INFO [train.py:894] (3/4) Epoch 15, batch 2550, loss[loss=0.1756, simple_loss=0.2559, pruned_loss=0.04759, over 18420.00 frames. ], tot_loss[loss=0.201, simple_loss=0.279, pruned_loss=0.06152, over 3713489.82 frames. ], batch size: 48, lr: 7.64e-03, grad_scale: 8.0 2022-12-23 08:15:01,044 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp0.9 from training. Duration: 25.988875 2022-12-23 08:15:10,455 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0001-134300-0_sp0.9 from training. Duration: 20.67225 2022-12-23 08:15:19,055 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=51650.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:15:57,982 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0 from training. Duration: 20.34 2022-12-23 08:16:15,223 INFO [train.py:894] (3/4) Epoch 15, batch 2600, loss[loss=0.202, simple_loss=0.2918, pruned_loss=0.05605, over 18711.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2792, pruned_loss=0.06121, over 3712891.52 frames. ], batch size: 52, lr: 7.64e-03, grad_scale: 8.0 2022-12-23 08:16:19,968 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5749, 1.7422, 0.8127, 1.8887, 2.4550, 1.6481, 2.0448, 2.2359], device='cuda:3'), covar=tensor([0.1305, 0.1903, 0.2471, 0.1427, 0.1494, 0.1643, 0.1365, 0.1491], device='cuda:3'), in_proj_covar=tensor([0.0093, 0.0098, 0.0119, 0.0096, 0.0114, 0.0091, 0.0099, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 08:17:03,261 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51718.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:17:07,844 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp0.9 from training. Duration: 25.061125 2022-12-23 08:17:17,095 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2022-12-23 08:17:20,422 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0 from training. Duration: 0.83 2022-12-23 08:17:24,695 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.474e+02 4.798e+02 5.976e+02 6.876e+02 1.380e+03, threshold=1.195e+03, percent-clipped=6.0 2022-12-23 08:17:32,150 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2022-12-23 08:17:32,248 INFO [train.py:894] (3/4) Epoch 15, batch 2650, loss[loss=0.218, simple_loss=0.29, pruned_loss=0.07302, over 18390.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2789, pruned_loss=0.06155, over 3713755.83 frames. ], batch size: 53, lr: 7.64e-03, grad_scale: 8.0 2022-12-23 08:17:34,642 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51738.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:17:45,806 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3877, 1.2963, 1.6633, 1.0390, 1.3844, 1.4600, 1.1901, 1.7407], device='cuda:3'), covar=tensor([0.0894, 0.1760, 0.0979, 0.1195, 0.0701, 0.0831, 0.2282, 0.0491], device='cuda:3'), in_proj_covar=tensor([0.0197, 0.0204, 0.0207, 0.0191, 0.0174, 0.0214, 0.0212, 0.0186], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 08:17:46,864 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0 from training. Duration: 24.73 2022-12-23 08:17:48,728 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51747.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:17:49,984 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51748.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:18:00,221 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0 from training. Duration: 23.965 2022-12-23 08:18:07,270 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0030-146996-0_sp0.9 from training. Duration: 22.088875 2022-12-23 08:18:16,976 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=51766.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:18:24,779 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0_sp0.9 from training. Duration: 23.6 2022-12-23 08:18:48,951 INFO [train.py:894] (3/4) Epoch 15, batch 2700, loss[loss=0.1678, simple_loss=0.2436, pruned_loss=0.046, over 18478.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2783, pruned_loss=0.0614, over 3714404.17 frames. ], batch size: 43, lr: 7.63e-03, grad_scale: 8.0 2022-12-23 08:19:20,157 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51808.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:19:50,350 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51828.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 08:19:55,520 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.453e+02 5.078e+02 6.132e+02 7.520e+02 1.991e+03, threshold=1.226e+03, percent-clipped=4.0 2022-12-23 08:20:03,365 INFO [train.py:894] (3/4) Epoch 15, batch 2750, loss[loss=0.1798, simple_loss=0.2619, pruned_loss=0.04884, over 18587.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2777, pruned_loss=0.06103, over 3714008.59 frames. ], batch size: 51, lr: 7.63e-03, grad_scale: 8.0 2022-12-23 08:20:09,888 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0 from training. Duration: 23.795 2022-12-23 08:20:23,891 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp1.1 from training. Duration: 21.5409375 2022-12-23 08:20:25,585 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51851.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:20:26,805 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp0.9 from training. Duration: 24.97775 2022-12-23 08:20:39,269 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0_sp0.9 from training. Duration: 23.3444375 2022-12-23 08:21:02,804 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=51876.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 08:21:06,763 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0_sp0.9 from training. Duration: 23.2 2022-12-23 08:21:13,923 WARNING [train.py:1060] (3/4) Exclude cut with ID 5653-46179-0060-117930-0_sp0.9 from training. Duration: 21.17225 2022-12-23 08:21:20,034 INFO [train.py:894] (3/4) Epoch 15, batch 2800, loss[loss=0.2212, simple_loss=0.2943, pruned_loss=0.07408, over 18651.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2777, pruned_loss=0.06101, over 3713484.03 frames. ], batch size: 175, lr: 7.63e-03, grad_scale: 16.0 2022-12-23 08:21:30,375 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2022-12-23 08:21:32,248 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp0.9 from training. Duration: 24.6555625 2022-12-23 08:22:28,033 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0 from training. Duration: 20.44 2022-12-23 08:22:29,759 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.494e+02 4.534e+02 5.436e+02 6.882e+02 1.456e+03, threshold=1.087e+03, percent-clipped=1.0 2022-12-23 08:22:35,860 INFO [train.py:894] (3/4) Epoch 15, batch 2850, loss[loss=0.2395, simple_loss=0.3113, pruned_loss=0.08386, over 18594.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.279, pruned_loss=0.06176, over 3714465.01 frames. ], batch size: 56, lr: 7.62e-03, grad_scale: 8.0 2022-12-23 08:22:43,553 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0_sp0.9 from training. Duration: 23.45 2022-12-23 08:23:12,468 WARNING [train.py:1060] (3/4) Exclude cut with ID 6945-60535-0076-12784-0_sp0.9 from training. Duration: 20.52225 2022-12-23 08:23:21,351 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0 from training. Duration: 22.19 2022-12-23 08:23:30,721 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp1.1 from training. Duration: 25.3818125 2022-12-23 08:23:48,172 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp0.9 from training. Duration: 28.0944375 2022-12-23 08:23:52,680 INFO [train.py:894] (3/4) Epoch 15, batch 2900, loss[loss=0.2333, simple_loss=0.3216, pruned_loss=0.07254, over 18398.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2788, pruned_loss=0.06158, over 3713718.22 frames. ], batch size: 53, lr: 7.62e-03, grad_scale: 8.0 2022-12-23 08:23:54,196 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0_sp0.9 from training. Duration: 22.9444375 2022-12-23 08:24:01,710 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp1.1 from training. Duration: 21.6318125 2022-12-23 08:24:23,310 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0 from training. Duration: 23.695 2022-12-23 08:24:47,581 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0 from training. Duration: 23.955 2022-12-23 08:25:07,018 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.195e+02 4.890e+02 5.936e+02 7.455e+02 1.185e+03, threshold=1.187e+03, percent-clipped=3.0 2022-12-23 08:25:13,447 INFO [train.py:894] (3/4) Epoch 15, batch 2950, loss[loss=0.1948, simple_loss=0.2769, pruned_loss=0.05637, over 18473.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2784, pruned_loss=0.06137, over 3713885.70 frames. ], batch size: 50, lr: 7.62e-03, grad_scale: 8.0 2022-12-23 08:25:15,380 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52038.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:25:18,229 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52040.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:25:22,735 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp0.9 from training. Duration: 26.438875 2022-12-23 08:25:30,055 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52048.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:26:06,653 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0021-26306-0_sp0.9 from training. Duration: 21.2444375 2022-12-23 08:26:06,680 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp0.9 from training. Duration: 31.02225 2022-12-23 08:26:17,410 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0 from training. Duration: 22.395 2022-12-23 08:26:27,554 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=52086.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:26:28,964 INFO [train.py:894] (3/4) Epoch 15, batch 3000, loss[loss=0.1888, simple_loss=0.2741, pruned_loss=0.05178, over 18544.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.279, pruned_loss=0.06159, over 3715042.08 frames. ], batch size: 55, lr: 7.61e-03, grad_scale: 8.0 2022-12-23 08:26:28,964 INFO [train.py:919] (3/4) Computing validation loss 2022-12-23 08:26:39,939 INFO [train.py:928] (3/4) Epoch 15, validation: loss=0.1652, simple_loss=0.2646, pruned_loss=0.03286, over 944034.00 frames. 2022-12-23 08:26:39,940 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-23 08:26:44,123 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0 from training. Duration: 21.075 2022-12-23 08:26:50,638 WARNING [train.py:1060] (3/4) Exclude cut with ID 6482-98857-0025-147532-0_sp0.9 from training. Duration: 20.0055625 2022-12-23 08:26:50,651 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0037-132304-0_sp0.9 from training. Duration: 22.05 2022-12-23 08:26:50,666 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0 from training. Duration: 26.8349375 2022-12-23 08:26:54,500 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp1.1 from training. Duration: 22.1090625 2022-12-23 08:26:54,623 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=52096.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:27:02,088 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp0.9 from training. Duration: 26.6166875 2022-12-23 08:27:02,477 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52101.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:27:05,016 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52103.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:27:19,130 WARNING [train.py:1060] (3/4) Exclude cut with ID 2046-178027-0000-53705-0_sp0.9 from training. Duration: 20.3055625 2022-12-23 08:27:27,082 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7329, 1.6489, 1.2600, 1.4984, 1.7876, 1.5912, 2.0531, 1.8771], device='cuda:3'), covar=tensor([0.0984, 0.1626, 0.2857, 0.1881, 0.1941, 0.0982, 0.1166, 0.1227], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0201, 0.0243, 0.0287, 0.0229, 0.0187, 0.0208, 0.0200], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 08:27:41,439 WARNING [train.py:1060] (3/4) Exclude cut with ID 7205-50138-0008-5373-0_sp0.9 from training. Duration: 20.7 2022-12-23 08:27:49,518 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.226e+02 4.827e+02 5.658e+02 7.296e+02 1.550e+03, threshold=1.132e+03, percent-clipped=2.0 2022-12-23 08:27:56,246 INFO [train.py:894] (3/4) Epoch 15, batch 3050, loss[loss=0.2174, simple_loss=0.2914, pruned_loss=0.07169, over 18660.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2799, pruned_loss=0.06197, over 3715843.97 frames. ], batch size: 60, lr: 7.61e-03, grad_scale: 8.0 2022-12-23 08:28:17,784 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52151.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:28:24,953 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0 from training. Duration: 22.48 2022-12-23 08:28:35,661 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7319, 1.6462, 1.2739, 1.5575, 1.7929, 1.5234, 2.0920, 1.8444], device='cuda:3'), covar=tensor([0.0931, 0.1453, 0.2618, 0.1643, 0.1746, 0.0964, 0.1021, 0.1120], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0200, 0.0241, 0.0285, 0.0227, 0.0185, 0.0206, 0.0199], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 08:28:39,799 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp0.9 from training. Duration: 29.816625 2022-12-23 08:28:57,350 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.3067, 5.3557, 4.8157, 2.7849, 5.4386, 4.2567, 1.0470, 3.8356], device='cuda:3'), covar=tensor([0.1585, 0.0999, 0.1139, 0.2854, 0.0632, 0.0704, 0.4745, 0.1115], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0133, 0.0156, 0.0123, 0.0136, 0.0111, 0.0143, 0.0112], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 08:29:00,321 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp1.1 from training. Duration: 22.7590625 2022-12-23 08:29:04,763 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6929, 3.5674, 3.6312, 1.4042, 3.7301, 2.7713, 0.6892, 2.5260], device='cuda:3'), covar=tensor([0.1871, 0.1147, 0.1379, 0.3945, 0.0991, 0.1045, 0.5190, 0.1589], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0133, 0.0155, 0.0123, 0.0135, 0.0111, 0.0143, 0.0112], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 08:29:05,891 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0 from training. Duration: 22.555 2022-12-23 08:29:11,938 INFO [train.py:894] (3/4) Epoch 15, batch 3100, loss[loss=0.1953, simple_loss=0.281, pruned_loss=0.05475, over 18574.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2792, pruned_loss=0.06152, over 3714690.61 frames. ], batch size: 56, lr: 7.60e-03, grad_scale: 8.0 2022-12-23 08:29:21,816 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52193.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:29:24,502 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0005-25975-0_sp0.9 from training. Duration: 21.688875 2022-12-23 08:29:29,292 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4815, 2.0507, 1.5691, 2.3473, 2.5345, 1.5228, 1.6146, 1.2803], device='cuda:3'), covar=tensor([0.2104, 0.1743, 0.1593, 0.0991, 0.1376, 0.1213, 0.2101, 0.1634], device='cuda:3'), in_proj_covar=tensor([0.0243, 0.0219, 0.0207, 0.0192, 0.0258, 0.0191, 0.0217, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 08:29:30,464 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=52199.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:29:44,846 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2022-12-23 08:30:00,905 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0_sp0.9 from training. Duration: 22.6 2022-12-23 08:30:22,175 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.672e+02 4.895e+02 6.130e+02 7.505e+02 1.385e+03, threshold=1.226e+03, percent-clipped=5.0 2022-12-23 08:30:27,883 INFO [train.py:894] (3/4) Epoch 15, batch 3150, loss[loss=0.1881, simple_loss=0.2705, pruned_loss=0.05283, over 18608.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2783, pruned_loss=0.06087, over 3714746.98 frames. ], batch size: 51, lr: 7.60e-03, grad_scale: 8.0 2022-12-23 08:30:39,797 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0 from training. Duration: 24.32 2022-12-23 08:30:53,396 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52254.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:31:38,176 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-276745-0093-13116-0_sp0.9 from training. Duration: 21.061125 2022-12-23 08:31:42,987 INFO [train.py:894] (3/4) Epoch 15, batch 3200, loss[loss=0.1835, simple_loss=0.2664, pruned_loss=0.05028, over 18529.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2774, pruned_loss=0.06017, over 3714622.79 frames. ], batch size: 58, lr: 7.60e-03, grad_scale: 8.0 2022-12-23 08:31:51,495 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0024-15855-0_sp0.9 from training. Duration: 20.32225 2022-12-23 08:32:03,694 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp1.1 from training. Duration: 0.7545625 2022-12-23 08:32:19,444 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0_sp0.9 from training. Duration: 23.9333125 2022-12-23 08:32:21,214 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52312.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:32:51,724 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp1.1 from training. Duration: 20.17275 2022-12-23 08:32:53,102 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.115e+02 4.671e+02 5.458e+02 7.644e+02 1.526e+03, threshold=1.092e+03, percent-clipped=4.0 2022-12-23 08:32:57,399 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp1.1 from training. Duration: 20.436375 2022-12-23 08:32:58,901 INFO [train.py:894] (3/4) Epoch 15, batch 3250, loss[loss=0.2475, simple_loss=0.314, pruned_loss=0.09055, over 18660.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2775, pruned_loss=0.06031, over 3714165.46 frames. ], batch size: 175, lr: 7.59e-03, grad_scale: 8.0 2022-12-23 08:33:54,372 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52373.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:34:15,744 INFO [train.py:894] (3/4) Epoch 15, batch 3300, loss[loss=0.1873, simple_loss=0.2772, pruned_loss=0.04864, over 18539.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2769, pruned_loss=0.05979, over 3712811.65 frames. ], batch size: 55, lr: 7.59e-03, grad_scale: 8.0 2022-12-23 08:34:20,217 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0_sp0.9 from training. Duration: 23.1055625 2022-12-23 08:34:20,291 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp1.1 from training. Duration: 23.5 2022-12-23 08:34:29,193 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52396.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:34:31,894 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp0.9 from training. Duration: 26.62775 2022-12-23 08:34:40,053 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52403.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:34:44,314 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0 from training. Duration: 21.105 2022-12-23 08:34:48,168 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0_sp0.9 from training. Duration: 24.411125 2022-12-23 08:35:17,012 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp1.1 from training. Duration: 21.263625 2022-12-23 08:35:27,536 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.913e+02 4.685e+02 5.736e+02 7.121e+02 1.192e+03, threshold=1.147e+03, percent-clipped=2.0 2022-12-23 08:35:33,513 INFO [train.py:894] (3/4) Epoch 15, batch 3350, loss[loss=0.1739, simple_loss=0.2498, pruned_loss=0.04898, over 18530.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2763, pruned_loss=0.05957, over 3713691.16 frames. ], batch size: 44, lr: 7.59e-03, grad_scale: 8.0 2022-12-23 08:35:39,792 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.3616, 3.7105, 3.9335, 4.2581, 4.0399, 3.9656, 4.4225, 2.4911], device='cuda:3'), covar=tensor([0.0656, 0.0644, 0.0593, 0.0774, 0.1155, 0.0980, 0.0749, 0.3538], device='cuda:3'), in_proj_covar=tensor([0.0328, 0.0216, 0.0226, 0.0252, 0.0307, 0.0259, 0.0275, 0.0274], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 08:35:49,089 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0 from training. Duration: 20.795 2022-12-23 08:35:55,726 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=52451.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:35:59,958 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0 from training. Duration: 24.76 2022-12-23 08:35:59,982 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0_sp0.9 from training. Duration: 22.25 2022-12-23 08:36:05,496 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2022-12-23 08:36:27,607 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp1.1 from training. Duration: 20.5045625 2022-12-23 08:36:31,798 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2022-12-23 08:36:49,784 INFO [train.py:894] (3/4) Epoch 15, batch 3400, loss[loss=0.2032, simple_loss=0.2828, pruned_loss=0.06176, over 18496.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2761, pruned_loss=0.05929, over 3713572.24 frames. ], batch size: 77, lr: 7.58e-03, grad_scale: 8.0 2022-12-23 08:36:57,908 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8757, 1.5560, 1.5544, 1.5546, 1.7575, 1.8775, 2.0271, 1.3736], device='cuda:3'), covar=tensor([0.0228, 0.0245, 0.0405, 0.0206, 0.0182, 0.0337, 0.0204, 0.0278], device='cuda:3'), in_proj_covar=tensor([0.0091, 0.0123, 0.0150, 0.0125, 0.0114, 0.0115, 0.0094, 0.0125], device='cuda:3'), out_proj_covar=tensor([7.4058e-05, 9.9153e-05, 1.2630e-04, 1.0129e-04, 9.4144e-05, 9.0018e-05, 7.4476e-05, 9.9713e-05], device='cuda:3') 2022-12-23 08:37:02,133 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5654, 1.1396, 1.6696, 2.8349, 2.0331, 2.3922, 0.7512, 2.0546], device='cuda:3'), covar=tensor([0.1814, 0.1825, 0.1490, 0.0758, 0.1172, 0.1097, 0.2278, 0.1156], device='cuda:3'), in_proj_covar=tensor([0.0103, 0.0116, 0.0132, 0.0140, 0.0106, 0.0136, 0.0129, 0.0112], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 08:37:57,104 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.148e+02 4.332e+02 5.107e+02 6.639e+02 1.156e+03, threshold=1.021e+03, percent-clipped=1.0 2022-12-23 08:38:02,850 INFO [train.py:894] (3/4) Epoch 15, batch 3450, loss[loss=0.1717, simple_loss=0.2562, pruned_loss=0.04356, over 18717.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2769, pruned_loss=0.05988, over 3713908.01 frames. ], batch size: 50, lr: 7.58e-03, grad_scale: 8.0 2022-12-23 08:38:20,355 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52549.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:38:29,013 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52555.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 08:38:57,285 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2022-12-23 08:39:16,292 INFO [train.py:894] (3/4) Epoch 15, batch 3500, loss[loss=0.2213, simple_loss=0.2923, pruned_loss=0.07517, over 18668.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2765, pruned_loss=0.05973, over 3714445.07 frames. ], batch size: 181, lr: 7.58e-03, grad_scale: 8.0 2022-12-23 08:39:37,573 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp1.1 from training. Duration: 22.2954375 2022-12-23 08:39:47,482 INFO [train.py:894] (3/4) Epoch 16, batch 0, loss[loss=0.176, simple_loss=0.2649, pruned_loss=0.0435, over 18401.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2649, pruned_loss=0.0435, over 18401.00 frames. ], batch size: 46, lr: 7.33e-03, grad_scale: 8.0 2022-12-23 08:39:47,482 INFO [train.py:919] (3/4) Computing validation loss 2022-12-23 08:39:58,260 INFO [train.py:928] (3/4) Epoch 16, validation: loss=0.1687, simple_loss=0.2674, pruned_loss=0.03497, over 944034.00 frames. 2022-12-23 08:39:58,262 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-23 08:40:03,353 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6346, 1.8016, 1.3959, 2.1363, 2.5180, 1.5253, 1.5417, 1.2998], device='cuda:3'), covar=tensor([0.1913, 0.1833, 0.1664, 0.1034, 0.1290, 0.1154, 0.2097, 0.1594], device='cuda:3'), in_proj_covar=tensor([0.0246, 0.0221, 0.0209, 0.0195, 0.0261, 0.0194, 0.0219, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 08:40:34,384 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52616.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 08:40:49,818 WARNING [train.py:1060] (3/4) Exclude cut with ID 298-126791-0067-24026-0_sp0.9 from training. Duration: 21.438875 2022-12-23 08:40:54,026 WARNING [train.py:1060] (3/4) Exclude cut with ID 5652-39938-0025-23684-0_sp0.9 from training. Duration: 22.2055625 2022-12-23 08:40:58,105 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.574e+02 3.834e+02 5.457e+02 6.710e+02 1.332e+03, threshold=1.091e+03, percent-clipped=4.0 2022-12-23 08:41:14,215 INFO [train.py:894] (3/4) Epoch 16, batch 50, loss[loss=0.1776, simple_loss=0.26, pruned_loss=0.04757, over 18634.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2757, pruned_loss=0.05176, over 837813.41 frames. ], batch size: 41, lr: 7.33e-03, grad_scale: 8.0 2022-12-23 08:41:52,363 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52668.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:42:29,862 INFO [train.py:894] (3/4) Epoch 16, batch 100, loss[loss=0.1647, simple_loss=0.24, pruned_loss=0.04464, over 18426.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2714, pruned_loss=0.05013, over 1475326.87 frames. ], batch size: 42, lr: 7.33e-03, grad_scale: 8.0 2022-12-23 08:42:34,309 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52696.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:43:14,302 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7588, 1.5551, 1.4891, 1.7366, 1.7207, 1.8097, 1.9269, 1.3569], device='cuda:3'), covar=tensor([0.0294, 0.0234, 0.0442, 0.0186, 0.0194, 0.0367, 0.0214, 0.0298], device='cuda:3'), in_proj_covar=tensor([0.0092, 0.0124, 0.0153, 0.0127, 0.0116, 0.0117, 0.0095, 0.0126], device='cuda:3'), out_proj_covar=tensor([7.4982e-05, 1.0037e-04, 1.2900e-04, 1.0263e-04, 9.5691e-05, 9.1853e-05, 7.5210e-05, 1.0110e-04], device='cuda:3') 2022-12-23 08:43:29,093 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.318e+02 3.254e+02 4.108e+02 5.312e+02 8.636e+02, threshold=8.216e+02, percent-clipped=0.0 2022-12-23 08:43:35,819 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2022-12-23 08:43:43,853 INFO [train.py:894] (3/4) Epoch 16, batch 150, loss[loss=0.1614, simple_loss=0.2463, pruned_loss=0.03832, over 18670.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2693, pruned_loss=0.04831, over 1971687.70 frames. ], batch size: 48, lr: 7.32e-03, grad_scale: 8.0 2022-12-23 08:43:45,431 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=52744.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:43:57,755 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52752.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:43:58,908 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0 from training. Duration: 24.525 2022-12-23 08:44:32,728 WARNING [train.py:1060] (3/4) Exclude cut with ID 3699-47246-0007-3408-0_sp0.9 from training. Duration: 20.26675 2022-12-23 08:44:45,470 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp0.9 from training. Duration: 27.25 2022-12-23 08:44:59,484 INFO [train.py:894] (3/4) Epoch 16, batch 200, loss[loss=0.1686, simple_loss=0.2666, pruned_loss=0.0353, over 18533.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2683, pruned_loss=0.04762, over 2358040.41 frames. ], batch size: 55, lr: 7.32e-03, grad_scale: 8.0 2022-12-23 08:45:31,807 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52813.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:45:59,672 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0 from training. Duration: 21.68 2022-12-23 08:46:00,914 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.133e+02 3.642e+02 4.189e+02 4.973e+02 1.073e+03, threshold=8.378e+02, percent-clipped=4.0 2022-12-23 08:46:13,119 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0 from training. Duration: 21.6300625 2022-12-23 08:46:15,855 INFO [train.py:894] (3/4) Epoch 16, batch 250, loss[loss=0.1645, simple_loss=0.2584, pruned_loss=0.03531, over 18540.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2659, pruned_loss=0.04656, over 2659128.59 frames. ], batch size: 55, lr: 7.32e-03, grad_scale: 8.0 2022-12-23 08:46:24,600 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.91 vs. limit=5.0 2022-12-23 08:46:25,663 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52849.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:46:36,052 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0_sp0.9 from training. Duration: 24.033375 2022-12-23 08:47:17,644 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.2097, 1.3553, 1.6634, 0.6978, 0.8982, 1.7792, 1.6615, 1.4971], device='cuda:3'), covar=tensor([0.0674, 0.0313, 0.0295, 0.0355, 0.0373, 0.0417, 0.0216, 0.0596], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0164, 0.0121, 0.0138, 0.0146, 0.0140, 0.0156, 0.0164], device='cuda:3'), out_proj_covar=tensor([1.1571e-04, 1.2966e-04, 9.3993e-05, 1.0577e-04, 1.1311e-04, 1.1060e-04, 1.2382e-04, 1.2821e-04], device='cuda:3') 2022-12-23 08:47:32,158 INFO [train.py:894] (3/4) Epoch 16, batch 300, loss[loss=0.1719, simple_loss=0.2634, pruned_loss=0.04014, over 18636.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2657, pruned_loss=0.04639, over 2893020.13 frames. ], batch size: 53, lr: 7.31e-03, grad_scale: 8.0 2022-12-23 08:47:33,905 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0 from training. Duration: 22.905 2022-12-23 08:47:33,956 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp1.1 from training. Duration: 23.4318125 2022-12-23 08:47:39,161 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=52897.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:47:46,662 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2022-12-23 08:48:00,886 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52911.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 08:48:31,999 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.517e+02 3.926e+02 4.602e+02 5.842e+02 9.674e+02, threshold=9.204e+02, percent-clipped=4.0 2022-12-23 08:48:38,532 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6915, 1.6118, 1.8642, 1.0513, 1.8667, 1.9077, 1.3710, 2.1606], device='cuda:3'), covar=tensor([0.1110, 0.1758, 0.1178, 0.1666, 0.0691, 0.1040, 0.2288, 0.0485], device='cuda:3'), in_proj_covar=tensor([0.0196, 0.0203, 0.0206, 0.0191, 0.0174, 0.0213, 0.0212, 0.0187], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 08:48:47,089 INFO [train.py:894] (3/4) Epoch 16, batch 350, loss[loss=0.1653, simple_loss=0.2555, pruned_loss=0.03756, over 18389.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2674, pruned_loss=0.04779, over 3075313.01 frames. ], batch size: 46, lr: 7.31e-03, grad_scale: 8.0 2022-12-23 08:49:24,375 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52968.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:49:29,925 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp1.1 from training. Duration: 20.82275 2022-12-23 08:49:31,424 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp0.9 from training. Duration: 25.45 2022-12-23 08:50:01,796 INFO [train.py:894] (3/4) Epoch 16, batch 400, loss[loss=0.1924, simple_loss=0.2859, pruned_loss=0.04943, over 18566.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2691, pruned_loss=0.04867, over 3217242.11 frames. ], batch size: 56, lr: 7.30e-03, grad_scale: 8.0 2022-12-23 08:50:04,396 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6117, 1.5096, 1.5808, 1.5176, 1.0816, 2.9976, 1.2540, 1.7234], device='cuda:3'), covar=tensor([0.3125, 0.1988, 0.1968, 0.2018, 0.1417, 0.0219, 0.1671, 0.0880], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0115, 0.0126, 0.0119, 0.0103, 0.0098, 0.0092, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 08:50:12,772 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7022, 1.6513, 1.2767, 1.6322, 1.8368, 1.5460, 2.0877, 1.8111], device='cuda:3'), covar=tensor([0.0932, 0.1774, 0.2700, 0.1707, 0.1745, 0.0941, 0.1036, 0.1222], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0201, 0.0240, 0.0285, 0.0227, 0.0185, 0.0204, 0.0198], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 08:50:32,586 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0 from training. Duration: 25.775 2022-12-23 08:50:38,333 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=53016.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:50:55,784 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0_sp0.9 from training. Duration: 22.25 2022-12-23 08:51:03,196 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.768e+02 3.604e+02 4.680e+02 6.020e+02 1.897e+03, threshold=9.359e+02, percent-clipped=5.0 2022-12-23 08:51:17,415 INFO [train.py:894] (3/4) Epoch 16, batch 450, loss[loss=0.1969, simple_loss=0.2857, pruned_loss=0.05407, over 18579.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2708, pruned_loss=0.04911, over 3327489.57 frames. ], batch size: 69, lr: 7.30e-03, grad_scale: 8.0 2022-12-23 08:51:21,855 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0 from training. Duration: 26.205 2022-12-23 08:51:38,051 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp0.9 from training. Duration: 30.1555625 2022-12-23 08:51:38,381 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53056.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:51:42,512 WARNING [train.py:1060] (3/4) Exclude cut with ID 1265-135635-0050-6781-0_sp0.9 from training. Duration: 21.8333125 2022-12-23 08:51:46,190 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2022-12-23 08:51:52,862 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp1.1 from training. Duration: 20.6545625 2022-12-23 08:52:02,786 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.16 vs. limit=5.0 2022-12-23 08:52:32,581 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0045-39920-0_sp0.9 from training. Duration: 20.52225 2022-12-23 08:52:35,691 INFO [train.py:894] (3/4) Epoch 16, batch 500, loss[loss=0.1798, simple_loss=0.268, pruned_loss=0.04578, over 18568.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2728, pruned_loss=0.05003, over 3412419.48 frames. ], batch size: 51, lr: 7.30e-03, grad_scale: 8.0 2022-12-23 08:52:52,235 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp0.9 from training. Duration: 29.1166875 2022-12-23 08:52:57,938 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53108.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:53:11,395 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53117.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:53:36,075 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.439e+02 3.611e+02 4.573e+02 5.633e+02 1.501e+03, threshold=9.146e+02, percent-clipped=5.0 2022-12-23 08:53:52,522 INFO [train.py:894] (3/4) Epoch 16, batch 550, loss[loss=0.2142, simple_loss=0.2945, pruned_loss=0.06693, over 18589.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2738, pruned_loss=0.05025, over 3478467.66 frames. ], batch size: 174, lr: 7.29e-03, grad_scale: 8.0 2022-12-23 08:53:55,464 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133211-0007-59831-0_sp0.9 from training. Duration: 21.388875 2022-12-23 08:54:02,847 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.6298, 3.1510, 2.2119, 1.7807, 3.7467, 4.0133, 3.3514, 2.8789], device='cuda:3'), covar=tensor([0.0357, 0.0362, 0.0591, 0.0743, 0.0185, 0.0296, 0.0387, 0.0622], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0123, 0.0128, 0.0119, 0.0092, 0.0121, 0.0133, 0.0152], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 08:54:32,168 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0 from training. Duration: 22.72 2022-12-23 08:54:32,213 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0_sp0.9 from training. Duration: 22.7444375 2022-12-23 08:54:38,914 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8657, 1.3016, 0.6836, 1.4016, 2.2599, 1.3563, 1.6261, 1.8979], device='cuda:3'), covar=tensor([0.1974, 0.2718, 0.2803, 0.1933, 0.1777, 0.2073, 0.1847, 0.2122], device='cuda:3'), in_proj_covar=tensor([0.0093, 0.0098, 0.0118, 0.0096, 0.0114, 0.0092, 0.0099, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 08:54:59,842 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2022-12-23 08:55:08,110 INFO [train.py:894] (3/4) Epoch 16, batch 600, loss[loss=0.1748, simple_loss=0.2571, pruned_loss=0.04622, over 18511.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2727, pruned_loss=0.04991, over 3530077.58 frames. ], batch size: 44, lr: 7.29e-03, grad_scale: 8.0 2022-12-23 08:55:15,545 WARNING [train.py:1060] (3/4) Exclude cut with ID 4133-6541-0027-40495-0_sp1.1 from training. Duration: 0.9681875 2022-12-23 08:55:18,631 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0_sp0.9 from training. Duration: 22.3166875 2022-12-23 08:55:24,741 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133212-0015-59917-0_sp0.9 from training. Duration: 21.8166875 2022-12-23 08:55:35,001 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53211.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 08:55:54,776 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7158, 2.4142, 1.9739, 1.3060, 3.0100, 2.7560, 2.5475, 1.9178], device='cuda:3'), covar=tensor([0.0322, 0.0346, 0.0509, 0.0721, 0.0189, 0.0318, 0.0393, 0.0692], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0123, 0.0128, 0.0118, 0.0092, 0.0121, 0.0133, 0.0152], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 08:56:07,404 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.754e+02 3.712e+02 4.400e+02 5.395e+02 1.040e+03, threshold=8.800e+02, percent-clipped=2.0 2022-12-23 08:56:21,050 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53241.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:56:23,462 INFO [train.py:894] (3/4) Epoch 16, batch 650, loss[loss=0.1786, simple_loss=0.2532, pruned_loss=0.052, over 18397.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2727, pruned_loss=0.04978, over 3570115.12 frames. ], batch size: 42, lr: 7.29e-03, grad_scale: 8.0 2022-12-23 08:56:46,973 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=53259.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 08:57:07,255 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.01 vs. limit=5.0 2022-12-23 08:57:07,891 WARNING [train.py:1060] (3/4) Exclude cut with ID 4957-30119-0041-23990-0_sp0.9 from training. Duration: 20.22775 2022-12-23 08:57:38,505 INFO [train.py:894] (3/4) Epoch 16, batch 700, loss[loss=0.1824, simple_loss=0.2548, pruned_loss=0.055, over 18400.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2732, pruned_loss=0.05023, over 3602295.09 frames. ], batch size: 42, lr: 7.28e-03, grad_scale: 8.0 2022-12-23 08:57:52,111 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53302.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:57:53,213 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp1.1 from training. Duration: 24.67275 2022-12-23 08:57:55,029 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53304.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:58:11,993 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53315.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:58:21,004 WARNING [train.py:1060] (3/4) Exclude cut with ID 3082-165428-0081-50734-0_sp0.9 from training. Duration: 21.8055625 2022-12-23 08:58:39,337 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.032e+02 3.701e+02 4.471e+02 5.554e+02 1.139e+03, threshold=8.941e+02, percent-clipped=3.0 2022-12-23 08:58:54,397 INFO [train.py:894] (3/4) Epoch 16, batch 750, loss[loss=0.1624, simple_loss=0.2403, pruned_loss=0.04227, over 18510.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2736, pruned_loss=0.05034, over 3627992.49 frames. ], batch size: 41, lr: 7.28e-03, grad_scale: 8.0 2022-12-23 08:59:00,155 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0_sp0.9 from training. Duration: 22.6666875 2022-12-23 08:59:26,819 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53365.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:59:29,877 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53367.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:59:43,792 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53376.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 08:59:46,746 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0551, 1.8328, 1.8333, 1.1102, 2.3905, 2.0469, 2.0065, 1.5421], device='cuda:3'), covar=tensor([0.0323, 0.0413, 0.0405, 0.0693, 0.0237, 0.0355, 0.0368, 0.0789], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0124, 0.0129, 0.0119, 0.0092, 0.0121, 0.0133, 0.0153], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 09:00:01,887 WARNING [train.py:1060] (3/4) Exclude cut with ID 2411-132532-0017-82279-0_sp1.1 from training. Duration: 0.9681875 2022-12-23 09:00:09,102 INFO [train.py:894] (3/4) Epoch 16, batch 800, loss[loss=0.1865, simple_loss=0.2766, pruned_loss=0.04817, over 18389.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2737, pruned_loss=0.05026, over 3647288.46 frames. ], batch size: 53, lr: 7.28e-03, grad_scale: 8.0 2022-12-23 09:00:11,138 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9074, 1.1323, 1.5877, 1.5381, 1.9102, 1.9549, 1.6692, 1.5291], device='cuda:3'), covar=tensor([0.1995, 0.2896, 0.2346, 0.2456, 0.1739, 0.0864, 0.2725, 0.1171], device='cuda:3'), in_proj_covar=tensor([0.0258, 0.0289, 0.0267, 0.0300, 0.0291, 0.0242, 0.0322, 0.0230], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 09:00:27,221 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0 from training. Duration: 22.485 2022-12-23 09:00:32,019 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53408.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 09:00:38,169 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53412.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 09:00:43,652 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3946, 2.6660, 3.2337, 0.8405, 2.6988, 3.5710, 2.4674, 2.7402], device='cuda:3'), covar=tensor([0.0847, 0.0375, 0.0300, 0.0509, 0.0445, 0.0319, 0.0376, 0.0675], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0164, 0.0121, 0.0137, 0.0145, 0.0139, 0.0156, 0.0165], device='cuda:3'), out_proj_covar=tensor([1.1536e-04, 1.2940e-04, 9.4005e-05, 1.0489e-04, 1.1269e-04, 1.0979e-04, 1.2357e-04, 1.2901e-04], device='cuda:3') 2022-12-23 09:00:56,891 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8278, 0.7707, 1.5984, 1.4468, 1.8353, 1.8433, 1.5381, 1.5275], device='cuda:3'), covar=tensor([0.1889, 0.2838, 0.2304, 0.2246, 0.1672, 0.0869, 0.2475, 0.1134], device='cuda:3'), in_proj_covar=tensor([0.0261, 0.0292, 0.0269, 0.0302, 0.0293, 0.0244, 0.0325, 0.0231], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 09:01:03,257 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53428.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 09:01:05,579 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp1.1 from training. Duration: 23.82275 2022-12-23 09:01:10,794 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.669e+02 3.732e+02 4.519e+02 5.573e+02 1.568e+03, threshold=9.039e+02, percent-clipped=4.0 2022-12-23 09:01:18,543 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0 from training. Duration: 20.77 2022-12-23 09:01:25,723 INFO [train.py:894] (3/4) Epoch 16, batch 850, loss[loss=0.1931, simple_loss=0.2622, pruned_loss=0.06202, over 18606.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2728, pruned_loss=0.04976, over 3662597.79 frames. ], batch size: 45, lr: 7.27e-03, grad_scale: 8.0 2022-12-23 09:01:27,231 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0_sp0.9 from training. Duration: 24.088875 2022-12-23 09:01:44,876 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=53456.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 09:01:55,786 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp1.1 from training. Duration: 20.4409375 2022-12-23 09:02:41,480 INFO [train.py:894] (3/4) Epoch 16, batch 900, loss[loss=0.1845, simple_loss=0.2652, pruned_loss=0.05187, over 18407.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2732, pruned_loss=0.04987, over 3673698.99 frames. ], batch size: 46, lr: 7.27e-03, grad_scale: 8.0 2022-12-23 09:02:43,743 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2022-12-23 09:02:55,601 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2589, 1.9396, 1.5971, 0.4857, 1.4389, 1.8258, 1.6139, 1.7913], device='cuda:3'), covar=tensor([0.0582, 0.0506, 0.1003, 0.1604, 0.1095, 0.1531, 0.1520, 0.0612], device='cuda:3'), in_proj_covar=tensor([0.0169, 0.0184, 0.0203, 0.0190, 0.0207, 0.0197, 0.0212, 0.0198], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 09:03:12,641 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0_sp0.9 from training. Duration: 22.511125 2022-12-23 09:03:12,674 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0 from training. Duration: 20.675 2022-12-23 09:03:42,941 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.459e+02 3.846e+02 4.469e+02 5.428e+02 1.357e+03, threshold=8.938e+02, percent-clipped=3.0 2022-12-23 09:03:57,837 INFO [train.py:894] (3/4) Epoch 16, batch 950, loss[loss=0.1833, simple_loss=0.2791, pruned_loss=0.04371, over 18518.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2732, pruned_loss=0.04993, over 3682741.19 frames. ], batch size: 55, lr: 7.27e-03, grad_scale: 8.0 2022-12-23 09:04:27,238 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.8319, 3.2249, 2.8682, 1.3153, 2.6056, 2.6090, 2.3565, 2.5970], device='cuda:3'), covar=tensor([0.0551, 0.0598, 0.1319, 0.1803, 0.1663, 0.1334, 0.1302, 0.0961], device='cuda:3'), in_proj_covar=tensor([0.0170, 0.0184, 0.0204, 0.0191, 0.0208, 0.0198, 0.0212, 0.0198], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 09:04:50,937 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp0.9 from training. Duration: 24.9833125 2022-12-23 09:05:04,498 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2819, 1.8845, 2.0667, 2.3569, 2.1865, 3.1839, 1.8022, 2.0038], device='cuda:3'), covar=tensor([0.0638, 0.1310, 0.1021, 0.0714, 0.1061, 0.0271, 0.1057, 0.1094], device='cuda:3'), in_proj_covar=tensor([0.0072, 0.0081, 0.0072, 0.0073, 0.0089, 0.0074, 0.0083, 0.0076], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 09:05:12,796 INFO [train.py:894] (3/4) Epoch 16, batch 1000, loss[loss=0.2342, simple_loss=0.312, pruned_loss=0.07816, over 18573.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2727, pruned_loss=0.0496, over 3689550.51 frames. ], batch size: 57, lr: 7.26e-03, grad_scale: 8.0 2022-12-23 09:05:18,877 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53597.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 09:05:21,588 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0 from training. Duration: 27.14 2022-12-23 09:05:25,195 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0262, 2.0959, 1.2927, 2.5470, 2.2007, 1.7677, 3.2318, 1.9875], device='cuda:3'), covar=tensor([0.0950, 0.1659, 0.2836, 0.1796, 0.1725, 0.1041, 0.0811, 0.1315], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0201, 0.0242, 0.0285, 0.0227, 0.0185, 0.0205, 0.0199], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 09:05:31,789 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2022-12-23 09:05:39,063 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0 from training. Duration: 22.44 2022-12-23 09:06:14,010 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.372e+02 3.452e+02 4.205e+02 5.209e+02 9.505e+02, threshold=8.411e+02, percent-clipped=1.0 2022-12-23 09:06:28,584 INFO [train.py:894] (3/4) Epoch 16, batch 1050, loss[loss=0.185, simple_loss=0.2752, pruned_loss=0.04741, over 18713.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.272, pruned_loss=0.04935, over 3694475.89 frames. ], batch size: 52, lr: 7.26e-03, grad_scale: 8.0 2022-12-23 09:06:53,784 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9974, 2.0119, 1.2153, 2.4623, 2.0850, 1.7451, 3.0837, 2.0064], device='cuda:3'), covar=tensor([0.1000, 0.1839, 0.3140, 0.2000, 0.1828, 0.1146, 0.0882, 0.1408], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0201, 0.0241, 0.0284, 0.0226, 0.0184, 0.0205, 0.0199], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 09:06:54,958 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53660.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 09:07:00,300 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0060-62364-0_sp0.9 from training. Duration: 21.361125 2022-12-23 09:07:07,040 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp1.1 from training. Duration: 27.0318125 2022-12-23 09:07:11,449 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53671.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 09:07:16,725 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp0.9 from training. Duration: 28.638875 2022-12-23 09:07:32,413 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.5573, 3.1781, 2.8055, 1.3936, 2.5510, 2.3402, 2.0752, 2.5292], device='cuda:3'), covar=tensor([0.0636, 0.0488, 0.1302, 0.1782, 0.1703, 0.1495, 0.1561, 0.0992], device='cuda:3'), in_proj_covar=tensor([0.0169, 0.0182, 0.0202, 0.0190, 0.0207, 0.0197, 0.0210, 0.0198], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 09:07:33,436 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0 from training. Duration: 20.4 2022-12-23 09:07:43,947 INFO [train.py:894] (3/4) Epoch 16, batch 1100, loss[loss=0.1842, simple_loss=0.2738, pruned_loss=0.04727, over 18561.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2723, pruned_loss=0.0495, over 3698827.99 frames. ], batch size: 57, lr: 7.26e-03, grad_scale: 8.0 2022-12-23 09:08:05,612 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0 from training. Duration: 20.025 2022-12-23 09:08:05,622 WARNING [train.py:1060] (3/4) Exclude cut with ID 2364-131735-0112-64612-0_sp0.9 from training. Duration: 20.488875 2022-12-23 09:08:11,791 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0 from training. Duration: 29.735 2022-12-23 09:08:13,735 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53712.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 09:08:30,134 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53723.0, num_to_drop=1, layers_to_drop={3} 2022-12-23 09:08:45,993 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.409e+02 3.829e+02 4.383e+02 5.525e+02 1.392e+03, threshold=8.765e+02, percent-clipped=4.0 2022-12-23 09:09:00,846 INFO [train.py:894] (3/4) Epoch 16, batch 1150, loss[loss=0.2121, simple_loss=0.2999, pruned_loss=0.06212, over 18669.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2714, pruned_loss=0.04905, over 3701866.21 frames. ], batch size: 60, lr: 7.25e-03, grad_scale: 8.0 2022-12-23 09:09:27,018 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=53760.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 09:09:31,795 WARNING [train.py:1060] (3/4) Exclude cut with ID 7276-92427-0014-12983-0_sp0.9 from training. Duration: 21.3055625 2022-12-23 09:09:34,771 WARNING [train.py:1060] (3/4) Exclude cut with ID 1025-75365-0008-79168-0_sp0.9 from training. Duration: 22.0666875 2022-12-23 09:09:47,270 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.11 vs. limit=5.0 2022-12-23 09:10:16,302 INFO [train.py:894] (3/4) Epoch 16, batch 1200, loss[loss=0.1577, simple_loss=0.2398, pruned_loss=0.03777, over 18659.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.271, pruned_loss=0.04897, over 3704314.50 frames. ], batch size: 41, lr: 7.25e-03, grad_scale: 8.0 2022-12-23 09:11:02,403 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.8188, 3.3414, 3.2867, 3.8180, 3.4589, 3.3998, 3.9530, 1.1512], device='cuda:3'), covar=tensor([0.0816, 0.0713, 0.0749, 0.0786, 0.1488, 0.1293, 0.0700, 0.4920], device='cuda:3'), in_proj_covar=tensor([0.0316, 0.0208, 0.0218, 0.0245, 0.0295, 0.0252, 0.0263, 0.0265], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 09:11:16,024 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.339e+02 3.426e+02 4.196e+02 5.027e+02 8.544e+02, threshold=8.392e+02, percent-clipped=0.0 2022-12-23 09:11:25,044 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0 from training. Duration: 20.26 2022-12-23 09:11:31,015 INFO [train.py:894] (3/4) Epoch 16, batch 1250, loss[loss=0.1749, simple_loss=0.2697, pruned_loss=0.04009, over 18468.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2711, pruned_loss=0.04854, over 3706975.33 frames. ], batch size: 64, lr: 7.25e-03, grad_scale: 8.0 2022-12-23 09:11:37,597 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0030-9324-0_sp0.9 from training. Duration: 21.3444375 2022-12-23 09:12:33,849 WARNING [train.py:1060] (3/4) Exclude cut with ID 497-129325-0061-62254-0_sp1.1 from training. Duration: 0.97725 2022-12-23 09:12:49,442 INFO [train.py:894] (3/4) Epoch 16, batch 1300, loss[loss=0.1712, simple_loss=0.2734, pruned_loss=0.03446, over 18577.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2712, pruned_loss=0.04869, over 3708821.34 frames. ], batch size: 56, lr: 7.24e-03, grad_scale: 8.0 2022-12-23 09:12:55,853 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53897.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 09:13:14,666 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0_sp0.9 from training. Duration: 22.97225 2022-12-23 09:13:43,645 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0047-39922-0_sp0.9 from training. Duration: 21.97775 2022-12-23 09:13:48,127 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.926e+02 3.573e+02 4.453e+02 5.683e+02 9.534e+02, threshold=8.906e+02, percent-clipped=1.0 2022-12-23 09:13:48,541 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4747, 2.2450, 1.7480, 0.7803, 1.6493, 1.9755, 1.6853, 1.9434], device='cuda:3'), covar=tensor([0.0571, 0.0472, 0.1140, 0.1562, 0.1199, 0.1382, 0.1461, 0.0703], device='cuda:3'), in_proj_covar=tensor([0.0170, 0.0184, 0.0204, 0.0191, 0.0209, 0.0198, 0.0212, 0.0201], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 09:13:57,556 WARNING [train.py:1060] (3/4) Exclude cut with ID 1112-1043-0006-89194-0_sp0.9 from training. Duration: 21.8333125 2022-12-23 09:14:04,502 INFO [train.py:894] (3/4) Epoch 16, batch 1350, loss[loss=0.1535, simple_loss=0.24, pruned_loss=0.03347, over 18516.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2715, pruned_loss=0.04877, over 3709806.97 frames. ], batch size: 41, lr: 7.24e-03, grad_scale: 16.0 2022-12-23 09:14:07,632 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=53945.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 09:14:07,910 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4786, 2.6650, 2.9712, 1.3139, 3.1618, 2.8910, 1.9692, 3.5504], device='cuda:3'), covar=tensor([0.1314, 0.1674, 0.1336, 0.2316, 0.0724, 0.1285, 0.2231, 0.0598], device='cuda:3'), in_proj_covar=tensor([0.0195, 0.0201, 0.0202, 0.0189, 0.0173, 0.0212, 0.0210, 0.0186], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 09:14:09,041 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0 from training. Duration: 20.47 2022-12-23 09:14:30,034 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53960.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 09:14:46,773 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53971.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 09:14:56,777 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9838, 1.6523, 0.9588, 1.5002, 2.1164, 1.6670, 1.8123, 2.0443], device='cuda:3'), covar=tensor([0.1395, 0.1774, 0.2312, 0.1456, 0.1708, 0.1510, 0.1315, 0.1386], device='cuda:3'), in_proj_covar=tensor([0.0092, 0.0098, 0.0117, 0.0096, 0.0115, 0.0091, 0.0098, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 09:15:16,451 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0037-39912-0_sp0.9 from training. Duration: 20.67225 2022-12-23 09:15:19,731 INFO [train.py:894] (3/4) Epoch 16, batch 1400, loss[loss=0.1803, simple_loss=0.2506, pruned_loss=0.05496, over 18689.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2708, pruned_loss=0.04839, over 3710577.22 frames. ], batch size: 46, lr: 7.24e-03, grad_scale: 16.0 2022-12-23 09:15:39,384 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp0.9 from training. Duration: 25.2444375 2022-12-23 09:15:45,864 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=54008.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 09:16:01,812 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=54019.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 09:16:03,090 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0021-76797-0_sp0.9 from training. Duration: 21.1445 2022-12-23 09:16:08,417 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=54023.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 09:16:22,367 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.191e+02 3.783e+02 4.326e+02 5.287e+02 1.206e+03, threshold=8.652e+02, percent-clipped=5.0 2022-12-23 09:16:28,605 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6995, 2.2382, 1.6482, 2.5495, 1.9696, 2.0975, 2.1224, 2.6417], device='cuda:3'), covar=tensor([0.1781, 0.2938, 0.1805, 0.2646, 0.3332, 0.0945, 0.2780, 0.0828], device='cuda:3'), in_proj_covar=tensor([0.0288, 0.0281, 0.0236, 0.0347, 0.0260, 0.0221, 0.0278, 0.0203], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 09:16:38,055 INFO [train.py:894] (3/4) Epoch 16, batch 1450, loss[loss=0.1894, simple_loss=0.2873, pruned_loss=0.04574, over 18726.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2697, pruned_loss=0.04781, over 3711048.29 frames. ], batch size: 54, lr: 7.23e-03, grad_scale: 16.0 2022-12-23 09:16:55,564 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2022-12-23 09:17:13,774 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp0.9 from training. Duration: 33.038875 2022-12-23 09:17:20,291 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=54071.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 09:17:36,168 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2022-12-23 09:17:52,842 INFO [train.py:894] (3/4) Epoch 16, batch 1500, loss[loss=0.2047, simple_loss=0.2972, pruned_loss=0.05609, over 18582.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2704, pruned_loss=0.04827, over 3711316.48 frames. ], batch size: 56, lr: 7.23e-03, grad_scale: 16.0 2022-12-23 09:17:52,933 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64291-0000-16059-0_sp0.9 from training. Duration: 20.0944375 2022-12-23 09:18:07,817 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp1.1 from training. Duration: 20.4 2022-12-23 09:18:15,961 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0 from training. Duration: 20.085 2022-12-23 09:18:20,778 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.4692, 3.3169, 2.5875, 1.9617, 3.7770, 3.9606, 3.1784, 2.7187], device='cuda:3'), covar=tensor([0.0302, 0.0353, 0.0475, 0.0651, 0.0206, 0.0281, 0.0426, 0.0734], device='cuda:3'), in_proj_covar=tensor([0.0125, 0.0127, 0.0131, 0.0122, 0.0096, 0.0125, 0.0136, 0.0157], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 09:18:26,441 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0_sp0.9 from training. Duration: 23.07775 2022-12-23 09:18:51,413 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.264e+02 4.118e+02 4.740e+02 5.968e+02 1.195e+03, threshold=9.480e+02, percent-clipped=4.0 2022-12-23 09:19:06,982 INFO [train.py:894] (3/4) Epoch 16, batch 1550, loss[loss=0.1788, simple_loss=0.2687, pruned_loss=0.04444, over 18717.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2704, pruned_loss=0.04819, over 3712461.88 frames. ], batch size: 52, lr: 7.23e-03, grad_scale: 16.0 2022-12-23 09:19:13,447 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp0.9 from training. Duration: 24.9333125 2022-12-23 09:19:58,559 WARNING [train.py:1060] (3/4) Exclude cut with ID 5118-111612-0016-124680-0_sp0.9 from training. Duration: 20.388875 2022-12-23 09:20:05,915 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp1.1 from training. Duration: 20.3590625 2022-12-23 09:20:15,490 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8948, 1.8755, 2.2000, 1.2133, 2.3720, 2.2278, 1.5085, 2.5851], device='cuda:3'), covar=tensor([0.1108, 0.1755, 0.1157, 0.1816, 0.0663, 0.1067, 0.2235, 0.0460], device='cuda:3'), in_proj_covar=tensor([0.0197, 0.0204, 0.0206, 0.0192, 0.0175, 0.0214, 0.0213, 0.0188], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 09:20:22,361 INFO [train.py:894] (3/4) Epoch 16, batch 1600, loss[loss=0.1791, simple_loss=0.2569, pruned_loss=0.05063, over 18529.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2707, pruned_loss=0.04836, over 3713257.54 frames. ], batch size: 44, lr: 7.22e-03, grad_scale: 16.0 2022-12-23 09:20:39,367 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2022-12-23 09:21:13,984 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0_sp1.1 from training. Duration: 0.836375 2022-12-23 09:21:18,994 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.1287, 2.4288, 1.8952, 3.1319, 2.2113, 2.5564, 2.4484, 3.5052], device='cuda:3'), covar=tensor([0.1834, 0.2961, 0.1829, 0.2831, 0.3733, 0.0965, 0.2990, 0.0675], device='cuda:3'), in_proj_covar=tensor([0.0289, 0.0282, 0.0236, 0.0347, 0.0261, 0.0221, 0.0279, 0.0202], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 09:21:23,105 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.634e+02 3.712e+02 4.661e+02 5.717e+02 1.218e+03, threshold=9.323e+02, percent-clipped=2.0 2022-12-23 09:21:38,405 INFO [train.py:894] (3/4) Epoch 16, batch 1650, loss[loss=0.2088, simple_loss=0.2843, pruned_loss=0.06662, over 18705.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2713, pruned_loss=0.04902, over 3713326.87 frames. ], batch size: 50, lr: 7.22e-03, grad_scale: 16.0 2022-12-23 09:21:58,416 WARNING [train.py:1060] (3/4) Exclude cut with ID 8565-290391-0049-67394-0_sp0.9 from training. Duration: 21.3166875 2022-12-23 09:22:27,287 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0029-104863-0_sp0.9 from training. Duration: 22.1055625 2022-12-23 09:22:38,836 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.1720, 3.5312, 3.4762, 4.1010, 3.7834, 3.6267, 4.3002, 1.4147], device='cuda:3'), covar=tensor([0.0664, 0.0721, 0.0694, 0.0712, 0.1274, 0.1163, 0.0587, 0.4470], device='cuda:3'), in_proj_covar=tensor([0.0318, 0.0211, 0.0219, 0.0244, 0.0299, 0.0254, 0.0263, 0.0266], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 09:22:40,135 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp1.1 from training. Duration: 21.77725 2022-12-23 09:22:55,260 INFO [train.py:894] (3/4) Epoch 16, batch 1700, loss[loss=0.1823, simple_loss=0.2558, pruned_loss=0.05433, over 18671.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2721, pruned_loss=0.05023, over 3713695.31 frames. ], batch size: 48, lr: 7.22e-03, grad_scale: 16.0 2022-12-23 09:22:59,842 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp0.9 from training. Duration: 27.8166875 2022-12-23 09:23:25,038 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp1.1 from training. Duration: 22.5090625 2022-12-23 09:23:25,369 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7664, 1.6805, 1.8272, 1.6848, 1.3318, 3.5067, 1.5552, 2.1007], device='cuda:3'), covar=tensor([0.2925, 0.1910, 0.1806, 0.1977, 0.1387, 0.0195, 0.1681, 0.0835], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0117, 0.0127, 0.0121, 0.0104, 0.0098, 0.0093, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 09:23:30,741 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0 from training. Duration: 25.035 2022-12-23 09:23:49,377 WARNING [train.py:1060] (3/4) Exclude cut with ID 774-127930-0014-10412-0_sp1.1 from training. Duration: 0.95 2022-12-23 09:23:55,485 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.933e+02 4.291e+02 5.331e+02 6.399e+02 1.011e+03, threshold=1.066e+03, percent-clipped=3.0 2022-12-23 09:23:57,795 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2022-12-23 09:24:10,672 INFO [train.py:894] (3/4) Epoch 16, batch 1750, loss[loss=0.1716, simple_loss=0.2433, pruned_loss=0.0499, over 18464.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.274, pruned_loss=0.05256, over 3713237.53 frames. ], batch size: 43, lr: 7.21e-03, grad_scale: 16.0 2022-12-23 09:24:10,690 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp0.9 from training. Duration: 0.92225 2022-12-23 09:24:34,853 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0 from training. Duration: 21.97 2022-12-23 09:24:55,992 WARNING [train.py:1060] (3/4) Exclude cut with ID 7492-105653-0055-62765-0_sp0.9 from training. Duration: 21.97225 2022-12-23 09:24:57,432 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp0.9 from training. Duration: 25.3333125 2022-12-23 09:24:59,394 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4938, 2.1869, 1.7569, 0.7971, 1.6249, 2.0789, 1.6819, 1.8714], device='cuda:3'), covar=tensor([0.0541, 0.0465, 0.1083, 0.1357, 0.1066, 0.1162, 0.1388, 0.0638], device='cuda:3'), in_proj_covar=tensor([0.0169, 0.0181, 0.0205, 0.0189, 0.0209, 0.0198, 0.0212, 0.0198], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 09:25:08,362 WARNING [train.py:1060] (3/4) Exclude cut with ID 5172-29468-0015-19128-0_sp0.9 from training. Duration: 21.5055625 2022-12-23 09:25:18,448 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp1.1 from training. Duration: 20.72725 2022-12-23 09:25:24,688 INFO [train.py:894] (3/4) Epoch 16, batch 1800, loss[loss=0.1781, simple_loss=0.2507, pruned_loss=0.05277, over 18480.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2747, pruned_loss=0.0544, over 3713646.13 frames. ], batch size: 43, lr: 7.21e-03, grad_scale: 16.0 2022-12-23 09:25:52,575 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp0.9 from training. Duration: 26.32775 2022-12-23 09:26:14,505 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2022-12-23 09:26:19,041 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2022-12-23 09:26:23,420 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0 from training. Duration: 20.025 2022-12-23 09:26:24,903 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.008e+02 5.203e+02 6.457e+02 7.874e+02 1.566e+03, threshold=1.291e+03, percent-clipped=4.0 2022-12-23 09:26:27,932 WARNING [train.py:1060] (3/4) Exclude cut with ID 6709-74022-0004-86860-0_sp1.1 from training. Duration: 0.9409375 2022-12-23 09:26:27,945 WARNING [train.py:1060] (3/4) Exclude cut with ID 4757-1811-0023-62229-0_sp0.9 from training. Duration: 21.37775 2022-12-23 09:26:40,518 INFO [train.py:894] (3/4) Epoch 16, batch 1850, loss[loss=0.1978, simple_loss=0.2839, pruned_loss=0.05587, over 18716.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.2756, pruned_loss=0.05583, over 3713191.57 frames. ], batch size: 69, lr: 7.21e-03, grad_scale: 8.0 2022-12-23 09:26:49,264 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0004-25974-0_sp0.9 from training. Duration: 21.17225 2022-12-23 09:26:49,277 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp0.9 from training. Duration: 27.511125 2022-12-23 09:27:20,785 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0 from training. Duration: 22.8 2022-12-23 09:27:25,305 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0 from training. Duration: 22.585 2022-12-23 09:27:55,995 INFO [train.py:894] (3/4) Epoch 16, batch 1900, loss[loss=0.2198, simple_loss=0.2914, pruned_loss=0.07411, over 18605.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2763, pruned_loss=0.05738, over 3713471.71 frames. ], batch size: 69, lr: 7.20e-03, grad_scale: 8.0 2022-12-23 09:27:56,086 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0001-146967-0_sp0.9 from training. Duration: 22.0166875 2022-12-23 09:28:14,016 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp1.1 from training. Duration: 24.395375 2022-12-23 09:28:19,776 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp0.9 from training. Duration: 27.47775 2022-12-23 09:28:24,196 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp0.9 from training. Duration: 24.8833125 2022-12-23 09:28:26,884 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0 from training. Duration: 23.39 2022-12-23 09:28:32,892 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp0.9 from training. Duration: 28.72225 2022-12-23 09:28:42,065 WARNING [train.py:1060] (3/4) Exclude cut with ID 585-294811-0110-133686-0_sp0.9 from training. Duration: 20.8944375 2022-12-23 09:28:47,517 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.2685, 1.0678, 1.4159, 2.0749, 1.5117, 2.1869, 0.9054, 1.5740], device='cuda:3'), covar=tensor([0.1771, 0.1674, 0.1236, 0.0807, 0.1185, 0.0972, 0.1778, 0.1340], device='cuda:3'), in_proj_covar=tensor([0.0102, 0.0115, 0.0131, 0.0141, 0.0105, 0.0135, 0.0130, 0.0111], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 09:28:55,340 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.0226, 4.6893, 4.4344, 2.4756, 4.8283, 3.6856, 1.2155, 3.1140], device='cuda:3'), covar=tensor([0.1936, 0.1014, 0.1262, 0.2994, 0.0748, 0.0814, 0.4604, 0.1453], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0134, 0.0155, 0.0123, 0.0135, 0.0110, 0.0145, 0.0114], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 09:28:58,200 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.390e+02 4.851e+02 5.792e+02 6.940e+02 1.812e+03, threshold=1.158e+03, percent-clipped=4.0 2022-12-23 09:28:58,295 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0_sp0.9 from training. Duration: 23.8444375 2022-12-23 09:29:11,814 INFO [train.py:894] (3/4) Epoch 16, batch 1950, loss[loss=0.1707, simple_loss=0.2464, pruned_loss=0.04745, over 18591.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2765, pruned_loss=0.05851, over 3713613.54 frames. ], batch size: 41, lr: 7.20e-03, grad_scale: 8.0 2022-12-23 09:29:22,600 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0 from training. Duration: 25.85 2022-12-23 09:29:22,612 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0 from training. Duration: 21.39 2022-12-23 09:29:34,234 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0 from training. Duration: 27.92 2022-12-23 09:30:02,598 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0039-130165-0_sp0.9 from training. Duration: 20.661125 2022-12-23 09:30:02,937 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8157, 1.2140, 0.7009, 1.3350, 2.2302, 1.1388, 1.4081, 1.7607], device='cuda:3'), covar=tensor([0.1548, 0.2146, 0.2465, 0.1480, 0.1670, 0.1714, 0.1516, 0.1559], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0098, 0.0118, 0.0096, 0.0116, 0.0091, 0.0098, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 09:30:05,654 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.10 vs. limit=5.0 2022-12-23 09:30:21,698 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2022-12-23 09:30:26,395 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0043-15874-0_sp0.9 from training. Duration: 20.07225 2022-12-23 09:30:27,846 INFO [train.py:894] (3/4) Epoch 16, batch 2000, loss[loss=0.1841, simple_loss=0.2646, pruned_loss=0.05179, over 18459.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2754, pruned_loss=0.05837, over 3714076.78 frames. ], batch size: 50, lr: 7.20e-03, grad_scale: 8.0 2022-12-23 09:30:35,032 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0 from training. Duration: 21.01 2022-12-23 09:31:30,804 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.471e+02 4.848e+02 5.721e+02 6.996e+02 1.149e+03, threshold=1.144e+03, percent-clipped=0.0 2022-12-23 09:31:42,496 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0 from training. Duration: 20.65 2022-12-23 09:31:43,883 INFO [train.py:894] (3/4) Epoch 16, batch 2050, loss[loss=0.2569, simple_loss=0.3143, pruned_loss=0.09975, over 18581.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2756, pruned_loss=0.05885, over 3713934.35 frames. ], batch size: 179, lr: 7.19e-03, grad_scale: 8.0 2022-12-23 09:31:49,251 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0 from training. Duration: 21.46 2022-12-23 09:32:18,275 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=54664.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 09:32:31,831 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0 from training. Duration: 0.92 2022-12-23 09:32:39,578 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0_sp0.9 from training. Duration: 23.7666875 2022-12-23 09:33:02,490 INFO [train.py:894] (3/4) Epoch 16, batch 2100, loss[loss=0.262, simple_loss=0.3174, pruned_loss=0.1033, over 18675.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2754, pruned_loss=0.05908, over 3713007.10 frames. ], batch size: 182, lr: 7.19e-03, grad_scale: 8.0 2022-12-23 09:33:17,044 WARNING [train.py:1060] (3/4) Exclude cut with ID 8544-281189-0060-101339-0_sp0.9 from training. Duration: 20.861125 2022-12-23 09:33:28,184 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0_sp0.9 from training. Duration: 22.711125 2022-12-23 09:33:32,784 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6854, 3.6997, 3.6548, 1.8232, 3.7772, 2.8688, 1.0632, 2.5479], device='cuda:3'), covar=tensor([0.2053, 0.1187, 0.1332, 0.3382, 0.0985, 0.0979, 0.4586, 0.1613], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0134, 0.0155, 0.0123, 0.0136, 0.0110, 0.0145, 0.0113], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 09:33:51,554 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=54725.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 09:34:04,007 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.766e+02 4.934e+02 6.087e+02 7.585e+02 2.101e+03, threshold=1.217e+03, percent-clipped=2.0 2022-12-23 09:34:09,223 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp1.1 from training. Duration: 22.986375 2022-12-23 09:34:18,059 INFO [train.py:894] (3/4) Epoch 16, batch 2150, loss[loss=0.2108, simple_loss=0.2862, pruned_loss=0.06765, over 18508.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2749, pruned_loss=0.0591, over 3714065.87 frames. ], batch size: 52, lr: 7.19e-03, grad_scale: 8.0 2022-12-23 09:34:25,421 WARNING [train.py:1060] (3/4) Exclude cut with ID 8040-260924-0003-80960-0_sp0.9 from training. Duration: 22.07225 2022-12-23 09:34:29,542 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0045-26330-0_sp0.9 from training. Duration: 20.3055625 2022-12-23 09:34:32,435 WARNING [train.py:1060] (3/4) Exclude cut with ID 6356-271890-0060-94317-0_sp0.9 from training. Duration: 20.72225 2022-12-23 09:34:51,099 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp1.1 from training. Duration: 22.4818125 2022-12-23 09:35:16,554 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp0.9 from training. Duration: 25.0944375 2022-12-23 09:35:18,649 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0 from training. Duration: 21.515 2022-12-23 09:35:26,161 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp0.9 from training. Duration: 27.02225 2022-12-23 09:35:30,733 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0010-62480-0_sp0.9 from training. Duration: 22.22225 2022-12-23 09:35:33,180 INFO [train.py:894] (3/4) Epoch 16, batch 2200, loss[loss=0.225, simple_loss=0.308, pruned_loss=0.071, over 18684.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2762, pruned_loss=0.05977, over 3714644.39 frames. ], batch size: 97, lr: 7.18e-03, grad_scale: 8.0 2022-12-23 09:35:37,647 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0085-44554-0_sp0.9 from training. Duration: 20.85 2022-12-23 09:36:03,068 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5960, 1.5655, 1.9028, 1.0637, 1.8950, 1.8737, 1.2243, 2.1857], device='cuda:3'), covar=tensor([0.1281, 0.2017, 0.1280, 0.1966, 0.0824, 0.1333, 0.2629, 0.0686], device='cuda:3'), in_proj_covar=tensor([0.0199, 0.0206, 0.0209, 0.0195, 0.0177, 0.0217, 0.0213, 0.0189], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 09:36:11,283 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0 from training. Duration: 21.54 2022-12-23 09:36:18,967 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp1.1 from training. Duration: 20.5318125 2022-12-23 09:36:29,043 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0012-134311-0_sp0.9 from training. Duration: 21.9333125 2022-12-23 09:36:38,351 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.727e+02 4.896e+02 6.033e+02 8.075e+02 1.619e+03, threshold=1.207e+03, percent-clipped=7.0 2022-12-23 09:36:51,279 INFO [train.py:894] (3/4) Epoch 16, batch 2250, loss[loss=0.1555, simple_loss=0.2325, pruned_loss=0.03924, over 18542.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.276, pruned_loss=0.05948, over 3713366.02 frames. ], batch size: 44, lr: 7.18e-03, grad_scale: 8.0 2022-12-23 09:37:20,764 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0025-130151-0_sp0.9 from training. Duration: 21.7944375 2022-12-23 09:37:21,142 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=54862.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 09:37:22,720 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0245, 1.9586, 1.4759, 2.0324, 2.1680, 1.8559, 2.6095, 2.0547], device='cuda:3'), covar=tensor([0.0814, 0.1465, 0.2599, 0.1578, 0.1563, 0.0815, 0.0880, 0.1066], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0204, 0.0245, 0.0286, 0.0230, 0.0186, 0.0208, 0.0200], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 09:37:33,937 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0_sp0.9 from training. Duration: 22.4666875 2022-12-23 09:37:41,006 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0 from training. Duration: 21.635 2022-12-23 09:37:42,767 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.3483, 3.7840, 3.7535, 4.2401, 3.9708, 3.8310, 4.4926, 1.4714], device='cuda:3'), covar=tensor([0.0693, 0.0649, 0.0659, 0.0895, 0.1304, 0.1242, 0.0581, 0.5060], device='cuda:3'), in_proj_covar=tensor([0.0328, 0.0215, 0.0229, 0.0252, 0.0309, 0.0262, 0.0272, 0.0273], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 09:37:46,730 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0_sp0.9 from training. Duration: 24.038875 2022-12-23 09:38:05,992 INFO [train.py:894] (3/4) Epoch 16, batch 2300, loss[loss=0.2084, simple_loss=0.2838, pruned_loss=0.06652, over 18512.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2758, pruned_loss=0.05938, over 3713601.42 frames. ], batch size: 52, lr: 7.18e-03, grad_scale: 8.0 2022-12-23 09:38:30,396 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp1.1 from training. Duration: 21.786375 2022-12-23 09:38:32,582 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2022-12-23 09:38:40,787 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0 from training. Duration: 20.22 2022-12-23 09:38:51,802 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=54923.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 09:39:08,574 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.948e+02 4.616e+02 5.200e+02 6.876e+02 1.499e+03, threshold=1.040e+03, percent-clipped=1.0 2022-12-23 09:39:22,502 INFO [train.py:894] (3/4) Epoch 16, batch 2350, loss[loss=0.2193, simple_loss=0.2961, pruned_loss=0.07126, over 18697.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2768, pruned_loss=0.05989, over 3714342.09 frames. ], batch size: 78, lr: 7.18e-03, grad_scale: 8.0 2022-12-23 09:39:50,929 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 2022-12-23 09:40:39,437 INFO [train.py:894] (3/4) Epoch 16, batch 2400, loss[loss=0.2245, simple_loss=0.3037, pruned_loss=0.07259, over 18395.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2765, pruned_loss=0.06029, over 3714192.98 frames. ], batch size: 53, lr: 7.17e-03, grad_scale: 8.0 2022-12-23 09:40:39,518 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0 from training. Duration: 25.285 2022-12-23 09:41:20,332 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55020.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 09:41:42,168 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.922e+02 4.598e+02 5.703e+02 7.103e+02 1.452e+03, threshold=1.141e+03, percent-clipped=8.0 2022-12-23 09:41:44,792 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2022-12-23 09:41:45,403 WARNING [train.py:1060] (3/4) Exclude cut with ID 811-130148-0001-63453-0_sp0.9 from training. Duration: 20.861125 2022-12-23 09:41:56,541 INFO [train.py:894] (3/4) Epoch 16, batch 2450, loss[loss=0.1992, simple_loss=0.2807, pruned_loss=0.05889, over 18702.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2769, pruned_loss=0.06064, over 3715445.49 frames. ], batch size: 60, lr: 7.17e-03, grad_scale: 8.0 2022-12-23 09:42:07,331 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0 from training. Duration: 20.88 2022-12-23 09:42:41,441 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0_sp0.9 from training. Duration: 23.4166875 2022-12-23 09:42:51,978 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2022-12-23 09:43:13,762 INFO [train.py:894] (3/4) Epoch 16, batch 2500, loss[loss=0.1736, simple_loss=0.2518, pruned_loss=0.04771, over 18528.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2767, pruned_loss=0.06045, over 3714730.52 frames. ], batch size: 44, lr: 7.17e-03, grad_scale: 8.0 2022-12-23 09:43:58,821 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0 from training. Duration: 21.24 2022-12-23 09:43:58,837 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0_sp0.9 from training. Duration: 23.9055625 2022-12-23 09:44:14,934 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.619e+02 4.431e+02 5.249e+02 6.433e+02 1.075e+03, threshold=1.050e+03, percent-clipped=0.0 2022-12-23 09:44:28,875 INFO [train.py:894] (3/4) Epoch 16, batch 2550, loss[loss=0.178, simple_loss=0.2581, pruned_loss=0.04896, over 18413.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2774, pruned_loss=0.06058, over 3714931.35 frames. ], batch size: 42, lr: 7.16e-03, grad_scale: 8.0 2022-12-23 09:44:31,685 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp0.9 from training. Duration: 25.988875 2022-12-23 09:44:36,165 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55148.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 09:44:40,376 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0001-134300-0_sp0.9 from training. Duration: 20.67225 2022-12-23 09:45:10,972 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4599, 1.7684, 1.2984, 2.1899, 2.4755, 1.5291, 1.4361, 1.2364], device='cuda:3'), covar=tensor([0.1942, 0.1759, 0.1693, 0.0926, 0.1181, 0.1118, 0.2047, 0.1551], device='cuda:3'), in_proj_covar=tensor([0.0244, 0.0220, 0.0210, 0.0194, 0.0258, 0.0194, 0.0218, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 09:45:32,476 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0 from training. Duration: 20.34 2022-12-23 09:45:44,370 INFO [train.py:894] (3/4) Epoch 16, batch 2600, loss[loss=0.1564, simple_loss=0.2375, pruned_loss=0.03768, over 18630.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2761, pruned_loss=0.05997, over 3715139.05 frames. ], batch size: 45, lr: 7.16e-03, grad_scale: 8.0 2022-12-23 09:46:09,186 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55209.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 09:46:12,297 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55211.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 09:46:22,208 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55218.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 09:46:42,905 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp0.9 from training. Duration: 25.061125 2022-12-23 09:46:47,524 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.767e+02 4.537e+02 5.819e+02 6.990e+02 1.532e+03, threshold=1.164e+03, percent-clipped=6.0 2022-12-23 09:46:53,373 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0 from training. Duration: 0.83 2022-12-23 09:47:01,080 INFO [train.py:894] (3/4) Epoch 16, batch 2650, loss[loss=0.1985, simple_loss=0.2869, pruned_loss=0.05505, over 18475.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2769, pruned_loss=0.06033, over 3714337.33 frames. ], batch size: 64, lr: 7.16e-03, grad_scale: 8.0 2022-12-23 09:47:17,798 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0 from training. Duration: 24.73 2022-12-23 09:47:20,605 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2022-12-23 09:47:30,114 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0 from training. Duration: 23.965 2022-12-23 09:47:39,823 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0030-146996-0_sp0.9 from training. Duration: 22.088875 2022-12-23 09:47:46,715 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55272.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 09:47:54,215 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0_sp0.9 from training. Duration: 23.6 2022-12-23 09:48:17,957 INFO [train.py:894] (3/4) Epoch 16, batch 2700, loss[loss=0.2119, simple_loss=0.2847, pruned_loss=0.06952, over 18582.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2763, pruned_loss=0.05954, over 3714337.89 frames. ], batch size: 181, lr: 7.15e-03, grad_scale: 8.0 2022-12-23 09:48:59,222 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2561, 1.9133, 1.4835, 0.5449, 1.4514, 1.9111, 1.5966, 1.7303], device='cuda:3'), covar=tensor([0.0585, 0.0506, 0.1074, 0.1484, 0.1027, 0.1493, 0.1591, 0.0680], device='cuda:3'), in_proj_covar=tensor([0.0170, 0.0184, 0.0206, 0.0190, 0.0209, 0.0199, 0.0213, 0.0201], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 09:49:00,647 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55320.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 09:49:20,788 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.908e+02 4.633e+02 5.450e+02 6.979e+02 1.469e+03, threshold=1.090e+03, percent-clipped=1.0 2022-12-23 09:49:27,201 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5047, 2.1780, 1.6912, 2.3370, 1.8882, 1.9806, 1.9324, 2.4020], device='cuda:3'), covar=tensor([0.1898, 0.2764, 0.1769, 0.2652, 0.3369, 0.1048, 0.2796, 0.0837], device='cuda:3'), in_proj_covar=tensor([0.0292, 0.0283, 0.0238, 0.0350, 0.0262, 0.0221, 0.0279, 0.0204], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 09:49:33,919 INFO [train.py:894] (3/4) Epoch 16, batch 2750, loss[loss=0.2138, simple_loss=0.3012, pruned_loss=0.06318, over 18599.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2764, pruned_loss=0.05935, over 3714224.18 frames. ], batch size: 69, lr: 7.15e-03, grad_scale: 8.0 2022-12-23 09:49:36,940 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0 from training. Duration: 23.795 2022-12-23 09:49:53,617 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp1.1 from training. Duration: 21.5409375 2022-12-23 09:49:57,131 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp0.9 from training. Duration: 24.97775 2022-12-23 09:50:08,131 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0_sp0.9 from training. Duration: 23.3444375 2022-12-23 09:50:14,026 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=55368.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 09:50:33,603 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0_sp0.9 from training. Duration: 23.2 2022-12-23 09:50:41,403 WARNING [train.py:1060] (3/4) Exclude cut with ID 5653-46179-0060-117930-0_sp0.9 from training. Duration: 21.17225 2022-12-23 09:50:51,459 INFO [train.py:894] (3/4) Epoch 16, batch 2800, loss[loss=0.2115, simple_loss=0.2913, pruned_loss=0.06591, over 18537.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2765, pruned_loss=0.05913, over 3713451.26 frames. ], batch size: 55, lr: 7.15e-03, grad_scale: 8.0 2022-12-23 09:50:59,456 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp0.9 from training. Duration: 24.6555625 2022-12-23 09:51:27,275 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1437, 2.2056, 1.7177, 2.5406, 2.4075, 2.0233, 3.0417, 2.1521], device='cuda:3'), covar=tensor([0.0847, 0.1689, 0.2609, 0.1684, 0.1589, 0.0881, 0.0834, 0.1166], device='cuda:3'), in_proj_covar=tensor([0.0177, 0.0203, 0.0246, 0.0286, 0.0231, 0.0186, 0.0209, 0.0201], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 09:51:43,774 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7833, 2.2902, 1.7751, 2.5102, 2.0488, 2.1709, 2.1369, 2.7319], device='cuda:3'), covar=tensor([0.1818, 0.3088, 0.1786, 0.3126, 0.3381, 0.1006, 0.2891, 0.0823], device='cuda:3'), in_proj_covar=tensor([0.0292, 0.0283, 0.0237, 0.0350, 0.0261, 0.0222, 0.0279, 0.0204], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 09:51:51,005 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8295, 1.7323, 2.1476, 1.1830, 2.0976, 2.0132, 1.4902, 2.3196], device='cuda:3'), covar=tensor([0.0980, 0.1655, 0.1080, 0.1702, 0.0588, 0.1104, 0.2160, 0.0528], device='cuda:3'), in_proj_covar=tensor([0.0198, 0.0207, 0.0210, 0.0197, 0.0178, 0.0217, 0.0213, 0.0189], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 09:51:55,064 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.964e+02 4.626e+02 5.479e+02 6.826e+02 2.106e+03, threshold=1.096e+03, percent-clipped=3.0 2022-12-23 09:51:58,285 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0 from training. Duration: 20.44 2022-12-23 09:52:08,916 INFO [train.py:894] (3/4) Epoch 16, batch 2850, loss[loss=0.1797, simple_loss=0.2549, pruned_loss=0.05222, over 18664.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2754, pruned_loss=0.05865, over 3714080.91 frames. ], batch size: 46, lr: 7.14e-03, grad_scale: 8.0 2022-12-23 09:52:13,492 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0_sp0.9 from training. Duration: 23.45 2022-12-23 09:52:22,884 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5564, 2.2297, 1.7191, 2.3270, 1.9395, 2.0817, 2.0039, 2.5235], device='cuda:3'), covar=tensor([0.1908, 0.2797, 0.1831, 0.2555, 0.3305, 0.1025, 0.2774, 0.0853], device='cuda:3'), in_proj_covar=tensor([0.0292, 0.0283, 0.0237, 0.0349, 0.0261, 0.0221, 0.0279, 0.0204], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 09:52:41,947 WARNING [train.py:1060] (3/4) Exclude cut with ID 6945-60535-0076-12784-0_sp0.9 from training. Duration: 20.52225 2022-12-23 09:52:49,429 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0 from training. Duration: 22.19 2022-12-23 09:52:58,357 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp1.1 from training. Duration: 25.3818125 2022-12-23 09:53:15,993 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp0.9 from training. Duration: 28.0944375 2022-12-23 09:53:24,585 INFO [train.py:894] (3/4) Epoch 16, batch 2900, loss[loss=0.1871, simple_loss=0.2631, pruned_loss=0.05552, over 18597.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2751, pruned_loss=0.05871, over 3712971.94 frames. ], batch size: 51, lr: 7.14e-03, grad_scale: 8.0 2022-12-23 09:53:24,587 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0_sp0.9 from training. Duration: 22.9444375 2022-12-23 09:53:31,894 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp1.1 from training. Duration: 21.6318125 2022-12-23 09:53:41,826 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55504.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 09:53:47,534 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0 from training. Duration: 23.695 2022-12-23 09:54:02,931 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55518.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 09:54:10,135 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3070, 2.0170, 2.0011, 2.2436, 2.0791, 2.0286, 2.3601, 0.9262], device='cuda:3'), covar=tensor([0.0818, 0.0681, 0.0733, 0.0877, 0.1285, 0.1159, 0.1017, 0.3383], device='cuda:3'), in_proj_covar=tensor([0.0327, 0.0216, 0.0227, 0.0253, 0.0309, 0.0262, 0.0275, 0.0274], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 09:54:14,183 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0 from training. Duration: 23.955 2022-12-23 09:54:26,972 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.388e+02 4.402e+02 5.592e+02 6.941e+02 1.570e+03, threshold=1.118e+03, percent-clipped=4.0 2022-12-23 09:54:40,716 INFO [train.py:894] (3/4) Epoch 16, batch 2950, loss[loss=0.2188, simple_loss=0.2969, pruned_loss=0.07034, over 18471.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2759, pruned_loss=0.05917, over 3713398.73 frames. ], batch size: 64, lr: 7.14e-03, grad_scale: 8.0 2022-12-23 09:54:46,617 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp0.9 from training. Duration: 26.438875 2022-12-23 09:55:15,024 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=55566.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 09:55:16,537 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55567.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 09:55:32,433 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0021-26306-0_sp0.9 from training. Duration: 21.2444375 2022-12-23 09:55:33,563 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp0.9 from training. Duration: 31.02225 2022-12-23 09:55:44,225 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0 from training. Duration: 22.395 2022-12-23 09:55:55,249 INFO [train.py:894] (3/4) Epoch 16, batch 3000, loss[loss=0.1912, simple_loss=0.2734, pruned_loss=0.05446, over 18512.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2755, pruned_loss=0.05887, over 3712654.91 frames. ], batch size: 52, lr: 7.13e-03, grad_scale: 8.0 2022-12-23 09:55:55,250 INFO [train.py:919] (3/4) Computing validation loss 2022-12-23 09:56:06,243 INFO [train.py:928] (3/4) Epoch 16, validation: loss=0.1656, simple_loss=0.2645, pruned_loss=0.03333, over 944034.00 frames. 2022-12-23 09:56:06,244 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-23 09:56:11,959 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0 from training. Duration: 21.075 2022-12-23 09:56:16,305 WARNING [train.py:1060] (3/4) Exclude cut with ID 6482-98857-0025-147532-0_sp0.9 from training. Duration: 20.0055625 2022-12-23 09:56:16,314 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0037-132304-0_sp0.9 from training. Duration: 22.05 2022-12-23 09:56:16,324 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0 from training. Duration: 26.8349375 2022-12-23 09:56:19,695 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp1.1 from training. Duration: 22.1090625 2022-12-23 09:56:27,265 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp0.9 from training. Duration: 26.6166875 2022-12-23 09:56:27,651 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.0490, 2.9145, 1.9787, 1.5792, 3.6634, 3.5125, 2.9740, 2.4379], device='cuda:3'), covar=tensor([0.0402, 0.0369, 0.0625, 0.0748, 0.0172, 0.0335, 0.0445, 0.0753], device='cuda:3'), in_proj_covar=tensor([0.0121, 0.0126, 0.0127, 0.0120, 0.0094, 0.0120, 0.0134, 0.0154], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 09:56:47,070 WARNING [train.py:1060] (3/4) Exclude cut with ID 2046-178027-0000-53705-0_sp0.9 from training. Duration: 20.3055625 2022-12-23 09:57:07,808 WARNING [train.py:1060] (3/4) Exclude cut with ID 7205-50138-0008-5373-0_sp0.9 from training. Duration: 20.7 2022-12-23 09:57:09,330 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.134e+02 4.889e+02 6.104e+02 7.743e+02 1.774e+03, threshold=1.221e+03, percent-clipped=6.0 2022-12-23 09:57:21,156 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2022-12-23 09:57:23,357 INFO [train.py:894] (3/4) Epoch 16, batch 3050, loss[loss=0.1879, simple_loss=0.2683, pruned_loss=0.05372, over 18723.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2757, pruned_loss=0.05926, over 3713346.77 frames. ], batch size: 54, lr: 7.13e-03, grad_scale: 8.0 2022-12-23 09:57:54,106 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0 from training. Duration: 22.48 2022-12-23 09:58:11,528 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp0.9 from training. Duration: 29.816625 2022-12-23 09:58:29,023 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.8749, 3.3944, 3.3045, 3.7597, 3.5004, 3.3525, 4.0001, 1.1275], device='cuda:3'), covar=tensor([0.0743, 0.0606, 0.0694, 0.0838, 0.1360, 0.1150, 0.0573, 0.4808], device='cuda:3'), in_proj_covar=tensor([0.0329, 0.0216, 0.0228, 0.0254, 0.0312, 0.0262, 0.0277, 0.0276], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 09:58:31,662 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp1.1 from training. Duration: 22.7590625 2022-12-23 09:58:36,155 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0 from training. Duration: 22.555 2022-12-23 09:58:40,481 INFO [train.py:894] (3/4) Epoch 16, batch 3100, loss[loss=0.1908, simple_loss=0.2717, pruned_loss=0.05492, over 18665.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.2747, pruned_loss=0.05839, over 3713235.43 frames. ], batch size: 60, lr: 7.13e-03, grad_scale: 8.0 2022-12-23 09:58:56,858 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0005-25975-0_sp0.9 from training. Duration: 21.688875 2022-12-23 09:59:33,192 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0_sp0.9 from training. Duration: 22.6 2022-12-23 09:59:43,767 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.787e+02 4.394e+02 5.359e+02 6.672e+02 1.594e+03, threshold=1.072e+03, percent-clipped=1.0 2022-12-23 09:59:56,469 INFO [train.py:894] (3/4) Epoch 16, batch 3150, loss[loss=0.1943, simple_loss=0.2715, pruned_loss=0.05854, over 18586.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2756, pruned_loss=0.0589, over 3713267.94 frames. ], batch size: 51, lr: 7.12e-03, grad_scale: 8.0 2022-12-23 10:00:04,188 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55748.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:00:10,588 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0 from training. Duration: 24.32 2022-12-23 10:00:32,143 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2022-12-23 10:01:08,981 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-276745-0093-13116-0_sp0.9 from training. Duration: 21.061125 2022-12-23 10:01:10,547 INFO [train.py:894] (3/4) Epoch 16, batch 3200, loss[loss=0.2175, simple_loss=0.3023, pruned_loss=0.06637, over 18726.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2758, pruned_loss=0.05887, over 3712878.01 frames. ], batch size: 54, lr: 7.12e-03, grad_scale: 8.0 2022-12-23 10:01:20,458 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0024-15855-0_sp0.9 from training. Duration: 20.32225 2022-12-23 10:01:27,871 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55804.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:01:34,050 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp1.1 from training. Duration: 0.7545625 2022-12-23 10:01:36,264 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55809.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:01:50,433 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0_sp0.9 from training. Duration: 23.9333125 2022-12-23 10:02:13,057 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.270e+02 4.851e+02 5.650e+02 7.307e+02 2.258e+03, threshold=1.130e+03, percent-clipped=5.0 2022-12-23 10:02:20,737 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp1.1 from training. Duration: 20.17275 2022-12-23 10:02:26,733 INFO [train.py:894] (3/4) Epoch 16, batch 3250, loss[loss=0.1793, simple_loss=0.2491, pruned_loss=0.05471, over 18701.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2761, pruned_loss=0.0592, over 3711919.77 frames. ], batch size: 46, lr: 7.12e-03, grad_scale: 8.0 2022-12-23 10:02:26,786 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp1.1 from training. Duration: 20.436375 2022-12-23 10:02:35,375 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55848.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:02:41,113 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=55852.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:03:05,086 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55867.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:03:39,172 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1681, 2.0429, 1.7025, 1.1894, 2.4662, 2.2467, 2.0416, 1.5977], device='cuda:3'), covar=tensor([0.0369, 0.0409, 0.0515, 0.0739, 0.0294, 0.0377, 0.0453, 0.0910], device='cuda:3'), in_proj_covar=tensor([0.0122, 0.0125, 0.0127, 0.0120, 0.0095, 0.0121, 0.0135, 0.0154], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 10:03:43,297 INFO [train.py:894] (3/4) Epoch 16, batch 3300, loss[loss=0.233, simple_loss=0.3007, pruned_loss=0.08267, over 18662.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2757, pruned_loss=0.05889, over 3712826.32 frames. ], batch size: 177, lr: 7.11e-03, grad_scale: 8.0 2022-12-23 10:03:48,689 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0_sp0.9 from training. Duration: 23.1055625 2022-12-23 10:03:52,312 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp1.1 from training. Duration: 23.5 2022-12-23 10:03:56,214 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4541, 1.4626, 1.1321, 1.7113, 1.5784, 2.9832, 1.4196, 1.4432], device='cuda:3'), covar=tensor([0.0849, 0.1707, 0.1105, 0.0839, 0.1388, 0.0279, 0.1310, 0.1488], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0083, 0.0073, 0.0074, 0.0091, 0.0074, 0.0085, 0.0077], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 10:03:58,157 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2022-12-23 10:04:00,761 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4489, 2.2157, 1.7654, 1.7218, 2.2711, 3.0196, 2.5670, 2.0217], device='cuda:3'), covar=tensor([0.0306, 0.0313, 0.0447, 0.0265, 0.0254, 0.0261, 0.0376, 0.0311], device='cuda:3'), in_proj_covar=tensor([0.0087, 0.0119, 0.0144, 0.0120, 0.0112, 0.0111, 0.0093, 0.0121], device='cuda:3'), out_proj_covar=tensor([7.0643e-05, 9.5313e-05, 1.2132e-04, 9.7119e-05, 9.1628e-05, 8.5794e-05, 7.3702e-05, 9.5900e-05], device='cuda:3') 2022-12-23 10:04:03,441 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp0.9 from training. Duration: 26.62775 2022-12-23 10:04:09,976 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55909.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:04:16,113 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0 from training. Duration: 21.105 2022-12-23 10:04:19,069 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=55915.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:04:20,491 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0_sp0.9 from training. Duration: 24.411125 2022-12-23 10:04:47,337 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.602e+02 4.449e+02 5.484e+02 6.704e+02 1.621e+03, threshold=1.097e+03, percent-clipped=1.0 2022-12-23 10:04:47,380 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp1.1 from training. Duration: 21.263625 2022-12-23 10:05:02,679 INFO [train.py:894] (3/4) Epoch 16, batch 3350, loss[loss=0.1715, simple_loss=0.2524, pruned_loss=0.04524, over 18383.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2752, pruned_loss=0.05851, over 3713053.34 frames. ], batch size: 46, lr: 7.11e-03, grad_scale: 8.0 2022-12-23 10:05:20,822 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0 from training. Duration: 20.795 2022-12-23 10:05:31,792 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0 from training. Duration: 24.76 2022-12-23 10:05:33,252 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0_sp0.9 from training. Duration: 22.25 2022-12-23 10:05:56,777 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp1.1 from training. Duration: 20.5045625 2022-12-23 10:06:03,608 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2022-12-23 10:06:19,228 INFO [train.py:894] (3/4) Epoch 16, batch 3400, loss[loss=0.1753, simple_loss=0.2592, pruned_loss=0.04573, over 18598.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2752, pruned_loss=0.05817, over 3712940.91 frames. ], batch size: 51, lr: 7.11e-03, grad_scale: 8.0 2022-12-23 10:07:01,557 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2022-12-23 10:07:06,254 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4353, 2.4352, 2.7730, 1.6111, 2.9402, 2.9679, 2.0780, 3.2730], device='cuda:3'), covar=tensor([0.1173, 0.1737, 0.1507, 0.2147, 0.0744, 0.1139, 0.2006, 0.0575], device='cuda:3'), in_proj_covar=tensor([0.0196, 0.0210, 0.0208, 0.0196, 0.0176, 0.0217, 0.0215, 0.0188], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 10:07:07,765 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8159, 1.6241, 1.4956, 1.4316, 1.9120, 1.9234, 2.0475, 1.3563], device='cuda:3'), covar=tensor([0.0282, 0.0255, 0.0422, 0.0224, 0.0180, 0.0320, 0.0221, 0.0286], device='cuda:3'), in_proj_covar=tensor([0.0088, 0.0119, 0.0144, 0.0121, 0.0112, 0.0111, 0.0092, 0.0122], device='cuda:3'), out_proj_covar=tensor([7.0881e-05, 9.5722e-05, 1.2103e-04, 9.7690e-05, 9.1899e-05, 8.6242e-05, 7.2811e-05, 9.7060e-05], device='cuda:3') 2022-12-23 10:07:23,261 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.742e+02 4.730e+02 5.600e+02 6.470e+02 1.614e+03, threshold=1.120e+03, percent-clipped=3.0 2022-12-23 10:07:35,978 INFO [train.py:894] (3/4) Epoch 16, batch 3450, loss[loss=0.2028, simple_loss=0.2871, pruned_loss=0.05922, over 18699.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2749, pruned_loss=0.05807, over 3713768.32 frames. ], batch size: 78, lr: 7.11e-03, grad_scale: 8.0 2022-12-23 10:08:04,448 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7435, 1.6969, 1.6979, 1.7339, 1.4899, 3.6322, 1.5927, 2.0390], device='cuda:3'), covar=tensor([0.3059, 0.1893, 0.1821, 0.1885, 0.1191, 0.0179, 0.1523, 0.0808], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0117, 0.0126, 0.0120, 0.0104, 0.0098, 0.0094, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 10:08:20,480 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7389, 1.7220, 1.6986, 1.7510, 1.2637, 3.9397, 1.6495, 2.1115], device='cuda:3'), covar=tensor([0.3192, 0.2080, 0.1972, 0.1981, 0.1414, 0.0183, 0.1538, 0.0884], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0117, 0.0126, 0.0120, 0.0104, 0.0098, 0.0094, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 10:08:48,639 INFO [train.py:894] (3/4) Epoch 16, batch 3500, loss[loss=0.228, simple_loss=0.2989, pruned_loss=0.07853, over 18584.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2745, pruned_loss=0.05838, over 3713923.51 frames. ], batch size: 178, lr: 7.10e-03, grad_scale: 8.0 2022-12-23 10:09:10,047 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp1.1 from training. Duration: 22.2954375 2022-12-23 10:09:18,376 INFO [train.py:894] (3/4) Epoch 17, batch 0, loss[loss=0.206, simple_loss=0.2944, pruned_loss=0.05878, over 18588.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2944, pruned_loss=0.05878, over 18588.00 frames. ], batch size: 56, lr: 6.89e-03, grad_scale: 8.0 2022-12-23 10:09:18,376 INFO [train.py:919] (3/4) Computing validation loss 2022-12-23 10:09:30,656 INFO [train.py:928] (3/4) Epoch 17, validation: loss=0.1676, simple_loss=0.2661, pruned_loss=0.03449, over 944034.00 frames. 2022-12-23 10:09:30,658 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-23 10:09:38,135 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56104.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:10:18,903 WARNING [train.py:1060] (3/4) Exclude cut with ID 298-126791-0067-24026-0_sp0.9 from training. Duration: 21.438875 2022-12-23 10:10:23,560 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.084e+02 4.090e+02 5.146e+02 6.475e+02 1.382e+03, threshold=1.029e+03, percent-clipped=5.0 2022-12-23 10:10:23,596 WARNING [train.py:1060] (3/4) Exclude cut with ID 5652-39938-0025-23684-0_sp0.9 from training. Duration: 22.2055625 2022-12-23 10:10:39,992 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56145.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:10:45,388 INFO [train.py:894] (3/4) Epoch 17, batch 50, loss[loss=0.172, simple_loss=0.2598, pruned_loss=0.04211, over 18707.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.2793, pruned_loss=0.05277, over 838328.00 frames. ], batch size: 50, lr: 6.88e-03, grad_scale: 8.0 2022-12-23 10:11:15,530 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9096, 1.1723, 1.6885, 1.5681, 1.9030, 1.8965, 1.6701, 1.5779], device='cuda:3'), covar=tensor([0.1909, 0.2835, 0.2234, 0.2429, 0.1835, 0.0839, 0.2722, 0.1157], device='cuda:3'), in_proj_covar=tensor([0.0263, 0.0295, 0.0274, 0.0308, 0.0297, 0.0246, 0.0330, 0.0235], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 10:11:45,254 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.3710, 3.6258, 2.3567, 1.7277, 3.8210, 4.0195, 3.2021, 2.9881], device='cuda:3'), covar=tensor([0.0337, 0.0267, 0.0534, 0.0722, 0.0157, 0.0274, 0.0447, 0.0593], device='cuda:3'), in_proj_covar=tensor([0.0122, 0.0126, 0.0128, 0.0121, 0.0095, 0.0121, 0.0136, 0.0155], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 10:11:57,888 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.3514, 2.4573, 1.9456, 3.1499, 2.2431, 2.3353, 2.2065, 3.4389], device='cuda:3'), covar=tensor([0.1775, 0.3207, 0.1803, 0.2805, 0.3823, 0.0982, 0.3260, 0.0687], device='cuda:3'), in_proj_covar=tensor([0.0292, 0.0282, 0.0237, 0.0349, 0.0263, 0.0223, 0.0278, 0.0206], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 10:12:00,276 INFO [train.py:894] (3/4) Epoch 17, batch 100, loss[loss=0.1685, simple_loss=0.2535, pruned_loss=0.04174, over 18701.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2726, pruned_loss=0.05008, over 1476233.27 frames. ], batch size: 46, lr: 6.88e-03, grad_scale: 8.0 2022-12-23 10:12:08,281 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56204.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:12:11,758 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56206.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:12:53,867 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.099e+02 3.400e+02 4.151e+02 4.945e+02 1.658e+03, threshold=8.301e+02, percent-clipped=1.0 2022-12-23 10:13:16,888 INFO [train.py:894] (3/4) Epoch 17, batch 150, loss[loss=0.1487, simple_loss=0.2339, pruned_loss=0.0318, over 18528.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2691, pruned_loss=0.04827, over 1972121.25 frames. ], batch size: 44, lr: 6.88e-03, grad_scale: 8.0 2022-12-23 10:13:21,324 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0 from training. Duration: 24.525 2022-12-23 10:13:43,045 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.51 vs. limit=5.0 2022-12-23 10:13:55,025 WARNING [train.py:1060] (3/4) Exclude cut with ID 3699-47246-0007-3408-0_sp0.9 from training. Duration: 20.26675 2022-12-23 10:14:08,237 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp0.9 from training. Duration: 27.25 2022-12-23 10:14:32,303 INFO [train.py:894] (3/4) Epoch 17, batch 200, loss[loss=0.2106, simple_loss=0.2904, pruned_loss=0.06537, over 18621.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2677, pruned_loss=0.04759, over 2358442.77 frames. ], batch size: 173, lr: 6.87e-03, grad_scale: 8.0 2022-12-23 10:15:05,702 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2022-12-23 10:15:24,836 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.339e+02 3.583e+02 4.534e+02 5.413e+02 1.730e+03, threshold=9.068e+02, percent-clipped=8.0 2022-12-23 10:15:26,199 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0 from training. Duration: 21.68 2022-12-23 10:15:39,526 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0 from training. Duration: 21.6300625 2022-12-23 10:15:46,998 INFO [train.py:894] (3/4) Epoch 17, batch 250, loss[loss=0.1539, simple_loss=0.2331, pruned_loss=0.03739, over 18463.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2666, pruned_loss=0.04707, over 2658225.15 frames. ], batch size: 43, lr: 6.87e-03, grad_scale: 8.0 2022-12-23 10:16:01,803 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0_sp0.9 from training. Duration: 24.033375 2022-12-23 10:16:54,699 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0 from training. Duration: 22.905 2022-12-23 10:16:56,124 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp1.1 from training. Duration: 23.4318125 2022-12-23 10:16:56,516 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3707, 1.4383, 1.2521, 1.7604, 1.7898, 1.4657, 1.0756, 1.2216], device='cuda:3'), covar=tensor([0.1959, 0.1886, 0.1745, 0.1097, 0.1145, 0.1171, 0.2173, 0.1575], device='cuda:3'), in_proj_covar=tensor([0.0242, 0.0220, 0.0211, 0.0194, 0.0256, 0.0193, 0.0218, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 10:17:02,088 INFO [train.py:894] (3/4) Epoch 17, batch 300, loss[loss=0.1772, simple_loss=0.2558, pruned_loss=0.04927, over 18524.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2672, pruned_loss=0.04709, over 2893275.84 frames. ], batch size: 47, lr: 6.87e-03, grad_scale: 8.0 2022-12-23 10:17:09,886 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56404.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:17:45,133 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0780, 1.3937, 2.0649, 2.5071, 1.9497, 4.6355, 1.4282, 1.6580], device='cuda:3'), covar=tensor([0.0735, 0.1752, 0.0872, 0.0833, 0.1351, 0.0151, 0.1401, 0.1540], device='cuda:3'), in_proj_covar=tensor([0.0072, 0.0081, 0.0073, 0.0073, 0.0090, 0.0074, 0.0084, 0.0077], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 10:17:53,854 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.464e+02 3.697e+02 4.364e+02 5.395e+02 9.387e+02, threshold=8.728e+02, percent-clipped=1.0 2022-12-23 10:18:16,092 INFO [train.py:894] (3/4) Epoch 17, batch 350, loss[loss=0.1606, simple_loss=0.2387, pruned_loss=0.04121, over 18551.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2676, pruned_loss=0.0474, over 3074821.14 frames. ], batch size: 41, lr: 6.87e-03, grad_scale: 16.0 2022-12-23 10:18:20,392 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=56452.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:18:39,006 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6317, 3.9360, 3.8544, 1.5345, 4.0853, 3.1115, 0.7086, 2.4215], device='cuda:3'), covar=tensor([0.2116, 0.0882, 0.1273, 0.3740, 0.0886, 0.0937, 0.5149, 0.1706], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0132, 0.0153, 0.0122, 0.0136, 0.0110, 0.0145, 0.0112], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 10:18:51,228 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.8194, 2.5586, 2.0820, 1.0200, 2.0596, 2.1069, 1.9702, 2.0925], device='cuda:3'), covar=tensor([0.0591, 0.0471, 0.1210, 0.1606, 0.1264, 0.1422, 0.1343, 0.0929], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0181, 0.0204, 0.0190, 0.0208, 0.0197, 0.0212, 0.0199], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 10:18:56,112 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp1.1 from training. Duration: 20.82275 2022-12-23 10:18:57,636 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp0.9 from training. Duration: 25.45 2022-12-23 10:19:30,097 INFO [train.py:894] (3/4) Epoch 17, batch 400, loss[loss=0.1875, simple_loss=0.2793, pruned_loss=0.0478, over 18510.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2688, pruned_loss=0.04804, over 3216474.30 frames. ], batch size: 52, lr: 6.86e-03, grad_scale: 16.0 2022-12-23 10:19:33,350 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56501.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:19:37,979 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56504.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:19:42,835 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8905, 1.8686, 1.2161, 1.9976, 2.0319, 1.6958, 2.7846, 1.9895], device='cuda:3'), covar=tensor([0.1007, 0.1667, 0.2985, 0.1731, 0.1816, 0.1082, 0.0805, 0.1372], device='cuda:3'), in_proj_covar=tensor([0.0177, 0.0204, 0.0246, 0.0286, 0.0231, 0.0186, 0.0207, 0.0202], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 10:19:54,100 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0 from training. Duration: 25.775 2022-12-23 10:20:07,478 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56523.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 10:20:15,566 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0_sp0.9 from training. Duration: 22.25 2022-12-23 10:20:23,541 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.488e+02 3.929e+02 4.700e+02 5.683e+02 9.730e+02, threshold=9.400e+02, percent-clipped=2.0 2022-12-23 10:20:43,055 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0 from training. Duration: 26.205 2022-12-23 10:20:45,861 INFO [train.py:894] (3/4) Epoch 17, batch 450, loss[loss=0.1955, simple_loss=0.2757, pruned_loss=0.05767, over 18451.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2703, pruned_loss=0.04867, over 3327313.13 frames. ], batch size: 50, lr: 6.86e-03, grad_scale: 16.0 2022-12-23 10:20:50,341 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=56552.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:20:59,319 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp0.9 from training. Duration: 30.1555625 2022-12-23 10:21:06,035 WARNING [train.py:1060] (3/4) Exclude cut with ID 1265-135635-0050-6781-0_sp0.9 from training. Duration: 21.8333125 2022-12-23 10:21:17,046 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp1.1 from training. Duration: 20.6545625 2022-12-23 10:21:40,100 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.2464, 1.0641, 1.3270, 2.1382, 1.6255, 2.1850, 0.6673, 1.5838], device='cuda:3'), covar=tensor([0.2037, 0.1773, 0.1398, 0.0774, 0.1156, 0.0938, 0.2027, 0.1302], device='cuda:3'), in_proj_covar=tensor([0.0101, 0.0114, 0.0130, 0.0140, 0.0104, 0.0136, 0.0127, 0.0110], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 10:21:40,259 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56584.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 10:21:57,683 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0045-39920-0_sp0.9 from training. Duration: 20.52225 2022-12-23 10:22:02,112 INFO [train.py:894] (3/4) Epoch 17, batch 500, loss[loss=0.1772, simple_loss=0.2716, pruned_loss=0.04146, over 18595.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2704, pruned_loss=0.04867, over 3411930.88 frames. ], batch size: 51, lr: 6.86e-03, grad_scale: 16.0 2022-12-23 10:22:19,252 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp0.9 from training. Duration: 29.1166875 2022-12-23 10:22:56,244 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.340e+02 3.538e+02 4.282e+02 5.017e+02 8.861e+02, threshold=8.564e+02, percent-clipped=0.0 2022-12-23 10:23:19,230 INFO [train.py:894] (3/4) Epoch 17, batch 550, loss[loss=0.1951, simple_loss=0.2854, pruned_loss=0.05239, over 18702.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2712, pruned_loss=0.04879, over 3479061.05 frames. ], batch size: 69, lr: 6.85e-03, grad_scale: 16.0 2022-12-23 10:23:20,949 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133211-0007-59831-0_sp0.9 from training. Duration: 21.388875 2022-12-23 10:23:56,095 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0 from training. Duration: 22.72 2022-12-23 10:23:57,484 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0_sp0.9 from training. Duration: 22.7444375 2022-12-23 10:24:34,986 INFO [train.py:894] (3/4) Epoch 17, batch 600, loss[loss=0.1668, simple_loss=0.2446, pruned_loss=0.04449, over 18548.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2712, pruned_loss=0.04877, over 3530635.85 frames. ], batch size: 44, lr: 6.85e-03, grad_scale: 16.0 2022-12-23 10:24:40,755 WARNING [train.py:1060] (3/4) Exclude cut with ID 4133-6541-0027-40495-0_sp1.1 from training. Duration: 0.9681875 2022-12-23 10:24:44,580 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0_sp0.9 from training. Duration: 22.3166875 2022-12-23 10:24:49,722 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133212-0015-59917-0_sp0.9 from training. Duration: 21.8166875 2022-12-23 10:25:29,262 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.559e+02 3.399e+02 4.174e+02 5.258e+02 9.875e+02, threshold=8.347e+02, percent-clipped=3.0 2022-12-23 10:25:51,054 INFO [train.py:894] (3/4) Epoch 17, batch 650, loss[loss=0.1585, simple_loss=0.2503, pruned_loss=0.0333, over 18665.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2714, pruned_loss=0.04867, over 3571499.95 frames. ], batch size: 48, lr: 6.85e-03, grad_scale: 16.0 2022-12-23 10:26:17,441 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2022-12-23 10:26:32,014 WARNING [train.py:1060] (3/4) Exclude cut with ID 4957-30119-0041-23990-0_sp0.9 from training. Duration: 20.22775 2022-12-23 10:27:07,833 INFO [train.py:894] (3/4) Epoch 17, batch 700, loss[loss=0.1535, simple_loss=0.2304, pruned_loss=0.03831, over 18426.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2707, pruned_loss=0.04825, over 3602723.93 frames. ], batch size: 42, lr: 6.84e-03, grad_scale: 16.0 2022-12-23 10:27:11,882 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56801.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:27:14,858 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56803.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:27:19,312 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp1.1 from training. Duration: 24.67275 2022-12-23 10:27:47,004 WARNING [train.py:1060] (3/4) Exclude cut with ID 3082-165428-0081-50734-0_sp0.9 from training. Duration: 21.8055625 2022-12-23 10:28:01,764 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.567e+02 3.551e+02 4.773e+02 5.686e+02 1.008e+03, threshold=9.547e+02, percent-clipped=4.0 2022-12-23 10:28:24,262 INFO [train.py:894] (3/4) Epoch 17, batch 750, loss[loss=0.1788, simple_loss=0.2782, pruned_loss=0.03974, over 18465.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2706, pruned_loss=0.04807, over 3626570.79 frames. ], batch size: 64, lr: 6.84e-03, grad_scale: 16.0 2022-12-23 10:28:24,293 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0_sp0.9 from training. Duration: 22.6666875 2022-12-23 10:28:24,995 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=56849.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:28:48,075 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56864.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:29:10,376 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56879.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 10:29:28,201 WARNING [train.py:1060] (3/4) Exclude cut with ID 2411-132532-0017-82279-0_sp1.1 from training. Duration: 0.9681875 2022-12-23 10:29:40,593 INFO [train.py:894] (3/4) Epoch 17, batch 800, loss[loss=0.1581, simple_loss=0.2423, pruned_loss=0.03688, over 18557.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2698, pruned_loss=0.04766, over 3645939.99 frames. ], batch size: 49, lr: 6.84e-03, grad_scale: 16.0 2022-12-23 10:29:44,465 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4930, 1.5555, 1.9235, 1.0870, 1.8119, 1.8833, 1.3178, 2.1814], device='cuda:3'), covar=tensor([0.1062, 0.1804, 0.1111, 0.1731, 0.0627, 0.1002, 0.2287, 0.0442], device='cuda:3'), in_proj_covar=tensor([0.0194, 0.0209, 0.0205, 0.0194, 0.0175, 0.0212, 0.0211, 0.0187], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 10:29:54,906 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0 from training. Duration: 22.485 2022-12-23 10:29:55,284 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56908.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:30:11,038 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5662, 3.7182, 3.5139, 1.1918, 3.7499, 2.7839, 0.6877, 2.2550], device='cuda:3'), covar=tensor([0.2081, 0.0977, 0.1448, 0.4132, 0.0876, 0.1026, 0.5032, 0.1801], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0130, 0.0151, 0.0120, 0.0133, 0.0108, 0.0142, 0.0111], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 10:30:33,215 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.478e+02 3.299e+02 4.044e+02 4.765e+02 1.058e+03, threshold=8.087e+02, percent-clipped=1.0 2022-12-23 10:30:34,474 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp1.1 from training. Duration: 23.82275 2022-12-23 10:30:48,828 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0 from training. Duration: 20.77 2022-12-23 10:30:54,926 INFO [train.py:894] (3/4) Epoch 17, batch 850, loss[loss=0.1715, simple_loss=0.2661, pruned_loss=0.03843, over 18581.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2697, pruned_loss=0.04734, over 3660896.43 frames. ], batch size: 51, lr: 6.84e-03, grad_scale: 16.0 2022-12-23 10:30:58,372 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0_sp0.9 from training. Duration: 24.088875 2022-12-23 10:31:25,903 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56969.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:31:26,910 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp1.1 from training. Duration: 20.4409375 2022-12-23 10:31:42,428 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7118, 1.6991, 1.5533, 1.6250, 2.0463, 1.9926, 2.0300, 1.3959], device='cuda:3'), covar=tensor([0.0351, 0.0256, 0.0427, 0.0209, 0.0169, 0.0350, 0.0235, 0.0288], device='cuda:3'), in_proj_covar=tensor([0.0090, 0.0122, 0.0148, 0.0123, 0.0116, 0.0115, 0.0096, 0.0124], device='cuda:3'), out_proj_covar=tensor([7.2633e-05, 9.8174e-05, 1.2401e-04, 9.9225e-05, 9.4810e-05, 8.9532e-05, 7.5773e-05, 9.8984e-05], device='cuda:3') 2022-12-23 10:32:01,257 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56993.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:32:04,488 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56995.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:32:10,250 INFO [train.py:894] (3/4) Epoch 17, batch 900, loss[loss=0.2214, simple_loss=0.3022, pruned_loss=0.07033, over 18558.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2701, pruned_loss=0.04753, over 3672098.72 frames. ], batch size: 49, lr: 6.83e-03, grad_scale: 16.0 2022-12-23 10:32:38,307 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8748, 1.4042, 0.8341, 1.3368, 1.9243, 1.3366, 1.6783, 1.9838], device='cuda:3'), covar=tensor([0.1626, 0.2041, 0.2479, 0.1550, 0.1982, 0.1737, 0.1396, 0.1528], device='cuda:3'), in_proj_covar=tensor([0.0093, 0.0097, 0.0117, 0.0095, 0.0116, 0.0091, 0.0098, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 10:32:43,753 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0_sp0.9 from training. Duration: 22.511125 2022-12-23 10:32:43,768 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0 from training. Duration: 20.675 2022-12-23 10:32:53,577 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([5.6047, 4.6965, 4.8460, 5.5580, 5.1601, 5.0405, 5.6600, 1.6577], device='cuda:3'), covar=tensor([0.0565, 0.0565, 0.0502, 0.0664, 0.1235, 0.0917, 0.0412, 0.5015], device='cuda:3'), in_proj_covar=tensor([0.0329, 0.0215, 0.0226, 0.0251, 0.0309, 0.0260, 0.0276, 0.0274], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 10:32:54,983 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.3589, 3.7284, 3.6769, 4.2845, 3.9403, 3.8650, 4.5201, 1.3608], device='cuda:3'), covar=tensor([0.0778, 0.0666, 0.0659, 0.0772, 0.1577, 0.1170, 0.0594, 0.5061], device='cuda:3'), in_proj_covar=tensor([0.0329, 0.0215, 0.0226, 0.0251, 0.0309, 0.0260, 0.0276, 0.0274], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 10:33:01,589 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.430e+02 4.029e+02 4.883e+02 6.473e+02 1.483e+03, threshold=9.767e+02, percent-clipped=10.0 2022-12-23 10:33:05,197 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9960, 1.9120, 1.7959, 1.0324, 2.3076, 2.0484, 1.8737, 1.5785], device='cuda:3'), covar=tensor([0.0321, 0.0397, 0.0411, 0.0772, 0.0264, 0.0341, 0.0437, 0.0800], device='cuda:3'), in_proj_covar=tensor([0.0121, 0.0123, 0.0127, 0.0119, 0.0094, 0.0120, 0.0134, 0.0154], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 10:33:10,465 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2022-12-23 10:33:15,577 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1964, 2.5954, 2.9721, 1.8390, 2.8162, 2.8692, 2.1739, 3.1680], device='cuda:3'), covar=tensor([0.1234, 0.1432, 0.1374, 0.1952, 0.0656, 0.1024, 0.1854, 0.0466], device='cuda:3'), in_proj_covar=tensor([0.0193, 0.0209, 0.0206, 0.0194, 0.0175, 0.0212, 0.0211, 0.0187], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 10:33:24,990 INFO [train.py:894] (3/4) Epoch 17, batch 950, loss[loss=0.1754, simple_loss=0.2664, pruned_loss=0.04219, over 18626.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2713, pruned_loss=0.04796, over 3681179.52 frames. ], batch size: 53, lr: 6.83e-03, grad_scale: 16.0 2022-12-23 10:33:25,454 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3987, 1.2561, 1.1633, 1.2804, 1.6473, 1.5078, 1.5330, 1.0957], device='cuda:3'), covar=tensor([0.0276, 0.0207, 0.0472, 0.0202, 0.0181, 0.0332, 0.0244, 0.0274], device='cuda:3'), in_proj_covar=tensor([0.0090, 0.0121, 0.0147, 0.0122, 0.0115, 0.0114, 0.0094, 0.0124], device='cuda:3'), out_proj_covar=tensor([7.2303e-05, 9.7194e-05, 1.2318e-04, 9.8476e-05, 9.4114e-05, 8.9021e-05, 7.4800e-05, 9.8445e-05], device='cuda:3') 2022-12-23 10:33:33,644 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57054.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:33:36,563 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57056.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:34:20,840 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp0.9 from training. Duration: 24.9833125 2022-12-23 10:34:25,786 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.1863, 1.4793, 1.1989, 1.7382, 1.6091, 1.4147, 0.9966, 1.2065], device='cuda:3'), covar=tensor([0.2102, 0.1884, 0.1727, 0.1125, 0.1382, 0.1250, 0.2175, 0.1618], device='cuda:3'), in_proj_covar=tensor([0.0243, 0.0217, 0.0209, 0.0193, 0.0253, 0.0192, 0.0215, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 10:34:41,011 INFO [train.py:894] (3/4) Epoch 17, batch 1000, loss[loss=0.158, simple_loss=0.2399, pruned_loss=0.038, over 18525.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2707, pruned_loss=0.04801, over 3687861.33 frames. ], batch size: 47, lr: 6.83e-03, grad_scale: 16.0 2022-12-23 10:34:53,960 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0 from training. Duration: 27.14 2022-12-23 10:35:10,608 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0 from training. Duration: 22.44 2022-12-23 10:35:33,881 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.085e+02 3.382e+02 4.120e+02 4.933e+02 1.139e+03, threshold=8.240e+02, percent-clipped=1.0 2022-12-23 10:35:56,744 INFO [train.py:894] (3/4) Epoch 17, batch 1050, loss[loss=0.1828, simple_loss=0.2648, pruned_loss=0.0504, over 18454.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2718, pruned_loss=0.04857, over 3693776.39 frames. ], batch size: 50, lr: 6.82e-03, grad_scale: 16.0 2022-12-23 10:36:11,700 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57159.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:36:18,836 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([5.8615, 4.9080, 5.1410, 5.9042, 5.4511, 5.2062, 5.9047, 1.7556], device='cuda:3'), covar=tensor([0.0548, 0.0570, 0.0461, 0.0522, 0.1151, 0.1000, 0.0386, 0.4670], device='cuda:3'), in_proj_covar=tensor([0.0324, 0.0212, 0.0222, 0.0249, 0.0302, 0.0256, 0.0272, 0.0270], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 10:36:29,430 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0060-62364-0_sp0.9 from training. Duration: 21.361125 2022-12-23 10:36:36,899 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp1.1 from training. Duration: 27.0318125 2022-12-23 10:36:41,927 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57179.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 10:36:44,743 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp0.9 from training. Duration: 28.638875 2022-12-23 10:36:52,234 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57186.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:37:01,131 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0 from training. Duration: 20.4 2022-12-23 10:37:08,469 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57197.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:37:10,971 INFO [train.py:894] (3/4) Epoch 17, batch 1100, loss[loss=0.1758, simple_loss=0.2626, pruned_loss=0.04453, over 18698.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2709, pruned_loss=0.04803, over 3697939.91 frames. ], batch size: 46, lr: 6.82e-03, grad_scale: 16.0 2022-12-23 10:37:33,362 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0 from training. Duration: 20.025 2022-12-23 10:37:33,379 WARNING [train.py:1060] (3/4) Exclude cut with ID 2364-131735-0112-64612-0_sp0.9 from training. Duration: 20.488875 2022-12-23 10:37:40,027 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0 from training. Duration: 29.735 2022-12-23 10:37:54,216 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57227.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 10:38:03,732 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.132e+02 3.502e+02 4.489e+02 5.299e+02 1.026e+03, threshold=8.977e+02, percent-clipped=3.0 2022-12-23 10:38:16,784 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6619, 2.3527, 1.8650, 1.1771, 2.8383, 2.7029, 2.4273, 1.8691], device='cuda:3'), covar=tensor([0.0288, 0.0366, 0.0543, 0.0836, 0.0244, 0.0311, 0.0399, 0.0786], device='cuda:3'), in_proj_covar=tensor([0.0122, 0.0124, 0.0128, 0.0119, 0.0095, 0.0120, 0.0135, 0.0156], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 10:38:19,635 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57244.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:38:24,345 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57247.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:38:26,750 INFO [train.py:894] (3/4) Epoch 17, batch 1150, loss[loss=0.1407, simple_loss=0.2217, pruned_loss=0.0299, over 18615.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2693, pruned_loss=0.0476, over 3701549.04 frames. ], batch size: 45, lr: 6.82e-03, grad_scale: 8.0 2022-12-23 10:38:41,330 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57258.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 10:38:50,905 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57264.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:39:02,684 WARNING [train.py:1060] (3/4) Exclude cut with ID 7276-92427-0014-12983-0_sp0.9 from training. Duration: 21.3055625 2022-12-23 10:39:02,741 WARNING [train.py:1060] (3/4) Exclude cut with ID 1025-75365-0008-79168-0_sp0.9 from training. Duration: 22.0666875 2022-12-23 10:39:43,497 INFO [train.py:894] (3/4) Epoch 17, batch 1200, loss[loss=0.2239, simple_loss=0.3002, pruned_loss=0.07383, over 18641.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2689, pruned_loss=0.04763, over 3704514.91 frames. ], batch size: 174, lr: 6.81e-03, grad_scale: 8.0 2022-12-23 10:39:52,728 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57305.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:40:07,755 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.4425, 3.8437, 3.8181, 4.3391, 4.1297, 3.9262, 4.5829, 1.4383], device='cuda:3'), covar=tensor([0.0686, 0.0723, 0.0640, 0.0766, 0.1302, 0.1222, 0.0632, 0.4987], device='cuda:3'), in_proj_covar=tensor([0.0323, 0.0212, 0.0222, 0.0249, 0.0303, 0.0256, 0.0272, 0.0269], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 10:40:30,577 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.82 vs. limit=5.0 2022-12-23 10:40:37,778 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.106e+02 3.673e+02 4.461e+02 5.439e+02 1.096e+03, threshold=8.923e+02, percent-clipped=3.0 2022-12-23 10:40:52,621 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0 from training. Duration: 20.26 2022-12-23 10:40:58,952 INFO [train.py:894] (3/4) Epoch 17, batch 1250, loss[loss=0.1796, simple_loss=0.2723, pruned_loss=0.04346, over 18445.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2694, pruned_loss=0.04759, over 3706748.50 frames. ], batch size: 64, lr: 6.81e-03, grad_scale: 8.0 2022-12-23 10:40:59,146 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57349.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:41:02,305 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57351.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:41:06,551 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0030-9324-0_sp0.9 from training. Duration: 21.3444375 2022-12-23 10:41:20,199 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.2627, 2.9129, 2.1473, 1.6545, 3.5687, 3.7279, 3.0730, 2.2376], device='cuda:3'), covar=tensor([0.0336, 0.0375, 0.0544, 0.0750, 0.0198, 0.0299, 0.0432, 0.0848], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0124, 0.0128, 0.0120, 0.0096, 0.0121, 0.0136, 0.0156], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 10:42:02,285 WARNING [train.py:1060] (3/4) Exclude cut with ID 497-129325-0061-62254-0_sp1.1 from training. Duration: 0.97725 2022-12-23 10:42:11,895 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57398.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:42:13,008 INFO [train.py:894] (3/4) Epoch 17, batch 1300, loss[loss=0.1466, simple_loss=0.2304, pruned_loss=0.0314, over 18703.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2693, pruned_loss=0.047, over 3707603.48 frames. ], batch size: 46, lr: 6.81e-03, grad_scale: 8.0 2022-12-23 10:42:45,615 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0_sp0.9 from training. Duration: 22.97225 2022-12-23 10:43:08,225 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.076e+02 3.308e+02 4.091e+02 4.945e+02 1.166e+03, threshold=8.182e+02, percent-clipped=4.0 2022-12-23 10:43:17,288 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0047-39922-0_sp0.9 from training. Duration: 21.97775 2022-12-23 10:43:28,430 WARNING [train.py:1060] (3/4) Exclude cut with ID 1112-1043-0006-89194-0_sp0.9 from training. Duration: 21.8333125 2022-12-23 10:43:29,889 INFO [train.py:894] (3/4) Epoch 17, batch 1350, loss[loss=0.1707, simple_loss=0.2642, pruned_loss=0.0386, over 18566.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2693, pruned_loss=0.04717, over 3708783.69 frames. ], batch size: 97, lr: 6.81e-03, grad_scale: 8.0 2022-12-23 10:43:37,648 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2935, 3.3076, 3.2836, 1.3275, 3.4242, 2.5290, 0.5451, 2.2302], device='cuda:3'), covar=tensor([0.2094, 0.1103, 0.1409, 0.3629, 0.0906, 0.0965, 0.4925, 0.1570], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0130, 0.0150, 0.0119, 0.0133, 0.0106, 0.0140, 0.0109], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 10:43:37,945 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4800, 1.7678, 1.4024, 2.0604, 2.1700, 1.5472, 1.2598, 1.3261], device='cuda:3'), covar=tensor([0.1950, 0.1729, 0.1601, 0.1018, 0.1157, 0.1180, 0.2231, 0.1562], device='cuda:3'), in_proj_covar=tensor([0.0247, 0.0221, 0.0212, 0.0196, 0.0258, 0.0195, 0.0219, 0.0197], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 10:43:38,932 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0 from training. Duration: 20.47 2022-12-23 10:43:45,296 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57459.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:43:45,465 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57459.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:43:49,679 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57462.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:43:55,405 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2022-12-23 10:44:39,435 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57495.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:44:43,934 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0037-39912-0_sp0.9 from training. Duration: 20.67225 2022-12-23 10:44:45,325 INFO [train.py:894] (3/4) Epoch 17, batch 1400, loss[loss=0.1841, simple_loss=0.2735, pruned_loss=0.04733, over 18567.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2685, pruned_loss=0.04715, over 3710596.07 frames. ], batch size: 56, lr: 6.80e-03, grad_scale: 8.0 2022-12-23 10:44:57,888 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57507.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:45:05,484 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp0.9 from training. Duration: 25.2444375 2022-12-23 10:45:22,336 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57523.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:45:28,181 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0021-76797-0_sp0.9 from training. Duration: 21.1445 2022-12-23 10:45:40,194 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.185e+02 3.513e+02 4.155e+02 4.979e+02 8.906e+02, threshold=8.311e+02, percent-clipped=2.0 2022-12-23 10:45:51,118 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57542.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:46:00,942 INFO [train.py:894] (3/4) Epoch 17, batch 1450, loss[loss=0.1603, simple_loss=0.2413, pruned_loss=0.03963, over 18664.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2684, pruned_loss=0.04704, over 3711148.33 frames. ], batch size: 46, lr: 6.80e-03, grad_scale: 8.0 2022-12-23 10:46:06,939 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57553.0, num_to_drop=1, layers_to_drop={3} 2022-12-23 10:46:12,060 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57556.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:46:23,912 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57564.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:46:44,474 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp0.9 from training. Duration: 33.038875 2022-12-23 10:47:15,628 INFO [train.py:894] (3/4) Epoch 17, batch 1500, loss[loss=0.1971, simple_loss=0.2809, pruned_loss=0.05661, over 18650.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2679, pruned_loss=0.0466, over 3711879.25 frames. ], batch size: 60, lr: 6.80e-03, grad_scale: 8.0 2022-12-23 10:47:17,661 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57600.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:47:22,864 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64291-0000-16059-0_sp0.9 from training. Duration: 20.0944375 2022-12-23 10:47:36,390 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp1.1 from training. Duration: 20.4 2022-12-23 10:47:36,530 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57612.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:47:44,324 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0 from training. Duration: 20.085 2022-12-23 10:47:55,618 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0_sp0.9 from training. Duration: 23.07775 2022-12-23 10:48:12,656 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.389e+02 3.464e+02 4.346e+02 5.489e+02 1.288e+03, threshold=8.692e+02, percent-clipped=8.0 2022-12-23 10:48:34,271 INFO [train.py:894] (3/4) Epoch 17, batch 1550, loss[loss=0.18, simple_loss=0.2651, pruned_loss=0.04745, over 18672.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2671, pruned_loss=0.04602, over 3712430.08 frames. ], batch size: 48, lr: 6.79e-03, grad_scale: 8.0 2022-12-23 10:48:35,224 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57649.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:48:37,874 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57651.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:48:42,052 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp0.9 from training. Duration: 24.9333125 2022-12-23 10:49:04,677 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.1757, 2.3630, 1.4841, 2.8735, 2.1164, 2.0027, 2.1760, 3.1205], device='cuda:3'), covar=tensor([0.1939, 0.3030, 0.2046, 0.2906, 0.3756, 0.1243, 0.3474, 0.0790], device='cuda:3'), in_proj_covar=tensor([0.0289, 0.0284, 0.0239, 0.0346, 0.0262, 0.0222, 0.0280, 0.0206], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 10:49:06,023 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5122, 1.9170, 1.5213, 2.3207, 2.2920, 1.5691, 1.4541, 1.3107], device='cuda:3'), covar=tensor([0.2092, 0.1784, 0.1711, 0.1004, 0.1338, 0.1293, 0.2291, 0.1720], device='cuda:3'), in_proj_covar=tensor([0.0245, 0.0220, 0.0212, 0.0195, 0.0257, 0.0194, 0.0218, 0.0196], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 10:49:28,321 WARNING [train.py:1060] (3/4) Exclude cut with ID 5118-111612-0016-124680-0_sp0.9 from training. Duration: 20.388875 2022-12-23 10:49:34,209 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp1.1 from training. Duration: 20.3590625 2022-12-23 10:49:43,680 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2022-12-23 10:49:47,400 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57697.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:49:50,132 INFO [train.py:894] (3/4) Epoch 17, batch 1600, loss[loss=0.1667, simple_loss=0.262, pruned_loss=0.03571, over 18623.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2671, pruned_loss=0.0458, over 3712692.80 frames. ], batch size: 53, lr: 6.79e-03, grad_scale: 8.0 2022-12-23 10:49:50,926 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57699.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:50:40,954 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57733.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:50:41,065 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.3412, 2.4682, 1.9070, 3.1577, 2.2909, 2.4037, 2.5097, 3.4805], device='cuda:3'), covar=tensor([0.1799, 0.3258, 0.1823, 0.2699, 0.3830, 0.0980, 0.3155, 0.0690], device='cuda:3'), in_proj_covar=tensor([0.0290, 0.0284, 0.0240, 0.0347, 0.0263, 0.0223, 0.0280, 0.0207], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 10:50:43,433 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.172e+02 3.678e+02 4.212e+02 5.309e+02 1.200e+03, threshold=8.424e+02, percent-clipped=1.0 2022-12-23 10:50:45,148 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0_sp1.1 from training. Duration: 0.836375 2022-12-23 10:51:04,906 INFO [train.py:894] (3/4) Epoch 17, batch 1650, loss[loss=0.1854, simple_loss=0.275, pruned_loss=0.04795, over 18674.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2673, pruned_loss=0.04653, over 3712268.10 frames. ], batch size: 79, lr: 6.79e-03, grad_scale: 8.0 2022-12-23 10:51:13,007 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57754.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:51:30,451 WARNING [train.py:1060] (3/4) Exclude cut with ID 8565-290391-0049-67394-0_sp0.9 from training. Duration: 21.3166875 2022-12-23 10:51:37,401 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.1020, 0.9803, 1.1690, 0.5481, 0.5924, 1.2295, 1.2416, 1.1836], device='cuda:3'), covar=tensor([0.0784, 0.0318, 0.0331, 0.0376, 0.0443, 0.0442, 0.0262, 0.0630], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0167, 0.0122, 0.0139, 0.0149, 0.0141, 0.0160, 0.0169], device='cuda:3'), out_proj_covar=tensor([1.1614e-04, 1.3073e-04, 9.3899e-05, 1.0530e-04, 1.1434e-04, 1.1067e-04, 1.2547e-04, 1.3092e-04], device='cuda:3') 2022-12-23 10:51:45,815 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2022-12-23 10:51:59,944 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0029-104863-0_sp0.9 from training. Duration: 22.1055625 2022-12-23 10:52:11,442 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp1.1 from training. Duration: 21.77725 2022-12-23 10:52:13,234 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57794.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:52:20,529 INFO [train.py:894] (3/4) Epoch 17, batch 1700, loss[loss=0.1875, simple_loss=0.2749, pruned_loss=0.05003, over 18516.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2698, pruned_loss=0.04882, over 3713515.46 frames. ], batch size: 52, lr: 6.79e-03, grad_scale: 8.0 2022-12-23 10:52:22,881 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7104, 1.6253, 1.4454, 1.4936, 1.8548, 1.8269, 1.8935, 1.2491], device='cuda:3'), covar=tensor([0.0336, 0.0232, 0.0489, 0.0216, 0.0199, 0.0397, 0.0258, 0.0340], device='cuda:3'), in_proj_covar=tensor([0.0091, 0.0123, 0.0151, 0.0125, 0.0118, 0.0117, 0.0096, 0.0126], device='cuda:3'), out_proj_covar=tensor([7.3528e-05, 9.8455e-05, 1.2632e-04, 1.0093e-04, 9.6233e-05, 9.0991e-05, 7.6207e-05, 1.0041e-04], device='cuda:3') 2022-12-23 10:52:32,523 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp0.9 from training. Duration: 27.8166875 2022-12-23 10:52:49,783 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57818.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:52:58,946 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp1.1 from training. Duration: 22.5090625 2022-12-23 10:53:02,091 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4151, 3.2068, 3.3065, 1.3028, 3.3668, 2.4846, 0.8389, 2.3120], device='cuda:3'), covar=tensor([0.2262, 0.1498, 0.1641, 0.4014, 0.1193, 0.1071, 0.4551, 0.1702], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0132, 0.0152, 0.0122, 0.0136, 0.0107, 0.0142, 0.0110], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 10:53:04,868 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0 from training. Duration: 25.035 2022-12-23 10:53:14,895 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.260e+02 4.546e+02 5.067e+02 6.355e+02 1.352e+03, threshold=1.013e+03, percent-clipped=7.0 2022-12-23 10:53:22,475 WARNING [train.py:1060] (3/4) Exclude cut with ID 774-127930-0014-10412-0_sp1.1 from training. Duration: 0.95 2022-12-23 10:53:25,493 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57842.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:53:35,706 INFO [train.py:894] (3/4) Epoch 17, batch 1750, loss[loss=0.2115, simple_loss=0.2943, pruned_loss=0.06439, over 18607.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2714, pruned_loss=0.05073, over 3713652.65 frames. ], batch size: 51, lr: 6.78e-03, grad_scale: 8.0 2022-12-23 10:53:38,893 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57851.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:53:40,228 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp0.9 from training. Duration: 0.92225 2022-12-23 10:53:42,169 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57853.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 10:54:07,145 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0 from training. Duration: 21.97 2022-12-23 10:54:07,495 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9545, 1.8270, 1.9645, 2.0788, 1.6442, 5.0709, 2.0168, 2.6407], device='cuda:3'), covar=tensor([0.3126, 0.1928, 0.1816, 0.1796, 0.1241, 0.0089, 0.1438, 0.0793], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0117, 0.0127, 0.0121, 0.0105, 0.0098, 0.0093, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 10:54:25,729 WARNING [train.py:1060] (3/4) Exclude cut with ID 7492-105653-0055-62765-0_sp0.9 from training. Duration: 21.97225 2022-12-23 10:54:25,778 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp0.9 from training. Duration: 25.3333125 2022-12-23 10:54:36,793 WARNING [train.py:1060] (3/4) Exclude cut with ID 5172-29468-0015-19128-0_sp0.9 from training. Duration: 21.5055625 2022-12-23 10:54:38,142 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57890.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:54:45,925 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp1.1 from training. Duration: 20.72725 2022-12-23 10:54:52,433 INFO [train.py:894] (3/4) Epoch 17, batch 1800, loss[loss=0.244, simple_loss=0.3099, pruned_loss=0.08901, over 18587.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2738, pruned_loss=0.05308, over 3713401.46 frames. ], batch size: 181, lr: 6.78e-03, grad_scale: 8.0 2022-12-23 10:54:54,317 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57900.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:54:55,904 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57901.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:55:18,252 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp0.9 from training. Duration: 26.32775 2022-12-23 10:55:47,101 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0 from training. Duration: 20.025 2022-12-23 10:55:49,147 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.792e+02 4.677e+02 5.758e+02 7.423e+02 3.377e+03, threshold=1.152e+03, percent-clipped=6.0 2022-12-23 10:55:53,724 WARNING [train.py:1060] (3/4) Exclude cut with ID 6709-74022-0004-86860-0_sp1.1 from training. Duration: 0.9409375 2022-12-23 10:55:55,116 WARNING [train.py:1060] (3/4) Exclude cut with ID 4757-1811-0023-62229-0_sp0.9 from training. Duration: 21.37775 2022-12-23 10:56:07,930 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57948.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:56:09,129 INFO [train.py:894] (3/4) Epoch 17, batch 1850, loss[loss=0.2158, simple_loss=0.3007, pruned_loss=0.06552, over 18717.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2743, pruned_loss=0.05455, over 3713155.11 frames. ], batch size: 54, lr: 6.78e-03, grad_scale: 8.0 2022-12-23 10:56:13,580 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0004-25974-0_sp0.9 from training. Duration: 21.17225 2022-12-23 10:56:15,433 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp0.9 from training. Duration: 27.511125 2022-12-23 10:56:25,503 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.8484, 3.3117, 3.2850, 3.7551, 3.5196, 3.3497, 3.9962, 1.1451], device='cuda:3'), covar=tensor([0.0791, 0.0737, 0.0755, 0.0858, 0.1497, 0.1244, 0.0658, 0.4907], device='cuda:3'), in_proj_covar=tensor([0.0328, 0.0215, 0.0227, 0.0253, 0.0310, 0.0260, 0.0277, 0.0273], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 10:56:31,128 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.53 vs. limit=5.0 2022-12-23 10:56:48,310 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0 from training. Duration: 22.8 2022-12-23 10:56:53,157 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0 from training. Duration: 22.585 2022-12-23 10:57:24,025 INFO [train.py:894] (3/4) Epoch 17, batch 1900, loss[loss=0.2489, simple_loss=0.3172, pruned_loss=0.09032, over 18632.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.275, pruned_loss=0.05604, over 3712620.62 frames. ], batch size: 99, lr: 6.77e-03, grad_scale: 8.0 2022-12-23 10:57:24,060 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0001-146967-0_sp0.9 from training. Duration: 22.0166875 2022-12-23 10:57:42,087 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp1.1 from training. Duration: 24.395375 2022-12-23 10:57:48,330 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp0.9 from training. Duration: 27.47775 2022-12-23 10:57:52,385 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp0.9 from training. Duration: 24.8833125 2022-12-23 10:57:54,198 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0 from training. Duration: 23.39 2022-12-23 10:58:00,632 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp0.9 from training. Duration: 28.72225 2022-12-23 10:58:11,254 WARNING [train.py:1060] (3/4) Exclude cut with ID 585-294811-0110-133686-0_sp0.9 from training. Duration: 20.8944375 2022-12-23 10:58:23,481 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.619e+02 4.765e+02 6.161e+02 7.873e+02 1.431e+03, threshold=1.232e+03, percent-clipped=2.0 2022-12-23 10:58:27,823 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0_sp0.9 from training. Duration: 23.8444375 2022-12-23 10:58:44,200 INFO [train.py:894] (3/4) Epoch 17, batch 1950, loss[loss=0.212, simple_loss=0.2871, pruned_loss=0.06839, over 18724.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2763, pruned_loss=0.05724, over 3713345.66 frames. ], batch size: 52, lr: 6.77e-03, grad_scale: 8.0 2022-12-23 10:58:52,454 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58054.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:58:53,651 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0 from training. Duration: 25.85 2022-12-23 10:58:53,662 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0 from training. Duration: 21.39 2022-12-23 10:59:05,161 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0 from training. Duration: 27.92 2022-12-23 10:59:29,279 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3311, 1.9681, 0.6004, 1.7310, 2.5324, 1.7768, 2.3105, 2.4259], device='cuda:3'), covar=tensor([0.1608, 0.1941, 0.2752, 0.1550, 0.1706, 0.1641, 0.1403, 0.1586], device='cuda:3'), in_proj_covar=tensor([0.0093, 0.0097, 0.0117, 0.0094, 0.0116, 0.0091, 0.0097, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 10:59:33,669 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0039-130165-0_sp0.9 from training. Duration: 20.661125 2022-12-23 10:59:45,967 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58089.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 10:59:56,158 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0043-15874-0_sp0.9 from training. Duration: 20.07225 2022-12-23 10:59:56,554 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7192, 1.6961, 1.7363, 1.7177, 1.2902, 3.6780, 1.6239, 2.2188], device='cuda:3'), covar=tensor([0.3037, 0.1912, 0.1876, 0.1952, 0.1416, 0.0187, 0.1649, 0.0811], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0117, 0.0126, 0.0120, 0.0104, 0.0098, 0.0093, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 11:00:01,379 INFO [train.py:894] (3/4) Epoch 17, batch 2000, loss[loss=0.1869, simple_loss=0.2604, pruned_loss=0.05672, over 18675.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.2748, pruned_loss=0.05692, over 3713618.74 frames. ], batch size: 46, lr: 6.77e-03, grad_scale: 8.0 2022-12-23 11:00:05,694 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0 from training. Duration: 21.01 2022-12-23 11:00:06,432 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=58102.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:00:30,816 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58118.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:00:57,070 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.188e+02 4.438e+02 5.445e+02 6.526e+02 1.482e+03, threshold=1.089e+03, percent-clipped=1.0 2022-12-23 11:01:00,214 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.0983, 0.8360, 0.8988, 1.1235, 1.2545, 1.1141, 1.0975, 0.8986], device='cuda:3'), covar=tensor([0.0265, 0.0242, 0.0521, 0.0198, 0.0235, 0.0376, 0.0271, 0.0289], device='cuda:3'), in_proj_covar=tensor([0.0091, 0.0125, 0.0150, 0.0124, 0.0118, 0.0117, 0.0096, 0.0126], device='cuda:3'), out_proj_covar=tensor([7.3464e-05, 1.0009e-04, 1.2594e-04, 1.0029e-04, 9.6273e-05, 9.0984e-05, 7.6173e-05, 1.0023e-04], device='cuda:3') 2022-12-23 11:01:04,602 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5771, 1.9323, 1.5523, 2.2611, 2.3687, 1.6022, 1.4732, 1.3367], device='cuda:3'), covar=tensor([0.1966, 0.1682, 0.1648, 0.0993, 0.1184, 0.1248, 0.2154, 0.1639], device='cuda:3'), in_proj_covar=tensor([0.0244, 0.0219, 0.0210, 0.0193, 0.0256, 0.0193, 0.0218, 0.0196], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 11:01:05,943 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58141.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:01:11,035 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58144.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:01:13,230 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0 from training. Duration: 20.65 2022-12-23 11:01:18,092 INFO [train.py:894] (3/4) Epoch 17, batch 2050, loss[loss=0.1935, simple_loss=0.2833, pruned_loss=0.05187, over 18717.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2734, pruned_loss=0.05656, over 3712985.37 frames. ], batch size: 54, lr: 6.77e-03, grad_scale: 8.0 2022-12-23 11:01:21,016 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0 from training. Duration: 21.46 2022-12-23 11:01:21,260 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58151.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:01:42,682 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=58166.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:01:42,755 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3037, 2.7879, 2.7682, 1.1206, 2.8740, 2.0499, 0.6142, 2.0061], device='cuda:3'), covar=tensor([0.1980, 0.1473, 0.1660, 0.3921, 0.1288, 0.1356, 0.4685, 0.1603], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0134, 0.0154, 0.0123, 0.0137, 0.0109, 0.0144, 0.0111], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 11:02:09,024 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0 from training. Duration: 0.92 2022-12-23 11:02:16,225 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0_sp0.9 from training. Duration: 23.7666875 2022-12-23 11:02:33,666 INFO [train.py:894] (3/4) Epoch 17, batch 2100, loss[loss=0.1855, simple_loss=0.2639, pruned_loss=0.0536, over 18711.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2733, pruned_loss=0.05677, over 3713759.34 frames. ], batch size: 52, lr: 6.76e-03, grad_scale: 8.0 2022-12-23 11:02:33,813 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=58199.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:02:38,690 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58202.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:02:41,606 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.1757, 5.5714, 4.9964, 2.7132, 5.6411, 4.2061, 1.1061, 4.1145], device='cuda:3'), covar=tensor([0.1691, 0.0876, 0.1295, 0.3205, 0.0585, 0.0803, 0.5160, 0.1126], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0133, 0.0154, 0.0122, 0.0136, 0.0108, 0.0143, 0.0111], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 11:02:43,260 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58205.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:02:53,285 WARNING [train.py:1060] (3/4) Exclude cut with ID 8544-281189-0060-101339-0_sp0.9 from training. Duration: 20.861125 2022-12-23 11:03:02,893 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0_sp0.9 from training. Duration: 22.711125 2022-12-23 11:03:12,985 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4549, 3.2788, 3.2424, 1.3439, 3.2977, 2.3584, 0.6583, 2.1206], device='cuda:3'), covar=tensor([0.2083, 0.1189, 0.1467, 0.3704, 0.1076, 0.1180, 0.4907, 0.1672], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0132, 0.0153, 0.0120, 0.0135, 0.0107, 0.0141, 0.0110], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 11:03:29,639 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.687e+02 4.700e+02 5.944e+02 7.223e+02 2.461e+03, threshold=1.189e+03, percent-clipped=2.0 2022-12-23 11:03:30,218 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7892, 1.6724, 1.3597, 1.5953, 1.9159, 1.7176, 2.0686, 1.8398], device='cuda:3'), covar=tensor([0.0909, 0.1619, 0.2694, 0.1687, 0.1748, 0.0907, 0.1072, 0.1250], device='cuda:3'), in_proj_covar=tensor([0.0177, 0.0205, 0.0245, 0.0288, 0.0233, 0.0188, 0.0208, 0.0201], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 11:03:44,102 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp1.1 from training. Duration: 22.986375 2022-12-23 11:03:51,314 INFO [train.py:894] (3/4) Epoch 17, batch 2150, loss[loss=0.1928, simple_loss=0.2648, pruned_loss=0.06044, over 18546.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2733, pruned_loss=0.05708, over 3712808.89 frames. ], batch size: 41, lr: 6.76e-03, grad_scale: 8.0 2022-12-23 11:04:00,497 WARNING [train.py:1060] (3/4) Exclude cut with ID 8040-260924-0003-80960-0_sp0.9 from training. Duration: 22.07225 2022-12-23 11:04:00,913 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5256, 1.4876, 1.5220, 1.6000, 1.1536, 3.5360, 1.4099, 2.0580], device='cuda:3'), covar=tensor([0.3191, 0.2099, 0.2131, 0.2071, 0.1482, 0.0190, 0.1647, 0.0845], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0116, 0.0126, 0.0120, 0.0103, 0.0097, 0.0092, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 11:04:04,953 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0045-26330-0_sp0.9 from training. Duration: 20.3055625 2022-12-23 11:04:06,411 WARNING [train.py:1060] (3/4) Exclude cut with ID 6356-271890-0060-94317-0_sp0.9 from training. Duration: 20.72225 2022-12-23 11:04:13,457 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0045, 2.0130, 1.4035, 2.1898, 2.1185, 1.9385, 2.7682, 2.0022], device='cuda:3'), covar=tensor([0.0941, 0.1678, 0.2960, 0.1902, 0.1858, 0.0902, 0.0978, 0.1265], device='cuda:3'), in_proj_covar=tensor([0.0177, 0.0205, 0.0244, 0.0287, 0.0233, 0.0188, 0.0208, 0.0201], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 11:04:25,641 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp1.1 from training. Duration: 22.4818125 2022-12-23 11:04:51,425 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp0.9 from training. Duration: 25.0944375 2022-12-23 11:04:54,726 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0 from training. Duration: 21.515 2022-12-23 11:05:00,391 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp0.9 from training. Duration: 27.02225 2022-12-23 11:05:06,308 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0010-62480-0_sp0.9 from training. Duration: 22.22225 2022-12-23 11:05:07,554 INFO [train.py:894] (3/4) Epoch 17, batch 2200, loss[loss=0.1957, simple_loss=0.2715, pruned_loss=0.06, over 18546.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.274, pruned_loss=0.05758, over 3712357.79 frames. ], batch size: 47, lr: 6.76e-03, grad_scale: 8.0 2022-12-23 11:05:13,358 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0085-44554-0_sp0.9 from training. Duration: 20.85 2022-12-23 11:05:46,348 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3849, 2.2105, 1.7046, 0.9891, 1.7222, 1.9711, 1.7621, 1.8953], device='cuda:3'), covar=tensor([0.0625, 0.0460, 0.1182, 0.1391, 0.1069, 0.1261, 0.1389, 0.0693], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0182, 0.0205, 0.0190, 0.0209, 0.0199, 0.0214, 0.0199], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 11:05:47,374 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0 from training. Duration: 21.54 2022-12-23 11:05:52,536 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.67 vs. limit=5.0 2022-12-23 11:05:53,210 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp1.1 from training. Duration: 20.5318125 2022-12-23 11:06:03,154 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.624e+02 4.367e+02 5.459e+02 6.850e+02 1.946e+03, threshold=1.092e+03, percent-clipped=2.0 2022-12-23 11:06:03,246 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0012-134311-0_sp0.9 from training. Duration: 21.9333125 2022-12-23 11:06:23,581 INFO [train.py:894] (3/4) Epoch 17, batch 2250, loss[loss=0.2329, simple_loss=0.2991, pruned_loss=0.08336, over 18487.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2746, pruned_loss=0.05829, over 3713582.34 frames. ], batch size: 64, lr: 6.75e-03, grad_scale: 8.0 2022-12-23 11:06:51,901 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0025-130151-0_sp0.9 from training. Duration: 21.7944375 2022-12-23 11:07:03,554 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0_sp0.9 from training. Duration: 22.4666875 2022-12-23 11:07:12,260 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0 from training. Duration: 21.635 2022-12-23 11:07:16,360 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0_sp0.9 from training. Duration: 24.038875 2022-12-23 11:07:25,595 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58389.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:07:29,310 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2022-12-23 11:07:40,066 INFO [train.py:894] (3/4) Epoch 17, batch 2300, loss[loss=0.1602, simple_loss=0.2431, pruned_loss=0.03862, over 18483.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.2745, pruned_loss=0.05841, over 3712326.75 frames. ], batch size: 50, lr: 6.75e-03, grad_scale: 8.0 2022-12-23 11:07:42,325 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5290, 1.4842, 1.6360, 1.5440, 1.1964, 2.6537, 1.5611, 1.8522], device='cuda:3'), covar=tensor([0.2489, 0.1733, 0.1466, 0.1610, 0.1238, 0.0261, 0.1785, 0.0765], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0117, 0.0127, 0.0121, 0.0104, 0.0098, 0.0093, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 11:08:02,225 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp1.1 from training. Duration: 21.786375 2022-12-23 11:08:14,061 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0 from training. Duration: 20.22 2022-12-23 11:08:35,940 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.729e+02 4.639e+02 5.586e+02 7.120e+02 1.338e+03, threshold=1.117e+03, percent-clipped=3.0 2022-12-23 11:08:38,866 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=58437.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:08:39,127 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9385, 1.8425, 2.2147, 1.3735, 2.2913, 2.2533, 1.4831, 2.5895], device='cuda:3'), covar=tensor([0.1086, 0.1758, 0.1290, 0.1810, 0.0692, 0.1111, 0.2313, 0.0513], device='cuda:3'), in_proj_covar=tensor([0.0198, 0.0211, 0.0207, 0.0196, 0.0177, 0.0216, 0.0214, 0.0191], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 11:08:56,501 INFO [train.py:894] (3/4) Epoch 17, batch 2350, loss[loss=0.1929, simple_loss=0.2811, pruned_loss=0.05238, over 18720.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2742, pruned_loss=0.05815, over 3712883.74 frames. ], batch size: 65, lr: 6.75e-03, grad_scale: 8.0 2022-12-23 11:09:06,301 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2022-12-23 11:09:58,046 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8554, 1.5199, 1.8244, 2.3229, 2.1802, 4.5365, 1.6944, 1.9091], device='cuda:3'), covar=tensor([0.0831, 0.1826, 0.1041, 0.0881, 0.1276, 0.0200, 0.1335, 0.1468], device='cuda:3'), in_proj_covar=tensor([0.0072, 0.0082, 0.0072, 0.0074, 0.0090, 0.0074, 0.0084, 0.0077], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 11:10:10,040 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58497.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:10:12,652 INFO [train.py:894] (3/4) Epoch 17, batch 2400, loss[loss=0.1962, simple_loss=0.2815, pruned_loss=0.05542, over 18532.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2738, pruned_loss=0.058, over 3712028.58 frames. ], batch size: 55, lr: 6.75e-03, grad_scale: 8.0 2022-12-23 11:10:15,511 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0 from training. Duration: 25.285 2022-12-23 11:10:15,610 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58500.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:10:23,195 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1339, 1.5030, 1.8274, 1.8502, 2.1440, 2.0495, 1.9119, 1.6698], device='cuda:3'), covar=tensor([0.2065, 0.3043, 0.2309, 0.2715, 0.1830, 0.0896, 0.2930, 0.1225], device='cuda:3'), in_proj_covar=tensor([0.0261, 0.0294, 0.0274, 0.0308, 0.0296, 0.0245, 0.0329, 0.0232], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 11:10:30,452 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1760, 2.2632, 1.5623, 2.6767, 2.4143, 2.0956, 3.0641, 2.2227], device='cuda:3'), covar=tensor([0.0826, 0.1588, 0.2695, 0.1713, 0.1620, 0.0878, 0.0863, 0.1146], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0203, 0.0243, 0.0284, 0.0232, 0.0187, 0.0207, 0.0200], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 11:10:52,830 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4712, 1.4139, 1.4097, 1.3848, 0.7759, 2.2612, 0.8859, 1.4098], device='cuda:3'), covar=tensor([0.3141, 0.2051, 0.2139, 0.2053, 0.1547, 0.0344, 0.1661, 0.0882], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0116, 0.0127, 0.0120, 0.0104, 0.0098, 0.0093, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 11:11:08,059 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.821e+02 4.687e+02 5.741e+02 6.736e+02 1.581e+03, threshold=1.148e+03, percent-clipped=3.0 2022-12-23 11:11:19,653 WARNING [train.py:1060] (3/4) Exclude cut with ID 811-130148-0001-63453-0_sp0.9 from training. Duration: 20.861125 2022-12-23 11:11:28,751 INFO [train.py:894] (3/4) Epoch 17, batch 2450, loss[loss=0.2198, simple_loss=0.3005, pruned_loss=0.06949, over 18492.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2734, pruned_loss=0.05767, over 3712983.81 frames. ], batch size: 77, lr: 6.74e-03, grad_scale: 8.0 2022-12-23 11:11:43,319 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0 from training. Duration: 20.88 2022-12-23 11:12:15,291 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0_sp0.9 from training. Duration: 23.4166875 2022-12-23 11:12:41,660 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2022-12-23 11:12:43,589 INFO [train.py:894] (3/4) Epoch 17, batch 2500, loss[loss=0.21, simple_loss=0.3016, pruned_loss=0.05917, over 18625.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.2733, pruned_loss=0.05763, over 3712909.63 frames. ], batch size: 53, lr: 6.74e-03, grad_scale: 8.0 2022-12-23 11:13:32,659 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0 from training. Duration: 21.24 2022-12-23 11:13:32,671 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0_sp0.9 from training. Duration: 23.9055625 2022-12-23 11:13:37,242 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.626e+02 4.362e+02 5.589e+02 6.894e+02 1.162e+03, threshold=1.118e+03, percent-clipped=1.0 2022-12-23 11:13:59,182 INFO [train.py:894] (3/4) Epoch 17, batch 2550, loss[loss=0.185, simple_loss=0.2722, pruned_loss=0.04886, over 18585.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2743, pruned_loss=0.05761, over 3712697.23 frames. ], batch size: 97, lr: 6.74e-03, grad_scale: 8.0 2022-12-23 11:14:05,600 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp0.9 from training. Duration: 25.988875 2022-12-23 11:14:12,063 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0001-134300-0_sp0.9 from training. Duration: 20.67225 2022-12-23 11:14:12,428 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5188, 1.5947, 1.8006, 1.0553, 1.7839, 1.7909, 1.3572, 2.0863], device='cuda:3'), covar=tensor([0.1051, 0.1828, 0.1092, 0.1513, 0.0690, 0.1022, 0.2328, 0.0542], device='cuda:3'), in_proj_covar=tensor([0.0197, 0.0210, 0.0205, 0.0195, 0.0175, 0.0216, 0.0213, 0.0191], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 11:14:48,463 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58681.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:14:59,719 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0 from training. Duration: 20.34 2022-12-23 11:15:17,178 INFO [train.py:894] (3/4) Epoch 17, batch 2600, loss[loss=0.1565, simple_loss=0.2393, pruned_loss=0.03687, over 18540.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2742, pruned_loss=0.05782, over 3712530.00 frames. ], batch size: 47, lr: 6.73e-03, grad_scale: 8.0 2022-12-23 11:16:07,738 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp0.9 from training. Duration: 25.061125 2022-12-23 11:16:12,108 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.664e+02 4.354e+02 5.588e+02 7.123e+02 1.427e+03, threshold=1.118e+03, percent-clipped=9.0 2022-12-23 11:16:18,685 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0 from training. Duration: 0.83 2022-12-23 11:16:23,858 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58742.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:16:34,626 INFO [train.py:894] (3/4) Epoch 17, batch 2650, loss[loss=0.16, simple_loss=0.2319, pruned_loss=0.04401, over 18402.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.274, pruned_loss=0.05763, over 3713232.78 frames. ], batch size: 42, lr: 6.73e-03, grad_scale: 8.0 2022-12-23 11:16:45,009 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0 from training. Duration: 24.73 2022-12-23 11:16:57,929 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0 from training. Duration: 23.965 2022-12-23 11:16:58,928 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 2022-12-23 11:17:06,691 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0030-146996-0_sp0.9 from training. Duration: 22.088875 2022-12-23 11:17:13,346 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58775.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:17:22,659 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0_sp0.9 from training. Duration: 23.6 2022-12-23 11:17:47,812 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58797.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:17:50,517 INFO [train.py:894] (3/4) Epoch 17, batch 2700, loss[loss=0.1978, simple_loss=0.2823, pruned_loss=0.05661, over 18608.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2738, pruned_loss=0.05732, over 3714625.07 frames. ], batch size: 97, lr: 6.73e-03, grad_scale: 8.0 2022-12-23 11:17:52,704 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58800.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:18:02,975 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.97 vs. limit=5.0 2022-12-23 11:18:30,490 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6503, 4.0361, 3.8556, 1.5956, 4.0742, 2.9809, 0.6690, 2.7173], device='cuda:3'), covar=tensor([0.2141, 0.1112, 0.1649, 0.3806, 0.1123, 0.1046, 0.5356, 0.1555], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0137, 0.0157, 0.0124, 0.0139, 0.0111, 0.0144, 0.0113], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 11:18:45,633 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.763e+02 4.526e+02 5.821e+02 7.032e+02 2.176e+03, threshold=1.164e+03, percent-clipped=2.0 2022-12-23 11:18:48,125 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58836.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:19:01,427 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=58845.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:19:05,894 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=58848.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:19:07,370 INFO [train.py:894] (3/4) Epoch 17, batch 2750, loss[loss=0.2276, simple_loss=0.2962, pruned_loss=0.07946, over 18565.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2741, pruned_loss=0.05747, over 3715586.86 frames. ], batch size: 175, lr: 6.73e-03, grad_scale: 8.0 2022-12-23 11:19:08,901 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0 from training. Duration: 23.795 2022-12-23 11:19:24,410 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp1.1 from training. Duration: 21.5409375 2022-12-23 11:19:27,297 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp0.9 from training. Duration: 24.97775 2022-12-23 11:19:38,152 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0_sp0.9 from training. Duration: 23.3444375 2022-12-23 11:20:06,123 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0_sp0.9 from training. Duration: 23.2 2022-12-23 11:20:13,626 WARNING [train.py:1060] (3/4) Exclude cut with ID 5653-46179-0060-117930-0_sp0.9 from training. Duration: 21.17225 2022-12-23 11:20:22,314 INFO [train.py:894] (3/4) Epoch 17, batch 2800, loss[loss=0.2094, simple_loss=0.289, pruned_loss=0.06493, over 18689.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2748, pruned_loss=0.0579, over 3714538.32 frames. ], batch size: 78, lr: 6.72e-03, grad_scale: 8.0 2022-12-23 11:20:30,135 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp0.9 from training. Duration: 24.6555625 2022-12-23 11:20:33,426 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58906.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:21:17,313 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.392e+02 4.668e+02 5.683e+02 7.379e+02 2.321e+03, threshold=1.137e+03, percent-clipped=4.0 2022-12-23 11:21:26,366 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0 from training. Duration: 20.44 2022-12-23 11:21:26,782 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4523, 1.3438, 1.3834, 1.2821, 0.9344, 2.3288, 0.8439, 1.4564], device='cuda:3'), covar=tensor([0.3366, 0.2160, 0.2132, 0.2188, 0.1447, 0.0330, 0.1712, 0.0857], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0116, 0.0126, 0.0120, 0.0104, 0.0097, 0.0093, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 11:21:38,062 INFO [train.py:894] (3/4) Epoch 17, batch 2850, loss[loss=0.1999, simple_loss=0.2823, pruned_loss=0.05875, over 18462.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2741, pruned_loss=0.05759, over 3714222.54 frames. ], batch size: 50, lr: 6.72e-03, grad_scale: 8.0 2022-12-23 11:21:39,653 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0_sp0.9 from training. Duration: 23.45 2022-12-23 11:22:07,274 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58967.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:22:11,656 WARNING [train.py:1060] (3/4) Exclude cut with ID 6945-60535-0076-12784-0_sp0.9 from training. Duration: 20.52225 2022-12-23 11:22:20,680 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0 from training. Duration: 22.19 2022-12-23 11:22:31,505 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp1.1 from training. Duration: 25.3818125 2022-12-23 11:22:47,228 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp0.9 from training. Duration: 28.0944375 2022-12-23 11:22:52,961 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0_sp0.9 from training. Duration: 22.9444375 2022-12-23 11:22:53,137 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2284, 2.8167, 2.7077, 1.2424, 2.9466, 2.1363, 0.6265, 1.7488], device='cuda:3'), covar=tensor([0.2092, 0.1395, 0.1797, 0.3749, 0.1156, 0.1151, 0.4796, 0.1707], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0138, 0.0157, 0.0125, 0.0139, 0.0112, 0.0145, 0.0113], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 11:22:54,347 INFO [train.py:894] (3/4) Epoch 17, batch 2900, loss[loss=0.2122, simple_loss=0.2949, pruned_loss=0.06472, over 18585.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2734, pruned_loss=0.05681, over 3714490.02 frames. ], batch size: 57, lr: 6.72e-03, grad_scale: 8.0 2022-12-23 11:23:02,114 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp1.1 from training. Duration: 21.6318125 2022-12-23 11:23:20,856 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0 from training. Duration: 23.695 2022-12-23 11:23:48,109 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0 from training. Duration: 23.955 2022-12-23 11:23:49,461 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.824e+02 4.704e+02 5.257e+02 6.358e+02 1.532e+03, threshold=1.051e+03, percent-clipped=1.0 2022-12-23 11:23:52,509 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59037.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:24:09,850 INFO [train.py:894] (3/4) Epoch 17, batch 2950, loss[loss=0.1735, simple_loss=0.2528, pruned_loss=0.04712, over 18695.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2724, pruned_loss=0.05637, over 3714864.24 frames. ], batch size: 50, lr: 6.71e-03, grad_scale: 8.0 2022-12-23 11:24:20,957 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp0.9 from training. Duration: 26.438875 2022-12-23 11:25:06,033 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0021-26306-0_sp0.9 from training. Duration: 21.2444375 2022-12-23 11:25:06,071 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp0.9 from training. Duration: 31.02225 2022-12-23 11:25:14,852 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0 from training. Duration: 22.395 2022-12-23 11:25:28,221 INFO [train.py:894] (3/4) Epoch 17, batch 3000, loss[loss=0.2225, simple_loss=0.3079, pruned_loss=0.06858, over 18627.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2729, pruned_loss=0.05688, over 3715103.12 frames. ], batch size: 60, lr: 6.71e-03, grad_scale: 8.0 2022-12-23 11:25:28,221 INFO [train.py:919] (3/4) Computing validation loss 2022-12-23 11:25:39,070 INFO [train.py:928] (3/4) Epoch 17, validation: loss=0.1653, simple_loss=0.2636, pruned_loss=0.03352, over 944034.00 frames. 2022-12-23 11:25:39,072 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-23 11:25:40,598 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0 from training. Duration: 21.075 2022-12-23 11:25:44,321 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4368, 2.1702, 1.7600, 1.1499, 2.8275, 2.6086, 2.2278, 1.6135], device='cuda:3'), covar=tensor([0.0329, 0.0428, 0.0534, 0.0769, 0.0218, 0.0330, 0.0451, 0.0959], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0125, 0.0127, 0.0121, 0.0097, 0.0122, 0.0135, 0.0156], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 11:25:47,129 WARNING [train.py:1060] (3/4) Exclude cut with ID 6482-98857-0025-147532-0_sp0.9 from training. Duration: 20.0055625 2022-12-23 11:25:47,143 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0037-132304-0_sp0.9 from training. Duration: 22.05 2022-12-23 11:25:47,154 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0 from training. Duration: 26.8349375 2022-12-23 11:25:48,090 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.2356, 3.0763, 1.8583, 1.4538, 3.6215, 3.8070, 3.0852, 2.3762], device='cuda:3'), covar=tensor([0.0341, 0.0323, 0.0575, 0.0722, 0.0186, 0.0276, 0.0436, 0.0725], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0125, 0.0128, 0.0121, 0.0097, 0.0122, 0.0135, 0.0156], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 11:25:50,602 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp1.1 from training. Duration: 22.1090625 2022-12-23 11:25:58,378 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp0.9 from training. Duration: 26.6166875 2022-12-23 11:26:14,724 WARNING [train.py:1060] (3/4) Exclude cut with ID 2046-178027-0000-53705-0_sp0.9 from training. Duration: 20.3055625 2022-12-23 11:26:28,237 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59131.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:26:34,377 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.258e+02 4.780e+02 5.724e+02 6.904e+02 1.208e+03, threshold=1.145e+03, percent-clipped=1.0 2022-12-23 11:26:40,157 WARNING [train.py:1060] (3/4) Exclude cut with ID 7205-50138-0008-5373-0_sp0.9 from training. Duration: 20.7 2022-12-23 11:26:56,020 INFO [train.py:894] (3/4) Epoch 17, batch 3050, loss[loss=0.1582, simple_loss=0.2413, pruned_loss=0.03752, over 18606.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2725, pruned_loss=0.05655, over 3713503.73 frames. ], batch size: 45, lr: 6.71e-03, grad_scale: 8.0 2022-12-23 11:26:56,434 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.8049, 2.5108, 2.0967, 0.9963, 1.9489, 2.1683, 1.8350, 2.1621], device='cuda:3'), covar=tensor([0.0628, 0.0557, 0.1326, 0.1784, 0.1366, 0.1393, 0.1587, 0.0900], device='cuda:3'), in_proj_covar=tensor([0.0168, 0.0180, 0.0201, 0.0189, 0.0206, 0.0197, 0.0210, 0.0198], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 11:27:18,596 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59164.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:27:22,934 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0 from training. Duration: 22.48 2022-12-23 11:27:38,544 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp0.9 from training. Duration: 29.816625 2022-12-23 11:27:58,608 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp1.1 from training. Duration: 22.7590625 2022-12-23 11:28:03,415 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0 from training. Duration: 22.555 2022-12-23 11:28:12,953 INFO [train.py:894] (3/4) Epoch 17, batch 3100, loss[loss=0.194, simple_loss=0.2806, pruned_loss=0.05371, over 18592.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2728, pruned_loss=0.05663, over 3713070.33 frames. ], batch size: 56, lr: 6.71e-03, grad_scale: 8.0 2022-12-23 11:28:26,561 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0005-25975-0_sp0.9 from training. Duration: 21.688875 2022-12-23 11:28:53,006 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59225.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:28:57,730 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0_sp0.9 from training. Duration: 22.6 2022-12-23 11:29:08,316 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.058e+02 4.490e+02 5.519e+02 6.910e+02 1.551e+03, threshold=1.104e+03, percent-clipped=2.0 2022-12-23 11:29:13,918 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5322, 2.0351, 1.4855, 2.5540, 2.5174, 1.5888, 1.6014, 1.2563], device='cuda:3'), covar=tensor([0.2132, 0.1745, 0.1754, 0.0948, 0.1524, 0.1303, 0.2201, 0.1712], device='cuda:3'), in_proj_covar=tensor([0.0244, 0.0222, 0.0209, 0.0194, 0.0258, 0.0193, 0.0220, 0.0196], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 11:29:30,672 INFO [train.py:894] (3/4) Epoch 17, batch 3150, loss[loss=0.1587, simple_loss=0.2351, pruned_loss=0.04121, over 18482.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2731, pruned_loss=0.05681, over 3714015.35 frames. ], batch size: 43, lr: 6.70e-03, grad_scale: 16.0 2022-12-23 11:29:35,420 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0 from training. Duration: 24.32 2022-12-23 11:29:40,149 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0069, 1.9004, 2.1707, 1.2838, 2.2925, 2.3349, 1.5691, 2.6860], device='cuda:3'), covar=tensor([0.1109, 0.1813, 0.1358, 0.1930, 0.0717, 0.1225, 0.2260, 0.0452], device='cuda:3'), in_proj_covar=tensor([0.0195, 0.0208, 0.0203, 0.0193, 0.0174, 0.0215, 0.0212, 0.0187], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 11:29:50,232 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59262.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:30:16,713 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8221, 1.5039, 1.8833, 2.2992, 1.9946, 4.4356, 1.5053, 1.7398], device='cuda:3'), covar=tensor([0.1039, 0.2459, 0.1199, 0.1114, 0.1719, 0.0245, 0.1986, 0.2167], device='cuda:3'), in_proj_covar=tensor([0.0072, 0.0081, 0.0071, 0.0073, 0.0089, 0.0074, 0.0084, 0.0076], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 11:30:35,059 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-276745-0093-13116-0_sp0.9 from training. Duration: 21.061125 2022-12-23 11:30:47,416 INFO [train.py:894] (3/4) Epoch 17, batch 3200, loss[loss=0.1944, simple_loss=0.2787, pruned_loss=0.05507, over 18556.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2728, pruned_loss=0.05666, over 3713800.76 frames. ], batch size: 57, lr: 6.70e-03, grad_scale: 16.0 2022-12-23 11:30:47,459 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0024-15855-0_sp0.9 from training. Duration: 20.32225 2022-12-23 11:30:59,438 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp1.1 from training. Duration: 0.7545625 2022-12-23 11:31:16,207 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0_sp0.9 from training. Duration: 23.9333125 2022-12-23 11:31:42,114 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.749e+02 4.792e+02 5.650e+02 6.896e+02 1.712e+03, threshold=1.130e+03, percent-clipped=1.0 2022-12-23 11:31:45,480 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59337.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:31:49,527 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp1.1 from training. Duration: 20.17275 2022-12-23 11:31:53,652 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp1.1 from training. Duration: 20.436375 2022-12-23 11:31:54,799 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2022-12-23 11:32:02,422 INFO [train.py:894] (3/4) Epoch 17, batch 3250, loss[loss=0.2193, simple_loss=0.3007, pruned_loss=0.06897, over 18467.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2726, pruned_loss=0.05636, over 3712813.72 frames. ], batch size: 54, lr: 6.70e-03, grad_scale: 16.0 2022-12-23 11:32:49,992 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9408, 0.6952, 1.6618, 1.5393, 1.8366, 1.8453, 1.4789, 1.5563], device='cuda:3'), covar=tensor([0.1923, 0.3014, 0.2260, 0.2466, 0.1694, 0.0929, 0.2652, 0.1218], device='cuda:3'), in_proj_covar=tensor([0.0263, 0.0296, 0.0275, 0.0310, 0.0299, 0.0247, 0.0331, 0.0235], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 11:32:58,996 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=59385.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:33:17,367 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2022-12-23 11:33:17,640 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0_sp0.9 from training. Duration: 23.1055625 2022-12-23 11:33:19,053 INFO [train.py:894] (3/4) Epoch 17, batch 3300, loss[loss=0.1693, simple_loss=0.2595, pruned_loss=0.03955, over 18447.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2711, pruned_loss=0.05578, over 3712672.67 frames. ], batch size: 50, lr: 6.69e-03, grad_scale: 16.0 2022-12-23 11:33:20,668 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp1.1 from training. Duration: 23.5 2022-12-23 11:33:31,105 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp0.9 from training. Duration: 26.62775 2022-12-23 11:33:43,530 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0 from training. Duration: 21.105 2022-12-23 11:33:48,320 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0_sp0.9 from training. Duration: 24.411125 2022-12-23 11:34:03,963 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.90 vs. limit=5.0 2022-12-23 11:34:09,096 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59431.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:34:14,484 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.836e+02 4.538e+02 5.968e+02 7.307e+02 1.624e+03, threshold=1.194e+03, percent-clipped=6.0 2022-12-23 11:34:17,545 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp1.1 from training. Duration: 21.263625 2022-12-23 11:34:35,003 INFO [train.py:894] (3/4) Epoch 17, batch 3350, loss[loss=0.2047, simple_loss=0.2795, pruned_loss=0.06493, over 18573.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2707, pruned_loss=0.05561, over 3713215.99 frames. ], batch size: 49, lr: 6.69e-03, grad_scale: 16.0 2022-12-23 11:34:36,951 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2829, 1.9819, 1.7707, 1.2861, 2.5660, 2.4000, 2.1643, 1.5891], device='cuda:3'), covar=tensor([0.0376, 0.0440, 0.0496, 0.0743, 0.0262, 0.0350, 0.0451, 0.0956], device='cuda:3'), in_proj_covar=tensor([0.0122, 0.0125, 0.0127, 0.0120, 0.0096, 0.0122, 0.0135, 0.0156], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 11:34:51,699 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0 from training. Duration: 20.795 2022-12-23 11:35:01,967 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0 from training. Duration: 24.76 2022-12-23 11:35:01,986 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0_sp0.9 from training. Duration: 22.25 2022-12-23 11:35:21,133 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=59479.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:35:28,643 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp1.1 from training. Duration: 20.5045625 2022-12-23 11:35:50,236 INFO [train.py:894] (3/4) Epoch 17, batch 3400, loss[loss=0.1999, simple_loss=0.2824, pruned_loss=0.05874, over 18494.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2705, pruned_loss=0.05531, over 3712927.48 frames. ], batch size: 52, lr: 6.69e-03, grad_scale: 16.0 2022-12-23 11:35:54,992 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59502.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:36:21,733 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59520.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:36:43,135 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.219e+02 4.503e+02 5.475e+02 7.433e+02 1.825e+03, threshold=1.095e+03, percent-clipped=2.0 2022-12-23 11:37:02,780 INFO [train.py:894] (3/4) Epoch 17, batch 3450, loss[loss=0.2151, simple_loss=0.2913, pruned_loss=0.06945, over 18539.00 frames. ], tot_loss[loss=0.1907, simple_loss=0.2704, pruned_loss=0.05545, over 3713386.33 frames. ], batch size: 58, lr: 6.69e-03, grad_scale: 16.0 2022-12-23 11:37:23,168 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59562.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:37:24,675 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59563.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:37:31,301 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4602, 1.1776, 1.6954, 2.6443, 1.9323, 2.2960, 0.8674, 1.8984], device='cuda:3'), covar=tensor([0.1831, 0.1841, 0.1382, 0.0724, 0.1098, 0.1059, 0.2099, 0.1232], device='cuda:3'), in_proj_covar=tensor([0.0102, 0.0116, 0.0133, 0.0145, 0.0106, 0.0139, 0.0129, 0.0112], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 11:38:17,578 INFO [train.py:894] (3/4) Epoch 17, batch 3500, loss[loss=0.2105, simple_loss=0.2856, pruned_loss=0.06766, over 18650.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2719, pruned_loss=0.0562, over 3713378.59 frames. ], batch size: 184, lr: 6.68e-03, grad_scale: 16.0 2022-12-23 11:38:39,698 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp1.1 from training. Duration: 22.2954375 2022-12-23 11:38:46,853 INFO [train.py:894] (3/4) Epoch 18, batch 0, loss[loss=0.1841, simple_loss=0.2772, pruned_loss=0.04547, over 18603.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2772, pruned_loss=0.04547, over 18603.00 frames. ], batch size: 56, lr: 6.49e-03, grad_scale: 16.0 2022-12-23 11:38:46,854 INFO [train.py:919] (3/4) Computing validation loss 2022-12-23 11:38:50,086 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.1806, 1.2586, 1.4232, 0.8381, 1.3320, 1.3559, 1.0848, 1.5174], device='cuda:3'), covar=tensor([0.1113, 0.2243, 0.1296, 0.1749, 0.0793, 0.1107, 0.2758, 0.0667], device='cuda:3'), in_proj_covar=tensor([0.0197, 0.0210, 0.0205, 0.0195, 0.0174, 0.0216, 0.0213, 0.0191], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 11:38:57,548 INFO [train.py:928] (3/4) Epoch 18, validation: loss=0.1653, simple_loss=0.2643, pruned_loss=0.03315, over 944034.00 frames. 2022-12-23 11:38:57,549 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-23 11:39:04,893 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=59610.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:39:09,466 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3345, 2.1557, 1.6906, 1.2600, 2.7379, 2.4855, 2.1487, 1.6495], device='cuda:3'), covar=tensor([0.0378, 0.0401, 0.0554, 0.0788, 0.0232, 0.0379, 0.0469, 0.0880], device='cuda:3'), in_proj_covar=tensor([0.0121, 0.0124, 0.0124, 0.0118, 0.0095, 0.0120, 0.0133, 0.0155], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 11:39:39,721 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2022-12-23 11:39:43,082 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.453e+02 4.121e+02 4.989e+02 6.015e+02 1.398e+03, threshold=9.978e+02, percent-clipped=2.0 2022-12-23 11:39:48,897 WARNING [train.py:1060] (3/4) Exclude cut with ID 298-126791-0067-24026-0_sp0.9 from training. Duration: 21.438875 2022-12-23 11:39:53,362 WARNING [train.py:1060] (3/4) Exclude cut with ID 5652-39938-0025-23684-0_sp0.9 from training. Duration: 22.2055625 2022-12-23 11:40:04,572 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4747, 2.1724, 1.9534, 1.9170, 2.5637, 2.9128, 2.7740, 2.1371], device='cuda:3'), covar=tensor([0.0318, 0.0254, 0.0363, 0.0242, 0.0199, 0.0320, 0.0280, 0.0289], device='cuda:3'), in_proj_covar=tensor([0.0092, 0.0123, 0.0149, 0.0123, 0.0115, 0.0117, 0.0096, 0.0125], device='cuda:3'), out_proj_covar=tensor([7.3665e-05, 9.8718e-05, 1.2434e-04, 9.9297e-05, 9.4290e-05, 9.0599e-05, 7.5684e-05, 9.9888e-05], device='cuda:3') 2022-12-23 11:40:12,953 INFO [train.py:894] (3/4) Epoch 18, batch 50, loss[loss=0.1662, simple_loss=0.2415, pruned_loss=0.04539, over 18472.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.266, pruned_loss=0.04805, over 838320.44 frames. ], batch size: 43, lr: 6.49e-03, grad_scale: 16.0 2022-12-23 11:40:52,773 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8715, 0.7037, 1.7402, 1.5254, 1.9105, 1.9193, 1.4589, 1.6568], device='cuda:3'), covar=tensor([0.2077, 0.3087, 0.2417, 0.2413, 0.1874, 0.0924, 0.2749, 0.1186], device='cuda:3'), in_proj_covar=tensor([0.0264, 0.0295, 0.0275, 0.0311, 0.0300, 0.0247, 0.0331, 0.0235], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 11:41:29,452 INFO [train.py:894] (3/4) Epoch 18, batch 100, loss[loss=0.1978, simple_loss=0.2784, pruned_loss=0.05864, over 18500.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2667, pruned_loss=0.04711, over 1475789.75 frames. ], batch size: 58, lr: 6.49e-03, grad_scale: 16.0 2022-12-23 11:42:02,143 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6925, 1.4550, 1.5196, 2.0020, 1.7958, 3.1516, 1.4250, 1.5594], device='cuda:3'), covar=tensor([0.0818, 0.1765, 0.1152, 0.0842, 0.1335, 0.0257, 0.1388, 0.1455], device='cuda:3'), in_proj_covar=tensor([0.0072, 0.0081, 0.0072, 0.0073, 0.0089, 0.0073, 0.0083, 0.0076], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 11:42:14,867 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.193e+02 3.350e+02 4.029e+02 4.665e+02 1.011e+03, threshold=8.057e+02, percent-clipped=1.0 2022-12-23 11:42:44,687 INFO [train.py:894] (3/4) Epoch 18, batch 150, loss[loss=0.1565, simple_loss=0.2479, pruned_loss=0.03257, over 18464.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2647, pruned_loss=0.04556, over 1971604.37 frames. ], batch size: 50, lr: 6.48e-03, grad_scale: 16.0 2022-12-23 11:42:46,667 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0399, 1.2989, 1.7786, 1.6483, 2.0212, 2.0429, 1.7807, 1.6271], device='cuda:3'), covar=tensor([0.2248, 0.3309, 0.2453, 0.2916, 0.1987, 0.0941, 0.3122, 0.1265], device='cuda:3'), in_proj_covar=tensor([0.0263, 0.0295, 0.0275, 0.0311, 0.0299, 0.0247, 0.0331, 0.0234], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 11:42:55,548 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0 from training. Duration: 24.525 2022-12-23 11:43:05,664 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7598, 2.1572, 2.5095, 2.3708, 2.5559, 2.6192, 2.5548, 2.0427], device='cuda:3'), covar=tensor([0.1936, 0.2956, 0.2016, 0.2636, 0.1803, 0.0793, 0.2993, 0.1133], device='cuda:3'), in_proj_covar=tensor([0.0263, 0.0295, 0.0274, 0.0311, 0.0299, 0.0247, 0.0331, 0.0235], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 11:43:30,690 WARNING [train.py:1060] (3/4) Exclude cut with ID 3699-47246-0007-3408-0_sp0.9 from training. Duration: 20.26675 2022-12-23 11:43:37,217 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2022-12-23 11:43:43,504 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp0.9 from training. Duration: 27.25 2022-12-23 11:44:01,081 INFO [train.py:894] (3/4) Epoch 18, batch 200, loss[loss=0.1608, simple_loss=0.2399, pruned_loss=0.04086, over 18595.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2632, pruned_loss=0.04529, over 2358331.36 frames. ], batch size: 45, lr: 6.48e-03, grad_scale: 16.0 2022-12-23 11:44:07,674 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2022-12-23 11:44:25,632 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59820.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:44:47,319 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.080e+02 3.472e+02 4.148e+02 5.017e+02 1.205e+03, threshold=8.296e+02, percent-clipped=2.0 2022-12-23 11:44:57,302 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0 from training. Duration: 21.68 2022-12-23 11:45:06,657 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59848.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:45:07,902 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0 from training. Duration: 21.6300625 2022-12-23 11:45:16,504 INFO [train.py:894] (3/4) Epoch 18, batch 250, loss[loss=0.1786, simple_loss=0.2678, pruned_loss=0.04466, over 18715.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2636, pruned_loss=0.04544, over 2658710.50 frames. ], batch size: 78, lr: 6.48e-03, grad_scale: 16.0 2022-12-23 11:45:21,531 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59858.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:45:32,278 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0_sp0.9 from training. Duration: 24.033375 2022-12-23 11:45:36,668 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=59868.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:46:01,631 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.3383, 3.7553, 3.7863, 4.3331, 3.9674, 3.8801, 4.5417, 1.4458], device='cuda:3'), covar=tensor([0.0776, 0.0751, 0.0673, 0.0727, 0.1444, 0.1161, 0.0591, 0.4944], device='cuda:3'), in_proj_covar=tensor([0.0333, 0.0217, 0.0229, 0.0254, 0.0310, 0.0258, 0.0278, 0.0271], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 11:46:07,742 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59889.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:46:27,794 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0 from training. Duration: 22.905 2022-12-23 11:46:29,288 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp1.1 from training. Duration: 23.4318125 2022-12-23 11:46:30,710 INFO [train.py:894] (3/4) Epoch 18, batch 300, loss[loss=0.1946, simple_loss=0.286, pruned_loss=0.05167, over 18720.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2633, pruned_loss=0.0447, over 2892898.97 frames. ], batch size: 62, lr: 6.48e-03, grad_scale: 16.0 2022-12-23 11:46:38,740 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59909.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:46:58,100 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7382, 1.6712, 1.7356, 0.9729, 1.7542, 1.8667, 1.4220, 2.1965], device='cuda:3'), covar=tensor([0.0995, 0.1964, 0.1243, 0.1830, 0.0709, 0.1165, 0.2247, 0.0520], device='cuda:3'), in_proj_covar=tensor([0.0194, 0.0207, 0.0203, 0.0193, 0.0172, 0.0215, 0.0209, 0.0188], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 11:47:05,355 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7589, 2.4212, 2.0666, 0.9936, 2.0075, 2.0390, 1.7523, 2.1883], device='cuda:3'), covar=tensor([0.0579, 0.0529, 0.1094, 0.1566, 0.1202, 0.1397, 0.1552, 0.0706], device='cuda:3'), in_proj_covar=tensor([0.0165, 0.0179, 0.0200, 0.0187, 0.0203, 0.0195, 0.0207, 0.0194], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 11:47:16,512 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.844e+02 3.499e+02 4.281e+02 5.083e+02 1.083e+03, threshold=8.562e+02, percent-clipped=4.0 2022-12-23 11:47:33,704 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0538, 1.1163, 1.8403, 1.6681, 2.0712, 2.0841, 1.7346, 1.7432], device='cuda:3'), covar=tensor([0.1970, 0.2913, 0.2315, 0.2557, 0.1926, 0.0839, 0.2849, 0.1157], device='cuda:3'), in_proj_covar=tensor([0.0260, 0.0293, 0.0272, 0.0308, 0.0298, 0.0245, 0.0330, 0.0232], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 11:47:39,592 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59950.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:47:47,291 INFO [train.py:894] (3/4) Epoch 18, batch 350, loss[loss=0.1972, simple_loss=0.2897, pruned_loss=0.05238, over 18496.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2649, pruned_loss=0.04529, over 3073883.41 frames. ], batch size: 77, lr: 6.47e-03, grad_scale: 16.0 2022-12-23 11:48:28,440 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp1.1 from training. Duration: 20.82275 2022-12-23 11:48:29,924 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp0.9 from training. Duration: 25.45 2022-12-23 11:48:48,014 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5826, 1.7324, 1.4549, 2.1203, 2.5352, 1.5352, 1.4173, 1.2963], device='cuda:3'), covar=tensor([0.1884, 0.1868, 0.1675, 0.1026, 0.1263, 0.1190, 0.2202, 0.1546], device='cuda:3'), in_proj_covar=tensor([0.0244, 0.0221, 0.0211, 0.0194, 0.0257, 0.0193, 0.0222, 0.0196], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 11:49:06,026 INFO [train.py:894] (3/4) Epoch 18, batch 400, loss[loss=0.1971, simple_loss=0.2913, pruned_loss=0.05144, over 18521.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2662, pruned_loss=0.04567, over 3216004.60 frames. ], batch size: 58, lr: 6.47e-03, grad_scale: 16.0 2022-12-23 11:49:33,114 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0 from training. Duration: 25.775 2022-12-23 11:49:50,716 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.438e+02 3.518e+02 4.134e+02 5.271e+02 1.666e+03, threshold=8.268e+02, percent-clipped=1.0 2022-12-23 11:49:55,870 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0_sp0.9 from training. Duration: 22.25 2022-12-23 11:49:59,084 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60040.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:50:17,507 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60052.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:50:21,183 INFO [train.py:894] (3/4) Epoch 18, batch 450, loss[loss=0.2387, simple_loss=0.3126, pruned_loss=0.08238, over 18580.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2671, pruned_loss=0.0462, over 3326483.91 frames. ], batch size: 57, lr: 6.47e-03, grad_scale: 16.0 2022-12-23 11:50:21,294 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0 from training. Duration: 26.205 2022-12-23 11:50:37,101 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60065.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:50:38,084 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp0.9 from training. Duration: 30.1555625 2022-12-23 11:50:42,910 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2022-12-23 11:50:43,773 WARNING [train.py:1060] (3/4) Exclude cut with ID 1265-135635-0050-6781-0_sp0.9 from training. Duration: 21.8333125 2022-12-23 11:50:52,470 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp1.1 from training. Duration: 20.6545625 2022-12-23 11:50:55,665 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60078.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:51:30,783 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60101.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 11:51:31,927 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0045-39920-0_sp0.9 from training. Duration: 20.52225 2022-12-23 11:51:37,528 INFO [train.py:894] (3/4) Epoch 18, batch 500, loss[loss=0.1911, simple_loss=0.2821, pruned_loss=0.05004, over 18680.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2685, pruned_loss=0.04679, over 3412971.98 frames. ], batch size: 78, lr: 6.46e-03, grad_scale: 16.0 2022-12-23 11:51:49,236 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60113.0, num_to_drop=1, layers_to_drop={3} 2022-12-23 11:51:54,235 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp0.9 from training. Duration: 29.1166875 2022-12-23 11:52:07,618 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60126.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:52:20,897 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.520e+02 3.872e+02 4.827e+02 5.955e+02 1.289e+03, threshold=9.654e+02, percent-clipped=8.0 2022-12-23 11:52:27,199 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60139.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:52:50,639 INFO [train.py:894] (3/4) Epoch 18, batch 550, loss[loss=0.2027, simple_loss=0.2896, pruned_loss=0.0579, over 18473.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2701, pruned_loss=0.04742, over 3480014.79 frames. ], batch size: 54, lr: 6.46e-03, grad_scale: 16.0 2022-12-23 11:52:53,593 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133211-0007-59831-0_sp0.9 from training. Duration: 21.388875 2022-12-23 11:52:55,477 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60158.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:53:29,331 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0 from training. Duration: 22.72 2022-12-23 11:53:30,824 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0_sp0.9 from training. Duration: 22.7444375 2022-12-23 11:53:55,207 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60198.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:54:04,702 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60204.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:54:06,006 INFO [train.py:894] (3/4) Epoch 18, batch 600, loss[loss=0.189, simple_loss=0.2811, pruned_loss=0.04851, over 18666.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2697, pruned_loss=0.0473, over 3531590.27 frames. ], batch size: 62, lr: 6.46e-03, grad_scale: 16.0 2022-12-23 11:54:07,613 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=60206.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:54:13,491 WARNING [train.py:1060] (3/4) Exclude cut with ID 4133-6541-0027-40495-0_sp1.1 from training. Duration: 0.9681875 2022-12-23 11:54:16,350 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0_sp0.9 from training. Duration: 22.3166875 2022-12-23 11:54:22,189 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133212-0015-59917-0_sp0.9 from training. Duration: 21.8166875 2022-12-23 11:54:50,621 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.513e+02 3.538e+02 4.331e+02 5.491e+02 1.076e+03, threshold=8.662e+02, percent-clipped=2.0 2022-12-23 11:55:05,749 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60245.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:55:21,585 INFO [train.py:894] (3/4) Epoch 18, batch 650, loss[loss=0.1718, simple_loss=0.2698, pruned_loss=0.03691, over 18584.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2697, pruned_loss=0.04739, over 3571654.92 frames. ], batch size: 51, lr: 6.46e-03, grad_scale: 16.0 2022-12-23 11:55:27,694 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60259.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:55:49,732 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9287, 2.4673, 1.7658, 2.8295, 3.4361, 1.9101, 2.1257, 1.4651], device='cuda:3'), covar=tensor([0.1737, 0.1502, 0.1406, 0.0839, 0.1077, 0.1012, 0.1689, 0.1412], device='cuda:3'), in_proj_covar=tensor([0.0242, 0.0219, 0.0210, 0.0193, 0.0255, 0.0193, 0.0219, 0.0194], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 11:55:59,603 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2022-12-23 11:56:04,641 WARNING [train.py:1060] (3/4) Exclude cut with ID 4957-30119-0041-23990-0_sp0.9 from training. Duration: 20.22775 2022-12-23 11:56:36,962 INFO [train.py:894] (3/4) Epoch 18, batch 700, loss[loss=0.1751, simple_loss=0.2643, pruned_loss=0.04288, over 18519.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2697, pruned_loss=0.04741, over 3602898.06 frames. ], batch size: 52, lr: 6.45e-03, grad_scale: 16.0 2022-12-23 11:56:46,982 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp1.1 from training. Duration: 24.67275 2022-12-23 11:56:47,222 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4962, 0.9933, 0.6956, 1.0621, 1.9457, 0.6417, 1.0995, 1.2221], device='cuda:3'), covar=tensor([0.1701, 0.2378, 0.2011, 0.1679, 0.1654, 0.1784, 0.1639, 0.1981], device='cuda:3'), in_proj_covar=tensor([0.0093, 0.0097, 0.0117, 0.0094, 0.0115, 0.0091, 0.0098, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 11:57:01,261 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.48 vs. limit=5.0 2022-12-23 11:57:15,071 WARNING [train.py:1060] (3/4) Exclude cut with ID 3082-165428-0081-50734-0_sp0.9 from training. Duration: 21.8055625 2022-12-23 11:57:20,766 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.961e+02 3.464e+02 4.226e+02 5.168e+02 1.038e+03, threshold=8.451e+02, percent-clipped=1.0 2022-12-23 11:57:22,627 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4171, 1.1444, 1.7504, 2.6994, 1.8886, 2.2539, 1.2163, 1.8021], device='cuda:3'), covar=tensor([0.1840, 0.1730, 0.1317, 0.0656, 0.1136, 0.1051, 0.1790, 0.1241], device='cuda:3'), in_proj_covar=tensor([0.0099, 0.0113, 0.0130, 0.0141, 0.0104, 0.0135, 0.0127, 0.0108], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 11:57:51,731 INFO [train.py:894] (3/4) Epoch 18, batch 750, loss[loss=0.1583, simple_loss=0.2327, pruned_loss=0.04193, over 18410.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2689, pruned_loss=0.04717, over 3626166.59 frames. ], batch size: 42, lr: 6.45e-03, grad_scale: 16.0 2022-12-23 11:57:51,764 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0_sp0.9 from training. Duration: 22.6666875 2022-12-23 11:58:22,960 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2022-12-23 11:58:39,367 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2022-12-23 11:58:52,754 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.0158, 5.4931, 4.8034, 2.4059, 5.5045, 4.0568, 0.7896, 3.6460], device='cuda:3'), covar=tensor([0.1766, 0.0933, 0.1204, 0.3152, 0.0584, 0.0776, 0.4908, 0.1282], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0136, 0.0155, 0.0122, 0.0138, 0.0110, 0.0141, 0.0112], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 11:58:52,762 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60396.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 11:58:56,806 WARNING [train.py:1060] (3/4) Exclude cut with ID 2411-132532-0017-82279-0_sp1.1 from training. Duration: 0.9681875 2022-12-23 11:59:05,837 INFO [train.py:894] (3/4) Epoch 18, batch 800, loss[loss=0.1718, simple_loss=0.2658, pruned_loss=0.03892, over 18530.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2686, pruned_loss=0.04689, over 3645123.77 frames. ], batch size: 55, lr: 6.45e-03, grad_scale: 16.0 2022-12-23 11:59:10,497 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60408.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 11:59:22,533 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9705, 1.9656, 1.5010, 2.1215, 2.1448, 1.9312, 2.6872, 2.0665], device='cuda:3'), covar=tensor([0.0888, 0.1665, 0.2668, 0.1635, 0.1685, 0.0874, 0.0891, 0.1262], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0202, 0.0244, 0.0283, 0.0231, 0.0187, 0.0206, 0.0199], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 11:59:23,576 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0 from training. Duration: 22.485 2022-12-23 11:59:29,448 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60421.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:59:48,841 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60434.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 11:59:50,044 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.393e+02 3.573e+02 4.387e+02 5.388e+02 1.358e+03, threshold=8.775e+02, percent-clipped=3.0 2022-12-23 12:00:02,377 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp1.1 from training. Duration: 23.82275 2022-12-23 12:00:14,288 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0 from training. Duration: 20.77 2022-12-23 12:00:16,170 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.5348, 3.0306, 2.5667, 1.3077, 2.6632, 2.1532, 1.6922, 2.3551], device='cuda:3'), covar=tensor([0.0663, 0.0696, 0.1891, 0.2248, 0.1788, 0.2039, 0.2349, 0.1318], device='cuda:3'), in_proj_covar=tensor([0.0170, 0.0182, 0.0205, 0.0190, 0.0208, 0.0201, 0.0212, 0.0198], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 12:00:20,325 INFO [train.py:894] (3/4) Epoch 18, batch 850, loss[loss=0.2094, simple_loss=0.2863, pruned_loss=0.06623, over 18680.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2679, pruned_loss=0.04657, over 3660194.62 frames. ], batch size: 186, lr: 6.45e-03, grad_scale: 16.0 2022-12-23 12:00:20,329 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0_sp0.9 from training. Duration: 24.088875 2022-12-23 12:00:43,015 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4729, 1.6510, 1.8268, 1.1004, 1.8460, 1.7853, 1.2760, 2.1248], device='cuda:3'), covar=tensor([0.1122, 0.1805, 0.1128, 0.1719, 0.0666, 0.1086, 0.2439, 0.0559], device='cuda:3'), in_proj_covar=tensor([0.0196, 0.0208, 0.0205, 0.0195, 0.0173, 0.0214, 0.0210, 0.0191], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 12:00:50,364 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp1.1 from training. Duration: 20.4409375 2022-12-23 12:01:33,955 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60504.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:01:35,266 INFO [train.py:894] (3/4) Epoch 18, batch 900, loss[loss=0.1786, simple_loss=0.2716, pruned_loss=0.04284, over 18708.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2682, pruned_loss=0.04652, over 3673176.12 frames. ], batch size: 52, lr: 6.44e-03, grad_scale: 16.0 2022-12-23 12:01:43,142 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5818, 3.2001, 3.1204, 1.7146, 3.2743, 2.4669, 1.1473, 2.2186], device='cuda:3'), covar=tensor([0.2314, 0.1268, 0.1425, 0.2934, 0.0896, 0.0911, 0.3847, 0.1491], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0137, 0.0155, 0.0123, 0.0138, 0.0110, 0.0143, 0.0113], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 12:02:07,155 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0_sp0.9 from training. Duration: 22.511125 2022-12-23 12:02:08,546 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0 from training. Duration: 20.675 2022-12-23 12:02:20,164 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.232e+02 3.736e+02 4.464e+02 5.276e+02 1.113e+03, threshold=8.928e+02, percent-clipped=1.0 2022-12-23 12:02:36,072 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60545.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:02:45,829 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=60552.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:02:48,660 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60554.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:02:49,904 INFO [train.py:894] (3/4) Epoch 18, batch 950, loss[loss=0.1587, simple_loss=0.245, pruned_loss=0.0362, over 18532.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2681, pruned_loss=0.04633, over 3682908.17 frames. ], batch size: 44, lr: 6.44e-03, grad_scale: 16.0 2022-12-23 12:03:47,425 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp0.9 from training. Duration: 24.9833125 2022-12-23 12:03:47,541 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=60593.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:04:05,108 INFO [train.py:894] (3/4) Epoch 18, batch 1000, loss[loss=0.19, simple_loss=0.2746, pruned_loss=0.05274, over 18710.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2687, pruned_loss=0.04672, over 3689010.97 frames. ], batch size: 52, lr: 6.44e-03, grad_scale: 16.0 2022-12-23 12:04:16,393 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0 from training. Duration: 27.14 2022-12-23 12:04:32,412 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0 from training. Duration: 22.44 2022-12-23 12:04:48,806 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.280e+02 3.696e+02 4.340e+02 5.163e+02 1.095e+03, threshold=8.681e+02, percent-clipped=1.0 2022-12-23 12:05:19,537 INFO [train.py:894] (3/4) Epoch 18, batch 1050, loss[loss=0.1625, simple_loss=0.2548, pruned_loss=0.03509, over 18679.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2683, pruned_loss=0.04626, over 3694208.50 frames. ], batch size: 48, lr: 6.44e-03, grad_scale: 16.0 2022-12-23 12:05:22,981 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7516, 1.9795, 2.1452, 1.2375, 2.2271, 2.3513, 1.5262, 2.4950], device='cuda:3'), covar=tensor([0.1218, 0.1842, 0.1271, 0.1995, 0.0681, 0.1049, 0.2383, 0.0566], device='cuda:3'), in_proj_covar=tensor([0.0198, 0.0209, 0.0205, 0.0196, 0.0173, 0.0214, 0.0212, 0.0190], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 12:05:30,729 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2022-12-23 12:05:45,268 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5534, 2.1564, 1.8438, 2.3379, 1.9463, 2.1455, 2.0367, 2.4655], device='cuda:3'), covar=tensor([0.1997, 0.3138, 0.1824, 0.2549, 0.3451, 0.1052, 0.2799, 0.0878], device='cuda:3'), in_proj_covar=tensor([0.0290, 0.0283, 0.0239, 0.0343, 0.0265, 0.0224, 0.0281, 0.0207], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 12:05:47,849 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0060-62364-0_sp0.9 from training. Duration: 21.361125 2022-12-23 12:05:52,281 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp1.1 from training. Duration: 27.0318125 2022-12-23 12:06:04,056 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp0.9 from training. Duration: 28.638875 2022-12-23 12:06:19,134 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0 from training. Duration: 20.4 2022-12-23 12:06:22,233 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60696.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:06:34,839 INFO [train.py:894] (3/4) Epoch 18, batch 1100, loss[loss=0.1822, simple_loss=0.2695, pruned_loss=0.04742, over 18686.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2676, pruned_loss=0.04615, over 3699177.87 frames. ], batch size: 46, lr: 6.43e-03, grad_scale: 16.0 2022-12-23 12:06:39,666 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60708.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:06:51,842 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0 from training. Duration: 20.025 2022-12-23 12:06:51,856 WARNING [train.py:1060] (3/4) Exclude cut with ID 2364-131735-0112-64612-0_sp0.9 from training. Duration: 20.488875 2022-12-23 12:06:55,256 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60718.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:06:57,669 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0 from training. Duration: 29.735 2022-12-23 12:06:59,118 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60721.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:07:19,392 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60734.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:07:20,615 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.164e+02 3.194e+02 3.933e+02 5.010e+02 1.354e+03, threshold=7.866e+02, percent-clipped=2.0 2022-12-23 12:07:32,462 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7375, 1.2549, 0.5938, 1.2428, 2.0979, 1.1294, 1.4875, 1.6335], device='cuda:3'), covar=tensor([0.1637, 0.1962, 0.2339, 0.1560, 0.1695, 0.1724, 0.1478, 0.1645], device='cuda:3'), in_proj_covar=tensor([0.0093, 0.0097, 0.0116, 0.0095, 0.0114, 0.0091, 0.0098, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 12:07:33,610 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=60744.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:07:49,972 INFO [train.py:894] (3/4) Epoch 18, batch 1150, loss[loss=0.1935, simple_loss=0.278, pruned_loss=0.05453, over 18597.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2683, pruned_loss=0.04649, over 3703047.48 frames. ], batch size: 51, lr: 6.43e-03, grad_scale: 16.0 2022-12-23 12:07:51,621 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=60756.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:08:10,464 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=60769.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:08:20,313 WARNING [train.py:1060] (3/4) Exclude cut with ID 7276-92427-0014-12983-0_sp0.9 from training. Duration: 21.3055625 2022-12-23 12:08:21,827 WARNING [train.py:1060] (3/4) Exclude cut with ID 1025-75365-0008-79168-0_sp0.9 from training. Duration: 22.0666875 2022-12-23 12:08:25,808 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60779.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:08:29,425 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=60782.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:08:43,813 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2022-12-23 12:09:03,245 INFO [train.py:894] (3/4) Epoch 18, batch 1200, loss[loss=0.1727, simple_loss=0.265, pruned_loss=0.04017, over 18609.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2684, pruned_loss=0.04631, over 3705683.42 frames. ], batch size: 69, lr: 6.43e-03, grad_scale: 16.0 2022-12-23 12:09:48,496 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.177e+02 3.323e+02 4.383e+02 5.760e+02 1.405e+03, threshold=8.765e+02, percent-clipped=5.0 2022-12-23 12:10:12,644 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0 from training. Duration: 20.26 2022-12-23 12:10:15,910 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60854.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:10:17,018 INFO [train.py:894] (3/4) Epoch 18, batch 1250, loss[loss=0.1668, simple_loss=0.2423, pruned_loss=0.04563, over 18433.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2676, pruned_loss=0.04599, over 3706654.09 frames. ], batch size: 42, lr: 6.43e-03, grad_scale: 16.0 2022-12-23 12:10:27,059 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0030-9324-0_sp0.9 from training. Duration: 21.3444375 2022-12-23 12:11:24,449 WARNING [train.py:1060] (3/4) Exclude cut with ID 497-129325-0061-62254-0_sp1.1 from training. Duration: 0.97725 2022-12-23 12:11:27,562 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=60902.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:11:32,086 INFO [train.py:894] (3/4) Epoch 18, batch 1300, loss[loss=0.2099, simple_loss=0.298, pruned_loss=0.06093, over 18690.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2676, pruned_loss=0.0459, over 3709451.13 frames. ], batch size: 97, lr: 6.42e-03, grad_scale: 16.0 2022-12-23 12:12:06,062 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6377, 2.1766, 1.6888, 2.4590, 1.9102, 2.1800, 2.0376, 2.7046], device='cuda:3'), covar=tensor([0.1901, 0.3227, 0.1860, 0.2637, 0.3646, 0.0942, 0.3021, 0.0799], device='cuda:3'), in_proj_covar=tensor([0.0290, 0.0284, 0.0239, 0.0343, 0.0266, 0.0223, 0.0282, 0.0207], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 12:12:06,931 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0_sp0.9 from training. Duration: 22.97225 2022-12-23 12:12:09,323 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9487, 1.8343, 1.4951, 1.7530, 1.6743, 1.7988, 1.6798, 1.8215], device='cuda:3'), covar=tensor([0.2086, 0.2946, 0.1925, 0.2256, 0.3175, 0.1042, 0.2630, 0.0936], device='cuda:3'), in_proj_covar=tensor([0.0290, 0.0284, 0.0239, 0.0343, 0.0266, 0.0223, 0.0282, 0.0207], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 12:12:16,807 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.396e+02 3.264e+02 4.038e+02 4.760e+02 9.694e+02, threshold=8.075e+02, percent-clipped=2.0 2022-12-23 12:12:36,646 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0047-39922-0_sp0.9 from training. Duration: 21.97775 2022-12-23 12:12:46,595 INFO [train.py:894] (3/4) Epoch 18, batch 1350, loss[loss=0.1976, simple_loss=0.2895, pruned_loss=0.05285, over 18575.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2682, pruned_loss=0.04615, over 3710992.63 frames. ], batch size: 57, lr: 6.42e-03, grad_scale: 16.0 2022-12-23 12:12:49,602 WARNING [train.py:1060] (3/4) Exclude cut with ID 1112-1043-0006-89194-0_sp0.9 from training. Duration: 21.8333125 2022-12-23 12:13:00,725 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0 from training. Duration: 20.47 2022-12-23 12:13:52,037 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2022-12-23 12:14:01,335 INFO [train.py:894] (3/4) Epoch 18, batch 1400, loss[loss=0.1771, simple_loss=0.2743, pruned_loss=0.03993, over 18724.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2666, pruned_loss=0.04541, over 3711090.60 frames. ], batch size: 54, lr: 6.42e-03, grad_scale: 16.0 2022-12-23 12:14:08,812 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0037-39912-0_sp0.9 from training. Duration: 20.67225 2022-12-23 12:14:27,317 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp0.9 from training. Duration: 25.2444375 2022-12-23 12:14:48,190 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.343e+02 3.490e+02 4.029e+02 5.313e+02 7.188e+02, threshold=8.059e+02, percent-clipped=0.0 2022-12-23 12:14:51,162 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0021-76797-0_sp0.9 from training. Duration: 21.1445 2022-12-23 12:15:17,586 INFO [train.py:894] (3/4) Epoch 18, batch 1450, loss[loss=0.2114, simple_loss=0.2946, pruned_loss=0.06415, over 18380.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2672, pruned_loss=0.04583, over 3711961.68 frames. ], batch size: 51, lr: 6.41e-03, grad_scale: 16.0 2022-12-23 12:15:27,196 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61061.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:15:45,979 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2022-12-23 12:15:46,560 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=61074.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:16:03,110 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2022-12-23 12:16:05,179 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp0.9 from training. Duration: 33.038875 2022-12-23 12:16:32,532 INFO [train.py:894] (3/4) Epoch 18, batch 1500, loss[loss=0.1593, simple_loss=0.2395, pruned_loss=0.03956, over 18666.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2664, pruned_loss=0.04565, over 3712165.30 frames. ], batch size: 48, lr: 6.41e-03, grad_scale: 16.0 2022-12-23 12:16:42,806 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64291-0000-16059-0_sp0.9 from training. Duration: 20.0944375 2022-12-23 12:16:55,146 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp1.1 from training. Duration: 20.4 2022-12-23 12:16:59,065 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61122.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:17:03,286 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0 from training. Duration: 20.085 2022-12-23 12:17:14,135 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0_sp0.9 from training. Duration: 23.07775 2022-12-23 12:17:18,337 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.189e+02 3.774e+02 4.544e+02 5.492e+02 1.131e+03, threshold=9.087e+02, percent-clipped=3.0 2022-12-23 12:17:28,938 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([5.8047, 4.9441, 5.0715, 5.8217, 5.3329, 5.1167, 5.8670, 1.6268], device='cuda:3'), covar=tensor([0.0590, 0.0568, 0.0520, 0.0610, 0.1222, 0.1071, 0.0425, 0.4843], device='cuda:3'), in_proj_covar=tensor([0.0330, 0.0215, 0.0229, 0.0251, 0.0309, 0.0260, 0.0278, 0.0272], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 12:17:48,325 INFO [train.py:894] (3/4) Epoch 18, batch 1550, loss[loss=0.1772, simple_loss=0.2682, pruned_loss=0.04315, over 18589.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2669, pruned_loss=0.04587, over 3713485.24 frames. ], batch size: 57, lr: 6.41e-03, grad_scale: 16.0 2022-12-23 12:17:59,721 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp0.9 from training. Duration: 24.9333125 2022-12-23 12:18:46,693 WARNING [train.py:1060] (3/4) Exclude cut with ID 5118-111612-0016-124680-0_sp0.9 from training. Duration: 20.388875 2022-12-23 12:18:52,556 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp1.1 from training. Duration: 20.3590625 2022-12-23 12:19:04,962 INFO [train.py:894] (3/4) Epoch 18, batch 1600, loss[loss=0.166, simple_loss=0.2439, pruned_loss=0.04405, over 18601.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2667, pruned_loss=0.04543, over 3712793.16 frames. ], batch size: 45, lr: 6.41e-03, grad_scale: 16.0 2022-12-23 12:19:06,950 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0581, 1.2669, 2.3624, 4.2280, 3.1796, 2.8016, 0.7187, 2.8512], device='cuda:3'), covar=tensor([0.1887, 0.1995, 0.1535, 0.0440, 0.0914, 0.1212, 0.2569, 0.1040], device='cuda:3'), in_proj_covar=tensor([0.0103, 0.0116, 0.0133, 0.0145, 0.0107, 0.0138, 0.0130, 0.0112], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 12:19:12,654 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4027, 1.3550, 1.1119, 1.7216, 1.6530, 2.9605, 1.2874, 1.4565], device='cuda:3'), covar=tensor([0.0874, 0.1753, 0.1107, 0.0873, 0.1436, 0.0273, 0.1391, 0.1558], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0082, 0.0073, 0.0075, 0.0091, 0.0075, 0.0085, 0.0077], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 12:19:48,909 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.328e+02 3.675e+02 4.504e+02 5.800e+02 1.343e+03, threshold=9.009e+02, percent-clipped=8.0 2022-12-23 12:20:02,131 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0_sp1.1 from training. Duration: 0.836375 2022-12-23 12:20:05,270 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6374, 2.3309, 1.8622, 0.9003, 1.8578, 2.0930, 1.6496, 2.0265], device='cuda:3'), covar=tensor([0.0625, 0.0515, 0.1206, 0.1588, 0.1224, 0.1446, 0.1673, 0.0789], device='cuda:3'), in_proj_covar=tensor([0.0170, 0.0183, 0.0206, 0.0189, 0.0210, 0.0199, 0.0211, 0.0200], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 12:20:18,441 INFO [train.py:894] (3/4) Epoch 18, batch 1650, loss[loss=0.1694, simple_loss=0.2536, pruned_loss=0.04258, over 18596.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2676, pruned_loss=0.04643, over 3712389.67 frames. ], batch size: 51, lr: 6.40e-03, grad_scale: 32.0 2022-12-23 12:20:45,150 WARNING [train.py:1060] (3/4) Exclude cut with ID 8565-290391-0049-67394-0_sp0.9 from training. Duration: 21.3166875 2022-12-23 12:21:14,373 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0029-104863-0_sp0.9 from training. Duration: 22.1055625 2022-12-23 12:21:25,651 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5133, 1.4159, 1.5285, 1.4542, 1.0322, 3.0148, 1.2100, 1.7090], device='cuda:3'), covar=tensor([0.3353, 0.2202, 0.2013, 0.2134, 0.1487, 0.0226, 0.1731, 0.0919], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0118, 0.0126, 0.0122, 0.0103, 0.0098, 0.0093, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 12:21:26,681 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp1.1 from training. Duration: 21.77725 2022-12-23 12:21:34,231 INFO [train.py:894] (3/4) Epoch 18, batch 1700, loss[loss=0.1814, simple_loss=0.2788, pruned_loss=0.04201, over 18449.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2691, pruned_loss=0.04817, over 3713025.27 frames. ], batch size: 50, lr: 6.40e-03, grad_scale: 16.0 2022-12-23 12:21:44,824 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp0.9 from training. Duration: 27.8166875 2022-12-23 12:21:55,427 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2022-12-23 12:22:11,125 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp1.1 from training. Duration: 22.5090625 2022-12-23 12:22:18,185 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0 from training. Duration: 25.035 2022-12-23 12:22:21,103 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.660e+02 4.296e+02 5.030e+02 5.943e+02 1.306e+03, threshold=1.006e+03, percent-clipped=4.0 2022-12-23 12:22:35,383 WARNING [train.py:1060] (3/4) Exclude cut with ID 774-127930-0014-10412-0_sp1.1 from training. Duration: 0.95 2022-12-23 12:22:38,333 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.10 vs. limit=5.0 2022-12-23 12:22:48,764 INFO [train.py:894] (3/4) Epoch 18, batch 1750, loss[loss=0.1601, simple_loss=0.2351, pruned_loss=0.04255, over 18412.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2693, pruned_loss=0.04916, over 3712659.53 frames. ], batch size: 42, lr: 6.40e-03, grad_scale: 16.0 2022-12-23 12:22:54,840 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp0.9 from training. Duration: 0.92225 2022-12-23 12:23:17,213 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=61374.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:23:21,298 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0 from training. Duration: 21.97 2022-12-23 12:23:40,875 WARNING [train.py:1060] (3/4) Exclude cut with ID 7492-105653-0055-62765-0_sp0.9 from training. Duration: 21.97225 2022-12-23 12:23:42,261 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp0.9 from training. Duration: 25.3333125 2022-12-23 12:23:51,981 WARNING [train.py:1060] (3/4) Exclude cut with ID 5172-29468-0015-19128-0_sp0.9 from training. Duration: 21.5055625 2022-12-23 12:23:55,519 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4631, 2.6252, 1.8788, 3.0675, 2.8409, 2.5663, 3.8584, 2.4583], device='cuda:3'), covar=tensor([0.0834, 0.1685, 0.2544, 0.1688, 0.1535, 0.0801, 0.0770, 0.1128], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0205, 0.0245, 0.0286, 0.0234, 0.0186, 0.0206, 0.0201], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 12:24:01,708 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp1.1 from training. Duration: 20.72725 2022-12-23 12:24:02,017 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7153, 1.2074, 0.7816, 1.2852, 2.1305, 0.9796, 1.3967, 1.6153], device='cuda:3'), covar=tensor([0.1602, 0.2065, 0.2204, 0.1537, 0.1671, 0.1762, 0.1543, 0.1576], device='cuda:3'), in_proj_covar=tensor([0.0093, 0.0097, 0.0116, 0.0095, 0.0115, 0.0091, 0.0099, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 12:24:04,530 INFO [train.py:894] (3/4) Epoch 18, batch 1800, loss[loss=0.1988, simple_loss=0.2787, pruned_loss=0.05947, over 18437.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2708, pruned_loss=0.05132, over 3713119.49 frames. ], batch size: 61, lr: 6.40e-03, grad_scale: 16.0 2022-12-23 12:24:15,702 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8751, 1.2849, 0.9162, 1.3639, 2.2413, 1.1816, 1.3274, 1.7331], device='cuda:3'), covar=tensor([0.1587, 0.2154, 0.2161, 0.1585, 0.1515, 0.1775, 0.1601, 0.1683], device='cuda:3'), in_proj_covar=tensor([0.0092, 0.0097, 0.0116, 0.0095, 0.0114, 0.0091, 0.0098, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 12:24:23,209 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=61417.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:24:30,565 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=61422.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:24:36,671 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp0.9 from training. Duration: 26.32775 2022-12-23 12:24:37,547 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2022-12-23 12:24:50,829 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.906e+02 4.710e+02 5.748e+02 7.409e+02 1.844e+03, threshold=1.150e+03, percent-clipped=6.0 2022-12-23 12:25:06,364 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0 from training. Duration: 20.025 2022-12-23 12:25:12,110 WARNING [train.py:1060] (3/4) Exclude cut with ID 6709-74022-0004-86860-0_sp1.1 from training. Duration: 0.9409375 2022-12-23 12:25:12,121 WARNING [train.py:1060] (3/4) Exclude cut with ID 4757-1811-0023-62229-0_sp0.9 from training. Duration: 21.37775 2022-12-23 12:25:18,727 INFO [train.py:894] (3/4) Epoch 18, batch 1850, loss[loss=0.2225, simple_loss=0.3017, pruned_loss=0.0716, over 18640.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2719, pruned_loss=0.05333, over 3713540.64 frames. ], batch size: 78, lr: 6.39e-03, grad_scale: 16.0 2022-12-23 12:25:31,311 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0004-25974-0_sp0.9 from training. Duration: 21.17225 2022-12-23 12:25:31,318 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp0.9 from training. Duration: 27.511125 2022-12-23 12:25:58,493 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2022-12-23 12:26:04,795 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0 from training. Duration: 22.8 2022-12-23 12:26:09,209 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0 from training. Duration: 22.585 2022-12-23 12:26:33,713 INFO [train.py:894] (3/4) Epoch 18, batch 1900, loss[loss=0.2076, simple_loss=0.2859, pruned_loss=0.06463, over 18448.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2731, pruned_loss=0.0547, over 3714412.68 frames. ], batch size: 50, lr: 6.39e-03, grad_scale: 16.0 2022-12-23 12:26:42,048 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0001-146967-0_sp0.9 from training. Duration: 22.0166875 2022-12-23 12:26:57,067 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp1.1 from training. Duration: 24.395375 2022-12-23 12:27:03,662 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp0.9 from training. Duration: 27.47775 2022-12-23 12:27:09,402 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp0.9 from training. Duration: 24.8833125 2022-12-23 12:27:12,288 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0 from training. Duration: 23.39 2022-12-23 12:27:18,034 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp0.9 from training. Duration: 28.72225 2022-12-23 12:27:20,884 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.500e+02 4.898e+02 5.639e+02 7.232e+02 1.821e+03, threshold=1.128e+03, percent-clipped=2.0 2022-12-23 12:27:27,550 WARNING [train.py:1060] (3/4) Exclude cut with ID 585-294811-0110-133686-0_sp0.9 from training. Duration: 20.8944375 2022-12-23 12:27:33,836 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.12 vs. limit=5.0 2022-12-23 12:27:42,085 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0_sp0.9 from training. Duration: 23.8444375 2022-12-23 12:27:42,517 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2168, 2.2671, 1.7255, 2.5705, 2.4086, 2.0932, 2.9553, 2.3314], device='cuda:3'), covar=tensor([0.0820, 0.1567, 0.2520, 0.1637, 0.1590, 0.0833, 0.0931, 0.1072], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0205, 0.0244, 0.0286, 0.0234, 0.0186, 0.0206, 0.0201], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 12:27:49,712 INFO [train.py:894] (3/4) Epoch 18, batch 1950, loss[loss=0.1886, simple_loss=0.2718, pruned_loss=0.05272, over 18390.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2722, pruned_loss=0.05477, over 3713481.10 frames. ], batch size: 53, lr: 6.39e-03, grad_scale: 16.0 2022-12-23 12:28:03,678 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6947, 3.7995, 3.7797, 1.4241, 3.9742, 2.8434, 0.8775, 2.5049], device='cuda:3'), covar=tensor([0.1954, 0.1196, 0.1306, 0.3704, 0.0884, 0.1019, 0.4920, 0.1488], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0140, 0.0155, 0.0124, 0.0140, 0.0112, 0.0145, 0.0113], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 12:28:06,353 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0 from training. Duration: 25.85 2022-12-23 12:28:06,361 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0 from training. Duration: 21.39 2022-12-23 12:28:17,422 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0 from training. Duration: 27.92 2022-12-23 12:28:45,175 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0039-130165-0_sp0.9 from training. Duration: 20.661125 2022-12-23 12:29:05,149 INFO [train.py:894] (3/4) Epoch 18, batch 2000, loss[loss=0.2004, simple_loss=0.2818, pruned_loss=0.05957, over 18712.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2728, pruned_loss=0.05532, over 3714728.69 frames. ], batch size: 99, lr: 6.39e-03, grad_scale: 16.0 2022-12-23 12:29:08,292 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0043-15874-0_sp0.9 from training. Duration: 20.07225 2022-12-23 12:29:15,921 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0 from training. Duration: 21.01 2022-12-23 12:29:52,146 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.258e+02 4.574e+02 5.379e+02 6.112e+02 1.044e+03, threshold=1.076e+03, percent-clipped=0.0 2022-12-23 12:30:21,756 INFO [train.py:894] (3/4) Epoch 18, batch 2050, loss[loss=0.225, simple_loss=0.3028, pruned_loss=0.07354, over 18576.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2743, pruned_loss=0.05634, over 3714724.63 frames. ], batch size: 56, lr: 6.38e-03, grad_scale: 16.0 2022-12-23 12:30:21,821 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0 from training. Duration: 20.65 2022-12-23 12:30:29,763 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0 from training. Duration: 21.46 2022-12-23 12:31:16,686 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0 from training. Duration: 0.92 2022-12-23 12:31:22,835 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0_sp0.9 from training. Duration: 23.7666875 2022-12-23 12:31:36,351 INFO [train.py:894] (3/4) Epoch 18, batch 2100, loss[loss=0.1951, simple_loss=0.2791, pruned_loss=0.05552, over 18623.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2731, pruned_loss=0.05594, over 3715362.16 frames. ], batch size: 78, lr: 6.38e-03, grad_scale: 16.0 2022-12-23 12:31:55,708 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=61717.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:31:58,254 WARNING [train.py:1060] (3/4) Exclude cut with ID 8544-281189-0060-101339-0_sp0.9 from training. Duration: 20.861125 2022-12-23 12:32:07,974 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0_sp0.9 from training. Duration: 22.711125 2022-12-23 12:32:18,456 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61732.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:32:24,052 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.133e+02 4.990e+02 6.097e+02 8.007e+02 2.031e+03, threshold=1.219e+03, percent-clipped=4.0 2022-12-23 12:32:36,075 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61743.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:32:53,503 INFO [train.py:894] (3/4) Epoch 18, batch 2150, loss[loss=0.2181, simple_loss=0.2936, pruned_loss=0.07136, over 18592.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2729, pruned_loss=0.05642, over 3715608.88 frames. ], batch size: 57, lr: 6.38e-03, grad_scale: 16.0 2022-12-23 12:32:53,550 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp1.1 from training. Duration: 22.986375 2022-12-23 12:33:08,585 WARNING [train.py:1060] (3/4) Exclude cut with ID 8040-260924-0003-80960-0_sp0.9 from training. Duration: 22.07225 2022-12-23 12:33:08,688 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=61765.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:33:13,350 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0045-26330-0_sp0.9 from training. Duration: 20.3055625 2022-12-23 12:33:16,168 WARNING [train.py:1060] (3/4) Exclude cut with ID 6356-271890-0060-94317-0_sp0.9 from training. Duration: 20.72225 2022-12-23 12:33:31,331 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp1.1 from training. Duration: 22.4818125 2022-12-23 12:33:52,249 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61793.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:33:57,858 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp0.9 from training. Duration: 25.0944375 2022-12-23 12:34:02,352 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0 from training. Duration: 21.515 2022-12-23 12:34:09,304 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp0.9 from training. Duration: 27.02225 2022-12-23 12:34:09,701 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61804.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:34:10,770 INFO [train.py:894] (3/4) Epoch 18, batch 2200, loss[loss=0.1814, simple_loss=0.2668, pruned_loss=0.04805, over 18710.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.2736, pruned_loss=0.05708, over 3713888.21 frames. ], batch size: 50, lr: 6.38e-03, grad_scale: 16.0 2022-12-23 12:34:15,202 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0010-62480-0_sp0.9 from training. Duration: 22.22225 2022-12-23 12:34:21,187 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0085-44554-0_sp0.9 from training. Duration: 20.85 2022-12-23 12:34:53,546 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0 from training. Duration: 21.54 2022-12-23 12:34:55,752 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2022-12-23 12:34:57,838 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.687e+02 4.531e+02 5.497e+02 7.224e+02 2.180e+03, threshold=1.099e+03, percent-clipped=3.0 2022-12-23 12:34:57,891 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp1.1 from training. Duration: 20.5318125 2022-12-23 12:35:08,010 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0012-134311-0_sp0.9 from training. Duration: 21.9333125 2022-12-23 12:35:22,746 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2022-12-23 12:35:26,414 INFO [train.py:894] (3/4) Epoch 18, batch 2250, loss[loss=0.1995, simple_loss=0.2878, pruned_loss=0.05557, over 18544.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.2732, pruned_loss=0.05701, over 3712410.83 frames. ], batch size: 55, lr: 6.37e-03, grad_scale: 16.0 2022-12-23 12:35:55,305 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0025-130151-0_sp0.9 from training. Duration: 21.7944375 2022-12-23 12:36:08,067 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0_sp0.9 from training. Duration: 22.4666875 2022-12-23 12:36:08,383 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8508, 1.4816, 1.8252, 2.3923, 1.6479, 4.4535, 1.4673, 1.5831], device='cuda:3'), covar=tensor([0.0919, 0.2034, 0.1204, 0.1001, 0.1698, 0.0227, 0.1636, 0.1835], device='cuda:3'), in_proj_covar=tensor([0.0072, 0.0081, 0.0073, 0.0074, 0.0089, 0.0074, 0.0084, 0.0077], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 12:36:12,556 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6508, 4.0908, 3.8747, 1.8144, 4.1735, 3.0404, 0.9297, 2.8140], device='cuda:3'), covar=tensor([0.2049, 0.1301, 0.1336, 0.3416, 0.0820, 0.0986, 0.5004, 0.1514], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0141, 0.0157, 0.0124, 0.0141, 0.0112, 0.0146, 0.0114], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 12:36:13,783 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0 from training. Duration: 21.635 2022-12-23 12:36:20,913 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0_sp0.9 from training. Duration: 24.038875 2022-12-23 12:36:35,404 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2488, 1.6330, 1.8587, 1.8473, 2.2255, 2.1699, 2.0293, 1.7465], device='cuda:3'), covar=tensor([0.1941, 0.2990, 0.2457, 0.2773, 0.1769, 0.0886, 0.2901, 0.1210], device='cuda:3'), in_proj_covar=tensor([0.0264, 0.0297, 0.0276, 0.0311, 0.0300, 0.0248, 0.0336, 0.0235], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 12:36:40,755 INFO [train.py:894] (3/4) Epoch 18, batch 2300, loss[loss=0.1627, simple_loss=0.2413, pruned_loss=0.04203, over 18383.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2739, pruned_loss=0.05753, over 3712717.63 frames. ], batch size: 46, lr: 6.37e-03, grad_scale: 16.0 2022-12-23 12:36:44,138 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6503, 1.2573, 1.4331, 1.9017, 1.4860, 3.3144, 1.1797, 1.3732], device='cuda:3'), covar=tensor([0.1052, 0.2553, 0.1335, 0.1082, 0.1895, 0.0310, 0.2010, 0.2303], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0082, 0.0073, 0.0075, 0.0090, 0.0075, 0.0084, 0.0078], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 12:36:55,263 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61914.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:37:02,409 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp1.1 from training. Duration: 21.786375 2022-12-23 12:37:14,179 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0 from training. Duration: 20.22 2022-12-23 12:37:19,068 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7553, 4.0882, 3.8958, 1.8543, 4.1919, 3.1524, 0.7801, 2.7058], device='cuda:3'), covar=tensor([0.2009, 0.1219, 0.1484, 0.3586, 0.0972, 0.0957, 0.5309, 0.1617], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0141, 0.0157, 0.0124, 0.0141, 0.0112, 0.0145, 0.0114], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 12:37:28,974 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.260e+02 4.610e+02 5.551e+02 6.861e+02 1.067e+03, threshold=1.110e+03, percent-clipped=0.0 2022-12-23 12:37:58,696 INFO [train.py:894] (3/4) Epoch 18, batch 2350, loss[loss=0.2443, simple_loss=0.31, pruned_loss=0.08928, over 18622.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.2739, pruned_loss=0.05809, over 3713554.80 frames. ], batch size: 189, lr: 6.37e-03, grad_scale: 16.0 2022-12-23 12:38:09,691 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.1076, 0.9759, 0.9570, 1.1563, 1.2887, 1.1949, 1.1279, 0.9858], device='cuda:3'), covar=tensor([0.0246, 0.0225, 0.0486, 0.0199, 0.0205, 0.0339, 0.0261, 0.0271], device='cuda:3'), in_proj_covar=tensor([0.0092, 0.0125, 0.0151, 0.0124, 0.0116, 0.0118, 0.0096, 0.0125], device='cuda:3'), out_proj_covar=tensor([7.4119e-05, 1.0048e-04, 1.2561e-04, 9.9454e-05, 9.4321e-05, 9.1620e-05, 7.5584e-05, 9.9667e-05], device='cuda:3') 2022-12-23 12:38:30,624 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61975.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:38:33,812 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2022-12-23 12:38:58,496 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.8152, 5.5661, 4.9677, 2.7277, 5.4584, 4.1736, 1.0787, 3.8219], device='cuda:3'), covar=tensor([0.2022, 0.0928, 0.1268, 0.3001, 0.0751, 0.0749, 0.5118, 0.1262], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0141, 0.0157, 0.0124, 0.0141, 0.0112, 0.0145, 0.0114], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 12:39:15,714 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0 from training. Duration: 25.285 2022-12-23 12:39:18,287 INFO [train.py:894] (3/4) Epoch 18, batch 2400, loss[loss=0.2174, simple_loss=0.2998, pruned_loss=0.06755, over 18468.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2743, pruned_loss=0.05808, over 3713241.15 frames. ], batch size: 50, lr: 6.37e-03, grad_scale: 16.0 2022-12-23 12:40:05,028 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.629e+02 4.370e+02 5.333e+02 6.634e+02 1.277e+03, threshold=1.067e+03, percent-clipped=2.0 2022-12-23 12:40:20,597 WARNING [train.py:1060] (3/4) Exclude cut with ID 811-130148-0001-63453-0_sp0.9 from training. Duration: 20.861125 2022-12-23 12:40:34,673 INFO [train.py:894] (3/4) Epoch 18, batch 2450, loss[loss=0.2219, simple_loss=0.2892, pruned_loss=0.07731, over 18649.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.2733, pruned_loss=0.05761, over 3713896.27 frames. ], batch size: 175, lr: 6.36e-03, grad_scale: 16.0 2022-12-23 12:40:45,013 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0 from training. Duration: 20.88 2022-12-23 12:41:06,250 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9455, 1.8676, 1.4707, 1.6751, 1.7015, 1.7361, 1.6564, 1.8845], device='cuda:3'), covar=tensor([0.2327, 0.2935, 0.1967, 0.2414, 0.3189, 0.1151, 0.2898, 0.1010], device='cuda:3'), in_proj_covar=tensor([0.0292, 0.0283, 0.0240, 0.0348, 0.0266, 0.0224, 0.0283, 0.0209], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 12:41:15,942 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0_sp0.9 from training. Duration: 23.4166875 2022-12-23 12:41:24,155 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62088.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:41:41,081 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62099.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:41:50,017 INFO [train.py:894] (3/4) Epoch 18, batch 2500, loss[loss=0.1792, simple_loss=0.2551, pruned_loss=0.05158, over 18611.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2733, pruned_loss=0.05754, over 3714271.04 frames. ], batch size: 45, lr: 6.36e-03, grad_scale: 16.0 2022-12-23 12:42:00,676 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.4153, 3.8036, 3.7378, 4.3380, 4.0079, 3.8839, 4.5358, 1.4141], device='cuda:3'), covar=tensor([0.0690, 0.0702, 0.0673, 0.0784, 0.1419, 0.1142, 0.0581, 0.4920], device='cuda:3'), in_proj_covar=tensor([0.0338, 0.0220, 0.0234, 0.0260, 0.0318, 0.0264, 0.0280, 0.0276], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 12:42:15,772 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62122.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:42:15,926 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4740, 2.0724, 1.5103, 2.2565, 1.8949, 1.9168, 1.9355, 2.4238], device='cuda:3'), covar=tensor([0.2058, 0.2998, 0.1969, 0.2583, 0.3379, 0.1151, 0.2961, 0.0868], device='cuda:3'), in_proj_covar=tensor([0.0290, 0.0281, 0.0239, 0.0346, 0.0264, 0.0223, 0.0280, 0.0207], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 12:42:33,530 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0 from training. Duration: 21.24 2022-12-23 12:42:34,915 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0_sp0.9 from training. Duration: 23.9055625 2022-12-23 12:42:36,305 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.067e+02 4.462e+02 5.167e+02 6.894e+02 1.650e+03, threshold=1.033e+03, percent-clipped=6.0 2022-12-23 12:42:38,580 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([5.6194, 4.8508, 4.9150, 5.5344, 5.1648, 4.9811, 5.6754, 1.7708], device='cuda:3'), covar=tensor([0.0487, 0.0528, 0.0492, 0.0613, 0.1128, 0.0975, 0.0393, 0.4661], device='cuda:3'), in_proj_covar=tensor([0.0339, 0.0222, 0.0235, 0.0261, 0.0320, 0.0267, 0.0281, 0.0278], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 12:42:53,780 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4944, 1.3961, 1.4871, 1.3346, 0.7356, 2.1888, 0.8238, 1.3801], device='cuda:3'), covar=tensor([0.3249, 0.2226, 0.2023, 0.2170, 0.1628, 0.0379, 0.1703, 0.0913], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0116, 0.0125, 0.0120, 0.0102, 0.0098, 0.0092, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 12:43:05,400 INFO [train.py:894] (3/4) Epoch 18, batch 2550, loss[loss=0.1786, simple_loss=0.2657, pruned_loss=0.04573, over 18621.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2723, pruned_loss=0.05662, over 3713878.30 frames. ], batch size: 53, lr: 6.36e-03, grad_scale: 16.0 2022-12-23 12:43:09,726 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp0.9 from training. Duration: 25.988875 2022-12-23 12:43:18,072 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0001-134300-0_sp0.9 from training. Duration: 20.67225 2022-12-23 12:43:30,672 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5008, 1.7051, 1.4383, 2.0850, 2.2221, 1.5251, 1.1640, 1.3342], device='cuda:3'), covar=tensor([0.1918, 0.1774, 0.1637, 0.1002, 0.1166, 0.1169, 0.2375, 0.1521], device='cuda:3'), in_proj_covar=tensor([0.0246, 0.0222, 0.0213, 0.0197, 0.0260, 0.0196, 0.0223, 0.0197], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 12:43:48,366 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62183.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:44:06,114 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0 from training. Duration: 20.34 2022-12-23 12:44:21,579 INFO [train.py:894] (3/4) Epoch 18, batch 2600, loss[loss=0.2094, simple_loss=0.2884, pruned_loss=0.06524, over 18628.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2721, pruned_loss=0.05693, over 3713768.77 frames. ], batch size: 98, lr: 6.36e-03, grad_scale: 16.0 2022-12-23 12:45:02,611 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7144, 2.1327, 1.7325, 2.4813, 2.4173, 1.7631, 1.6214, 1.5258], device='cuda:3'), covar=tensor([0.1766, 0.1386, 0.1395, 0.0874, 0.1231, 0.1075, 0.2105, 0.1452], device='cuda:3'), in_proj_covar=tensor([0.0246, 0.0221, 0.0212, 0.0196, 0.0259, 0.0195, 0.0223, 0.0197], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 12:45:07,937 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.177e+02 4.427e+02 5.358e+02 6.353e+02 1.453e+03, threshold=1.072e+03, percent-clipped=5.0 2022-12-23 12:45:15,309 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp0.9 from training. Duration: 25.061125 2022-12-23 12:45:27,834 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0 from training. Duration: 0.83 2022-12-23 12:45:37,362 INFO [train.py:894] (3/4) Epoch 18, batch 2650, loss[loss=0.1937, simple_loss=0.2819, pruned_loss=0.0528, over 18463.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2729, pruned_loss=0.05682, over 3714146.29 frames. ], batch size: 54, lr: 6.35e-03, grad_scale: 16.0 2022-12-23 12:45:52,009 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0 from training. Duration: 24.73 2022-12-23 12:46:00,072 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62270.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:46:05,188 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62273.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:46:07,766 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0 from training. Duration: 23.965 2022-12-23 12:46:15,110 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0030-146996-0_sp0.9 from training. Duration: 22.088875 2022-12-23 12:46:31,030 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0_sp0.9 from training. Duration: 23.6 2022-12-23 12:46:52,865 INFO [train.py:894] (3/4) Epoch 18, batch 2700, loss[loss=0.2065, simple_loss=0.2915, pruned_loss=0.06074, over 18525.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2723, pruned_loss=0.05643, over 3713804.47 frames. ], batch size: 98, lr: 6.35e-03, grad_scale: 16.0 2022-12-23 12:47:36,961 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62334.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:47:40,001 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.427e+02 4.380e+02 5.434e+02 6.774e+02 1.412e+03, threshold=1.087e+03, percent-clipped=1.0 2022-12-23 12:48:02,184 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2022-12-23 12:48:08,468 INFO [train.py:894] (3/4) Epoch 18, batch 2750, loss[loss=0.2028, simple_loss=0.2674, pruned_loss=0.0691, over 18407.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2711, pruned_loss=0.05573, over 3714362.38 frames. ], batch size: 46, lr: 6.35e-03, grad_scale: 16.0 2022-12-23 12:48:09,861 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0 from training. Duration: 23.795 2022-12-23 12:48:27,335 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp1.1 from training. Duration: 21.5409375 2022-12-23 12:48:28,755 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp0.9 from training. Duration: 24.97775 2022-12-23 12:48:37,948 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0_sp0.9 from training. Duration: 23.3444375 2022-12-23 12:49:00,658 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62388.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:49:05,398 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62391.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:49:06,770 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0_sp0.9 from training. Duration: 23.2 2022-12-23 12:49:12,471 WARNING [train.py:1060] (3/4) Exclude cut with ID 5653-46179-0060-117930-0_sp0.9 from training. Duration: 21.17225 2022-12-23 12:49:12,918 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.1413, 1.0984, 1.2393, 0.6186, 0.7396, 1.1872, 1.2543, 1.1948], device='cuda:3'), covar=tensor([0.0715, 0.0289, 0.0341, 0.0338, 0.0427, 0.0505, 0.0255, 0.0570], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0168, 0.0125, 0.0137, 0.0147, 0.0141, 0.0159, 0.0169], device='cuda:3'), out_proj_covar=tensor([1.1484e-04, 1.2989e-04, 9.5416e-05, 1.0310e-04, 1.1171e-04, 1.0972e-04, 1.2394e-04, 1.3047e-04], device='cuda:3') 2022-12-23 12:49:16,814 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.41 vs. limit=5.0 2022-12-23 12:49:17,538 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62399.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:49:25,894 INFO [train.py:894] (3/4) Epoch 18, batch 2800, loss[loss=0.1797, simple_loss=0.2613, pruned_loss=0.04906, over 18477.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2711, pruned_loss=0.05578, over 3714775.68 frames. ], batch size: 64, lr: 6.35e-03, grad_scale: 16.0 2022-12-23 12:49:33,374 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp0.9 from training. Duration: 24.6555625 2022-12-23 12:50:14,069 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.980e+02 4.461e+02 5.405e+02 6.441e+02 1.681e+03, threshold=1.081e+03, percent-clipped=2.0 2022-12-23 12:50:14,262 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=62436.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:50:25,679 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0 from training. Duration: 20.44 2022-12-23 12:50:30,093 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=62447.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:50:37,700 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62452.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:50:42,219 INFO [train.py:894] (3/4) Epoch 18, batch 2850, loss[loss=0.2268, simple_loss=0.2979, pruned_loss=0.07788, over 18688.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2722, pruned_loss=0.05656, over 3714873.09 frames. ], batch size: 60, lr: 6.34e-03, grad_scale: 16.0 2022-12-23 12:50:42,248 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0_sp0.9 from training. Duration: 23.45 2022-12-23 12:51:11,458 WARNING [train.py:1060] (3/4) Exclude cut with ID 6945-60535-0076-12784-0_sp0.9 from training. Duration: 20.52225 2022-12-23 12:51:17,463 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62478.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:51:20,400 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0 from training. Duration: 22.19 2022-12-23 12:51:29,000 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp1.1 from training. Duration: 25.3818125 2022-12-23 12:51:45,518 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp0.9 from training. Duration: 28.0944375 2022-12-23 12:51:51,432 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0_sp0.9 from training. Duration: 22.9444375 2022-12-23 12:51:57,907 INFO [train.py:894] (3/4) Epoch 18, batch 2900, loss[loss=0.1825, simple_loss=0.2653, pruned_loss=0.04987, over 18697.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2723, pruned_loss=0.05643, over 3714913.03 frames. ], batch size: 48, lr: 6.34e-03, grad_scale: 16.0 2022-12-23 12:52:00,862 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp1.1 from training. Duration: 21.6318125 2022-12-23 12:52:18,951 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0 from training. Duration: 23.695 2022-12-23 12:52:44,255 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.769e+02 4.491e+02 5.311e+02 7.328e+02 1.636e+03, threshold=1.062e+03, percent-clipped=8.0 2022-12-23 12:52:45,914 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0 from training. Duration: 23.955 2022-12-23 12:52:58,239 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62545.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:53:13,517 INFO [train.py:894] (3/4) Epoch 18, batch 2950, loss[loss=0.2498, simple_loss=0.3167, pruned_loss=0.09141, over 18676.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2711, pruned_loss=0.05599, over 3714376.68 frames. ], batch size: 179, lr: 6.34e-03, grad_scale: 16.0 2022-12-23 12:53:13,978 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.6070, 3.1121, 2.6807, 1.1459, 2.4972, 2.4172, 2.0660, 2.1646], device='cuda:3'), covar=tensor([0.0632, 0.0672, 0.1408, 0.2055, 0.1607, 0.1450, 0.1755, 0.1315], device='cuda:3'), in_proj_covar=tensor([0.0173, 0.0185, 0.0209, 0.0194, 0.0211, 0.0201, 0.0217, 0.0203], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 12:53:19,238 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp0.9 from training. Duration: 26.438875 2022-12-23 12:53:37,039 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62570.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:53:58,856 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62585.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:54:04,376 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0021-26306-0_sp0.9 from training. Duration: 21.2444375 2022-12-23 12:54:04,402 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp0.9 from training. Duration: 31.02225 2022-12-23 12:54:14,594 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0 from training. Duration: 22.395 2022-12-23 12:54:29,662 INFO [train.py:894] (3/4) Epoch 18, batch 3000, loss[loss=0.1794, simple_loss=0.2613, pruned_loss=0.04875, over 18521.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2716, pruned_loss=0.05609, over 3714288.75 frames. ], batch size: 47, lr: 6.34e-03, grad_scale: 16.0 2022-12-23 12:54:29,662 INFO [train.py:919] (3/4) Computing validation loss 2022-12-23 12:54:35,096 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9980, 1.7832, 0.9768, 1.6602, 2.2379, 1.6198, 1.8892, 1.9834], device='cuda:3'), covar=tensor([0.1686, 0.1976, 0.2338, 0.1559, 0.1893, 0.1602, 0.1447, 0.1697], device='cuda:3'), in_proj_covar=tensor([0.0093, 0.0097, 0.0115, 0.0094, 0.0115, 0.0090, 0.0098, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 12:54:40,507 INFO [train.py:928] (3/4) Epoch 18, validation: loss=0.164, simple_loss=0.2628, pruned_loss=0.0326, over 944034.00 frames. 2022-12-23 12:54:40,508 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-23 12:54:42,638 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62606.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:54:43,767 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0 from training. Duration: 21.075 2022-12-23 12:54:48,058 WARNING [train.py:1060] (3/4) Exclude cut with ID 6482-98857-0025-147532-0_sp0.9 from training. Duration: 20.0055625 2022-12-23 12:54:48,074 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0037-132304-0_sp0.9 from training. Duration: 22.05 2022-12-23 12:54:48,088 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0 from training. Duration: 26.8349375 2022-12-23 12:54:51,358 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp1.1 from training. Duration: 22.1090625 2022-12-23 12:54:59,161 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp0.9 from training. Duration: 26.6166875 2022-12-23 12:55:00,658 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=62618.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:55:15,832 WARNING [train.py:1060] (3/4) Exclude cut with ID 2046-178027-0000-53705-0_sp0.9 from training. Duration: 20.3055625 2022-12-23 12:55:18,429 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62629.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:55:28,401 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.034e+02 4.348e+02 5.116e+02 6.564e+02 1.235e+03, threshold=1.023e+03, percent-clipped=2.0 2022-12-23 12:55:34,585 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3593, 2.5428, 3.0921, 0.9245, 2.5123, 3.3949, 2.2872, 2.6233], device='cuda:3'), covar=tensor([0.0854, 0.0439, 0.0348, 0.0543, 0.0462, 0.0400, 0.0417, 0.0685], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0167, 0.0125, 0.0136, 0.0147, 0.0141, 0.0159, 0.0168], device='cuda:3'), out_proj_covar=tensor([1.1437e-04, 1.2958e-04, 9.5390e-05, 1.0243e-04, 1.1124e-04, 1.0923e-04, 1.2359e-04, 1.2982e-04], device='cuda:3') 2022-12-23 12:55:40,203 WARNING [train.py:1060] (3/4) Exclude cut with ID 7205-50138-0008-5373-0_sp0.9 from training. Duration: 20.7 2022-12-23 12:55:44,083 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62646.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:55:57,533 INFO [train.py:894] (3/4) Epoch 18, batch 3050, loss[loss=0.1701, simple_loss=0.2604, pruned_loss=0.03989, over 18572.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2722, pruned_loss=0.05588, over 3714313.11 frames. ], batch size: 56, lr: 6.33e-03, grad_scale: 16.0 2022-12-23 12:56:14,076 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62666.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:56:22,816 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0 from training. Duration: 22.48 2022-12-23 12:56:39,845 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp0.9 from training. Duration: 29.816625 2022-12-23 12:56:59,044 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp1.1 from training. Duration: 22.7590625 2022-12-23 12:57:05,061 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0 from training. Duration: 22.555 2022-12-23 12:57:13,176 INFO [train.py:894] (3/4) Epoch 18, batch 3100, loss[loss=0.2125, simple_loss=0.2852, pruned_loss=0.06994, over 18521.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2717, pruned_loss=0.05573, over 3714254.13 frames. ], batch size: 58, lr: 6.33e-03, grad_scale: 8.0 2022-12-23 12:57:26,257 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0005-25975-0_sp0.9 from training. Duration: 21.688875 2022-12-23 12:57:46,941 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62727.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:58:00,961 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.728e+02 4.372e+02 5.368e+02 7.060e+02 1.480e+03, threshold=1.074e+03, percent-clipped=5.0 2022-12-23 12:58:04,398 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0_sp0.9 from training. Duration: 22.6 2022-12-23 12:58:16,175 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62747.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:58:27,804 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4241, 2.8551, 3.3642, 1.2483, 2.7150, 3.6632, 2.5528, 2.9134], device='cuda:3'), covar=tensor([0.0806, 0.0319, 0.0259, 0.0449, 0.0378, 0.0384, 0.0343, 0.0620], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0167, 0.0125, 0.0136, 0.0146, 0.0141, 0.0159, 0.0167], device='cuda:3'), out_proj_covar=tensor([1.1469e-04, 1.2958e-04, 9.5023e-05, 1.0226e-04, 1.1091e-04, 1.0898e-04, 1.2363e-04, 1.2916e-04], device='cuda:3') 2022-12-23 12:58:28,828 INFO [train.py:894] (3/4) Epoch 18, batch 3150, loss[loss=0.1563, simple_loss=0.2373, pruned_loss=0.03765, over 18704.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2721, pruned_loss=0.05593, over 3713747.91 frames. ], batch size: 46, lr: 6.33e-03, grad_scale: 8.0 2022-12-23 12:58:38,998 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0 from training. Duration: 24.32 2022-12-23 12:58:55,969 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62773.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:59:03,143 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62778.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 12:59:34,707 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-276745-0093-13116-0_sp0.9 from training. Duration: 21.061125 2022-12-23 12:59:44,258 INFO [train.py:894] (3/4) Epoch 18, batch 3200, loss[loss=0.2177, simple_loss=0.2954, pruned_loss=0.07002, over 18683.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2731, pruned_loss=0.05637, over 3714677.27 frames. ], batch size: 69, lr: 6.33e-03, grad_scale: 8.0 2022-12-23 12:59:48,805 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0024-15855-0_sp0.9 from training. Duration: 20.32225 2022-12-23 13:00:02,558 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp1.1 from training. Duration: 0.7545625 2022-12-23 13:00:16,787 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0_sp0.9 from training. Duration: 23.9333125 2022-12-23 13:00:16,902 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=62826.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:00:29,816 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62834.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:00:33,649 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.693e+02 4.436e+02 5.531e+02 6.977e+02 1.575e+03, threshold=1.106e+03, percent-clipped=1.0 2022-12-23 13:00:49,210 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp1.1 from training. Duration: 20.17275 2022-12-23 13:00:49,501 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8842, 1.4376, 0.7484, 1.4424, 2.0880, 1.2853, 1.6588, 1.8190], device='cuda:3'), covar=tensor([0.1514, 0.1940, 0.2218, 0.1364, 0.1806, 0.1767, 0.1354, 0.1503], device='cuda:3'), in_proj_covar=tensor([0.0093, 0.0097, 0.0115, 0.0094, 0.0115, 0.0090, 0.0097, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 13:00:54,925 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp1.1 from training. Duration: 20.436375 2022-12-23 13:01:00,870 INFO [train.py:894] (3/4) Epoch 18, batch 3250, loss[loss=0.2051, simple_loss=0.2862, pruned_loss=0.06203, over 18396.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2716, pruned_loss=0.05528, over 3713388.93 frames. ], batch size: 53, lr: 6.32e-03, grad_scale: 8.0 2022-12-23 13:02:10,835 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62901.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:02:16,425 INFO [train.py:894] (3/4) Epoch 18, batch 3300, loss[loss=0.1767, simple_loss=0.251, pruned_loss=0.05118, over 18583.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2719, pruned_loss=0.0554, over 3713309.86 frames. ], batch size: 45, lr: 6.32e-03, grad_scale: 8.0 2022-12-23 13:02:18,476 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0_sp0.9 from training. Duration: 23.1055625 2022-12-23 13:02:19,798 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp1.1 from training. Duration: 23.5 2022-12-23 13:02:32,434 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp0.9 from training. Duration: 26.62775 2022-12-23 13:02:45,682 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0 from training. Duration: 21.105 2022-12-23 13:02:48,751 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0_sp0.9 from training. Duration: 24.411125 2022-12-23 13:02:53,465 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62929.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:03:05,529 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.669e+02 4.438e+02 5.623e+02 6.897e+02 1.767e+03, threshold=1.125e+03, percent-clipped=3.0 2022-12-23 13:03:11,437 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62941.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:03:15,343 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp1.1 from training. Duration: 21.263625 2022-12-23 13:03:21,577 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5258, 3.6477, 3.5703, 1.4502, 3.7060, 2.8148, 1.0933, 2.2654], device='cuda:3'), covar=tensor([0.2097, 0.1218, 0.1447, 0.3758, 0.1108, 0.0934, 0.4269, 0.1695], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0140, 0.0158, 0.0124, 0.0143, 0.0112, 0.0144, 0.0114], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 13:03:23,075 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4860, 1.0696, 0.7902, 1.2030, 1.9877, 0.6735, 1.1424, 1.2368], device='cuda:3'), covar=tensor([0.1800, 0.2426, 0.2039, 0.1585, 0.1857, 0.1907, 0.1749, 0.2024], device='cuda:3'), in_proj_covar=tensor([0.0093, 0.0096, 0.0115, 0.0094, 0.0115, 0.0090, 0.0098, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 13:03:31,688 INFO [train.py:894] (3/4) Epoch 18, batch 3350, loss[loss=0.182, simple_loss=0.2703, pruned_loss=0.04692, over 18459.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2721, pruned_loss=0.05584, over 3712712.87 frames. ], batch size: 64, lr: 6.32e-03, grad_scale: 8.0 2022-12-23 13:03:48,520 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0 from training. Duration: 20.795 2022-12-23 13:03:59,266 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0 from training. Duration: 24.76 2022-12-23 13:03:59,286 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0_sp0.9 from training. Duration: 22.25 2022-12-23 13:04:05,353 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=62977.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:04:23,835 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp1.1 from training. Duration: 20.5045625 2022-12-23 13:04:48,414 INFO [train.py:894] (3/4) Epoch 18, batch 3400, loss[loss=0.1787, simple_loss=0.2714, pruned_loss=0.04298, over 18568.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.272, pruned_loss=0.05564, over 3713911.74 frames. ], batch size: 56, lr: 6.32e-03, grad_scale: 8.0 2022-12-23 13:05:13,723 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63022.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:05:36,154 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.859e+02 4.260e+02 5.426e+02 6.876e+02 1.239e+03, threshold=1.085e+03, percent-clipped=1.0 2022-12-23 13:05:50,636 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63047.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:06:02,087 INFO [train.py:894] (3/4) Epoch 18, batch 3450, loss[loss=0.2091, simple_loss=0.2977, pruned_loss=0.06025, over 18472.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2712, pruned_loss=0.05524, over 3713577.11 frames. ], batch size: 54, lr: 6.31e-03, grad_scale: 8.0 2022-12-23 13:06:41,366 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.88 vs. limit=5.0 2022-12-23 13:06:46,632 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2022-12-23 13:06:58,865 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=63095.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:07:14,061 INFO [train.py:894] (3/4) Epoch 18, batch 3500, loss[loss=0.1973, simple_loss=0.2802, pruned_loss=0.05722, over 18564.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2722, pruned_loss=0.0556, over 3714853.26 frames. ], batch size: 77, lr: 6.31e-03, grad_scale: 8.0 2022-12-23 13:07:18,030 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2151, 1.9217, 1.6183, 2.0783, 1.7676, 1.8066, 1.7609, 2.1393], device='cuda:3'), covar=tensor([0.1634, 0.2390, 0.1427, 0.2131, 0.2604, 0.0882, 0.2520, 0.0753], device='cuda:3'), in_proj_covar=tensor([0.0296, 0.0287, 0.0243, 0.0350, 0.0269, 0.0227, 0.0285, 0.0211], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 13:07:35,627 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp1.1 from training. Duration: 22.2954375 2022-12-23 13:07:45,951 INFO [train.py:894] (3/4) Epoch 19, batch 0, loss[loss=0.1867, simple_loss=0.2775, pruned_loss=0.04793, over 18491.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2775, pruned_loss=0.04793, over 18491.00 frames. ], batch size: 52, lr: 6.14e-03, grad_scale: 8.0 2022-12-23 13:07:45,952 INFO [train.py:919] (3/4) Computing validation loss 2022-12-23 13:07:56,916 INFO [train.py:928] (3/4) Epoch 19, validation: loss=0.1656, simple_loss=0.2641, pruned_loss=0.03356, over 944034.00 frames. 2022-12-23 13:07:56,917 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-23 13:08:24,357 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63129.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:08:26,184 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.85 vs. limit=5.0 2022-12-23 13:08:35,789 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.218e+02 4.228e+02 5.499e+02 7.176e+02 2.026e+03, threshold=1.100e+03, percent-clipped=7.0 2022-12-23 13:08:53,074 WARNING [train.py:1060] (3/4) Exclude cut with ID 298-126791-0067-24026-0_sp0.9 from training. Duration: 21.438875 2022-12-23 13:08:57,626 WARNING [train.py:1060] (3/4) Exclude cut with ID 5652-39938-0025-23684-0_sp0.9 from training. Duration: 22.2055625 2022-12-23 13:09:12,659 INFO [train.py:894] (3/4) Epoch 19, batch 50, loss[loss=0.1663, simple_loss=0.2466, pruned_loss=0.04302, over 18605.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2651, pruned_loss=0.04505, over 838044.70 frames. ], batch size: 41, lr: 6.14e-03, grad_scale: 8.0 2022-12-23 13:10:15,232 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63201.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:10:29,631 INFO [train.py:894] (3/4) Epoch 19, batch 100, loss[loss=0.1595, simple_loss=0.232, pruned_loss=0.04344, over 18468.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2649, pruned_loss=0.04552, over 1475646.54 frames. ], batch size: 43, lr: 6.13e-03, grad_scale: 8.0 2022-12-23 13:10:45,512 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63221.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:11:07,783 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.052e+02 3.397e+02 4.166e+02 5.229e+02 9.928e+02, threshold=8.333e+02, percent-clipped=0.0 2022-12-23 13:11:14,362 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63241.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:11:26,422 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=63249.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:11:43,018 INFO [train.py:894] (3/4) Epoch 19, batch 150, loss[loss=0.1982, simple_loss=0.2828, pruned_loss=0.05682, over 18513.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2664, pruned_loss=0.04566, over 1971863.01 frames. ], batch size: 77, lr: 6.13e-03, grad_scale: 8.0 2022-12-23 13:12:01,538 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0 from training. Duration: 24.525 2022-12-23 13:12:14,865 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63282.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:12:25,564 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=63289.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:12:32,680 WARNING [train.py:1060] (3/4) Exclude cut with ID 3699-47246-0007-3408-0_sp0.9 from training. Duration: 20.26675 2022-12-23 13:12:46,704 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp0.9 from training. Duration: 27.25 2022-12-23 13:12:59,142 INFO [train.py:894] (3/4) Epoch 19, batch 200, loss[loss=0.1923, simple_loss=0.2841, pruned_loss=0.05027, over 18658.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2659, pruned_loss=0.04514, over 2356873.68 frames. ], batch size: 60, lr: 6.13e-03, grad_scale: 8.0 2022-12-23 13:13:15,742 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63322.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:13:29,272 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6159, 1.6659, 1.8419, 1.0592, 1.7729, 1.8864, 1.3849, 2.1081], device='cuda:3'), covar=tensor([0.1082, 0.1790, 0.1134, 0.1772, 0.0800, 0.1039, 0.2279, 0.0530], device='cuda:3'), in_proj_covar=tensor([0.0197, 0.0210, 0.0204, 0.0194, 0.0176, 0.0215, 0.0212, 0.0187], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 13:13:38,110 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.297e+02 3.501e+02 4.294e+02 5.225e+02 1.195e+03, threshold=8.588e+02, percent-clipped=3.0 2022-12-23 13:14:02,326 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0 from training. Duration: 21.68 2022-12-23 13:14:13,270 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0 from training. Duration: 21.6300625 2022-12-23 13:14:14,636 INFO [train.py:894] (3/4) Epoch 19, batch 250, loss[loss=0.197, simple_loss=0.2886, pruned_loss=0.05267, over 18639.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2625, pruned_loss=0.04371, over 2657662.00 frames. ], batch size: 53, lr: 6.13e-03, grad_scale: 8.0 2022-12-23 13:14:28,135 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=63370.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:14:31,278 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.1789, 3.5951, 3.6088, 4.0922, 3.7996, 3.6595, 4.3174, 1.3325], device='cuda:3'), covar=tensor([0.0801, 0.0773, 0.0706, 0.0837, 0.1520, 0.1305, 0.0673, 0.5108], device='cuda:3'), in_proj_covar=tensor([0.0342, 0.0222, 0.0236, 0.0263, 0.0321, 0.0265, 0.0286, 0.0278], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 13:14:35,411 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0_sp0.9 from training. Duration: 24.033375 2022-12-23 13:15:30,906 INFO [train.py:894] (3/4) Epoch 19, batch 300, loss[loss=0.1684, simple_loss=0.2626, pruned_loss=0.03709, over 18665.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2634, pruned_loss=0.04406, over 2892803.73 frames. ], batch size: 96, lr: 6.12e-03, grad_scale: 8.0 2022-12-23 13:15:30,975 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0 from training. Duration: 22.905 2022-12-23 13:15:32,352 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp1.1 from training. Duration: 23.4318125 2022-12-23 13:15:37,748 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([5.9634, 4.9332, 5.1713, 5.9309, 5.4994, 5.2187, 5.9697, 1.6902], device='cuda:3'), covar=tensor([0.0530, 0.0592, 0.0545, 0.0581, 0.1174, 0.1076, 0.0411, 0.4946], device='cuda:3'), in_proj_covar=tensor([0.0339, 0.0221, 0.0235, 0.0262, 0.0319, 0.0264, 0.0284, 0.0276], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 13:15:42,806 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2022-12-23 13:15:44,736 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63420.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:15:47,498 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63422.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:15:58,074 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63429.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:16:10,932 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.316e+02 3.457e+02 4.109e+02 5.023e+02 1.116e+03, threshold=8.218e+02, percent-clipped=3.0 2022-12-23 13:16:46,629 INFO [train.py:894] (3/4) Epoch 19, batch 350, loss[loss=0.1887, simple_loss=0.2749, pruned_loss=0.05123, over 18380.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2636, pruned_loss=0.04429, over 3075366.36 frames. ], batch size: 51, lr: 6.12e-03, grad_scale: 8.0 2022-12-23 13:17:09,850 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8047, 1.9069, 2.1778, 1.2141, 2.1089, 2.2116, 1.5061, 2.5358], device='cuda:3'), covar=tensor([0.1212, 0.1771, 0.1218, 0.1911, 0.0797, 0.1123, 0.2221, 0.0536], device='cuda:3'), in_proj_covar=tensor([0.0197, 0.0211, 0.0204, 0.0193, 0.0176, 0.0215, 0.0213, 0.0188], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 13:17:11,753 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=63477.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:17:17,978 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63481.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:17:21,407 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63483.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:17:31,483 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp1.1 from training. Duration: 20.82275 2022-12-23 13:17:33,132 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp0.9 from training. Duration: 25.45 2022-12-23 13:18:02,714 INFO [train.py:894] (3/4) Epoch 19, batch 400, loss[loss=0.1847, simple_loss=0.2821, pruned_loss=0.0437, over 18468.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2632, pruned_loss=0.04413, over 3216848.85 frames. ], batch size: 54, lr: 6.12e-03, grad_scale: 8.0 2022-12-23 13:18:28,555 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0 from training. Duration: 25.775 2022-12-23 13:18:42,197 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.168e+02 3.327e+02 4.266e+02 5.162e+02 9.583e+02, threshold=8.532e+02, percent-clipped=3.0 2022-12-23 13:18:50,591 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0_sp0.9 from training. Duration: 22.25 2022-12-23 13:19:16,824 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.2342, 2.5121, 1.8602, 3.2444, 2.2568, 2.2888, 2.4905, 3.3070], device='cuda:3'), covar=tensor([0.1794, 0.3076, 0.1831, 0.2522, 0.3686, 0.1041, 0.3139, 0.0764], device='cuda:3'), in_proj_covar=tensor([0.0291, 0.0286, 0.0240, 0.0345, 0.0267, 0.0224, 0.0285, 0.0207], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 13:19:17,714 INFO [train.py:894] (3/4) Epoch 19, batch 450, loss[loss=0.174, simple_loss=0.2678, pruned_loss=0.04014, over 18605.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2648, pruned_loss=0.04495, over 3328139.51 frames. ], batch size: 51, lr: 6.12e-03, grad_scale: 8.0 2022-12-23 13:19:17,763 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0 from training. Duration: 26.205 2022-12-23 13:19:29,479 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63569.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:19:33,692 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp0.9 from training. Duration: 30.1555625 2022-12-23 13:19:40,222 WARNING [train.py:1060] (3/4) Exclude cut with ID 1265-135635-0050-6781-0_sp0.9 from training. Duration: 21.8333125 2022-12-23 13:19:41,644 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63577.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:19:49,035 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp1.1 from training. Duration: 20.6545625 2022-12-23 13:19:55,514 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63586.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:20:30,980 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0045-39920-0_sp0.9 from training. Duration: 20.52225 2022-12-23 13:20:32,292 INFO [train.py:894] (3/4) Epoch 19, batch 500, loss[loss=0.2055, simple_loss=0.2942, pruned_loss=0.05844, over 18678.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2656, pruned_loss=0.04519, over 3414709.19 frames. ], batch size: 60, lr: 6.11e-03, grad_scale: 8.0 2022-12-23 13:20:52,439 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp0.9 from training. Duration: 29.1166875 2022-12-23 13:21:01,476 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63630.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:21:07,695 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1368, 1.6596, 1.8698, 2.3739, 2.0445, 4.7302, 1.8111, 2.1437], device='cuda:3'), covar=tensor([0.0778, 0.1732, 0.1002, 0.0906, 0.1308, 0.0162, 0.1279, 0.1435], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0082, 0.0073, 0.0074, 0.0090, 0.0075, 0.0085, 0.0077], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 13:21:11,693 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.157e+02 3.552e+02 4.116e+02 4.841e+02 8.719e+02, threshold=8.233e+02, percent-clipped=1.0 2022-12-23 13:21:27,167 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63647.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:21:48,211 INFO [train.py:894] (3/4) Epoch 19, batch 550, loss[loss=0.199, simple_loss=0.2843, pruned_loss=0.05689, over 18489.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2658, pruned_loss=0.04555, over 3480553.53 frames. ], batch size: 77, lr: 6.11e-03, grad_scale: 8.0 2022-12-23 13:21:51,257 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133211-0007-59831-0_sp0.9 from training. Duration: 21.388875 2022-12-23 13:22:28,614 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0 from training. Duration: 22.72 2022-12-23 13:22:30,068 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0_sp0.9 from training. Duration: 22.7444375 2022-12-23 13:23:04,733 INFO [train.py:894] (3/4) Epoch 19, batch 600, loss[loss=0.1752, simple_loss=0.2732, pruned_loss=0.03861, over 18535.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2666, pruned_loss=0.04572, over 3533535.37 frames. ], batch size: 55, lr: 6.11e-03, grad_scale: 8.0 2022-12-23 13:23:14,949 WARNING [train.py:1060] (3/4) Exclude cut with ID 4133-6541-0027-40495-0_sp1.1 from training. Duration: 0.9681875 2022-12-23 13:23:18,806 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0_sp0.9 from training. Duration: 22.3166875 2022-12-23 13:23:19,046 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4262, 4.0538, 3.8597, 1.5053, 4.2503, 3.0295, 0.5267, 2.5283], device='cuda:3'), covar=tensor([0.2186, 0.0863, 0.1259, 0.3642, 0.0677, 0.0867, 0.5173, 0.1485], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0134, 0.0152, 0.0120, 0.0138, 0.0109, 0.0140, 0.0111], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 13:23:25,119 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133212-0015-59917-0_sp0.9 from training. Duration: 21.8166875 2022-12-23 13:23:43,163 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6754, 1.4783, 1.6832, 1.5340, 0.9041, 3.0873, 1.3054, 1.8411], device='cuda:3'), covar=tensor([0.3023, 0.2095, 0.1859, 0.2009, 0.1553, 0.0215, 0.1537, 0.0798], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0117, 0.0126, 0.0121, 0.0102, 0.0097, 0.0092, 0.0088], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 13:23:46,947 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.867e+02 3.213e+02 3.819e+02 4.458e+02 9.536e+02, threshold=7.638e+02, percent-clipped=0.0 2022-12-23 13:24:22,566 INFO [train.py:894] (3/4) Epoch 19, batch 650, loss[loss=0.1865, simple_loss=0.2802, pruned_loss=0.0464, over 18475.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2679, pruned_loss=0.04635, over 3573161.82 frames. ], batch size: 54, lr: 6.11e-03, grad_scale: 8.0 2022-12-23 13:24:45,349 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63776.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:24:48,847 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63778.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:25:10,798 WARNING [train.py:1060] (3/4) Exclude cut with ID 4957-30119-0041-23990-0_sp0.9 from training. Duration: 20.22775 2022-12-23 13:25:28,850 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8378, 0.7069, 1.6443, 1.4762, 1.8521, 1.8843, 1.5915, 1.5814], device='cuda:3'), covar=tensor([0.2060, 0.3206, 0.2489, 0.2560, 0.1903, 0.0923, 0.2731, 0.1245], device='cuda:3'), in_proj_covar=tensor([0.0266, 0.0297, 0.0276, 0.0312, 0.0305, 0.0249, 0.0337, 0.0237], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 13:25:35,827 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63810.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:25:36,899 INFO [train.py:894] (3/4) Epoch 19, batch 700, loss[loss=0.1729, simple_loss=0.2682, pruned_loss=0.03879, over 18659.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2685, pruned_loss=0.04647, over 3604478.57 frames. ], batch size: 62, lr: 6.11e-03, grad_scale: 8.0 2022-12-23 13:25:56,206 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp1.1 from training. Duration: 24.67275 2022-12-23 13:26:15,080 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.104e+02 3.773e+02 4.476e+02 5.280e+02 9.303e+02, threshold=8.951e+02, percent-clipped=10.0 2022-12-23 13:26:25,108 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4705, 2.2289, 1.7877, 1.3673, 2.8637, 2.7270, 2.2934, 1.8552], device='cuda:3'), covar=tensor([0.0388, 0.0428, 0.0587, 0.0780, 0.0261, 0.0316, 0.0470, 0.0825], device='cuda:3'), in_proj_covar=tensor([0.0125, 0.0127, 0.0130, 0.0122, 0.0098, 0.0123, 0.0137, 0.0159], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 13:26:26,149 WARNING [train.py:1060] (3/4) Exclude cut with ID 3082-165428-0081-50734-0_sp0.9 from training. Duration: 21.8055625 2022-12-23 13:26:27,795 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63845.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:26:51,612 INFO [train.py:894] (3/4) Epoch 19, batch 750, loss[loss=0.1664, simple_loss=0.2615, pruned_loss=0.03568, over 18456.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2681, pruned_loss=0.04598, over 3628425.83 frames. ], batch size: 54, lr: 6.10e-03, grad_scale: 8.0 2022-12-23 13:27:00,992 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0_sp0.9 from training. Duration: 22.6666875 2022-12-23 13:27:07,454 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63871.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:27:16,847 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63877.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:28:00,188 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63906.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:28:04,209 WARNING [train.py:1060] (3/4) Exclude cut with ID 2411-132532-0017-82279-0_sp1.1 from training. Duration: 0.9681875 2022-12-23 13:28:06,918 INFO [train.py:894] (3/4) Epoch 19, batch 800, loss[loss=0.1455, simple_loss=0.2351, pruned_loss=0.02797, over 18618.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2678, pruned_loss=0.04606, over 3647141.34 frames. ], batch size: 45, lr: 6.10e-03, grad_scale: 8.0 2022-12-23 13:28:15,435 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63916.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:28:29,379 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=63925.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:28:29,448 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63925.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:28:32,222 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0 from training. Duration: 22.485 2022-12-23 13:28:32,572 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9132, 1.5019, 2.1055, 2.4519, 2.1236, 4.6546, 1.7815, 2.0104], device='cuda:3'), covar=tensor([0.0838, 0.1811, 0.0886, 0.0878, 0.1285, 0.0168, 0.1242, 0.1424], device='cuda:3'), in_proj_covar=tensor([0.0074, 0.0083, 0.0074, 0.0075, 0.0091, 0.0076, 0.0085, 0.0078], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 13:28:47,233 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.190e+02 3.396e+02 4.192e+02 4.908e+02 9.047e+02, threshold=8.383e+02, percent-clipped=1.0 2022-12-23 13:28:47,751 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.3821, 2.4919, 1.9983, 3.2278, 2.3218, 2.4966, 2.5244, 3.5135], device='cuda:3'), covar=tensor([0.1752, 0.3463, 0.1927, 0.2787, 0.3935, 0.0999, 0.3212, 0.0762], device='cuda:3'), in_proj_covar=tensor([0.0293, 0.0288, 0.0243, 0.0348, 0.0269, 0.0225, 0.0287, 0.0210], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 13:28:54,656 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63942.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:29:08,721 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp1.1 from training. Duration: 23.82275 2022-12-23 13:29:19,108 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0 from training. Duration: 20.77 2022-12-23 13:29:23,353 INFO [train.py:894] (3/4) Epoch 19, batch 850, loss[loss=0.1978, simple_loss=0.2812, pruned_loss=0.05724, over 18523.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2674, pruned_loss=0.04592, over 3662575.96 frames. ], batch size: 58, lr: 6.10e-03, grad_scale: 8.0 2022-12-23 13:29:26,967 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0_sp0.9 from training. Duration: 24.088875 2022-12-23 13:29:48,326 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63977.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:29:53,772 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp1.1 from training. Duration: 20.4409375 2022-12-23 13:30:43,306 INFO [train.py:894] (3/4) Epoch 19, batch 900, loss[loss=0.1888, simple_loss=0.2742, pruned_loss=0.05164, over 18670.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2674, pruned_loss=0.04604, over 3674459.08 frames. ], batch size: 62, lr: 6.10e-03, grad_scale: 8.0 2022-12-23 13:31:01,900 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64023.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:31:12,227 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0_sp0.9 from training. Duration: 22.511125 2022-12-23 13:31:13,707 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0 from training. Duration: 20.675 2022-12-23 13:31:22,423 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.188e+02 3.638e+02 4.436e+02 5.508e+02 1.633e+03, threshold=8.872e+02, percent-clipped=5.0 2022-12-23 13:31:58,436 INFO [train.py:894] (3/4) Epoch 19, batch 950, loss[loss=0.1619, simple_loss=0.2421, pruned_loss=0.04086, over 18705.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2674, pruned_loss=0.04599, over 3682654.84 frames. ], batch size: 50, lr: 6.09e-03, grad_scale: 8.0 2022-12-23 13:31:59,910 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2022-12-23 13:32:02,768 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.8015, 2.2851, 1.7505, 2.5792, 2.0263, 2.1787, 2.1370, 2.7839], device='cuda:3'), covar=tensor([0.1804, 0.3275, 0.1834, 0.2700, 0.3610, 0.1021, 0.3043, 0.0805], device='cuda:3'), in_proj_covar=tensor([0.0291, 0.0286, 0.0241, 0.0345, 0.0267, 0.0223, 0.0284, 0.0208], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 13:32:21,463 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64076.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:32:24,282 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64078.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:32:32,946 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64084.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:32:51,801 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp0.9 from training. Duration: 24.9833125 2022-12-23 13:33:12,932 INFO [train.py:894] (3/4) Epoch 19, batch 1000, loss[loss=0.1639, simple_loss=0.2462, pruned_loss=0.04085, over 18427.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2675, pruned_loss=0.04606, over 3690313.11 frames. ], batch size: 42, lr: 6.09e-03, grad_scale: 8.0 2022-12-23 13:33:24,006 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0 from training. Duration: 27.14 2022-12-23 13:33:33,171 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=64124.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:33:36,279 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=64126.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:33:37,740 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0 from training. Duration: 22.44 2022-12-23 13:33:52,924 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.707e+02 3.878e+02 4.599e+02 5.845e+02 2.549e+03, threshold=9.198e+02, percent-clipped=7.0 2022-12-23 13:34:29,462 INFO [train.py:894] (3/4) Epoch 19, batch 1050, loss[loss=0.1515, simple_loss=0.2303, pruned_loss=0.03637, over 18472.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2671, pruned_loss=0.04601, over 3696146.41 frames. ], batch size: 43, lr: 6.09e-03, grad_scale: 8.0 2022-12-23 13:34:37,103 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64166.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:34:55,622 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0060-62364-0_sp0.9 from training. Duration: 21.361125 2022-12-23 13:35:01,869 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp1.1 from training. Duration: 27.0318125 2022-12-23 13:35:06,438 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.7531, 4.1071, 4.0408, 4.6605, 4.4082, 4.2031, 4.8940, 1.4071], device='cuda:3'), covar=tensor([0.0576, 0.0587, 0.0607, 0.0671, 0.1099, 0.1060, 0.0444, 0.4825], device='cuda:3'), in_proj_covar=tensor([0.0333, 0.0218, 0.0230, 0.0255, 0.0315, 0.0260, 0.0279, 0.0272], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 13:35:12,232 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp0.9 from training. Duration: 28.638875 2022-12-23 13:35:27,773 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0 from training. Duration: 20.4 2022-12-23 13:35:29,576 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64201.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:35:44,105 INFO [train.py:894] (3/4) Epoch 19, batch 1100, loss[loss=0.177, simple_loss=0.2662, pruned_loss=0.04392, over 18590.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2666, pruned_loss=0.04603, over 3700270.00 frames. ], batch size: 51, lr: 6.09e-03, grad_scale: 8.0 2022-12-23 13:36:01,214 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0 from training. Duration: 20.025 2022-12-23 13:36:01,228 WARNING [train.py:1060] (3/4) Exclude cut with ID 2364-131735-0112-64612-0_sp0.9 from training. Duration: 20.488875 2022-12-23 13:36:05,779 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0 from training. Duration: 29.735 2022-12-23 13:36:06,124 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64225.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:36:17,520 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64232.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:36:24,143 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.216e+02 3.480e+02 4.012e+02 4.856e+02 7.809e+02, threshold=8.025e+02, percent-clipped=0.0 2022-12-23 13:36:31,624 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64242.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:36:55,170 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.9202, 5.4008, 4.9347, 2.4226, 5.4719, 4.1568, 0.5643, 3.5105], device='cuda:3'), covar=tensor([0.1784, 0.0741, 0.1053, 0.3119, 0.0545, 0.0695, 0.5053, 0.1334], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0134, 0.0151, 0.0120, 0.0136, 0.0109, 0.0139, 0.0110], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 13:36:59,350 INFO [train.py:894] (3/4) Epoch 19, batch 1150, loss[loss=0.1634, simple_loss=0.245, pruned_loss=0.04085, over 18435.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2664, pruned_loss=0.04549, over 3704000.55 frames. ], batch size: 48, lr: 6.08e-03, grad_scale: 8.0 2022-12-23 13:37:16,522 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64272.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:37:17,959 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=64273.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:37:25,999 WARNING [train.py:1060] (3/4) Exclude cut with ID 7276-92427-0014-12983-0_sp0.9 from training. Duration: 21.3055625 2022-12-23 13:37:27,642 WARNING [train.py:1060] (3/4) Exclude cut with ID 1025-75365-0008-79168-0_sp0.9 from training. Duration: 22.0666875 2022-12-23 13:37:43,692 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=64290.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:37:48,636 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64293.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:37:58,110 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5212, 1.4013, 1.2385, 1.6300, 1.6051, 3.0200, 1.2942, 1.5874], device='cuda:3'), covar=tensor([0.0810, 0.1728, 0.1016, 0.0897, 0.1389, 0.0249, 0.1334, 0.1484], device='cuda:3'), in_proj_covar=tensor([0.0072, 0.0081, 0.0071, 0.0073, 0.0089, 0.0074, 0.0084, 0.0076], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 13:38:15,992 INFO [train.py:894] (3/4) Epoch 19, batch 1200, loss[loss=0.1948, simple_loss=0.2848, pruned_loss=0.05238, over 18428.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2657, pruned_loss=0.04478, over 3706042.51 frames. ], batch size: 48, lr: 6.08e-03, grad_scale: 8.0 2022-12-23 13:38:22,193 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2022-12-23 13:38:54,836 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.279e+02 3.709e+02 4.310e+02 5.274e+02 1.425e+03, threshold=8.620e+02, percent-clipped=2.0 2022-12-23 13:39:07,266 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64345.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:39:17,329 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0 from training. Duration: 20.26 2022-12-23 13:39:29,917 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0030-9324-0_sp0.9 from training. Duration: 21.3444375 2022-12-23 13:39:31,050 INFO [train.py:894] (3/4) Epoch 19, batch 1250, loss[loss=0.2028, simple_loss=0.2873, pruned_loss=0.05914, over 18725.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2657, pruned_loss=0.04462, over 3707797.25 frames. ], batch size: 52, lr: 6.08e-03, grad_scale: 8.0 2022-12-23 13:39:58,526 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64379.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:40:26,907 WARNING [train.py:1060] (3/4) Exclude cut with ID 497-129325-0061-62254-0_sp1.1 from training. Duration: 0.97725 2022-12-23 13:40:39,193 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64406.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:40:47,314 INFO [train.py:894] (3/4) Epoch 19, batch 1300, loss[loss=0.1846, simple_loss=0.271, pruned_loss=0.04913, over 18584.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2659, pruned_loss=0.04442, over 3709651.08 frames. ], batch size: 96, lr: 6.08e-03, grad_scale: 8.0 2022-12-23 13:40:58,151 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2022-12-23 13:41:09,917 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0_sp0.9 from training. Duration: 22.97225 2022-12-23 13:41:24,360 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.334e+02 3.220e+02 4.296e+02 5.683e+02 9.700e+02, threshold=8.592e+02, percent-clipped=5.0 2022-12-23 13:41:40,687 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0047-39922-0_sp0.9 from training. Duration: 21.97775 2022-12-23 13:41:55,147 WARNING [train.py:1060] (3/4) Exclude cut with ID 1112-1043-0006-89194-0_sp0.9 from training. Duration: 21.8333125 2022-12-23 13:42:00,555 INFO [train.py:894] (3/4) Epoch 19, batch 1350, loss[loss=0.1695, simple_loss=0.259, pruned_loss=0.03995, over 18574.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.265, pruned_loss=0.04402, over 3711322.77 frames. ], batch size: 49, lr: 6.07e-03, grad_scale: 8.0 2022-12-23 13:42:06,439 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0 from training. Duration: 20.47 2022-12-23 13:42:06,757 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64465.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:42:08,051 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64466.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:42:39,631 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2022-12-23 13:42:59,923 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64501.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:43:11,906 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0037-39912-0_sp0.9 from training. Duration: 20.67225 2022-12-23 13:43:15,489 INFO [train.py:894] (3/4) Epoch 19, batch 1400, loss[loss=0.2031, simple_loss=0.2953, pruned_loss=0.05545, over 18699.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2643, pruned_loss=0.04384, over 3712412.86 frames. ], batch size: 65, lr: 6.07e-03, grad_scale: 8.0 2022-12-23 13:43:19,651 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=64514.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:43:34,259 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp0.9 from training. Duration: 25.2444375 2022-12-23 13:43:40,624 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64526.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:43:55,840 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.643e+02 3.417e+02 3.883e+02 4.818e+02 7.997e+02, threshold=7.766e+02, percent-clipped=0.0 2022-12-23 13:43:57,322 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0021-76797-0_sp0.9 from training. Duration: 21.1445 2022-12-23 13:44:13,694 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=64549.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:44:31,857 INFO [train.py:894] (3/4) Epoch 19, batch 1450, loss[loss=0.1505, simple_loss=0.2335, pruned_loss=0.03371, over 18680.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2643, pruned_loss=0.04408, over 3713103.73 frames. ], batch size: 48, lr: 6.07e-03, grad_scale: 8.0 2022-12-23 13:44:48,596 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64572.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:45:11,650 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp0.9 from training. Duration: 33.038875 2022-12-23 13:45:11,814 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64588.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:45:13,394 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64589.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:45:23,752 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7907, 2.1754, 1.9338, 1.4456, 2.8019, 2.7992, 2.3674, 1.8501], device='cuda:3'), covar=tensor([0.0293, 0.0472, 0.0574, 0.0803, 0.0264, 0.0338, 0.0430, 0.0882], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0126, 0.0130, 0.0121, 0.0097, 0.0121, 0.0135, 0.0157], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 13:45:46,438 INFO [train.py:894] (3/4) Epoch 19, batch 1500, loss[loss=0.1587, simple_loss=0.2461, pruned_loss=0.03567, over 18541.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2651, pruned_loss=0.04446, over 3713835.41 frames. ], batch size: 47, lr: 6.07e-03, grad_scale: 8.0 2022-12-23 13:45:47,988 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64291-0000-16059-0_sp0.9 from training. Duration: 20.0944375 2022-12-23 13:46:00,458 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=64620.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:46:03,104 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp1.1 from training. Duration: 20.4 2022-12-23 13:46:12,106 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0 from training. Duration: 20.085 2022-12-23 13:46:22,215 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0_sp0.9 from training. Duration: 23.07775 2022-12-23 13:46:24,999 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.225e+02 3.474e+02 4.325e+02 5.384e+02 8.400e+02, threshold=8.650e+02, percent-clipped=2.0 2022-12-23 13:46:42,647 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64648.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:46:45,635 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64650.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:47:02,832 INFO [train.py:894] (3/4) Epoch 19, batch 1550, loss[loss=0.1869, simple_loss=0.2849, pruned_loss=0.04441, over 18671.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2653, pruned_loss=0.04447, over 3714021.21 frames. ], batch size: 78, lr: 6.07e-03, grad_scale: 8.0 2022-12-23 13:47:07,203 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp0.9 from training. Duration: 24.9333125 2022-12-23 13:47:29,574 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64679.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:47:41,357 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7077, 2.4460, 2.1398, 0.9609, 2.0063, 2.1110, 1.9706, 2.1768], device='cuda:3'), covar=tensor([0.0653, 0.0586, 0.1168, 0.1781, 0.1431, 0.1464, 0.1468, 0.0819], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0181, 0.0201, 0.0188, 0.0206, 0.0195, 0.0210, 0.0197], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 13:47:51,533 WARNING [train.py:1060] (3/4) Exclude cut with ID 5118-111612-0016-124680-0_sp0.9 from training. Duration: 20.388875 2022-12-23 13:47:59,110 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp1.1 from training. Duration: 20.3590625 2022-12-23 13:48:02,707 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64701.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:48:05,808 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64703.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:48:15,355 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64709.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:48:17,777 INFO [train.py:894] (3/4) Epoch 19, batch 1600, loss[loss=0.1639, simple_loss=0.2582, pruned_loss=0.03475, over 18598.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2655, pruned_loss=0.04472, over 3714131.85 frames. ], batch size: 51, lr: 6.06e-03, grad_scale: 16.0 2022-12-23 13:48:41,817 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=64727.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:48:57,058 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.475e+02 3.648e+02 4.518e+02 5.484e+02 9.669e+02, threshold=9.035e+02, percent-clipped=5.0 2022-12-23 13:49:07,591 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0_sp1.1 from training. Duration: 0.836375 2022-12-23 13:49:09,524 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64745.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:49:19,909 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64752.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:49:33,375 INFO [train.py:894] (3/4) Epoch 19, batch 1650, loss[loss=0.2552, simple_loss=0.3228, pruned_loss=0.09381, over 18623.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2656, pruned_loss=0.04571, over 3713476.41 frames. ], batch size: 179, lr: 6.06e-03, grad_scale: 16.0 2022-12-23 13:49:38,517 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64764.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:49:50,904 WARNING [train.py:1060] (3/4) Exclude cut with ID 8565-290391-0049-67394-0_sp0.9 from training. Duration: 21.3166875 2022-12-23 13:50:21,087 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0029-104863-0_sp0.9 from training. Duration: 22.1055625 2022-12-23 13:50:31,331 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp1.1 from training. Duration: 21.77725 2022-12-23 13:50:42,084 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64806.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:50:49,673 INFO [train.py:894] (3/4) Epoch 19, batch 1700, loss[loss=0.1916, simple_loss=0.2845, pruned_loss=0.0494, over 18714.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2669, pruned_loss=0.04711, over 3713170.13 frames. ], batch size: 54, lr: 6.06e-03, grad_scale: 16.0 2022-12-23 13:50:52,522 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp0.9 from training. Duration: 27.8166875 2022-12-23 13:50:52,899 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64813.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:51:04,283 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64821.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:51:18,631 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp1.1 from training. Duration: 22.5090625 2022-12-23 13:51:26,322 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0 from training. Duration: 25.035 2022-12-23 13:51:29,131 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.637e+02 3.847e+02 4.863e+02 6.028e+02 1.438e+03, threshold=9.726e+02, percent-clipped=3.0 2022-12-23 13:51:40,318 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6209, 1.8606, 1.4331, 2.1295, 2.3478, 1.5903, 1.3974, 1.2479], device='cuda:3'), covar=tensor([0.1826, 0.1724, 0.1596, 0.1008, 0.1178, 0.1059, 0.2111, 0.1475], device='cuda:3'), in_proj_covar=tensor([0.0241, 0.0220, 0.0211, 0.0196, 0.0257, 0.0192, 0.0218, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 13:51:44,042 WARNING [train.py:1060] (3/4) Exclude cut with ID 774-127930-0014-10412-0_sp1.1 from training. Duration: 0.95 2022-12-23 13:52:03,753 INFO [train.py:894] (3/4) Epoch 19, batch 1750, loss[loss=0.1739, simple_loss=0.2572, pruned_loss=0.04532, over 18450.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2675, pruned_loss=0.04818, over 3713465.08 frames. ], batch size: 48, lr: 6.06e-03, grad_scale: 8.0 2022-12-23 13:52:03,778 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp0.9 from training. Duration: 0.92225 2022-12-23 13:52:08,069 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2022-12-23 13:52:10,739 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3813, 1.1742, 1.1874, 1.5517, 1.3551, 2.1365, 1.1834, 1.2419], device='cuda:3'), covar=tensor([0.0753, 0.1518, 0.1155, 0.0775, 0.1181, 0.0417, 0.1229, 0.1343], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0082, 0.0072, 0.0074, 0.0089, 0.0075, 0.0084, 0.0077], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 13:52:29,289 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0 from training. Duration: 21.97 2022-12-23 13:52:44,608 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64888.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:52:47,327 WARNING [train.py:1060] (3/4) Exclude cut with ID 7492-105653-0055-62765-0_sp0.9 from training. Duration: 21.97225 2022-12-23 13:52:48,694 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp0.9 from training. Duration: 25.3333125 2022-12-23 13:52:58,723 WARNING [train.py:1060] (3/4) Exclude cut with ID 5172-29468-0015-19128-0_sp0.9 from training. Duration: 21.5055625 2022-12-23 13:53:08,102 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp1.1 from training. Duration: 20.72725 2022-12-23 13:53:20,447 INFO [train.py:894] (3/4) Epoch 19, batch 1800, loss[loss=0.177, simple_loss=0.2502, pruned_loss=0.05187, over 18605.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2692, pruned_loss=0.05029, over 3713346.50 frames. ], batch size: 45, lr: 6.05e-03, grad_scale: 8.0 2022-12-23 13:53:41,068 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp0.9 from training. Duration: 26.32775 2022-12-23 13:53:57,611 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=64936.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:54:00,265 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.796e+02 4.805e+02 5.504e+02 6.542e+02 1.511e+03, threshold=1.101e+03, percent-clipped=5.0 2022-12-23 13:54:11,118 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64945.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:54:15,888 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0 from training. Duration: 20.025 2022-12-23 13:54:20,127 WARNING [train.py:1060] (3/4) Exclude cut with ID 6709-74022-0004-86860-0_sp1.1 from training. Duration: 0.9409375 2022-12-23 13:54:20,139 WARNING [train.py:1060] (3/4) Exclude cut with ID 4757-1811-0023-62229-0_sp0.9 from training. Duration: 21.37775 2022-12-23 13:54:35,440 INFO [train.py:894] (3/4) Epoch 19, batch 1850, loss[loss=0.2264, simple_loss=0.2989, pruned_loss=0.07694, over 18480.00 frames. ], tot_loss[loss=0.1869, simple_loss=0.2702, pruned_loss=0.05182, over 3714286.23 frames. ], batch size: 54, lr: 6.05e-03, grad_scale: 8.0 2022-12-23 13:54:41,261 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0004-25974-0_sp0.9 from training. Duration: 21.17225 2022-12-23 13:54:42,603 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp0.9 from training. Duration: 27.511125 2022-12-23 13:54:57,594 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64976.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:55:14,983 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0 from training. Duration: 22.8 2022-12-23 13:55:19,215 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0 from training. Duration: 22.585 2022-12-23 13:55:35,653 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65001.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:55:39,918 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65004.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:55:50,582 INFO [train.py:894] (3/4) Epoch 19, batch 1900, loss[loss=0.2353, simple_loss=0.3013, pruned_loss=0.08461, over 18599.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2704, pruned_loss=0.05247, over 3712867.74 frames. ], batch size: 176, lr: 6.05e-03, grad_scale: 8.0 2022-12-23 13:55:50,610 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0001-146967-0_sp0.9 from training. Duration: 22.0166875 2022-12-23 13:56:05,548 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp1.1 from training. Duration: 24.395375 2022-12-23 13:56:11,674 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp0.9 from training. Duration: 27.47775 2022-12-23 13:56:16,179 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp0.9 from training. Duration: 24.8833125 2022-12-23 13:56:19,191 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0 from training. Duration: 23.39 2022-12-23 13:56:25,166 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp0.9 from training. Duration: 28.72225 2022-12-23 13:56:29,857 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65037.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 13:56:30,838 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.978e+02 4.683e+02 6.050e+02 7.570e+02 2.134e+03, threshold=1.210e+03, percent-clipped=7.0 2022-12-23 13:56:32,157 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2022-12-23 13:56:35,771 WARNING [train.py:1060] (3/4) Exclude cut with ID 585-294811-0110-133686-0_sp0.9 from training. Duration: 20.8944375 2022-12-23 13:56:49,106 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0_sp0.9 from training. Duration: 23.8444375 2022-12-23 13:56:49,253 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65049.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:56:58,564 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65055.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:57:04,424 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65059.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:57:07,088 INFO [train.py:894] (3/4) Epoch 19, batch 1950, loss[loss=0.221, simple_loss=0.2898, pruned_loss=0.07611, over 18577.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2713, pruned_loss=0.05394, over 3713741.67 frames. ], batch size: 49, lr: 6.05e-03, grad_scale: 8.0 2022-12-23 13:57:12,676 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0 from training. Duration: 25.85 2022-12-23 13:57:12,683 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0 from training. Duration: 21.39 2022-12-23 13:57:23,887 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0 from training. Duration: 27.92 2022-12-23 13:57:52,654 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2022-12-23 13:57:53,308 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0039-130165-0_sp0.9 from training. Duration: 20.661125 2022-12-23 13:58:08,886 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65101.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:58:18,558 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0043-15874-0_sp0.9 from training. Duration: 20.07225 2022-12-23 13:58:18,762 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65108.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:58:23,139 INFO [train.py:894] (3/4) Epoch 19, batch 2000, loss[loss=0.1596, simple_loss=0.231, pruned_loss=0.04406, over 18583.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.271, pruned_loss=0.05432, over 3714119.54 frames. ], batch size: 41, lr: 6.04e-03, grad_scale: 8.0 2022-12-23 13:58:24,884 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0 from training. Duration: 21.01 2022-12-23 13:58:31,461 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65116.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:58:38,878 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65121.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 13:58:40,370 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1631, 2.7970, 2.7094, 1.1007, 2.8988, 2.0894, 0.7550, 1.8078], device='cuda:3'), covar=tensor([0.2301, 0.1505, 0.1697, 0.3827, 0.1323, 0.1134, 0.4416, 0.1806], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0138, 0.0155, 0.0123, 0.0141, 0.0112, 0.0144, 0.0112], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 13:59:05,244 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.657e+02 4.505e+02 5.231e+02 6.101e+02 1.148e+03, threshold=1.046e+03, percent-clipped=0.0 2022-12-23 13:59:17,858 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.3355, 1.6872, 1.9300, 1.0902, 1.1340, 2.0575, 1.7489, 1.6581], device='cuda:3'), covar=tensor([0.0725, 0.0312, 0.0318, 0.0353, 0.0371, 0.0421, 0.0245, 0.0659], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0168, 0.0125, 0.0138, 0.0148, 0.0143, 0.0161, 0.0169], device='cuda:3'), out_proj_covar=tensor([1.1402e-04, 1.2971e-04, 9.5270e-05, 1.0358e-04, 1.1174e-04, 1.1045e-04, 1.2482e-04, 1.3015e-04], device='cuda:3') 2022-12-23 13:59:21,985 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6589, 1.3414, 1.5448, 2.0578, 1.6415, 3.4794, 1.3137, 1.4521], device='cuda:3'), covar=tensor([0.1056, 0.2411, 0.1241, 0.1046, 0.1692, 0.0309, 0.1896, 0.2119], device='cuda:3'), in_proj_covar=tensor([0.0072, 0.0081, 0.0072, 0.0074, 0.0089, 0.0074, 0.0084, 0.0076], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 13:59:35,719 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0 from training. Duration: 20.65 2022-12-23 13:59:38,664 INFO [train.py:894] (3/4) Epoch 19, batch 2050, loss[loss=0.1999, simple_loss=0.2908, pruned_loss=0.05447, over 18470.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2719, pruned_loss=0.0551, over 3714232.68 frames. ], batch size: 54, lr: 6.04e-03, grad_scale: 8.0 2022-12-23 13:59:42,318 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0 from training. Duration: 21.46 2022-12-23 13:59:51,030 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65169.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 14:00:13,746 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0053, 1.9667, 2.3226, 1.2840, 2.3714, 2.2891, 1.6832, 2.6399], device='cuda:3'), covar=tensor([0.1144, 0.1826, 0.1284, 0.1926, 0.0708, 0.1209, 0.2203, 0.0610], device='cuda:3'), in_proj_covar=tensor([0.0200, 0.0213, 0.0208, 0.0195, 0.0178, 0.0218, 0.0216, 0.0192], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 14:00:28,415 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0 from training. Duration: 0.92 2022-12-23 14:00:35,607 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0_sp0.9 from training. Duration: 23.7666875 2022-12-23 14:00:54,016 INFO [train.py:894] (3/4) Epoch 19, batch 2100, loss[loss=0.1982, simple_loss=0.2818, pruned_loss=0.05731, over 18700.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2724, pruned_loss=0.05589, over 3714653.09 frames. ], batch size: 65, lr: 6.04e-03, grad_scale: 8.0 2022-12-23 14:01:07,512 WARNING [train.py:1060] (3/4) Exclude cut with ID 8544-281189-0060-101339-0_sp0.9 from training. Duration: 20.861125 2022-12-23 14:01:18,856 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0_sp0.9 from training. Duration: 22.711125 2022-12-23 14:01:36,006 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.467e+02 4.618e+02 5.775e+02 7.305e+02 1.300e+03, threshold=1.155e+03, percent-clipped=2.0 2022-12-23 14:01:46,557 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65245.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 14:02:00,340 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp1.1 from training. Duration: 22.986375 2022-12-23 14:02:10,569 INFO [train.py:894] (3/4) Epoch 19, batch 2150, loss[loss=0.1793, simple_loss=0.249, pruned_loss=0.05475, over 18417.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.272, pruned_loss=0.05587, over 3713440.81 frames. ], batch size: 42, lr: 6.04e-03, grad_scale: 8.0 2022-12-23 14:02:15,193 WARNING [train.py:1060] (3/4) Exclude cut with ID 8040-260924-0003-80960-0_sp0.9 from training. Duration: 22.07225 2022-12-23 14:02:19,377 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0045-26330-0_sp0.9 from training. Duration: 20.3055625 2022-12-23 14:02:22,682 WARNING [train.py:1060] (3/4) Exclude cut with ID 6356-271890-0060-94317-0_sp0.9 from training. Duration: 20.72225 2022-12-23 14:02:40,985 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp1.1 from training. Duration: 22.4818125 2022-12-23 14:02:47,943 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65285.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 14:02:59,880 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65293.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 14:03:07,070 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp0.9 from training. Duration: 25.0944375 2022-12-23 14:03:11,870 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0 from training. Duration: 21.515 2022-12-23 14:03:13,998 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2022-12-23 14:03:16,493 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65304.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 14:03:17,743 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp0.9 from training. Duration: 27.02225 2022-12-23 14:03:24,815 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0010-62480-0_sp0.9 from training. Duration: 22.22225 2022-12-23 14:03:27,402 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.96 vs. limit=5.0 2022-12-23 14:03:27,776 INFO [train.py:894] (3/4) Epoch 19, batch 2200, loss[loss=0.1738, simple_loss=0.2554, pruned_loss=0.04612, over 18539.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2709, pruned_loss=0.05505, over 3714078.41 frames. ], batch size: 47, lr: 6.04e-03, grad_scale: 8.0 2022-12-23 14:03:30,668 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0085-44554-0_sp0.9 from training. Duration: 20.85 2022-12-23 14:04:00,729 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65332.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 14:04:04,886 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0 from training. Duration: 21.54 2022-12-23 14:04:09,154 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.710e+02 4.175e+02 5.130e+02 6.233e+02 1.122e+03, threshold=1.026e+03, percent-clipped=0.0 2022-12-23 14:04:09,188 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp1.1 from training. Duration: 20.5318125 2022-12-23 14:04:17,687 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0012-134311-0_sp0.9 from training. Duration: 21.9333125 2022-12-23 14:04:21,775 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65346.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 14:04:30,290 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65352.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 14:04:40,344 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65359.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 14:04:43,095 INFO [train.py:894] (3/4) Epoch 19, batch 2250, loss[loss=0.194, simple_loss=0.277, pruned_loss=0.0555, over 18553.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2718, pruned_loss=0.05532, over 3713877.82 frames. ], batch size: 97, lr: 6.03e-03, grad_scale: 8.0 2022-12-23 14:05:09,253 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0025-130151-0_sp0.9 from training. Duration: 21.7944375 2022-12-23 14:05:20,856 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0_sp0.9 from training. Duration: 22.4666875 2022-12-23 14:05:29,233 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0 from training. Duration: 21.635 2022-12-23 14:05:35,149 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0_sp0.9 from training. Duration: 24.038875 2022-12-23 14:05:44,174 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65401.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 14:05:52,553 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65407.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 14:05:54,028 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65408.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 14:05:58,207 INFO [train.py:894] (3/4) Epoch 19, batch 2300, loss[loss=0.1721, simple_loss=0.253, pruned_loss=0.04562, over 18392.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2722, pruned_loss=0.05531, over 3713825.92 frames. ], batch size: 46, lr: 6.03e-03, grad_scale: 8.0 2022-12-23 14:05:58,395 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65411.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 14:06:20,065 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp1.1 from training. Duration: 21.786375 2022-12-23 14:06:30,495 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0 from training. Duration: 20.22 2022-12-23 14:06:39,237 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.902e+02 4.557e+02 5.632e+02 7.135e+02 1.827e+03, threshold=1.126e+03, percent-clipped=4.0 2022-12-23 14:06:56,386 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65449.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 14:07:06,643 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65456.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 14:07:07,413 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.82 vs. limit=5.0 2022-12-23 14:07:11,302 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8828, 1.4953, 1.7849, 2.3036, 1.9994, 4.5518, 1.7770, 1.7672], device='cuda:3'), covar=tensor([0.0956, 0.2056, 0.1212, 0.1091, 0.1492, 0.0203, 0.1518, 0.1708], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0082, 0.0073, 0.0074, 0.0089, 0.0075, 0.0085, 0.0077], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 14:07:14,132 INFO [train.py:894] (3/4) Epoch 19, batch 2350, loss[loss=0.179, simple_loss=0.251, pruned_loss=0.05354, over 18543.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2706, pruned_loss=0.05478, over 3713956.97 frames. ], batch size: 44, lr: 6.03e-03, grad_scale: 8.0 2022-12-23 14:07:57,167 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2022-12-23 14:07:58,538 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9322, 1.9368, 1.4853, 2.0707, 2.1171, 1.8385, 2.6053, 2.0226], device='cuda:3'), covar=tensor([0.0906, 0.1565, 0.2779, 0.1706, 0.1784, 0.0939, 0.0947, 0.1234], device='cuda:3'), in_proj_covar=tensor([0.0177, 0.0207, 0.0250, 0.0290, 0.0237, 0.0191, 0.0208, 0.0204], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 14:08:00,140 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2147, 2.0790, 1.4078, 2.1063, 1.8161, 1.7397, 1.8183, 2.2334], device='cuda:3'), covar=tensor([0.1969, 0.2469, 0.1896, 0.2186, 0.2710, 0.1183, 0.2498, 0.0805], device='cuda:3'), in_proj_covar=tensor([0.0288, 0.0284, 0.0240, 0.0343, 0.0266, 0.0223, 0.0281, 0.0207], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 14:08:15,799 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.46 vs. limit=5.0 2022-12-23 14:08:30,815 INFO [train.py:894] (3/4) Epoch 19, batch 2400, loss[loss=0.1546, simple_loss=0.2325, pruned_loss=0.03839, over 18532.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2712, pruned_loss=0.05528, over 3714856.82 frames. ], batch size: 44, lr: 6.03e-03, grad_scale: 8.0 2022-12-23 14:08:30,859 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0 from training. Duration: 25.285 2022-12-23 14:09:02,811 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65532.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 14:09:10,949 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.058e+02 4.637e+02 5.730e+02 7.137e+02 1.631e+03, threshold=1.146e+03, percent-clipped=2.0 2022-12-23 14:09:35,691 WARNING [train.py:1060] (3/4) Exclude cut with ID 811-130148-0001-63453-0_sp0.9 from training. Duration: 20.861125 2022-12-23 14:09:46,575 INFO [train.py:894] (3/4) Epoch 19, batch 2450, loss[loss=0.194, simple_loss=0.2686, pruned_loss=0.05972, over 18418.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2705, pruned_loss=0.05475, over 3713888.64 frames. ], batch size: 48, lr: 6.02e-03, grad_scale: 8.0 2022-12-23 14:09:57,720 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0 from training. Duration: 20.88 2022-12-23 14:10:05,716 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7019, 1.5434, 1.7260, 1.7392, 1.1432, 3.8783, 1.6265, 2.0712], device='cuda:3'), covar=tensor([0.3088, 0.2073, 0.1893, 0.2026, 0.1500, 0.0175, 0.1555, 0.0847], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0117, 0.0126, 0.0120, 0.0102, 0.0097, 0.0092, 0.0088], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 14:10:31,125 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0_sp0.9 from training. Duration: 23.4166875 2022-12-23 14:10:35,935 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.5270, 2.8609, 2.6383, 1.3310, 2.6015, 2.4045, 2.2101, 2.5424], device='cuda:3'), covar=tensor([0.0569, 0.0589, 0.1404, 0.1691, 0.1341, 0.1272, 0.1367, 0.1005], device='cuda:3'), in_proj_covar=tensor([0.0173, 0.0183, 0.0205, 0.0190, 0.0209, 0.0199, 0.0212, 0.0201], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 14:10:35,949 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65593.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 14:11:04,066 INFO [train.py:894] (3/4) Epoch 19, batch 2500, loss[loss=0.1934, simple_loss=0.2764, pruned_loss=0.05524, over 18459.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2694, pruned_loss=0.05438, over 3713432.75 frames. ], batch size: 54, lr: 6.02e-03, grad_scale: 8.0 2022-12-23 14:11:36,389 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65632.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 14:11:44,822 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.293e+02 4.318e+02 5.164e+02 6.581e+02 1.348e+03, threshold=1.033e+03, percent-clipped=4.0 2022-12-23 14:11:44,928 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0 from training. Duration: 21.24 2022-12-23 14:11:46,310 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0_sp0.9 from training. Duration: 23.9055625 2022-12-23 14:11:49,452 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65641.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 14:12:19,911 INFO [train.py:894] (3/4) Epoch 19, batch 2550, loss[loss=0.1807, simple_loss=0.2619, pruned_loss=0.04979, over 18435.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2695, pruned_loss=0.0544, over 3712757.20 frames. ], batch size: 48, lr: 6.02e-03, grad_scale: 8.0 2022-12-23 14:12:20,008 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp0.9 from training. Duration: 25.988875 2022-12-23 14:12:28,991 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0001-134300-0_sp0.9 from training. Duration: 20.67225 2022-12-23 14:12:48,992 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65680.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 14:12:49,293 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8305, 1.8310, 1.5834, 1.8863, 2.0511, 1.9793, 2.1197, 1.4200], device='cuda:3'), covar=tensor([0.0323, 0.0279, 0.0462, 0.0181, 0.0190, 0.0412, 0.0257, 0.0347], device='cuda:3'), in_proj_covar=tensor([0.0093, 0.0127, 0.0152, 0.0124, 0.0117, 0.0119, 0.0097, 0.0127], device='cuda:3'), out_proj_covar=tensor([7.4751e-05, 1.0137e-04, 1.2605e-04, 9.9615e-05, 9.5047e-05, 9.1652e-05, 7.6212e-05, 1.0001e-04], device='cuda:3') 2022-12-23 14:13:17,340 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0 from training. Duration: 20.34 2022-12-23 14:13:35,176 INFO [train.py:894] (3/4) Epoch 19, batch 2600, loss[loss=0.1956, simple_loss=0.276, pruned_loss=0.0576, over 18430.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2695, pruned_loss=0.05444, over 3713731.08 frames. ], batch size: 48, lr: 6.02e-03, grad_scale: 8.0 2022-12-23 14:13:35,580 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65711.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 14:14:16,262 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.847e+02 4.482e+02 5.428e+02 6.828e+02 2.111e+03, threshold=1.086e+03, percent-clipped=5.0 2022-12-23 14:14:24,351 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.1569, 5.2209, 4.6479, 2.3078, 5.3076, 4.1078, 0.9193, 3.6669], device='cuda:3'), covar=tensor([0.1757, 0.0925, 0.1400, 0.3142, 0.0649, 0.0653, 0.5184, 0.1134], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0141, 0.0158, 0.0124, 0.0142, 0.0113, 0.0145, 0.0114], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 14:14:29,848 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp0.9 from training. Duration: 25.061125 2022-12-23 14:14:33,509 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.6181, 3.9946, 3.9855, 4.4700, 4.2330, 4.0942, 4.7332, 1.3611], device='cuda:3'), covar=tensor([0.0677, 0.0623, 0.0606, 0.0842, 0.1303, 0.1111, 0.0532, 0.5129], device='cuda:3'), in_proj_covar=tensor([0.0344, 0.0226, 0.0236, 0.0267, 0.0327, 0.0272, 0.0290, 0.0283], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 14:14:42,553 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0 from training. Duration: 0.83 2022-12-23 14:14:48,294 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65759.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 14:14:50,977 INFO [train.py:894] (3/4) Epoch 19, batch 2650, loss[loss=0.1945, simple_loss=0.2857, pruned_loss=0.05162, over 18620.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2698, pruned_loss=0.05446, over 3714072.02 frames. ], batch size: 53, lr: 6.01e-03, grad_scale: 8.0 2022-12-23 14:15:06,286 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0 from training. Duration: 24.73 2022-12-23 14:15:17,865 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0 from training. Duration: 23.965 2022-12-23 14:15:23,367 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5602, 2.3444, 1.8853, 0.9073, 1.8122, 1.9586, 1.6775, 2.0809], device='cuda:3'), covar=tensor([0.0630, 0.0526, 0.1234, 0.1674, 0.1233, 0.1505, 0.1654, 0.0809], device='cuda:3'), in_proj_covar=tensor([0.0173, 0.0182, 0.0204, 0.0190, 0.0208, 0.0200, 0.0213, 0.0201], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 14:15:25,883 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0030-146996-0_sp0.9 from training. Duration: 22.088875 2022-12-23 14:15:43,008 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0_sp0.9 from training. Duration: 23.6 2022-12-23 14:16:06,595 INFO [train.py:894] (3/4) Epoch 19, batch 2700, loss[loss=0.1713, simple_loss=0.2451, pruned_loss=0.04879, over 18473.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2701, pruned_loss=0.05443, over 3714321.58 frames. ], batch size: 43, lr: 6.01e-03, grad_scale: 8.0 2022-12-23 14:16:47,921 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.953e+02 4.141e+02 5.298e+02 6.563e+02 1.106e+03, threshold=1.060e+03, percent-clipped=1.0 2022-12-23 14:16:48,421 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65838.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 14:17:23,085 INFO [train.py:894] (3/4) Epoch 19, batch 2750, loss[loss=0.1803, simple_loss=0.2565, pruned_loss=0.05207, over 18552.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2709, pruned_loss=0.05509, over 3714319.22 frames. ], batch size: 49, lr: 6.01e-03, grad_scale: 8.0 2022-12-23 14:17:24,565 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0 from training. Duration: 23.795 2022-12-23 14:17:39,895 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp1.1 from training. Duration: 21.5409375 2022-12-23 14:17:42,756 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp0.9 from training. Duration: 24.97775 2022-12-23 14:17:53,635 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0_sp0.9 from training. Duration: 23.3444375 2022-12-23 14:18:01,780 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2483, 2.7837, 2.7596, 1.1540, 2.8540, 2.1381, 0.5327, 1.6849], device='cuda:3'), covar=tensor([0.2093, 0.1296, 0.1542, 0.3476, 0.1189, 0.1077, 0.4631, 0.1660], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0142, 0.0158, 0.0124, 0.0144, 0.0115, 0.0146, 0.0115], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 14:18:04,910 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65888.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 14:18:20,210 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0_sp0.9 from training. Duration: 23.2 2022-12-23 14:18:22,547 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65899.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 14:18:27,467 WARNING [train.py:1060] (3/4) Exclude cut with ID 5653-46179-0060-117930-0_sp0.9 from training. Duration: 21.17225 2022-12-23 14:18:39,156 INFO [train.py:894] (3/4) Epoch 19, batch 2800, loss[loss=0.2267, simple_loss=0.3021, pruned_loss=0.07564, over 18625.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2705, pruned_loss=0.05494, over 3713993.64 frames. ], batch size: 53, lr: 6.01e-03, grad_scale: 8.0 2022-12-23 14:18:46,042 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2022-12-23 14:18:48,135 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp0.9 from training. Duration: 24.6555625 2022-12-23 14:19:20,967 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.932e+02 4.793e+02 6.010e+02 7.523e+02 1.271e+03, threshold=1.202e+03, percent-clipped=4.0 2022-12-23 14:19:24,466 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7744, 1.6715, 1.3587, 1.5502, 1.8421, 1.6352, 2.1142, 1.8880], device='cuda:3'), covar=tensor([0.0919, 0.1673, 0.2753, 0.1786, 0.1867, 0.0968, 0.1042, 0.1209], device='cuda:3'), in_proj_covar=tensor([0.0178, 0.0208, 0.0251, 0.0291, 0.0239, 0.0191, 0.0209, 0.0205], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 14:19:25,609 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65941.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 14:19:27,955 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2022-12-23 14:19:43,694 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0 from training. Duration: 20.44 2022-12-23 14:19:56,142 INFO [train.py:894] (3/4) Epoch 19, batch 2850, loss[loss=0.147, simple_loss=0.2265, pruned_loss=0.03378, over 18645.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2708, pruned_loss=0.05537, over 3713494.09 frames. ], batch size: 41, lr: 6.01e-03, grad_scale: 8.0 2022-12-23 14:19:57,831 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0_sp0.9 from training. Duration: 23.45 2022-12-23 14:20:02,557 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.7079, 0.6139, 0.5795, 0.6894, 0.8442, 0.7728, 0.7612, 0.6829], device='cuda:3'), covar=tensor([0.0202, 0.0207, 0.0427, 0.0183, 0.0200, 0.0241, 0.0178, 0.0231], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0127, 0.0153, 0.0124, 0.0116, 0.0119, 0.0097, 0.0127], device='cuda:3'), out_proj_covar=tensor([7.5234e-05, 1.0120e-04, 1.2669e-04, 9.9672e-05, 9.4433e-05, 9.1729e-05, 7.6167e-05, 1.0046e-04], device='cuda:3') 2022-12-23 14:20:27,921 WARNING [train.py:1060] (3/4) Exclude cut with ID 6945-60535-0076-12784-0_sp0.9 from training. Duration: 20.52225 2022-12-23 14:20:35,687 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0 from training. Duration: 22.19 2022-12-23 14:20:39,878 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65989.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 14:20:45,986 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp1.1 from training. Duration: 25.3818125 2022-12-23 14:21:03,312 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp0.9 from training. Duration: 28.0944375 2022-12-23 14:21:09,548 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0_sp0.9 from training. Duration: 22.9444375 2022-12-23 14:21:15,611 INFO [train.py:894] (3/4) Epoch 19, batch 2900, loss[loss=0.1912, simple_loss=0.274, pruned_loss=0.05421, over 18528.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2713, pruned_loss=0.05566, over 3713535.31 frames. ], batch size: 55, lr: 6.00e-03, grad_scale: 8.0 2022-12-23 14:21:19,067 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp1.1 from training. Duration: 21.6318125 2022-12-23 14:21:36,722 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0 from training. Duration: 23.695 2022-12-23 14:21:55,734 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.668e+02 4.307e+02 5.420e+02 6.494e+02 1.228e+03, threshold=1.084e+03, percent-clipped=1.0 2022-12-23 14:22:02,320 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0 from training. Duration: 23.955 2022-12-23 14:22:30,196 INFO [train.py:894] (3/4) Epoch 19, batch 2950, loss[loss=0.2324, simple_loss=0.2961, pruned_loss=0.08439, over 18696.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2709, pruned_loss=0.05552, over 3712576.63 frames. ], batch size: 50, lr: 6.00e-03, grad_scale: 8.0 2022-12-23 14:22:36,412 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp0.9 from training. Duration: 26.438875 2022-12-23 14:23:03,154 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5398, 1.5220, 1.6060, 1.5320, 0.9405, 3.5853, 1.4537, 1.9769], device='cuda:3'), covar=tensor([0.3257, 0.2042, 0.2030, 0.2140, 0.1630, 0.0186, 0.1660, 0.0880], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0117, 0.0127, 0.0121, 0.0103, 0.0098, 0.0092, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 14:23:21,800 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0021-26306-0_sp0.9 from training. Duration: 21.2444375 2022-12-23 14:23:21,827 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp0.9 from training. Duration: 31.02225 2022-12-23 14:23:31,042 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.61 vs. limit=5.0 2022-12-23 14:23:31,791 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0 from training. Duration: 22.395 2022-12-23 14:23:44,318 INFO [train.py:894] (3/4) Epoch 19, batch 3000, loss[loss=0.1904, simple_loss=0.278, pruned_loss=0.05146, over 18560.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2709, pruned_loss=0.05504, over 3713517.82 frames. ], batch size: 56, lr: 6.00e-03, grad_scale: 8.0 2022-12-23 14:23:44,318 INFO [train.py:919] (3/4) Computing validation loss 2022-12-23 14:23:55,322 INFO [train.py:928] (3/4) Epoch 19, validation: loss=0.1652, simple_loss=0.263, pruned_loss=0.03365, over 944034.00 frames. 2022-12-23 14:23:55,323 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-23 14:24:00,919 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0 from training. Duration: 21.075 2022-12-23 14:24:05,218 WARNING [train.py:1060] (3/4) Exclude cut with ID 6482-98857-0025-147532-0_sp0.9 from training. Duration: 20.0055625 2022-12-23 14:24:07,096 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0037-132304-0_sp0.9 from training. Duration: 22.05 2022-12-23 14:24:07,105 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0 from training. Duration: 26.8349375 2022-12-23 14:24:09,935 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp1.1 from training. Duration: 22.1090625 2022-12-23 14:24:17,259 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp0.9 from training. Duration: 26.6166875 2022-12-23 14:24:20,774 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2022-12-23 14:24:33,094 WARNING [train.py:1060] (3/4) Exclude cut with ID 2046-178027-0000-53705-0_sp0.9 from training. Duration: 20.3055625 2022-12-23 14:24:34,463 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.524e+02 4.579e+02 5.252e+02 6.316e+02 1.044e+03, threshold=1.050e+03, percent-clipped=0.0 2022-12-23 14:24:45,291 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66144.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 14:24:57,047 WARNING [train.py:1060] (3/4) Exclude cut with ID 7205-50138-0008-5373-0_sp0.9 from training. Duration: 20.7 2022-12-23 14:25:09,581 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66160.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 14:25:10,760 INFO [train.py:894] (3/4) Epoch 19, batch 3050, loss[loss=0.2338, simple_loss=0.2961, pruned_loss=0.08577, over 18644.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.2701, pruned_loss=0.05461, over 3714921.31 frames. ], batch size: 174, lr: 6.00e-03, grad_scale: 8.0 2022-12-23 14:25:38,989 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0 from training. Duration: 22.48 2022-12-23 14:25:51,617 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66188.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 14:25:54,780 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp0.9 from training. Duration: 29.816625 2022-12-23 14:26:00,786 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66194.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 14:26:12,391 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp1.1 from training. Duration: 22.7590625 2022-12-23 14:26:18,802 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0 from training. Duration: 22.555 2022-12-23 14:26:19,075 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66205.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 14:26:27,929 INFO [train.py:894] (3/4) Epoch 19, batch 3100, loss[loss=0.1781, simple_loss=0.2501, pruned_loss=0.05301, over 18593.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.271, pruned_loss=0.05516, over 3715209.21 frames. ], batch size: 41, lr: 5.99e-03, grad_scale: 8.0 2022-12-23 14:26:38,796 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0005-25975-0_sp0.9 from training. Duration: 21.688875 2022-12-23 14:26:43,391 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66221.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 14:27:05,969 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=66236.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 14:27:08,465 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.484e+02 4.466e+02 5.278e+02 6.458e+02 1.421e+03, threshold=1.056e+03, percent-clipped=3.0 2022-12-23 14:27:15,251 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0_sp0.9 from training. Duration: 22.6 2022-12-23 14:27:17,953 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2022-12-23 14:27:39,448 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66258.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 14:27:43,446 INFO [train.py:894] (3/4) Epoch 19, batch 3150, loss[loss=0.1868, simple_loss=0.2681, pruned_loss=0.05271, over 18565.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2698, pruned_loss=0.05444, over 3715705.97 frames. ], batch size: 97, lr: 5.99e-03, grad_scale: 8.0 2022-12-23 14:27:54,048 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0 from training. Duration: 24.32 2022-12-23 14:28:50,561 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-276745-0093-13116-0_sp0.9 from training. Duration: 21.061125 2022-12-23 14:28:50,931 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2048, 2.2455, 2.8779, 1.2857, 2.8492, 2.7215, 1.6877, 2.9468], device='cuda:3'), covar=tensor([0.1564, 0.2001, 0.1396, 0.2471, 0.1002, 0.1589, 0.2553, 0.0766], device='cuda:3'), in_proj_covar=tensor([0.0199, 0.0213, 0.0206, 0.0194, 0.0177, 0.0219, 0.0216, 0.0190], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 14:28:59,665 INFO [train.py:894] (3/4) Epoch 19, batch 3200, loss[loss=0.1976, simple_loss=0.2816, pruned_loss=0.05682, over 18453.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2701, pruned_loss=0.05472, over 3715332.66 frames. ], batch size: 64, lr: 5.99e-03, grad_scale: 8.0 2022-12-23 14:29:02,948 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0024-15855-0_sp0.9 from training. Duration: 20.32225 2022-12-23 14:29:05,499 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2022-12-23 14:29:12,281 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66319.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 14:29:16,400 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp1.1 from training. Duration: 0.7545625 2022-12-23 14:29:32,223 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0_sp0.9 from training. Duration: 23.9333125 2022-12-23 14:29:32,517 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9532, 1.4552, 2.5801, 4.3817, 3.3858, 2.8752, 0.6739, 3.1631], device='cuda:3'), covar=tensor([0.1804, 0.1715, 0.1368, 0.0552, 0.0848, 0.1089, 0.2379, 0.0856], device='cuda:3'), in_proj_covar=tensor([0.0102, 0.0116, 0.0135, 0.0148, 0.0106, 0.0140, 0.0130, 0.0113], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 14:29:40,641 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.023e+02 4.549e+02 5.620e+02 6.931e+02 1.616e+03, threshold=1.124e+03, percent-clipped=4.0 2022-12-23 14:30:00,680 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2022-12-23 14:30:04,018 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp1.1 from training. Duration: 20.17275 2022-12-23 14:30:11,111 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp1.1 from training. Duration: 20.436375 2022-12-23 14:30:13,764 INFO [train.py:894] (3/4) Epoch 19, batch 3250, loss[loss=0.1996, simple_loss=0.2885, pruned_loss=0.05538, over 18718.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2699, pruned_loss=0.05448, over 3716210.90 frames. ], batch size: 60, lr: 5.99e-03, grad_scale: 8.0 2022-12-23 14:30:56,705 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2022-12-23 14:31:28,824 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0_sp0.9 from training. Duration: 23.1055625 2022-12-23 14:31:30,474 INFO [train.py:894] (3/4) Epoch 19, batch 3300, loss[loss=0.2264, simple_loss=0.2972, pruned_loss=0.07778, over 18601.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2702, pruned_loss=0.05474, over 3716309.19 frames. ], batch size: 179, lr: 5.99e-03, grad_scale: 8.0 2022-12-23 14:31:30,541 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp1.1 from training. Duration: 23.5 2022-12-23 14:31:41,425 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp0.9 from training. Duration: 26.62775 2022-12-23 14:31:56,709 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0 from training. Duration: 21.105 2022-12-23 14:32:01,029 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0_sp0.9 from training. Duration: 24.411125 2022-12-23 14:32:02,758 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6519, 1.4332, 1.4337, 1.8768, 1.6514, 3.4930, 1.4708, 1.5931], device='cuda:3'), covar=tensor([0.0880, 0.1880, 0.1135, 0.0952, 0.1494, 0.0263, 0.1439, 0.1569], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0083, 0.0073, 0.0075, 0.0091, 0.0076, 0.0085, 0.0078], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 14:32:07,016 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7860, 2.3762, 1.6991, 2.5434, 3.0692, 1.7659, 1.9099, 1.5778], device='cuda:3'), covar=tensor([0.1857, 0.1540, 0.1540, 0.0920, 0.1306, 0.1088, 0.1847, 0.1442], device='cuda:3'), in_proj_covar=tensor([0.0244, 0.0223, 0.0213, 0.0198, 0.0260, 0.0194, 0.0221, 0.0196], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 14:32:10,910 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.913e+02 4.378e+02 5.345e+02 6.831e+02 1.350e+03, threshold=1.069e+03, percent-clipped=1.0 2022-12-23 14:32:28,240 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp1.1 from training. Duration: 21.263625 2022-12-23 14:32:45,773 INFO [train.py:894] (3/4) Epoch 19, batch 3350, loss[loss=0.2096, simple_loss=0.2796, pruned_loss=0.06979, over 18582.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2694, pruned_loss=0.05419, over 3715352.22 frames. ], batch size: 183, lr: 5.98e-03, grad_scale: 8.0 2022-12-23 14:32:59,720 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0 from training. Duration: 20.795 2022-12-23 14:33:01,535 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2226, 1.6055, 1.7789, 1.8269, 2.1834, 2.1231, 2.0741, 1.7194], device='cuda:3'), covar=tensor([0.2177, 0.3290, 0.2601, 0.2860, 0.1929, 0.0953, 0.3077, 0.1311], device='cuda:3'), in_proj_covar=tensor([0.0269, 0.0302, 0.0280, 0.0315, 0.0305, 0.0251, 0.0341, 0.0240], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 14:33:10,336 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0 from training. Duration: 24.76 2022-12-23 14:33:10,350 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0_sp0.9 from training. Duration: 22.25 2022-12-23 14:33:34,404 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp1.1 from training. Duration: 20.5045625 2022-12-23 14:33:37,607 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66494.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 14:33:46,491 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66500.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 14:33:54,337 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6983, 1.4803, 1.3870, 1.8526, 1.6488, 3.3263, 1.3625, 1.5968], device='cuda:3'), covar=tensor([0.0804, 0.1722, 0.1043, 0.0912, 0.1425, 0.0245, 0.1408, 0.1497], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0083, 0.0073, 0.0075, 0.0091, 0.0076, 0.0085, 0.0077], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 14:34:03,484 INFO [train.py:894] (3/4) Epoch 19, batch 3400, loss[loss=0.1901, simple_loss=0.2715, pruned_loss=0.05437, over 18561.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2696, pruned_loss=0.054, over 3715485.92 frames. ], batch size: 57, lr: 5.98e-03, grad_scale: 8.0 2022-12-23 14:34:10,366 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66516.0, num_to_drop=1, layers_to_drop={3} 2022-12-23 14:34:17,788 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66521.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 14:34:40,813 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.665e+02 4.804e+02 5.705e+02 7.377e+02 1.281e+03, threshold=1.141e+03, percent-clipped=2.0 2022-12-23 14:34:44,335 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2022-12-23 14:34:46,522 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=66542.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 14:35:13,231 INFO [train.py:894] (3/4) Epoch 19, batch 3450, loss[loss=0.1925, simple_loss=0.2814, pruned_loss=0.0518, over 18577.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2699, pruned_loss=0.05418, over 3715364.27 frames. ], batch size: 57, lr: 5.98e-03, grad_scale: 8.0 2022-12-23 14:35:32,362 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6983, 1.3312, 0.8448, 1.2448, 2.1696, 1.0283, 1.4873, 1.6178], device='cuda:3'), covar=tensor([0.1626, 0.2013, 0.2128, 0.1543, 0.1620, 0.1689, 0.1494, 0.1651], device='cuda:3'), in_proj_covar=tensor([0.0093, 0.0097, 0.0116, 0.0095, 0.0115, 0.0090, 0.0097, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 14:35:42,659 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66582.0, num_to_drop=1, layers_to_drop={3} 2022-12-23 14:36:26,014 INFO [train.py:894] (3/4) Epoch 19, batch 3500, loss[loss=0.1939, simple_loss=0.2742, pruned_loss=0.05683, over 18640.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2692, pruned_loss=0.05402, over 3715385.75 frames. ], batch size: 174, lr: 5.98e-03, grad_scale: 8.0 2022-12-23 14:36:31,108 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66614.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 14:36:46,658 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp1.1 from training. Duration: 22.2954375 2022-12-23 14:36:56,210 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2022-12-23 14:36:56,766 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2022-12-23 14:36:56,854 INFO [train.py:894] (3/4) Epoch 20, batch 0, loss[loss=0.1672, simple_loss=0.247, pruned_loss=0.04371, over 18424.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.247, pruned_loss=0.04371, over 18424.00 frames. ], batch size: 42, lr: 5.82e-03, grad_scale: 8.0 2022-12-23 14:36:56,855 INFO [train.py:919] (3/4) Computing validation loss 2022-12-23 14:37:07,740 INFO [train.py:928] (3/4) Epoch 20, validation: loss=0.1637, simple_loss=0.2619, pruned_loss=0.03275, over 944034.00 frames. 2022-12-23 14:37:07,742 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-23 14:37:31,885 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3381, 1.8689, 0.7862, 1.5805, 2.3168, 1.6676, 2.1275, 2.1522], device='cuda:3'), covar=tensor([0.1520, 0.1911, 0.2523, 0.1643, 0.1732, 0.1752, 0.1365, 0.1718], device='cuda:3'), in_proj_covar=tensor([0.0093, 0.0096, 0.0116, 0.0095, 0.0114, 0.0090, 0.0097, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 14:37:39,091 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.159e+02 4.005e+02 5.121e+02 6.959e+02 1.364e+03, threshold=1.024e+03, percent-clipped=1.0 2022-12-23 14:37:52,786 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9855, 1.8221, 2.3868, 1.3540, 2.3014, 2.2295, 1.6277, 2.4634], device='cuda:3'), covar=tensor([0.1403, 0.2112, 0.1292, 0.2179, 0.0874, 0.1362, 0.2501, 0.0794], device='cuda:3'), in_proj_covar=tensor([0.0197, 0.0211, 0.0206, 0.0193, 0.0176, 0.0217, 0.0216, 0.0190], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 14:37:57,996 WARNING [train.py:1060] (3/4) Exclude cut with ID 298-126791-0067-24026-0_sp0.9 from training. Duration: 21.438875 2022-12-23 14:38:02,629 WARNING [train.py:1060] (3/4) Exclude cut with ID 5652-39938-0025-23684-0_sp0.9 from training. Duration: 22.2055625 2022-12-23 14:38:23,333 INFO [train.py:894] (3/4) Epoch 20, batch 50, loss[loss=0.1496, simple_loss=0.2344, pruned_loss=0.03234, over 18504.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2659, pruned_loss=0.04645, over 838863.65 frames. ], batch size: 41, lr: 5.82e-03, grad_scale: 8.0 2022-12-23 14:38:42,159 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5284, 2.6423, 1.8305, 3.3117, 2.9974, 2.3823, 3.8975, 2.4415], device='cuda:3'), covar=tensor([0.0771, 0.1740, 0.2737, 0.1649, 0.1574, 0.0844, 0.0716, 0.1177], device='cuda:3'), in_proj_covar=tensor([0.0178, 0.0208, 0.0251, 0.0292, 0.0238, 0.0191, 0.0209, 0.0204], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 14:39:03,255 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.6555, 1.9413, 2.2889, 1.0461, 1.5728, 2.4310, 2.1067, 1.8066], device='cuda:3'), covar=tensor([0.0773, 0.0333, 0.0302, 0.0430, 0.0361, 0.0383, 0.0260, 0.0678], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0170, 0.0126, 0.0140, 0.0148, 0.0142, 0.0162, 0.0171], device='cuda:3'), out_proj_covar=tensor([1.1545e-04, 1.3118e-04, 9.5413e-05, 1.0494e-04, 1.1183e-04, 1.0945e-04, 1.2573e-04, 1.3078e-04], device='cuda:3') 2022-12-23 14:39:37,539 INFO [train.py:894] (3/4) Epoch 20, batch 100, loss[loss=0.2058, simple_loss=0.2855, pruned_loss=0.06305, over 18659.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2673, pruned_loss=0.04661, over 1477230.17 frames. ], batch size: 179, lr: 5.82e-03, grad_scale: 8.0 2022-12-23 14:39:59,304 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66731.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 14:40:09,559 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.446e+02 3.334e+02 3.825e+02 4.671e+02 9.107e+02, threshold=7.651e+02, percent-clipped=0.0 2022-12-23 14:40:45,457 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.37 vs. limit=5.0 2022-12-23 14:40:53,549 INFO [train.py:894] (3/4) Epoch 20, batch 150, loss[loss=0.1836, simple_loss=0.2746, pruned_loss=0.04632, over 18439.00 frames. ], tot_loss[loss=0.178, simple_loss=0.265, pruned_loss=0.04544, over 1973522.80 frames. ], batch size: 64, lr: 5.82e-03, grad_scale: 8.0 2022-12-23 14:41:02,669 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0 from training. Duration: 24.525 2022-12-23 14:41:18,920 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9943, 1.9169, 1.3925, 2.1506, 2.2136, 1.9587, 2.7619, 2.0598], device='cuda:3'), covar=tensor([0.1072, 0.1926, 0.3173, 0.1994, 0.1943, 0.1111, 0.1059, 0.1480], device='cuda:3'), in_proj_covar=tensor([0.0177, 0.0207, 0.0248, 0.0290, 0.0236, 0.0189, 0.0207, 0.0202], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 14:41:30,337 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66792.0, num_to_drop=1, layers_to_drop={3} 2022-12-23 14:41:37,591 WARNING [train.py:1060] (3/4) Exclude cut with ID 3699-47246-0007-3408-0_sp0.9 from training. Duration: 20.26675 2022-12-23 14:41:42,882 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66800.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 14:41:52,188 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp0.9 from training. Duration: 27.25 2022-12-23 14:42:07,280 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66816.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 14:42:08,301 INFO [train.py:894] (3/4) Epoch 20, batch 200, loss[loss=0.1937, simple_loss=0.2834, pruned_loss=0.05201, over 18528.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2635, pruned_loss=0.04436, over 2358830.93 frames. ], batch size: 58, lr: 5.81e-03, grad_scale: 8.0 2022-12-23 14:42:08,762 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6224, 1.5340, 1.4853, 1.4353, 1.8951, 1.7813, 1.7417, 1.2154], device='cuda:3'), covar=tensor([0.0331, 0.0255, 0.0472, 0.0215, 0.0182, 0.0371, 0.0284, 0.0335], device='cuda:3'), in_proj_covar=tensor([0.0096, 0.0128, 0.0155, 0.0125, 0.0116, 0.0120, 0.0099, 0.0128], device='cuda:3'), out_proj_covar=tensor([7.6400e-05, 1.0158e-04, 1.2814e-04, 1.0038e-04, 9.4348e-05, 9.2665e-05, 7.7678e-05, 1.0120e-04], device='cuda:3') 2022-12-23 14:42:39,484 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.187e+02 3.272e+02 3.815e+02 5.495e+02 8.905e+02, threshold=7.631e+02, percent-clipped=2.0 2022-12-23 14:42:54,454 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=66848.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 14:43:07,114 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0 from training. Duration: 21.68 2022-12-23 14:43:18,228 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0 from training. Duration: 21.6300625 2022-12-23 14:43:19,964 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=66864.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 14:43:24,561 INFO [train.py:894] (3/4) Epoch 20, batch 250, loss[loss=0.1717, simple_loss=0.2681, pruned_loss=0.03768, over 18531.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2632, pruned_loss=0.04409, over 2657707.69 frames. ], batch size: 55, lr: 5.81e-03, grad_scale: 16.0 2022-12-23 14:43:27,253 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2022-12-23 14:43:39,168 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66877.0, num_to_drop=1, layers_to_drop={3} 2022-12-23 14:43:41,657 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0_sp0.9 from training. Duration: 24.033375 2022-12-23 14:43:42,022 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6305, 1.6119, 1.7047, 1.6186, 1.4442, 3.6456, 1.6451, 2.1135], device='cuda:3'), covar=tensor([0.3093, 0.2022, 0.1880, 0.1943, 0.1255, 0.0168, 0.1563, 0.0830], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0117, 0.0125, 0.0120, 0.0102, 0.0096, 0.0091, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 14:44:34,335 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66914.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 14:44:38,953 INFO [train.py:894] (3/4) Epoch 20, batch 300, loss[loss=0.1596, simple_loss=0.2332, pruned_loss=0.04304, over 18587.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2621, pruned_loss=0.0437, over 2892087.60 frames. ], batch size: 41, lr: 5.81e-03, grad_scale: 16.0 2022-12-23 14:44:39,671 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.99 vs. limit=5.0 2022-12-23 14:44:43,196 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0 from training. Duration: 22.905 2022-12-23 14:44:43,246 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp1.1 from training. Duration: 23.4318125 2022-12-23 14:44:46,439 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.1765, 3.6129, 3.5863, 4.1451, 3.7782, 3.6730, 4.2881, 1.2251], device='cuda:3'), covar=tensor([0.0753, 0.0773, 0.0714, 0.0813, 0.1498, 0.1298, 0.0691, 0.5512], device='cuda:3'), in_proj_covar=tensor([0.0340, 0.0222, 0.0231, 0.0263, 0.0321, 0.0267, 0.0286, 0.0278], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 14:45:09,776 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.301e+02 3.375e+02 4.196e+02 5.009e+02 1.030e+03, threshold=8.392e+02, percent-clipped=5.0 2022-12-23 14:45:39,670 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66957.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 14:45:47,460 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=66962.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 14:45:55,245 INFO [train.py:894] (3/4) Epoch 20, batch 350, loss[loss=0.1784, simple_loss=0.276, pruned_loss=0.04045, over 18698.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2628, pruned_loss=0.04377, over 3074704.35 frames. ], batch size: 60, lr: 5.81e-03, grad_scale: 16.0 2022-12-23 14:46:15,560 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.4696, 3.3326, 2.2141, 1.7505, 3.6771, 3.9356, 3.4108, 2.7803], device='cuda:3'), covar=tensor([0.0335, 0.0346, 0.0547, 0.0737, 0.0172, 0.0280, 0.0407, 0.0647], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0126, 0.0129, 0.0120, 0.0097, 0.0121, 0.0134, 0.0155], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 14:46:39,851 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp1.1 from training. Duration: 20.82275 2022-12-23 14:46:41,276 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp0.9 from training. Duration: 25.45 2022-12-23 14:46:43,044 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6202, 2.2094, 0.9013, 1.7833, 2.5309, 1.8759, 2.3475, 2.3903], device='cuda:3'), covar=tensor([0.1292, 0.1618, 0.2286, 0.1370, 0.1486, 0.1463, 0.1274, 0.1520], device='cuda:3'), in_proj_covar=tensor([0.0093, 0.0097, 0.0117, 0.0096, 0.0115, 0.0092, 0.0098, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 14:47:02,916 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4597, 1.3889, 1.4675, 1.4672, 0.9330, 3.0496, 1.2466, 1.7261], device='cuda:3'), covar=tensor([0.3695, 0.2476, 0.2282, 0.2292, 0.1694, 0.0240, 0.1684, 0.0943], device='cuda:3'), in_proj_covar=tensor([0.0131, 0.0115, 0.0124, 0.0119, 0.0101, 0.0096, 0.0090, 0.0088], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 14:47:11,580 INFO [train.py:894] (3/4) Epoch 20, batch 400, loss[loss=0.173, simple_loss=0.2702, pruned_loss=0.03793, over 18387.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.263, pruned_loss=0.04397, over 3216486.48 frames. ], batch size: 53, lr: 5.80e-03, grad_scale: 16.0 2022-12-23 14:47:13,247 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67018.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 14:47:42,052 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.086e+02 3.535e+02 4.114e+02 5.180e+02 1.001e+03, threshold=8.228e+02, percent-clipped=4.0 2022-12-23 14:47:43,637 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0 from training. Duration: 25.775 2022-12-23 14:48:05,686 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0_sp0.9 from training. Duration: 22.25 2022-12-23 14:48:26,123 INFO [train.py:894] (3/4) Epoch 20, batch 450, loss[loss=0.1926, simple_loss=0.2831, pruned_loss=0.05103, over 18571.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2638, pruned_loss=0.04449, over 3326750.63 frames. ], batch size: 56, lr: 5.80e-03, grad_scale: 16.0 2022-12-23 14:48:34,732 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0 from training. Duration: 26.205 2022-12-23 14:48:47,799 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2022-12-23 14:48:50,567 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67083.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 14:48:51,736 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp0.9 from training. Duration: 30.1555625 2022-12-23 14:48:56,063 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67087.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 14:48:58,843 WARNING [train.py:1060] (3/4) Exclude cut with ID 1265-135635-0050-6781-0_sp0.9 from training. Duration: 21.8333125 2022-12-23 14:49:06,641 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp1.1 from training. Duration: 20.6545625 2022-12-23 14:49:24,899 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([6.0086, 5.0323, 5.2744, 5.9973, 5.5745, 5.3374, 6.0739, 1.7621], device='cuda:3'), covar=tensor([0.0602, 0.0650, 0.0523, 0.0798, 0.1100, 0.1048, 0.0399, 0.5119], device='cuda:3'), in_proj_covar=tensor([0.0340, 0.0222, 0.0232, 0.0263, 0.0321, 0.0268, 0.0285, 0.0278], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 14:49:42,234 INFO [train.py:894] (3/4) Epoch 20, batch 500, loss[loss=0.1634, simple_loss=0.2578, pruned_loss=0.03452, over 18538.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2637, pruned_loss=0.04461, over 3412558.07 frames. ], batch size: 55, lr: 5.80e-03, grad_scale: 16.0 2022-12-23 14:49:46,795 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0045-39920-0_sp0.9 from training. Duration: 20.52225 2022-12-23 14:50:00,335 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2022-12-23 14:50:04,224 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8457, 1.8152, 1.9217, 2.4411, 2.1696, 4.6594, 2.2270, 2.0628], device='cuda:3'), covar=tensor([0.0876, 0.1656, 0.1014, 0.0905, 0.1311, 0.0183, 0.1163, 0.1406], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0083, 0.0073, 0.0075, 0.0090, 0.0076, 0.0085, 0.0077], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 14:50:06,692 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp0.9 from training. Duration: 29.1166875 2022-12-23 14:50:12,592 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.117e+02 3.572e+02 4.412e+02 5.360e+02 1.378e+03, threshold=8.823e+02, percent-clipped=5.0 2022-12-23 14:50:22,438 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67144.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 14:50:56,642 INFO [train.py:894] (3/4) Epoch 20, batch 550, loss[loss=0.2029, simple_loss=0.2962, pruned_loss=0.05477, over 18685.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2646, pruned_loss=0.04486, over 3479139.07 frames. ], batch size: 62, lr: 5.80e-03, grad_scale: 16.0 2022-12-23 14:51:05,147 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133211-0007-59831-0_sp0.9 from training. Duration: 21.388875 2022-12-23 14:51:11,869 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67177.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 14:51:40,399 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0 from training. Duration: 22.72 2022-12-23 14:51:41,891 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0_sp0.9 from training. Duration: 22.7444375 2022-12-23 14:52:10,862 INFO [train.py:894] (3/4) Epoch 20, batch 600, loss[loss=0.1966, simple_loss=0.2896, pruned_loss=0.05185, over 18445.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2664, pruned_loss=0.04562, over 3532160.57 frames. ], batch size: 50, lr: 5.80e-03, grad_scale: 16.0 2022-12-23 14:52:22,874 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=67225.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 14:52:24,138 WARNING [train.py:1060] (3/4) Exclude cut with ID 4133-6541-0027-40495-0_sp1.1 from training. Duration: 0.9681875 2022-12-23 14:52:27,547 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1131, 1.4177, 1.7797, 1.7738, 2.1257, 2.1347, 1.9283, 1.7156], device='cuda:3'), covar=tensor([0.2227, 0.3422, 0.2698, 0.2929, 0.2095, 0.0964, 0.3488, 0.1339], device='cuda:3'), in_proj_covar=tensor([0.0264, 0.0296, 0.0275, 0.0312, 0.0302, 0.0249, 0.0338, 0.0237], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 14:52:28,495 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0_sp0.9 from training. Duration: 22.3166875 2022-12-23 14:52:34,041 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133212-0015-59917-0_sp0.9 from training. Duration: 21.8166875 2022-12-23 14:52:41,651 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.645e+02 3.735e+02 4.222e+02 4.991e+02 1.201e+03, threshold=8.444e+02, percent-clipped=1.0 2022-12-23 14:52:48,443 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2022-12-23 14:53:09,942 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0000, 2.2764, 1.9211, 2.4189, 2.6851, 1.8869, 2.0188, 1.7994], device='cuda:3'), covar=tensor([0.1435, 0.1336, 0.1294, 0.0850, 0.1141, 0.0942, 0.1910, 0.1185], device='cuda:3'), in_proj_covar=tensor([0.0242, 0.0222, 0.0212, 0.0196, 0.0256, 0.0193, 0.0220, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 14:53:25,781 INFO [train.py:894] (3/4) Epoch 20, batch 650, loss[loss=0.1857, simple_loss=0.2743, pruned_loss=0.04852, over 18717.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2661, pruned_loss=0.04515, over 3573017.85 frames. ], batch size: 54, lr: 5.79e-03, grad_scale: 16.0 2022-12-23 14:54:16,832 WARNING [train.py:1060] (3/4) Exclude cut with ID 4957-30119-0041-23990-0_sp0.9 from training. Duration: 20.22775 2022-12-23 14:54:27,331 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.64 vs. limit=5.0 2022-12-23 14:54:28,829 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2022-12-23 14:54:32,745 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2674, 2.1049, 1.9666, 1.2934, 2.6126, 2.4076, 2.3159, 1.7169], device='cuda:3'), covar=tensor([0.0355, 0.0421, 0.0429, 0.0728, 0.0258, 0.0340, 0.0378, 0.0848], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0124, 0.0127, 0.0119, 0.0096, 0.0120, 0.0133, 0.0154], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 14:54:33,990 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67313.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 14:54:39,641 INFO [train.py:894] (3/4) Epoch 20, batch 700, loss[loss=0.1502, simple_loss=0.228, pruned_loss=0.03619, over 18433.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2658, pruned_loss=0.04487, over 3603220.42 frames. ], batch size: 42, lr: 5.79e-03, grad_scale: 8.0 2022-12-23 14:55:00,205 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp1.1 from training. Duration: 24.67275 2022-12-23 14:55:12,712 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.463e+02 3.535e+02 4.247e+02 5.530e+02 8.059e+02, threshold=8.494e+02, percent-clipped=0.0 2022-12-23 14:55:28,707 WARNING [train.py:1060] (3/4) Exclude cut with ID 3082-165428-0081-50734-0_sp0.9 from training. Duration: 21.8055625 2022-12-23 14:55:54,231 INFO [train.py:894] (3/4) Epoch 20, batch 750, loss[loss=0.1753, simple_loss=0.2699, pruned_loss=0.04039, over 18723.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2657, pruned_loss=0.04457, over 3628003.19 frames. ], batch size: 52, lr: 5.79e-03, grad_scale: 8.0 2022-12-23 14:56:05,673 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0_sp0.9 from training. Duration: 22.6666875 2022-12-23 14:56:23,998 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67387.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 14:57:08,164 INFO [train.py:894] (3/4) Epoch 20, batch 800, loss[loss=0.1618, simple_loss=0.2416, pruned_loss=0.041, over 18694.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2659, pruned_loss=0.045, over 3646697.07 frames. ], batch size: 46, lr: 5.79e-03, grad_scale: 8.0 2022-12-23 14:57:09,683 WARNING [train.py:1060] (3/4) Exclude cut with ID 2411-132532-0017-82279-0_sp1.1 from training. Duration: 0.9681875 2022-12-23 14:57:35,226 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0 from training. Duration: 22.485 2022-12-23 14:57:35,333 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=67435.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 14:57:37,166 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2022-12-23 14:57:40,535 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.476e+02 3.510e+02 4.303e+02 5.243e+02 9.947e+02, threshold=8.606e+02, percent-clipped=5.0 2022-12-23 14:57:40,836 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67439.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 14:58:10,550 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp1.1 from training. Duration: 23.82275 2022-12-23 14:58:15,338 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([5.9990, 5.0323, 5.2699, 5.9727, 5.5448, 5.2916, 6.0247, 1.6926], device='cuda:3'), covar=tensor([0.0553, 0.0564, 0.0499, 0.0639, 0.1295, 0.1032, 0.0418, 0.5235], device='cuda:3'), in_proj_covar=tensor([0.0340, 0.0224, 0.0234, 0.0264, 0.0321, 0.0268, 0.0285, 0.0279], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 14:58:22,211 INFO [train.py:894] (3/4) Epoch 20, batch 850, loss[loss=0.1798, simple_loss=0.2745, pruned_loss=0.04258, over 18581.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2653, pruned_loss=0.04437, over 3660988.94 frames. ], batch size: 56, lr: 5.79e-03, grad_scale: 8.0 2022-12-23 14:58:22,252 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0 from training. Duration: 20.77 2022-12-23 14:58:29,338 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0_sp0.9 from training. Duration: 24.088875 2022-12-23 14:58:56,577 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp1.1 from training. Duration: 20.4409375 2022-12-23 14:59:32,031 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67513.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 14:59:37,802 INFO [train.py:894] (3/4) Epoch 20, batch 900, loss[loss=0.1587, simple_loss=0.2549, pruned_loss=0.03124, over 18727.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2659, pruned_loss=0.04451, over 3672976.85 frames. ], batch size: 52, lr: 5.78e-03, grad_scale: 8.0 2022-12-23 15:00:11,359 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.522e+02 3.439e+02 4.457e+02 5.643e+02 1.323e+03, threshold=8.915e+02, percent-clipped=3.0 2022-12-23 15:00:11,428 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0_sp0.9 from training. Duration: 22.511125 2022-12-23 15:00:12,782 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0 from training. Duration: 20.675 2022-12-23 15:00:49,684 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3705, 2.0792, 1.7481, 2.0181, 1.8612, 2.1097, 1.9213, 2.2055], device='cuda:3'), covar=tensor([0.2056, 0.3135, 0.1881, 0.2539, 0.3415, 0.1002, 0.2988, 0.0963], device='cuda:3'), in_proj_covar=tensor([0.0293, 0.0288, 0.0244, 0.0347, 0.0269, 0.0227, 0.0288, 0.0212], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 15:00:53,647 INFO [train.py:894] (3/4) Epoch 20, batch 950, loss[loss=0.1849, simple_loss=0.2756, pruned_loss=0.04703, over 18645.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2651, pruned_loss=0.04428, over 3681416.33 frames. ], batch size: 62, lr: 5.78e-03, grad_scale: 8.0 2022-12-23 15:01:04,522 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67574.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 15:01:48,829 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp0.9 from training. Duration: 24.9833125 2022-12-23 15:02:03,415 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67613.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:02:05,055 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2078, 2.0580, 1.7901, 1.2057, 2.4836, 2.3699, 2.1398, 1.7497], device='cuda:3'), covar=tensor([0.0351, 0.0421, 0.0491, 0.0753, 0.0274, 0.0347, 0.0407, 0.0845], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0124, 0.0127, 0.0119, 0.0096, 0.0120, 0.0132, 0.0153], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 15:02:08,870 INFO [train.py:894] (3/4) Epoch 20, batch 1000, loss[loss=0.1983, simple_loss=0.2906, pruned_loss=0.05299, over 18636.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2651, pruned_loss=0.04395, over 3688242.67 frames. ], batch size: 62, lr: 5.78e-03, grad_scale: 8.0 2022-12-23 15:02:22,240 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0 from training. Duration: 27.14 2022-12-23 15:02:36,367 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0 from training. Duration: 22.44 2022-12-23 15:02:42,945 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.353e+02 3.421e+02 4.027e+02 4.985e+02 1.255e+03, threshold=8.054e+02, percent-clipped=1.0 2022-12-23 15:03:16,105 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=67661.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:03:24,773 INFO [train.py:894] (3/4) Epoch 20, batch 1050, loss[loss=0.1703, simple_loss=0.257, pruned_loss=0.04181, over 18679.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2648, pruned_loss=0.04386, over 3693918.51 frames. ], batch size: 48, lr: 5.78e-03, grad_scale: 8.0 2022-12-23 15:03:44,621 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67680.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:03:54,974 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0060-62364-0_sp0.9 from training. Duration: 21.361125 2022-12-23 15:03:58,501 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([5.9792, 5.0385, 5.2450, 6.0366, 5.5633, 5.2661, 6.0560, 1.6800], device='cuda:3'), covar=tensor([0.0484, 0.0577, 0.0474, 0.0583, 0.1097, 0.0963, 0.0337, 0.4990], device='cuda:3'), in_proj_covar=tensor([0.0338, 0.0222, 0.0233, 0.0263, 0.0319, 0.0267, 0.0284, 0.0277], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 15:03:59,680 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp1.1 from training. Duration: 27.0318125 2022-12-23 15:04:03,119 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67692.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:04:12,017 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp0.9 from training. Duration: 28.638875 2022-12-23 15:04:19,723 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5430, 2.3003, 1.8650, 1.3999, 2.8930, 2.7436, 2.4373, 1.7949], device='cuda:3'), covar=tensor([0.0327, 0.0386, 0.0521, 0.0719, 0.0211, 0.0327, 0.0410, 0.0808], device='cuda:3'), in_proj_covar=tensor([0.0122, 0.0124, 0.0127, 0.0118, 0.0095, 0.0120, 0.0132, 0.0153], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 15:04:28,179 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0 from training. Duration: 20.4 2022-12-23 15:04:40,541 INFO [train.py:894] (3/4) Epoch 20, batch 1100, loss[loss=0.1781, simple_loss=0.2542, pruned_loss=0.051, over 18616.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.265, pruned_loss=0.04396, over 3698461.37 frames. ], batch size: 41, lr: 5.77e-03, grad_scale: 8.0 2022-12-23 15:05:00,033 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0 from training. Duration: 20.025 2022-12-23 15:05:01,344 WARNING [train.py:1060] (3/4) Exclude cut with ID 2364-131735-0112-64612-0_sp0.9 from training. Duration: 20.488875 2022-12-23 15:05:04,641 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0 from training. Duration: 29.735 2022-12-23 15:05:14,323 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.101e+02 3.349e+02 4.308e+02 5.408e+02 2.073e+03, threshold=8.617e+02, percent-clipped=3.0 2022-12-23 15:05:14,635 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67739.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:05:17,740 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67741.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:05:31,885 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.3595, 3.8328, 3.8121, 4.3955, 3.9455, 3.9188, 4.5419, 1.3853], device='cuda:3'), covar=tensor([0.0717, 0.0649, 0.0616, 0.0674, 0.1434, 0.1017, 0.0536, 0.4987], device='cuda:3'), in_proj_covar=tensor([0.0334, 0.0220, 0.0230, 0.0258, 0.0315, 0.0263, 0.0280, 0.0273], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 15:05:34,828 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67753.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 15:05:55,179 INFO [train.py:894] (3/4) Epoch 20, batch 1150, loss[loss=0.1897, simple_loss=0.2745, pruned_loss=0.0525, over 18668.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2647, pruned_loss=0.04405, over 3701201.89 frames. ], batch size: 60, lr: 5.77e-03, grad_scale: 8.0 2022-12-23 15:06:25,333 WARNING [train.py:1060] (3/4) Exclude cut with ID 7276-92427-0014-12983-0_sp0.9 from training. Duration: 21.3055625 2022-12-23 15:06:25,455 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=67787.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:06:26,745 WARNING [train.py:1060] (3/4) Exclude cut with ID 1025-75365-0008-79168-0_sp0.9 from training. Duration: 22.0666875 2022-12-23 15:06:31,712 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2022-12-23 15:07:05,590 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2022-12-23 15:07:10,100 INFO [train.py:894] (3/4) Epoch 20, batch 1200, loss[loss=0.1768, simple_loss=0.2695, pruned_loss=0.04207, over 18683.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2648, pruned_loss=0.04416, over 3704206.94 frames. ], batch size: 60, lr: 5.77e-03, grad_scale: 8.0 2022-12-23 15:07:25,580 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4788, 2.1662, 1.6605, 1.5852, 2.1187, 2.9731, 2.7431, 1.9422], device='cuda:3'), covar=tensor([0.0346, 0.0353, 0.0549, 0.0325, 0.0304, 0.0411, 0.0335, 0.0372], device='cuda:3'), in_proj_covar=tensor([0.0095, 0.0127, 0.0154, 0.0125, 0.0116, 0.0120, 0.0097, 0.0126], device='cuda:3'), out_proj_covar=tensor([7.5606e-05, 1.0094e-04, 1.2724e-04, 1.0014e-04, 9.3949e-05, 9.2302e-05, 7.6542e-05, 9.9268e-05], device='cuda:3') 2022-12-23 15:07:43,805 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.358e+02 3.704e+02 4.374e+02 5.678e+02 1.404e+03, threshold=8.749e+02, percent-clipped=7.0 2022-12-23 15:08:11,531 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7391, 2.6110, 1.9902, 2.0599, 2.5250, 3.2271, 3.1448, 2.3389], device='cuda:3'), covar=tensor([0.0263, 0.0320, 0.0418, 0.0240, 0.0202, 0.0333, 0.0258, 0.0311], device='cuda:3'), in_proj_covar=tensor([0.0095, 0.0127, 0.0154, 0.0125, 0.0116, 0.0120, 0.0098, 0.0126], device='cuda:3'), out_proj_covar=tensor([7.5557e-05, 1.0086e-04, 1.2720e-04, 1.0021e-04, 9.3968e-05, 9.2594e-05, 7.6767e-05, 9.9275e-05], device='cuda:3') 2022-12-23 15:08:13,830 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0 from training. Duration: 20.26 2022-12-23 15:08:24,719 INFO [train.py:894] (3/4) Epoch 20, batch 1250, loss[loss=0.1764, simple_loss=0.2685, pruned_loss=0.0422, over 18513.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2654, pruned_loss=0.04415, over 3706580.88 frames. ], batch size: 52, lr: 5.77e-03, grad_scale: 8.0 2022-12-23 15:08:26,075 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0030-9324-0_sp0.9 from training. Duration: 21.3444375 2022-12-23 15:08:27,805 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67869.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 15:09:22,803 WARNING [train.py:1060] (3/4) Exclude cut with ID 497-129325-0061-62254-0_sp1.1 from training. Duration: 0.97725 2022-12-23 15:09:40,765 INFO [train.py:894] (3/4) Epoch 20, batch 1300, loss[loss=0.1596, simple_loss=0.2446, pruned_loss=0.03729, over 18496.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2652, pruned_loss=0.04422, over 3708599.77 frames. ], batch size: 41, lr: 5.77e-03, grad_scale: 8.0 2022-12-23 15:10:05,604 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0_sp0.9 from training. Duration: 22.97225 2022-12-23 15:10:15,143 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.448e+02 3.360e+02 4.182e+02 5.041e+02 9.316e+02, threshold=8.363e+02, percent-clipped=1.0 2022-12-23 15:10:27,297 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6724, 2.3845, 1.7271, 2.7055, 2.9040, 1.7542, 1.9133, 1.2913], device='cuda:3'), covar=tensor([0.2012, 0.1690, 0.1569, 0.0924, 0.1544, 0.1164, 0.1997, 0.1641], device='cuda:3'), in_proj_covar=tensor([0.0244, 0.0224, 0.0212, 0.0198, 0.0260, 0.0195, 0.0222, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 15:10:38,153 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0047-39922-0_sp0.9 from training. Duration: 21.97775 2022-12-23 15:10:49,873 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2022-12-23 15:10:50,506 WARNING [train.py:1060] (3/4) Exclude cut with ID 1112-1043-0006-89194-0_sp0.9 from training. Duration: 21.8333125 2022-12-23 15:10:56,998 INFO [train.py:894] (3/4) Epoch 20, batch 1350, loss[loss=0.1555, simple_loss=0.2459, pruned_loss=0.03256, over 18572.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2638, pruned_loss=0.04352, over 3708836.14 frames. ], batch size: 51, lr: 5.76e-03, grad_scale: 8.0 2022-12-23 15:11:01,565 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0 from training. Duration: 20.47 2022-12-23 15:11:18,872 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67981.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:11:29,616 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8566, 1.3601, 0.6495, 1.4562, 2.2333, 1.1839, 1.6563, 1.7771], device='cuda:3'), covar=tensor([0.1656, 0.2094, 0.2455, 0.1588, 0.1668, 0.1840, 0.1477, 0.1713], device='cuda:3'), in_proj_covar=tensor([0.0095, 0.0099, 0.0117, 0.0098, 0.0117, 0.0092, 0.0099, 0.0095], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 15:11:53,603 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5506, 1.3230, 0.6559, 1.2247, 2.0021, 0.6344, 1.3497, 1.4801], device='cuda:3'), covar=tensor([0.1642, 0.1877, 0.2005, 0.1564, 0.1693, 0.1759, 0.1401, 0.1680], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0099, 0.0117, 0.0097, 0.0117, 0.0092, 0.0098, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 15:12:09,906 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0037-39912-0_sp0.9 from training. Duration: 20.67225 2022-12-23 15:12:15,618 INFO [train.py:894] (3/4) Epoch 20, batch 1400, loss[loss=0.1567, simple_loss=0.2429, pruned_loss=0.03523, over 18461.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2639, pruned_loss=0.04347, over 3710121.94 frames. ], batch size: 50, lr: 5.76e-03, grad_scale: 8.0 2022-12-23 15:12:28,930 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp0.9 from training. Duration: 25.2444375 2022-12-23 15:12:41,884 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2022-12-23 15:12:44,185 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68036.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:12:48,511 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.229e+02 4.016e+02 4.665e+02 5.451e+02 1.427e+03, threshold=9.330e+02, percent-clipped=6.0 2022-12-23 15:12:52,774 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0021-76797-0_sp0.9 from training. Duration: 21.1445 2022-12-23 15:12:53,039 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68042.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:13:01,335 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68048.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 15:13:30,353 INFO [train.py:894] (3/4) Epoch 20, batch 1450, loss[loss=0.1661, simple_loss=0.2585, pruned_loss=0.03684, over 18592.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2635, pruned_loss=0.04337, over 3711863.61 frames. ], batch size: 51, lr: 5.76e-03, grad_scale: 8.0 2022-12-23 15:13:37,824 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5427, 1.1114, 0.6822, 1.1711, 1.8581, 0.5867, 1.2527, 1.3745], device='cuda:3'), covar=tensor([0.1648, 0.2077, 0.2001, 0.1525, 0.1935, 0.1908, 0.1492, 0.1717], device='cuda:3'), in_proj_covar=tensor([0.0095, 0.0099, 0.0117, 0.0098, 0.0117, 0.0092, 0.0099, 0.0095], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 15:14:00,930 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68087.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 15:14:03,852 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6088, 2.1793, 1.7136, 2.4377, 1.9327, 2.1784, 2.0239, 2.3621], device='cuda:3'), covar=tensor([0.1942, 0.3262, 0.1925, 0.2659, 0.3526, 0.1026, 0.3111, 0.0981], device='cuda:3'), in_proj_covar=tensor([0.0295, 0.0291, 0.0247, 0.0348, 0.0272, 0.0229, 0.0290, 0.0215], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 15:14:10,527 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp0.9 from training. Duration: 33.038875 2022-12-23 15:14:27,313 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7009, 1.6981, 1.8502, 1.7106, 1.2670, 3.8837, 1.8631, 2.1705], device='cuda:3'), covar=tensor([0.3060, 0.1973, 0.1826, 0.1968, 0.1383, 0.0147, 0.1509, 0.0805], device='cuda:3'), in_proj_covar=tensor([0.0131, 0.0117, 0.0125, 0.0120, 0.0103, 0.0096, 0.0091, 0.0088], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 15:14:45,603 INFO [train.py:894] (3/4) Epoch 20, batch 1500, loss[loss=0.1843, simple_loss=0.2712, pruned_loss=0.04866, over 18491.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2638, pruned_loss=0.04322, over 3712403.51 frames. ], batch size: 54, lr: 5.76e-03, grad_scale: 8.0 2022-12-23 15:14:49,997 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64291-0000-16059-0_sp0.9 from training. Duration: 20.0944375 2022-12-23 15:15:05,416 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp1.1 from training. Duration: 20.4 2022-12-23 15:15:12,682 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0 from training. Duration: 20.085 2022-12-23 15:15:18,477 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.391e+02 3.554e+02 4.254e+02 5.043e+02 9.798e+02, threshold=8.509e+02, percent-clipped=1.0 2022-12-23 15:15:22,864 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0_sp0.9 from training. Duration: 23.07775 2022-12-23 15:15:32,131 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68148.0, num_to_drop=1, layers_to_drop={3} 2022-12-23 15:16:00,685 INFO [train.py:894] (3/4) Epoch 20, batch 1550, loss[loss=0.1762, simple_loss=0.2718, pruned_loss=0.04026, over 18664.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2646, pruned_loss=0.04353, over 3712823.35 frames. ], batch size: 62, lr: 5.76e-03, grad_scale: 8.0 2022-12-23 15:16:03,845 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68169.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:16:10,337 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp0.9 from training. Duration: 24.9333125 2022-12-23 15:16:56,216 WARNING [train.py:1060] (3/4) Exclude cut with ID 5118-111612-0016-124680-0_sp0.9 from training. Duration: 20.388875 2022-12-23 15:17:03,305 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp1.1 from training. Duration: 20.3590625 2022-12-23 15:17:15,379 INFO [train.py:894] (3/4) Epoch 20, batch 1600, loss[loss=0.1846, simple_loss=0.2742, pruned_loss=0.04752, over 18408.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2648, pruned_loss=0.0437, over 3712161.12 frames. ], batch size: 48, lr: 5.75e-03, grad_scale: 8.0 2022-12-23 15:17:15,560 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=68217.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:17:43,297 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2022-12-23 15:17:43,847 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8119, 1.2959, 0.9693, 1.3706, 2.1207, 1.3330, 1.3327, 1.6307], device='cuda:3'), covar=tensor([0.2250, 0.2874, 0.2532, 0.2200, 0.1983, 0.2196, 0.2133, 0.2659], device='cuda:3'), in_proj_covar=tensor([0.0095, 0.0099, 0.0117, 0.0097, 0.0117, 0.0092, 0.0098, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 15:17:49,162 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.373e+02 3.652e+02 4.480e+02 5.377e+02 1.144e+03, threshold=8.960e+02, percent-clipped=5.0 2022-12-23 15:18:08,624 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0_sp1.1 from training. Duration: 0.836375 2022-12-23 15:18:30,945 INFO [train.py:894] (3/4) Epoch 20, batch 1650, loss[loss=0.1959, simple_loss=0.2871, pruned_loss=0.05237, over 18584.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2649, pruned_loss=0.04443, over 3712700.50 frames. ], batch size: 57, lr: 5.75e-03, grad_scale: 8.0 2022-12-23 15:18:41,666 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.4133, 3.8377, 3.7611, 4.2518, 4.0382, 3.9805, 4.5887, 1.3684], device='cuda:3'), covar=tensor([0.0750, 0.0687, 0.0660, 0.0861, 0.1448, 0.1147, 0.0581, 0.4901], device='cuda:3'), in_proj_covar=tensor([0.0335, 0.0219, 0.0230, 0.0258, 0.0316, 0.0264, 0.0281, 0.0275], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 15:18:54,995 WARNING [train.py:1060] (3/4) Exclude cut with ID 8565-290391-0049-67394-0_sp0.9 from training. Duration: 21.3166875 2022-12-23 15:19:24,432 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0029-104863-0_sp0.9 from training. Duration: 22.1055625 2022-12-23 15:19:26,235 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5158, 2.3358, 2.0591, 1.3347, 2.8779, 2.6581, 2.3639, 1.8644], device='cuda:3'), covar=tensor([0.0346, 0.0393, 0.0493, 0.0768, 0.0266, 0.0342, 0.0430, 0.0768], device='cuda:3'), in_proj_covar=tensor([0.0122, 0.0125, 0.0128, 0.0119, 0.0096, 0.0121, 0.0134, 0.0154], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 15:19:35,076 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp1.1 from training. Duration: 21.77725 2022-12-23 15:19:42,642 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8560, 1.2352, 0.6749, 1.3716, 2.2185, 1.0128, 1.5359, 1.8189], device='cuda:3'), covar=tensor([0.1504, 0.2080, 0.2204, 0.1478, 0.1632, 0.1695, 0.1445, 0.1479], device='cuda:3'), in_proj_covar=tensor([0.0095, 0.0099, 0.0117, 0.0097, 0.0117, 0.0092, 0.0099, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 15:19:45,561 INFO [train.py:894] (3/4) Epoch 20, batch 1700, loss[loss=0.1929, simple_loss=0.2736, pruned_loss=0.0561, over 18671.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2662, pruned_loss=0.0457, over 3712865.33 frames. ], batch size: 178, lr: 5.75e-03, grad_scale: 8.0 2022-12-23 15:19:57,125 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp0.9 from training. Duration: 27.8166875 2022-12-23 15:20:14,647 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68336.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:20:15,837 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68337.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:20:18,388 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.300e+02 4.261e+02 4.999e+02 6.266e+02 1.831e+03, threshold=9.998e+02, percent-clipped=7.0 2022-12-23 15:20:21,355 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp1.1 from training. Duration: 22.5090625 2022-12-23 15:20:28,589 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0 from training. Duration: 25.035 2022-12-23 15:20:31,686 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68348.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:20:46,675 WARNING [train.py:1060] (3/4) Exclude cut with ID 774-127930-0014-10412-0_sp1.1 from training. Duration: 0.95 2022-12-23 15:20:52,823 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2714, 1.8913, 1.4281, 0.5177, 1.4022, 1.9863, 1.6630, 1.7987], device='cuda:3'), covar=tensor([0.0586, 0.0570, 0.1073, 0.1571, 0.1142, 0.1407, 0.1584, 0.0651], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0183, 0.0204, 0.0190, 0.0209, 0.0200, 0.0214, 0.0202], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 15:20:59,563 INFO [train.py:894] (3/4) Epoch 20, batch 1750, loss[loss=0.212, simple_loss=0.2884, pruned_loss=0.0678, over 18605.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2668, pruned_loss=0.04687, over 3712165.27 frames. ], batch size: 51, lr: 5.75e-03, grad_scale: 8.0 2022-12-23 15:21:05,889 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp0.9 from training. Duration: 0.92225 2022-12-23 15:21:17,216 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4830, 2.0850, 1.5109, 2.4690, 2.6393, 1.6012, 1.6235, 1.2114], device='cuda:3'), covar=tensor([0.2146, 0.1851, 0.1762, 0.0983, 0.1353, 0.1319, 0.2218, 0.1768], device='cuda:3'), in_proj_covar=tensor([0.0242, 0.0222, 0.0211, 0.0197, 0.0258, 0.0194, 0.0220, 0.0194], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 15:21:25,519 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=68384.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:21:32,774 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0 from training. Duration: 21.97 2022-12-23 15:21:43,038 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=68396.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:21:48,179 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68399.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:21:53,998 WARNING [train.py:1060] (3/4) Exclude cut with ID 7492-105653-0055-62765-0_sp0.9 from training. Duration: 21.97225 2022-12-23 15:21:55,336 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp0.9 from training. Duration: 25.3333125 2022-12-23 15:22:06,844 WARNING [train.py:1060] (3/4) Exclude cut with ID 5172-29468-0015-19128-0_sp0.9 from training. Duration: 21.5055625 2022-12-23 15:22:14,874 INFO [train.py:894] (3/4) Epoch 20, batch 1800, loss[loss=0.1659, simple_loss=0.2401, pruned_loss=0.04586, over 18443.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2671, pruned_loss=0.04828, over 3712644.03 frames. ], batch size: 42, lr: 5.75e-03, grad_scale: 8.0 2022-12-23 15:22:16,421 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp1.1 from training. Duration: 20.72725 2022-12-23 15:22:28,482 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7108, 1.7314, 1.7177, 1.6822, 1.3655, 3.7800, 1.7508, 2.1328], device='cuda:3'), covar=tensor([0.3089, 0.1935, 0.1911, 0.2038, 0.1397, 0.0174, 0.1626, 0.0848], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0117, 0.0126, 0.0121, 0.0103, 0.0096, 0.0092, 0.0088], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 15:22:46,866 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp0.9 from training. Duration: 26.32775 2022-12-23 15:22:48,148 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.646e+02 4.362e+02 5.315e+02 6.425e+02 1.643e+03, threshold=1.063e+03, percent-clipped=4.0 2022-12-23 15:22:54,446 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68443.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 15:23:20,159 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0 from training. Duration: 20.025 2022-12-23 15:23:20,547 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68460.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:23:24,933 WARNING [train.py:1060] (3/4) Exclude cut with ID 6709-74022-0004-86860-0_sp1.1 from training. Duration: 0.9409375 2022-12-23 15:23:24,945 WARNING [train.py:1060] (3/4) Exclude cut with ID 4757-1811-0023-62229-0_sp0.9 from training. Duration: 21.37775 2022-12-23 15:23:30,996 INFO [train.py:894] (3/4) Epoch 20, batch 1850, loss[loss=0.1837, simple_loss=0.2736, pruned_loss=0.04696, over 18580.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2677, pruned_loss=0.04945, over 3712092.01 frames. ], batch size: 56, lr: 5.74e-03, grad_scale: 8.0 2022-12-23 15:23:45,462 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0004-25974-0_sp0.9 from training. Duration: 21.17225 2022-12-23 15:23:45,477 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp0.9 from training. Duration: 27.511125 2022-12-23 15:24:00,307 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2591, 1.9557, 1.5001, 0.5694, 1.4805, 1.9299, 1.6402, 1.9036], device='cuda:3'), covar=tensor([0.0611, 0.0486, 0.1068, 0.1529, 0.1063, 0.1399, 0.1533, 0.0602], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0184, 0.0205, 0.0191, 0.0210, 0.0202, 0.0216, 0.0202], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 15:24:19,265 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0 from training. Duration: 22.8 2022-12-23 15:24:23,692 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0 from training. Duration: 22.585 2022-12-23 15:24:47,602 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3786, 2.6369, 2.8864, 0.9755, 2.5989, 3.1728, 2.2114, 2.4835], device='cuda:3'), covar=tensor([0.0711, 0.0396, 0.0378, 0.0479, 0.0386, 0.0313, 0.0402, 0.0648], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0169, 0.0124, 0.0139, 0.0147, 0.0141, 0.0162, 0.0171], device='cuda:3'), out_proj_covar=tensor([1.1412e-04, 1.2964e-04, 9.3631e-05, 1.0368e-04, 1.1099e-04, 1.0838e-04, 1.2484e-04, 1.3043e-04], device='cuda:3') 2022-12-23 15:24:48,438 INFO [train.py:894] (3/4) Epoch 20, batch 1900, loss[loss=0.1838, simple_loss=0.2696, pruned_loss=0.04904, over 18582.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.268, pruned_loss=0.05035, over 3712848.95 frames. ], batch size: 57, lr: 5.74e-03, grad_scale: 8.0 2022-12-23 15:24:55,775 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0001-146967-0_sp0.9 from training. Duration: 22.0166875 2022-12-23 15:25:12,922 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp1.1 from training. Duration: 24.395375 2022-12-23 15:25:16,890 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp0.9 from training. Duration: 27.47775 2022-12-23 15:25:19,916 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.101e+02 4.390e+02 5.434e+02 6.998e+02 1.389e+03, threshold=1.087e+03, percent-clipped=7.0 2022-12-23 15:25:19,994 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp0.9 from training. Duration: 24.8833125 2022-12-23 15:25:23,693 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0 from training. Duration: 23.39 2022-12-23 15:25:29,728 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp0.9 from training. Duration: 28.72225 2022-12-23 15:25:39,443 WARNING [train.py:1060] (3/4) Exclude cut with ID 585-294811-0110-133686-0_sp0.9 from training. Duration: 20.8944375 2022-12-23 15:25:55,747 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0_sp0.9 from training. Duration: 23.8444375 2022-12-23 15:26:04,397 INFO [train.py:894] (3/4) Epoch 20, batch 1950, loss[loss=0.1762, simple_loss=0.2645, pruned_loss=0.04392, over 18524.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2679, pruned_loss=0.0511, over 3712439.52 frames. ], batch size: 96, lr: 5.74e-03, grad_scale: 8.0 2022-12-23 15:26:20,189 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0 from training. Duration: 25.85 2022-12-23 15:26:20,201 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0 from training. Duration: 21.39 2022-12-23 15:26:20,589 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68578.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:26:31,422 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0 from training. Duration: 27.92 2022-12-23 15:26:57,764 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0039-130165-0_sp0.9 from training. Duration: 20.661125 2022-12-23 15:27:19,092 INFO [train.py:894] (3/4) Epoch 20, batch 2000, loss[loss=0.1637, simple_loss=0.2458, pruned_loss=0.04078, over 18694.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.269, pruned_loss=0.05214, over 3713958.85 frames. ], batch size: 48, lr: 5.74e-03, grad_scale: 8.0 2022-12-23 15:27:22,153 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0043-15874-0_sp0.9 from training. Duration: 20.07225 2022-12-23 15:27:27,566 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.56 vs. limit=5.0 2022-12-23 15:27:29,610 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0 from training. Duration: 21.01 2022-12-23 15:27:50,188 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68637.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:27:52,678 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.139e+02 4.584e+02 5.306e+02 6.807e+02 1.087e+03, threshold=1.061e+03, percent-clipped=0.0 2022-12-23 15:27:53,205 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68639.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:28:24,095 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.9931, 2.4260, 1.8482, 2.9339, 2.2140, 2.2770, 2.3493, 3.3004], device='cuda:3'), covar=tensor([0.1926, 0.3137, 0.1899, 0.2818, 0.3670, 0.1046, 0.3179, 0.0749], device='cuda:3'), in_proj_covar=tensor([0.0295, 0.0290, 0.0246, 0.0348, 0.0270, 0.0228, 0.0288, 0.0214], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 15:28:35,210 INFO [train.py:894] (3/4) Epoch 20, batch 2050, loss[loss=0.2048, simple_loss=0.28, pruned_loss=0.0648, over 18567.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2694, pruned_loss=0.05291, over 3714900.10 frames. ], batch size: 49, lr: 5.74e-03, grad_scale: 8.0 2022-12-23 15:28:38,163 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0 from training. Duration: 20.65 2022-12-23 15:28:40,020 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1858, 1.6015, 2.5897, 4.3228, 3.0100, 2.7349, 0.6620, 3.2237], device='cuda:3'), covar=tensor([0.1662, 0.1492, 0.1228, 0.0457, 0.0966, 0.1191, 0.2310, 0.0762], device='cuda:3'), in_proj_covar=tensor([0.0102, 0.0118, 0.0135, 0.0148, 0.0106, 0.0141, 0.0130, 0.0113], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 15:28:45,683 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0 from training. Duration: 21.46 2022-12-23 15:29:03,105 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=68685.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:29:03,370 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68685.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:29:28,000 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2022-12-23 15:29:30,053 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0 from training. Duration: 0.92 2022-12-23 15:29:36,641 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0_sp0.9 from training. Duration: 23.7666875 2022-12-23 15:29:47,239 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6110, 1.6390, 1.6581, 1.6715, 1.0173, 3.7004, 1.6169, 1.9918], device='cuda:3'), covar=tensor([0.3139, 0.2046, 0.1878, 0.1972, 0.1556, 0.0184, 0.1534, 0.0883], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0117, 0.0125, 0.0120, 0.0103, 0.0097, 0.0092, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 15:29:51,482 INFO [train.py:894] (3/4) Epoch 20, batch 2100, loss[loss=0.2354, simple_loss=0.317, pruned_loss=0.07686, over 18501.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2704, pruned_loss=0.05349, over 3715588.23 frames. ], batch size: 58, lr: 5.73e-03, grad_scale: 8.0 2022-12-23 15:29:53,309 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7666, 1.4642, 1.7019, 2.0568, 1.7706, 3.6619, 1.3822, 1.5849], device='cuda:3'), covar=tensor([0.1046, 0.2280, 0.1169, 0.1027, 0.1740, 0.0263, 0.1839, 0.1986], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0083, 0.0073, 0.0075, 0.0091, 0.0077, 0.0085, 0.0078], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2022-12-23 15:30:13,740 WARNING [train.py:1060] (3/4) Exclude cut with ID 8544-281189-0060-101339-0_sp0.9 from training. Duration: 20.861125 2022-12-23 15:30:22,628 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0_sp0.9 from training. Duration: 22.711125 2022-12-23 15:30:24,146 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.984e+02 4.264e+02 4.924e+02 6.073e+02 1.296e+03, threshold=9.848e+02, percent-clipped=2.0 2022-12-23 15:30:30,281 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68743.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 15:30:35,384 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68746.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:30:49,346 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68755.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:31:02,350 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp1.1 from training. Duration: 22.986375 2022-12-23 15:31:06,592 INFO [train.py:894] (3/4) Epoch 20, batch 2150, loss[loss=0.1782, simple_loss=0.2594, pruned_loss=0.04854, over 18455.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2692, pruned_loss=0.05325, over 3714659.06 frames. ], batch size: 50, lr: 5.73e-03, grad_scale: 8.0 2022-12-23 15:31:18,256 WARNING [train.py:1060] (3/4) Exclude cut with ID 8040-260924-0003-80960-0_sp0.9 from training. Duration: 22.07225 2022-12-23 15:31:21,243 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0045-26330-0_sp0.9 from training. Duration: 20.3055625 2022-12-23 15:31:24,174 WARNING [train.py:1060] (3/4) Exclude cut with ID 6356-271890-0060-94317-0_sp0.9 from training. Duration: 20.72225 2022-12-23 15:31:24,829 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2022-12-23 15:31:42,096 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp1.1 from training. Duration: 22.4818125 2022-12-23 15:31:44,328 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=68791.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 15:32:02,493 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2022-12-23 15:32:10,105 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp0.9 from training. Duration: 25.0944375 2022-12-23 15:32:13,097 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0 from training. Duration: 21.515 2022-12-23 15:32:20,596 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp0.9 from training. Duration: 27.02225 2022-12-23 15:32:23,512 INFO [train.py:894] (3/4) Epoch 20, batch 2200, loss[loss=0.1865, simple_loss=0.2738, pruned_loss=0.04962, over 18465.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2688, pruned_loss=0.0532, over 3714310.64 frames. ], batch size: 50, lr: 5.73e-03, grad_scale: 8.0 2022-12-23 15:32:23,641 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0010-62480-0_sp0.9 from training. Duration: 22.22225 2022-12-23 15:32:33,522 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0085-44554-0_sp0.9 from training. Duration: 20.85 2022-12-23 15:32:34,313 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2022-12-23 15:32:57,608 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.534e+02 4.308e+02 5.383e+02 7.138e+02 1.577e+03, threshold=1.077e+03, percent-clipped=5.0 2022-12-23 15:33:03,707 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0 from training. Duration: 21.54 2022-12-23 15:33:10,638 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp1.1 from training. Duration: 20.5318125 2022-12-23 15:33:20,769 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0012-134311-0_sp0.9 from training. Duration: 21.9333125 2022-12-23 15:33:40,882 INFO [train.py:894] (3/4) Epoch 20, batch 2250, loss[loss=0.2146, simple_loss=0.2907, pruned_loss=0.06921, over 18588.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2686, pruned_loss=0.05352, over 3714355.31 frames. ], batch size: 174, lr: 5.73e-03, grad_scale: 8.0 2022-12-23 15:33:50,815 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2022-12-23 15:34:03,818 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2022-12-23 15:34:08,715 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0025-130151-0_sp0.9 from training. Duration: 21.7944375 2022-12-23 15:34:20,876 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0_sp0.9 from training. Duration: 22.4666875 2022-12-23 15:34:27,000 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0 from training. Duration: 21.635 2022-12-23 15:34:33,080 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0_sp0.9 from training. Duration: 24.038875 2022-12-23 15:34:57,026 INFO [train.py:894] (3/4) Epoch 20, batch 2300, loss[loss=0.1715, simple_loss=0.2591, pruned_loss=0.04193, over 18716.00 frames. ], tot_loss[loss=0.188, simple_loss=0.269, pruned_loss=0.05346, over 3714132.70 frames. ], batch size: 60, lr: 5.72e-03, grad_scale: 8.0 2022-12-23 15:35:15,154 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp1.1 from training. Duration: 21.786375 2022-12-23 15:35:22,773 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68934.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:35:26,337 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0 from training. Duration: 20.22 2022-12-23 15:35:30,638 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.085e+02 4.536e+02 5.472e+02 6.588e+02 1.270e+03, threshold=1.094e+03, percent-clipped=1.0 2022-12-23 15:36:13,470 INFO [train.py:894] (3/4) Epoch 20, batch 2350, loss[loss=0.2012, simple_loss=0.2866, pruned_loss=0.05794, over 18665.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2686, pruned_loss=0.05341, over 3715149.14 frames. ], batch size: 60, lr: 5.72e-03, grad_scale: 8.0 2022-12-23 15:36:29,234 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5208, 1.4618, 1.5459, 1.4840, 1.0390, 3.5971, 1.4978, 1.9942], device='cuda:3'), covar=tensor([0.3274, 0.2151, 0.2075, 0.2159, 0.1609, 0.0199, 0.1609, 0.0866], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0118, 0.0126, 0.0121, 0.0103, 0.0097, 0.0092, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 15:37:22,354 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0 from training. Duration: 25.285 2022-12-23 15:37:27,087 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.2434, 1.2280, 0.8950, 1.2980, 1.5016, 2.4126, 1.2398, 1.4731], device='cuda:3'), covar=tensor([0.0938, 0.1990, 0.1187, 0.0977, 0.1517, 0.0345, 0.1561, 0.1605], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0083, 0.0073, 0.0075, 0.0091, 0.0076, 0.0085, 0.0078], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 15:37:30,431 INFO [train.py:894] (3/4) Epoch 20, batch 2400, loss[loss=0.1997, simple_loss=0.2801, pruned_loss=0.05963, over 18716.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2695, pruned_loss=0.05409, over 3715078.93 frames. ], batch size: 78, lr: 5.72e-03, grad_scale: 8.0 2022-12-23 15:38:02,923 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.677e+02 4.455e+02 5.012e+02 6.194e+02 1.142e+03, threshold=1.002e+03, percent-clipped=1.0 2022-12-23 15:38:05,980 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69041.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:38:09,005 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69043.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:38:10,677 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7335, 1.5200, 1.8344, 1.1399, 1.7940, 1.8323, 1.3782, 2.1871], device='cuda:3'), covar=tensor([0.0927, 0.1788, 0.1102, 0.1474, 0.0655, 0.0981, 0.2342, 0.0475], device='cuda:3'), in_proj_covar=tensor([0.0199, 0.0215, 0.0212, 0.0197, 0.0178, 0.0219, 0.0219, 0.0194], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 15:38:27,588 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69055.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:38:30,179 WARNING [train.py:1060] (3/4) Exclude cut with ID 811-130148-0001-63453-0_sp0.9 from training. Duration: 20.861125 2022-12-23 15:38:45,651 INFO [train.py:894] (3/4) Epoch 20, batch 2450, loss[loss=0.2049, simple_loss=0.2865, pruned_loss=0.06167, over 18513.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2695, pruned_loss=0.05371, over 3716146.25 frames. ], batch size: 99, lr: 5.72e-03, grad_scale: 8.0 2022-12-23 15:38:51,262 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0 from training. Duration: 20.88 2022-12-23 15:39:26,583 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0_sp0.9 from training. Duration: 23.4166875 2022-12-23 15:39:34,197 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0928, 2.1874, 1.7345, 2.5003, 2.2660, 2.0055, 2.6162, 2.2376], device='cuda:3'), covar=tensor([0.0782, 0.1429, 0.2204, 0.1388, 0.1398, 0.0761, 0.0924, 0.0983], device='cuda:3'), in_proj_covar=tensor([0.0177, 0.0208, 0.0250, 0.0290, 0.0236, 0.0189, 0.0209, 0.0202], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 15:39:39,597 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69103.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:39:41,333 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69104.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:40:00,770 INFO [train.py:894] (3/4) Epoch 20, batch 2500, loss[loss=0.1752, simple_loss=0.2591, pruned_loss=0.04572, over 18578.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.269, pruned_loss=0.05325, over 3715565.92 frames. ], batch size: 51, lr: 5.72e-03, grad_scale: 8.0 2022-12-23 15:40:33,622 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.874e+02 4.273e+02 5.084e+02 6.265e+02 1.046e+03, threshold=1.017e+03, percent-clipped=1.0 2022-12-23 15:40:41,378 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0 from training. Duration: 21.24 2022-12-23 15:40:41,399 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0_sp0.9 from training. Duration: 23.9055625 2022-12-23 15:41:09,468 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69162.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:41:15,270 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp0.9 from training. Duration: 25.988875 2022-12-23 15:41:16,785 INFO [train.py:894] (3/4) Epoch 20, batch 2550, loss[loss=0.2103, simple_loss=0.2823, pruned_loss=0.06918, over 18586.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2691, pruned_loss=0.0536, over 3714203.06 frames. ], batch size: 176, lr: 5.71e-03, grad_scale: 8.0 2022-12-23 15:41:19,429 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69168.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:41:22,082 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0001-134300-0_sp0.9 from training. Duration: 20.67225 2022-12-23 15:42:05,919 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3591, 2.7644, 3.1417, 1.2097, 2.7498, 3.4861, 2.5690, 2.6561], device='cuda:3'), covar=tensor([0.0937, 0.0424, 0.0277, 0.0521, 0.0382, 0.0360, 0.0373, 0.0785], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0171, 0.0126, 0.0140, 0.0148, 0.0143, 0.0163, 0.0173], device='cuda:3'), out_proj_covar=tensor([1.1576e-04, 1.3101e-04, 9.5349e-05, 1.0464e-04, 1.1131e-04, 1.0974e-04, 1.2573e-04, 1.3190e-04], device='cuda:3') 2022-12-23 15:42:12,361 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0 from training. Duration: 20.34 2022-12-23 15:42:33,140 INFO [train.py:894] (3/4) Epoch 20, batch 2600, loss[loss=0.172, simple_loss=0.2637, pruned_loss=0.04014, over 18662.00 frames. ], tot_loss[loss=0.187, simple_loss=0.2684, pruned_loss=0.05283, over 3713861.18 frames. ], batch size: 60, lr: 5.71e-03, grad_scale: 8.0 2022-12-23 15:42:42,351 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69223.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:42:45,339 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69225.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:42:51,389 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69229.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:42:59,369 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69234.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:43:06,599 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.608e+02 4.487e+02 5.367e+02 6.345e+02 1.020e+03, threshold=1.073e+03, percent-clipped=1.0 2022-12-23 15:43:15,816 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69245.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 15:43:22,394 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp0.9 from training. Duration: 25.061125 2022-12-23 15:43:35,414 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0 from training. Duration: 0.83 2022-12-23 15:43:48,923 INFO [train.py:894] (3/4) Epoch 20, batch 2650, loss[loss=0.1802, simple_loss=0.2704, pruned_loss=0.04498, over 18563.00 frames. ], tot_loss[loss=0.1869, simple_loss=0.2683, pruned_loss=0.05274, over 3713651.56 frames. ], batch size: 98, lr: 5.71e-03, grad_scale: 8.0 2022-12-23 15:43:58,108 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69273.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:44:01,040 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0 from training. Duration: 24.73 2022-12-23 15:44:12,020 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69282.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:44:13,324 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0 from training. Duration: 23.965 2022-12-23 15:44:18,302 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69286.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:44:21,005 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0030-146996-0_sp0.9 from training. Duration: 22.088875 2022-12-23 15:44:36,680 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0_sp0.9 from training. Duration: 23.6 2022-12-23 15:44:38,699 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5298, 1.4315, 1.3032, 1.3936, 1.7178, 1.5900, 1.5732, 1.1630], device='cuda:3'), covar=tensor([0.0244, 0.0217, 0.0467, 0.0187, 0.0176, 0.0339, 0.0229, 0.0288], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0127, 0.0152, 0.0125, 0.0116, 0.0119, 0.0097, 0.0126], device='cuda:3'), out_proj_covar=tensor([7.5051e-05, 1.0090e-04, 1.2600e-04, 1.0008e-04, 9.3928e-05, 9.1732e-05, 7.6021e-05, 9.9485e-05], device='cuda:3') 2022-12-23 15:44:49,845 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69306.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 15:44:50,213 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2022-12-23 15:45:06,097 INFO [train.py:894] (3/4) Epoch 20, batch 2700, loss[loss=0.2041, simple_loss=0.2748, pruned_loss=0.06667, over 18727.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2685, pruned_loss=0.0532, over 3714308.16 frames. ], batch size: 46, lr: 5.71e-03, grad_scale: 16.0 2022-12-23 15:45:11,043 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69320.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:45:32,422 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69334.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 15:45:39,091 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.969e+02 4.616e+02 5.478e+02 6.566e+02 1.187e+03, threshold=1.096e+03, percent-clipped=2.0 2022-12-23 15:45:42,700 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69341.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:46:17,248 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.00 vs. limit=5.0 2022-12-23 15:46:17,815 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0 from training. Duration: 23.795 2022-12-23 15:46:20,603 INFO [train.py:894] (3/4) Epoch 20, batch 2750, loss[loss=0.1846, simple_loss=0.254, pruned_loss=0.05765, over 18530.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2683, pruned_loss=0.05344, over 3714681.80 frames. ], batch size: 44, lr: 5.71e-03, grad_scale: 16.0 2022-12-23 15:46:22,390 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69368.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:46:35,057 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp1.1 from training. Duration: 21.5409375 2022-12-23 15:46:38,265 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp0.9 from training. Duration: 24.97775 2022-12-23 15:46:42,762 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69381.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:46:49,430 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0_sp0.9 from training. Duration: 23.3444375 2022-12-23 15:46:54,605 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69389.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:47:01,953 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69394.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:47:08,906 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69399.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:47:18,622 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0_sp0.9 from training. Duration: 23.2 2022-12-23 15:47:23,142 WARNING [train.py:1060] (3/4) Exclude cut with ID 5653-46179-0060-117930-0_sp0.9 from training. Duration: 21.17225 2022-12-23 15:47:36,459 INFO [train.py:894] (3/4) Epoch 20, batch 2800, loss[loss=0.2071, simple_loss=0.2855, pruned_loss=0.06434, over 18653.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2689, pruned_loss=0.05361, over 3714619.27 frames. ], batch size: 62, lr: 5.70e-03, grad_scale: 16.0 2022-12-23 15:47:43,201 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp0.9 from training. Duration: 24.6555625 2022-12-23 15:47:54,901 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69429.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:48:09,263 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.532e+02 4.423e+02 5.251e+02 6.577e+02 9.250e+02, threshold=1.050e+03, percent-clipped=0.0 2022-12-23 15:48:23,275 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7093, 1.6525, 1.8074, 1.0593, 1.8372, 1.8581, 1.4268, 2.2374], device='cuda:3'), covar=tensor([0.1042, 0.1783, 0.1112, 0.1608, 0.0675, 0.0999, 0.2385, 0.0522], device='cuda:3'), in_proj_covar=tensor([0.0201, 0.0216, 0.0211, 0.0199, 0.0178, 0.0221, 0.0221, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 15:48:33,297 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69455.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:48:38,918 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0 from training. Duration: 20.44 2022-12-23 15:48:50,729 INFO [train.py:894] (3/4) Epoch 20, batch 2850, loss[loss=0.2245, simple_loss=0.2928, pruned_loss=0.07805, over 18690.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2686, pruned_loss=0.05359, over 3714264.60 frames. ], batch size: 170, lr: 5.70e-03, grad_scale: 16.0 2022-12-23 15:48:53,582 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2022-12-23 15:48:54,357 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0_sp0.9 from training. Duration: 23.45 2022-12-23 15:49:24,420 WARNING [train.py:1060] (3/4) Exclude cut with ID 6945-60535-0076-12784-0_sp0.9 from training. Duration: 20.52225 2022-12-23 15:49:31,638 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0 from training. Duration: 22.19 2022-12-23 15:49:41,312 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp1.1 from training. Duration: 25.3818125 2022-12-23 15:49:58,831 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp0.9 from training. Duration: 28.0944375 2022-12-23 15:49:59,775 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8953, 1.7815, 1.8597, 2.0688, 1.3397, 5.0824, 1.8904, 2.3773], device='cuda:3'), covar=tensor([0.3056, 0.1978, 0.1848, 0.1796, 0.1361, 0.0109, 0.1432, 0.0813], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0117, 0.0125, 0.0120, 0.0103, 0.0097, 0.0092, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 15:50:05,680 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0_sp0.9 from training. Duration: 22.9444375 2022-12-23 15:50:06,390 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2022-12-23 15:50:07,296 INFO [train.py:894] (3/4) Epoch 20, batch 2900, loss[loss=0.2077, simple_loss=0.2916, pruned_loss=0.06192, over 18566.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2686, pruned_loss=0.05364, over 3714242.85 frames. ], batch size: 57, lr: 5.70e-03, grad_scale: 16.0 2022-12-23 15:50:08,961 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69518.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:50:12,368 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp1.1 from training. Duration: 21.6318125 2022-12-23 15:50:18,156 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69524.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:50:31,971 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0 from training. Duration: 23.695 2022-12-23 15:50:40,707 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.134e+02 4.802e+02 5.771e+02 6.924e+02 1.235e+03, threshold=1.154e+03, percent-clipped=4.0 2022-12-23 15:50:44,865 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5209, 2.1673, 1.7287, 0.6924, 1.5616, 2.0311, 1.8389, 1.9367], device='cuda:3'), covar=tensor([0.0581, 0.0550, 0.1109, 0.1662, 0.1114, 0.1378, 0.1365, 0.0732], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0185, 0.0208, 0.0193, 0.0212, 0.0203, 0.0216, 0.0204], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 15:50:56,220 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0 from training. Duration: 23.955 2022-12-23 15:51:23,310 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.7694, 2.0181, 2.2383, 1.2689, 1.6255, 2.3517, 2.0369, 1.8466], device='cuda:3'), covar=tensor([0.0753, 0.0299, 0.0293, 0.0406, 0.0360, 0.0414, 0.0250, 0.0666], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0170, 0.0126, 0.0140, 0.0149, 0.0142, 0.0163, 0.0172], device='cuda:3'), out_proj_covar=tensor([1.1463e-04, 1.3081e-04, 9.5137e-05, 1.0447e-04, 1.1185e-04, 1.0892e-04, 1.2531e-04, 1.3120e-04], device='cuda:3') 2022-12-23 15:51:24,256 INFO [train.py:894] (3/4) Epoch 20, batch 2950, loss[loss=0.1858, simple_loss=0.2723, pruned_loss=0.04968, over 18645.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2683, pruned_loss=0.05333, over 3714174.02 frames. ], batch size: 78, lr: 5.70e-03, grad_scale: 16.0 2022-12-23 15:51:29,890 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp0.9 from training. Duration: 26.438875 2022-12-23 15:51:44,968 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69581.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:51:52,199 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0829, 1.5244, 1.8516, 1.7705, 2.0600, 2.0586, 1.9197, 1.7909], device='cuda:3'), covar=tensor([0.2018, 0.2931, 0.2432, 0.2672, 0.1924, 0.0987, 0.2956, 0.1242], device='cuda:3'), in_proj_covar=tensor([0.0269, 0.0301, 0.0280, 0.0317, 0.0306, 0.0253, 0.0344, 0.0241], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 15:52:14,209 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69601.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 15:52:15,540 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0021-26306-0_sp0.9 from training. Duration: 21.2444375 2022-12-23 15:52:16,972 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp0.9 from training. Duration: 31.02225 2022-12-23 15:52:26,621 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0 from training. Duration: 22.395 2022-12-23 15:52:38,571 INFO [train.py:894] (3/4) Epoch 20, batch 3000, loss[loss=0.1648, simple_loss=0.2478, pruned_loss=0.04094, over 18573.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2673, pruned_loss=0.05273, over 3714973.52 frames. ], batch size: 49, lr: 5.70e-03, grad_scale: 16.0 2022-12-23 15:52:38,571 INFO [train.py:919] (3/4) Computing validation loss 2022-12-23 15:52:48,257 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4333, 1.7487, 0.9189, 1.7144, 2.3460, 1.8548, 2.0081, 2.2507], device='cuda:3'), covar=tensor([0.1550, 0.1986, 0.2437, 0.1524, 0.1746, 0.1564, 0.1364, 0.1644], device='cuda:3'), in_proj_covar=tensor([0.0095, 0.0099, 0.0117, 0.0097, 0.0119, 0.0093, 0.0098, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 15:52:49,540 INFO [train.py:928] (3/4) Epoch 20, validation: loss=0.1661, simple_loss=0.2631, pruned_loss=0.03452, over 944034.00 frames. 2022-12-23 15:52:49,541 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-23 15:52:55,198 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0 from training. Duration: 21.075 2022-12-23 15:53:01,719 WARNING [train.py:1060] (3/4) Exclude cut with ID 6482-98857-0025-147532-0_sp0.9 from training. Duration: 20.0055625 2022-12-23 15:53:01,732 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0037-132304-0_sp0.9 from training. Duration: 22.05 2022-12-23 15:53:01,747 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0 from training. Duration: 26.8349375 2022-12-23 15:53:04,735 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp1.1 from training. Duration: 22.1090625 2022-12-23 15:53:07,652 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69629.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 15:53:11,788 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp0.9 from training. Duration: 26.6166875 2022-12-23 15:53:22,195 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.117e+02 4.417e+02 5.463e+02 6.727e+02 1.419e+03, threshold=1.093e+03, percent-clipped=3.0 2022-12-23 15:53:30,675 WARNING [train.py:1060] (3/4) Exclude cut with ID 2046-178027-0000-53705-0_sp0.9 from training. Duration: 20.3055625 2022-12-23 15:53:37,160 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5443, 2.3103, 1.8805, 1.2733, 2.8483, 2.6775, 2.3442, 1.8954], device='cuda:3'), covar=tensor([0.0414, 0.0456, 0.0592, 0.0829, 0.0273, 0.0383, 0.0484, 0.0874], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0129, 0.0129, 0.0121, 0.0099, 0.0123, 0.0136, 0.0158], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 15:53:53,480 WARNING [train.py:1060] (3/4) Exclude cut with ID 7205-50138-0008-5373-0_sp0.9 from training. Duration: 20.7 2022-12-23 15:54:02,320 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=2.14 vs. limit=2.0 2022-12-23 15:54:05,305 INFO [train.py:894] (3/4) Epoch 20, batch 3050, loss[loss=0.1681, simple_loss=0.2498, pruned_loss=0.04322, over 18615.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2667, pruned_loss=0.05227, over 3714421.09 frames. ], batch size: 45, lr: 5.69e-03, grad_scale: 16.0 2022-12-23 15:54:15,094 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3845, 1.6959, 1.2805, 2.0177, 1.9176, 1.4708, 1.1700, 1.1835], device='cuda:3'), covar=tensor([0.2152, 0.1944, 0.1914, 0.1141, 0.1499, 0.1297, 0.2505, 0.1770], device='cuda:3'), in_proj_covar=tensor([0.0244, 0.0223, 0.0211, 0.0197, 0.0258, 0.0194, 0.0221, 0.0196], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 15:54:16,335 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69674.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 15:54:18,878 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69676.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:54:33,806 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0 from training. Duration: 22.48 2022-12-23 15:54:49,341 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp0.9 from training. Duration: 29.816625 2022-12-23 15:54:54,006 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69699.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:55:09,437 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp1.1 from training. Duration: 22.7590625 2022-12-23 15:55:13,787 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0 from training. Duration: 22.555 2022-12-23 15:55:21,877 INFO [train.py:894] (3/4) Epoch 20, batch 3100, loss[loss=0.207, simple_loss=0.2841, pruned_loss=0.06495, over 18645.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2678, pruned_loss=0.05272, over 3715189.31 frames. ], batch size: 178, lr: 5.69e-03, grad_scale: 16.0 2022-12-23 15:55:24,238 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2022-12-23 15:55:31,909 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69724.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:55:34,750 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0005-25975-0_sp0.9 from training. Duration: 21.688875 2022-12-23 15:55:49,033 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69735.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 15:55:54,821 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.420e+02 4.377e+02 5.437e+02 7.323e+02 2.028e+03, threshold=1.087e+03, percent-clipped=6.0 2022-12-23 15:56:06,073 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0_sp0.9 from training. Duration: 22.6 2022-12-23 15:56:07,662 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69747.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:56:12,592 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69750.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:56:32,652 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7865, 3.9253, 3.7445, 1.6079, 3.9198, 2.9523, 0.9541, 2.4632], device='cuda:3'), covar=tensor([0.1690, 0.1077, 0.1278, 0.3258, 0.0893, 0.0926, 0.4337, 0.1562], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0144, 0.0160, 0.0125, 0.0145, 0.0115, 0.0146, 0.0116], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 15:56:38,090 INFO [train.py:894] (3/4) Epoch 20, batch 3150, loss[loss=0.1863, simple_loss=0.2727, pruned_loss=0.04994, over 18708.00 frames. ], tot_loss[loss=0.187, simple_loss=0.2682, pruned_loss=0.05288, over 3714442.34 frames. ], batch size: 54, lr: 5.69e-03, grad_scale: 16.0 2022-12-23 15:56:45,344 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0 from training. Duration: 24.32 2022-12-23 15:57:45,985 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-276745-0093-13116-0_sp0.9 from training. Duration: 21.061125 2022-12-23 15:57:55,013 INFO [train.py:894] (3/4) Epoch 20, batch 3200, loss[loss=0.2289, simple_loss=0.2992, pruned_loss=0.07931, over 18550.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2682, pruned_loss=0.05266, over 3714189.03 frames. ], batch size: 176, lr: 5.69e-03, grad_scale: 16.0 2022-12-23 15:57:56,710 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69818.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:58:00,893 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0024-15855-0_sp0.9 from training. Duration: 20.32225 2022-12-23 15:58:02,611 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8756, 1.8022, 2.1072, 1.2845, 2.0170, 2.0096, 1.5320, 2.3403], device='cuda:3'), covar=tensor([0.0975, 0.1598, 0.1232, 0.1601, 0.0627, 0.1057, 0.2037, 0.0504], device='cuda:3'), in_proj_covar=tensor([0.0198, 0.0216, 0.0210, 0.0196, 0.0177, 0.0219, 0.0218, 0.0194], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 15:58:05,183 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69824.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:58:12,808 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp1.1 from training. Duration: 0.7545625 2022-12-23 15:58:14,602 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.4393, 3.2395, 2.0552, 1.5184, 3.7398, 3.6082, 3.1480, 2.7986], device='cuda:3'), covar=tensor([0.0339, 0.0353, 0.0599, 0.0786, 0.0196, 0.0329, 0.0415, 0.0645], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0128, 0.0129, 0.0120, 0.0099, 0.0123, 0.0135, 0.0157], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 15:58:27,640 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.717e+02 4.162e+02 5.372e+02 7.071e+02 1.448e+03, threshold=1.074e+03, percent-clipped=3.0 2022-12-23 15:58:29,437 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0_sp0.9 from training. Duration: 23.9333125 2022-12-23 15:58:34,108 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9289, 1.5734, 1.8751, 2.2392, 1.8751, 3.3900, 1.5439, 1.7476], device='cuda:3'), covar=tensor([0.0809, 0.1684, 0.1045, 0.0812, 0.1308, 0.0271, 0.1300, 0.1380], device='cuda:3'), in_proj_covar=tensor([0.0072, 0.0082, 0.0073, 0.0074, 0.0091, 0.0075, 0.0084, 0.0077], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 15:58:36,268 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2022-12-23 15:59:01,155 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp1.1 from training. Duration: 20.17275 2022-12-23 15:59:08,756 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp1.1 from training. Duration: 20.436375 2022-12-23 15:59:08,896 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69866.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:59:10,195 INFO [train.py:894] (3/4) Epoch 20, batch 3250, loss[loss=0.2014, simple_loss=0.2882, pruned_loss=0.05728, over 18527.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2672, pruned_loss=0.05205, over 3714086.73 frames. ], batch size: 58, lr: 5.69e-03, grad_scale: 16.0 2022-12-23 15:59:18,244 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69872.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:59:32,312 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69881.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 15:59:32,520 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9911, 2.3970, 1.6941, 2.6879, 3.1441, 1.8188, 1.9513, 1.4709], device='cuda:3'), covar=tensor([0.1733, 0.1513, 0.1486, 0.0846, 0.1211, 0.1055, 0.1814, 0.1391], device='cuda:3'), in_proj_covar=tensor([0.0244, 0.0223, 0.0211, 0.0197, 0.0258, 0.0194, 0.0221, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 16:00:02,384 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7305, 1.1264, 0.8340, 1.2718, 2.0083, 0.9209, 1.4035, 1.4826], device='cuda:3'), covar=tensor([0.1664, 0.2232, 0.2156, 0.1586, 0.1843, 0.1814, 0.1559, 0.1785], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0098, 0.0116, 0.0095, 0.0117, 0.0091, 0.0097, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 16:00:02,398 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69901.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 16:00:25,362 INFO [train.py:894] (3/4) Epoch 20, batch 3300, loss[loss=0.172, simple_loss=0.2521, pruned_loss=0.04595, over 18420.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2685, pruned_loss=0.05282, over 3714766.17 frames. ], batch size: 48, lr: 5.68e-03, grad_scale: 16.0 2022-12-23 16:00:25,427 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0_sp0.9 from training. Duration: 23.1055625 2022-12-23 16:00:28,835 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp1.1 from training. Duration: 23.5 2022-12-23 16:00:41,293 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp0.9 from training. Duration: 26.62775 2022-12-23 16:00:45,033 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69929.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:00:45,166 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69929.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 16:00:55,360 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0 from training. Duration: 21.105 2022-12-23 16:00:59,526 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.216e+02 4.161e+02 4.947e+02 6.395e+02 1.283e+03, threshold=9.894e+02, percent-clipped=2.0 2022-12-23 16:00:59,596 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0_sp0.9 from training. Duration: 24.411125 2022-12-23 16:01:14,918 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69949.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 16:01:25,482 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp1.1 from training. Duration: 21.263625 2022-12-23 16:01:42,436 INFO [train.py:894] (3/4) Epoch 20, batch 3350, loss[loss=0.1724, simple_loss=0.245, pruned_loss=0.04991, over 18672.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2666, pruned_loss=0.05183, over 3714285.59 frames. ], batch size: 48, lr: 5.68e-03, grad_scale: 16.0 2022-12-23 16:01:57,070 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69976.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:01:58,269 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69977.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:02:00,977 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0 from training. Duration: 20.795 2022-12-23 16:02:11,152 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0 from training. Duration: 24.76 2022-12-23 16:02:11,166 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0_sp0.9 from training. Duration: 22.25 2022-12-23 16:02:38,276 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp1.1 from training. Duration: 20.5045625 2022-12-23 16:03:00,767 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4954, 1.9387, 1.4655, 2.2985, 2.3922, 1.6680, 1.5169, 1.3094], device='cuda:3'), covar=tensor([0.1967, 0.1699, 0.1626, 0.0941, 0.1275, 0.1091, 0.2135, 0.1532], device='cuda:3'), in_proj_covar=tensor([0.0246, 0.0226, 0.0214, 0.0198, 0.0260, 0.0196, 0.0224, 0.0198], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 16:03:03,977 INFO [train.py:894] (3/4) Epoch 20, batch 3400, loss[loss=0.1611, simple_loss=0.2518, pruned_loss=0.03513, over 18515.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2666, pruned_loss=0.05201, over 3713579.57 frames. ], batch size: 55, lr: 5.68e-03, grad_scale: 16.0 2022-12-23 16:03:14,441 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=70024.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:03:14,549 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70024.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:03:23,153 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=70030.0, num_to_drop=1, layers_to_drop={3} 2022-12-23 16:03:28,994 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.3898, 1.8708, 2.1729, 1.1666, 1.4206, 2.2431, 2.0457, 1.8029], device='cuda:3'), covar=tensor([0.0778, 0.0325, 0.0303, 0.0368, 0.0364, 0.0428, 0.0240, 0.0668], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0170, 0.0126, 0.0139, 0.0148, 0.0141, 0.0162, 0.0172], device='cuda:3'), out_proj_covar=tensor([1.1327e-04, 1.3011e-04, 9.5445e-05, 1.0393e-04, 1.1099e-04, 1.0839e-04, 1.2478e-04, 1.3090e-04], device='cuda:3') 2022-12-23 16:03:36,037 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.836e+02 4.387e+02 5.410e+02 6.591e+02 1.191e+03, threshold=1.082e+03, percent-clipped=3.0 2022-12-23 16:03:52,171 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70050.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:03:55,197 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3529, 1.2813, 1.4557, 0.9421, 1.3516, 1.4377, 1.1704, 1.6928], device='cuda:3'), covar=tensor([0.0920, 0.1968, 0.1178, 0.1413, 0.0780, 0.1026, 0.2514, 0.0585], device='cuda:3'), in_proj_covar=tensor([0.0198, 0.0216, 0.0210, 0.0196, 0.0177, 0.0218, 0.0219, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 16:04:15,713 INFO [train.py:894] (3/4) Epoch 20, batch 3450, loss[loss=0.1908, simple_loss=0.2758, pruned_loss=0.05286, over 18531.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2675, pruned_loss=0.05262, over 3714004.60 frames. ], batch size: 55, lr: 5.68e-03, grad_scale: 16.0 2022-12-23 16:04:22,925 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=70072.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:04:42,043 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5258, 2.8434, 3.3274, 1.1766, 2.7274, 3.5509, 2.4538, 2.7832], device='cuda:3'), covar=tensor([0.0812, 0.0370, 0.0275, 0.0530, 0.0427, 0.0447, 0.0444, 0.0695], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0169, 0.0126, 0.0139, 0.0147, 0.0141, 0.0161, 0.0171], device='cuda:3'), out_proj_covar=tensor([1.1251e-04, 1.2943e-04, 9.4938e-05, 1.0360e-04, 1.1034e-04, 1.0799e-04, 1.2397e-04, 1.3075e-04], device='cuda:3') 2022-12-23 16:05:01,364 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=70098.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:05:01,678 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4560, 1.3766, 1.3432, 1.4045, 1.7024, 1.5722, 1.5717, 1.1584], device='cuda:3'), covar=tensor([0.0323, 0.0215, 0.0466, 0.0184, 0.0181, 0.0357, 0.0253, 0.0294], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0126, 0.0151, 0.0124, 0.0115, 0.0119, 0.0097, 0.0125], device='cuda:3'), out_proj_covar=tensor([7.5162e-05, 9.9934e-05, 1.2430e-04, 9.8848e-05, 9.3380e-05, 9.1644e-05, 7.5821e-05, 9.8719e-05], device='cuda:3') 2022-12-23 16:05:08,793 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4552, 0.8827, 0.7460, 1.1023, 1.8762, 0.6595, 1.0179, 1.1717], device='cuda:3'), covar=tensor([0.2324, 0.3216, 0.2513, 0.2115, 0.2064, 0.2504, 0.2300, 0.2697], device='cuda:3'), in_proj_covar=tensor([0.0093, 0.0097, 0.0116, 0.0094, 0.0116, 0.0090, 0.0097, 0.0092], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 16:05:18,914 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70110.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:05:30,004 INFO [train.py:894] (3/4) Epoch 20, batch 3500, loss[loss=0.1761, simple_loss=0.2655, pruned_loss=0.04335, over 18606.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2676, pruned_loss=0.05276, over 3714303.27 frames. ], batch size: 98, lr: 5.68e-03, grad_scale: 16.0 2022-12-23 16:05:51,872 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp1.1 from training. Duration: 22.2954375 2022-12-23 16:06:02,624 INFO [train.py:894] (3/4) Epoch 21, batch 0, loss[loss=0.1721, simple_loss=0.2523, pruned_loss=0.046, over 18590.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2523, pruned_loss=0.046, over 18590.00 frames. ], batch size: 45, lr: 5.54e-03, grad_scale: 16.0 2022-12-23 16:06:02,625 INFO [train.py:919] (3/4) Computing validation loss 2022-12-23 16:06:13,593 INFO [train.py:928] (3/4) Epoch 21, validation: loss=0.1647, simple_loss=0.2623, pruned_loss=0.03359, over 944034.00 frames. 2022-12-23 16:06:13,594 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-23 16:06:36,621 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.766e+02 4.209e+02 5.633e+02 7.242e+02 1.131e+03, threshold=1.127e+03, percent-clipped=2.0 2022-12-23 16:07:04,062 WARNING [train.py:1060] (3/4) Exclude cut with ID 298-126791-0067-24026-0_sp0.9 from training. Duration: 21.438875 2022-12-23 16:07:09,042 WARNING [train.py:1060] (3/4) Exclude cut with ID 5652-39938-0025-23684-0_sp0.9 from training. Duration: 22.2055625 2022-12-23 16:07:26,352 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=70171.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:07:29,008 INFO [train.py:894] (3/4) Epoch 21, batch 50, loss[loss=0.2362, simple_loss=0.3099, pruned_loss=0.08122, over 18686.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2714, pruned_loss=0.04854, over 839043.95 frames. ], batch size: 176, lr: 5.53e-03, grad_scale: 16.0 2022-12-23 16:08:44,902 INFO [train.py:894] (3/4) Epoch 21, batch 100, loss[loss=0.1808, simple_loss=0.2674, pruned_loss=0.0471, over 18535.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2674, pruned_loss=0.04591, over 1476940.84 frames. ], batch size: 176, lr: 5.53e-03, grad_scale: 16.0 2022-12-23 16:09:07,227 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.079e+02 3.457e+02 4.152e+02 4.912e+02 1.134e+03, threshold=8.304e+02, percent-clipped=1.0 2022-12-23 16:09:10,582 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8316, 1.3966, 0.7668, 1.3951, 2.1983, 1.3167, 1.6362, 1.6907], device='cuda:3'), covar=tensor([0.1744, 0.2141, 0.2455, 0.1569, 0.1701, 0.1724, 0.1484, 0.1773], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0098, 0.0116, 0.0094, 0.0117, 0.0091, 0.0097, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 16:09:14,298 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5702, 1.5476, 1.5250, 1.6161, 1.2319, 3.4082, 1.4519, 2.0584], device='cuda:3'), covar=tensor([0.3188, 0.2164, 0.2051, 0.1966, 0.1404, 0.0190, 0.1667, 0.0824], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0118, 0.0126, 0.0121, 0.0104, 0.0098, 0.0092, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 16:09:40,109 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70260.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:09:59,305 INFO [train.py:894] (3/4) Epoch 21, batch 150, loss[loss=0.1707, simple_loss=0.2624, pruned_loss=0.03955, over 18717.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2651, pruned_loss=0.0444, over 1973904.55 frames. ], batch size: 52, lr: 5.53e-03, grad_scale: 16.0 2022-12-23 16:10:08,375 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0 from training. Duration: 24.525 2022-12-23 16:10:10,228 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9021, 1.9351, 2.2973, 1.2788, 2.2480, 2.3397, 1.7726, 2.7598], device='cuda:3'), covar=tensor([0.1378, 0.2131, 0.1378, 0.2171, 0.0837, 0.1379, 0.2343, 0.0543], device='cuda:3'), in_proj_covar=tensor([0.0197, 0.0215, 0.0207, 0.0195, 0.0176, 0.0216, 0.0216, 0.0192], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 16:10:41,521 WARNING [train.py:1060] (3/4) Exclude cut with ID 3699-47246-0007-3408-0_sp0.9 from training. Duration: 20.26675 2022-12-23 16:10:54,068 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp0.9 from training. Duration: 27.25 2022-12-23 16:11:13,246 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=70321.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:11:15,744 INFO [train.py:894] (3/4) Epoch 21, batch 200, loss[loss=0.1677, simple_loss=0.2448, pruned_loss=0.04535, over 18685.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2647, pruned_loss=0.04416, over 2359439.79 frames. ], batch size: 46, lr: 5.53e-03, grad_scale: 16.0 2022-12-23 16:11:25,765 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70330.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 16:11:39,250 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.182e+02 3.326e+02 4.110e+02 5.021e+02 8.767e+02, threshold=8.220e+02, percent-clipped=2.0 2022-12-23 16:11:41,640 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5610, 1.4464, 1.8604, 1.1401, 1.7739, 1.7457, 1.3497, 2.0731], device='cuda:3'), covar=tensor([0.1116, 0.2139, 0.1144, 0.1610, 0.0754, 0.1105, 0.2325, 0.0558], device='cuda:3'), in_proj_covar=tensor([0.0197, 0.0214, 0.0207, 0.0195, 0.0175, 0.0217, 0.0215, 0.0192], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 16:12:06,346 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0 from training. Duration: 21.68 2022-12-23 16:12:16,410 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0 from training. Duration: 21.6300625 2022-12-23 16:12:29,194 INFO [train.py:894] (3/4) Epoch 21, batch 250, loss[loss=0.1678, simple_loss=0.2603, pruned_loss=0.03763, over 18722.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2641, pruned_loss=0.04371, over 2659425.53 frames. ], batch size: 52, lr: 5.53e-03, grad_scale: 16.0 2022-12-23 16:12:36,716 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=70378.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 16:12:39,396 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0_sp0.9 from training. Duration: 24.033375 2022-12-23 16:13:38,002 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0 from training. Duration: 22.905 2022-12-23 16:13:39,486 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp1.1 from training. Duration: 23.4318125 2022-12-23 16:13:44,120 INFO [train.py:894] (3/4) Epoch 21, batch 300, loss[loss=0.1727, simple_loss=0.2491, pruned_loss=0.04817, over 18483.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2626, pruned_loss=0.04325, over 2893541.16 frames. ], batch size: 43, lr: 5.52e-03, grad_scale: 16.0 2022-12-23 16:14:08,537 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.163e+02 3.548e+02 4.548e+02 5.706e+02 1.035e+03, threshold=9.095e+02, percent-clipped=5.0 2022-12-23 16:14:49,710 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=70466.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:14:59,952 INFO [train.py:894] (3/4) Epoch 21, batch 350, loss[loss=0.2, simple_loss=0.297, pruned_loss=0.05146, over 18469.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2623, pruned_loss=0.043, over 3075307.08 frames. ], batch size: 54, lr: 5.52e-03, grad_scale: 16.0 2022-12-23 16:15:10,696 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.9745, 5.1577, 4.6814, 2.6309, 5.1195, 4.0909, 0.9583, 3.4088], device='cuda:3'), covar=tensor([0.1899, 0.0868, 0.1116, 0.2871, 0.0551, 0.0612, 0.4800, 0.1278], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0142, 0.0158, 0.0124, 0.0143, 0.0114, 0.0144, 0.0116], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 16:15:37,933 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0406, 2.0963, 2.3964, 1.3500, 2.3997, 2.4432, 1.8821, 2.8664], device='cuda:3'), covar=tensor([0.1127, 0.1785, 0.1205, 0.2038, 0.0704, 0.1193, 0.2063, 0.0515], device='cuda:3'), in_proj_covar=tensor([0.0196, 0.0214, 0.0206, 0.0195, 0.0176, 0.0216, 0.0215, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 16:15:40,620 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp1.1 from training. Duration: 20.82275 2022-12-23 16:15:40,671 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp0.9 from training. Duration: 25.45 2022-12-23 16:16:16,679 INFO [train.py:894] (3/4) Epoch 21, batch 400, loss[loss=0.1808, simple_loss=0.2523, pruned_loss=0.0547, over 18471.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2622, pruned_loss=0.04333, over 3217288.61 frames. ], batch size: 43, lr: 5.52e-03, grad_scale: 16.0 2022-12-23 16:16:40,730 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.333e+02 3.461e+02 4.220e+02 5.034e+02 8.271e+02, threshold=8.440e+02, percent-clipped=0.0 2022-12-23 16:16:40,749 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0 from training. Duration: 25.775 2022-12-23 16:17:02,595 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0_sp0.9 from training. Duration: 22.25 2022-12-23 16:17:29,945 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0 from training. Duration: 26.205 2022-12-23 16:17:31,366 INFO [train.py:894] (3/4) Epoch 21, batch 450, loss[loss=0.1968, simple_loss=0.2828, pruned_loss=0.05541, over 18391.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2634, pruned_loss=0.04394, over 3326929.24 frames. ], batch size: 46, lr: 5.52e-03, grad_scale: 16.0 2022-12-23 16:17:47,467 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp0.9 from training. Duration: 30.1555625 2022-12-23 16:17:52,000 WARNING [train.py:1060] (3/4) Exclude cut with ID 1265-135635-0050-6781-0_sp0.9 from training. Duration: 21.8333125 2022-12-23 16:18:01,268 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp1.1 from training. Duration: 20.6545625 2022-12-23 16:18:33,238 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2022-12-23 16:18:38,247 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=70616.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:18:45,344 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0045-39920-0_sp0.9 from training. Duration: 20.52225 2022-12-23 16:18:48,355 INFO [train.py:894] (3/4) Epoch 21, batch 500, loss[loss=0.1616, simple_loss=0.2419, pruned_loss=0.04067, over 18481.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2631, pruned_loss=0.04379, over 3413028.63 frames. ], batch size: 43, lr: 5.52e-03, grad_scale: 16.0 2022-12-23 16:18:57,048 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2022-12-23 16:19:07,567 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp0.9 from training. Duration: 29.1166875 2022-12-23 16:19:13,337 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.177e+02 3.381e+02 4.074e+02 5.237e+02 1.211e+03, threshold=8.147e+02, percent-clipped=2.0 2022-12-23 16:19:31,670 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1778, 1.3727, 1.8249, 1.8128, 2.1279, 2.1928, 1.9858, 1.8052], device='cuda:3'), covar=tensor([0.2143, 0.3294, 0.2643, 0.2797, 0.1933, 0.0926, 0.3089, 0.1290], device='cuda:3'), in_proj_covar=tensor([0.0267, 0.0297, 0.0278, 0.0317, 0.0306, 0.0251, 0.0342, 0.0241], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 16:19:58,526 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6286, 1.5557, 1.4997, 1.4698, 1.8195, 1.7798, 1.7752, 1.2937], device='cuda:3'), covar=tensor([0.0316, 0.0243, 0.0460, 0.0216, 0.0198, 0.0341, 0.0261, 0.0324], device='cuda:3'), in_proj_covar=tensor([0.0095, 0.0128, 0.0151, 0.0125, 0.0116, 0.0120, 0.0098, 0.0127], device='cuda:3'), out_proj_covar=tensor([7.5639e-05, 1.0162e-04, 1.2484e-04, 9.9706e-05, 9.3773e-05, 9.2173e-05, 7.6509e-05, 9.9993e-05], device='cuda:3') 2022-12-23 16:20:05,326 INFO [train.py:894] (3/4) Epoch 21, batch 550, loss[loss=0.1817, simple_loss=0.2701, pruned_loss=0.04663, over 18654.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2637, pruned_loss=0.04381, over 3479895.78 frames. ], batch size: 53, lr: 5.51e-03, grad_scale: 16.0 2022-12-23 16:20:05,369 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133211-0007-59831-0_sp0.9 from training. Duration: 21.388875 2022-12-23 16:20:42,369 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0 from training. Duration: 22.72 2022-12-23 16:20:43,821 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0_sp0.9 from training. Duration: 22.7444375 2022-12-23 16:20:45,465 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6334, 0.9983, 0.6569, 1.2191, 1.9516, 1.2904, 1.1897, 1.6712], device='cuda:3'), covar=tensor([0.2472, 0.3285, 0.2987, 0.2326, 0.2454, 0.2480, 0.2326, 0.2562], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0098, 0.0116, 0.0095, 0.0117, 0.0091, 0.0097, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 16:20:51,714 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2022-12-23 16:21:07,997 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.27 vs. limit=5.0 2022-12-23 16:21:20,564 INFO [train.py:894] (3/4) Epoch 21, batch 600, loss[loss=0.155, simple_loss=0.2488, pruned_loss=0.03059, over 18589.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2647, pruned_loss=0.04462, over 3531313.73 frames. ], batch size: 97, lr: 5.51e-03, grad_scale: 16.0 2022-12-23 16:21:25,176 WARNING [train.py:1060] (3/4) Exclude cut with ID 4133-6541-0027-40495-0_sp1.1 from training. Duration: 0.9681875 2022-12-23 16:21:27,907 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0_sp0.9 from training. Duration: 22.3166875 2022-12-23 16:21:33,882 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133212-0015-59917-0_sp0.9 from training. Duration: 21.8166875 2022-12-23 16:21:44,623 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.038e+02 3.325e+02 3.856e+02 4.616e+02 9.252e+02, threshold=7.711e+02, percent-clipped=1.0 2022-12-23 16:22:25,023 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70766.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:22:35,291 INFO [train.py:894] (3/4) Epoch 21, batch 650, loss[loss=0.1571, simple_loss=0.244, pruned_loss=0.03516, over 18532.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2644, pruned_loss=0.04417, over 3571927.03 frames. ], batch size: 44, lr: 5.51e-03, grad_scale: 16.0 2022-12-23 16:23:16,797 WARNING [train.py:1060] (3/4) Exclude cut with ID 4957-30119-0041-23990-0_sp0.9 from training. Duration: 20.22775 2022-12-23 16:23:37,707 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=70814.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:23:45,988 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6244, 1.6703, 1.6415, 1.5853, 1.0631, 3.4372, 1.4046, 2.0135], device='cuda:3'), covar=tensor([0.3105, 0.2006, 0.1881, 0.2015, 0.1539, 0.0187, 0.1554, 0.0799], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0118, 0.0125, 0.0121, 0.0104, 0.0097, 0.0092, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 16:23:47,175 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.1364, 5.5623, 5.0020, 2.8945, 5.7372, 4.5581, 0.7477, 3.8697], device='cuda:3'), covar=tensor([0.1842, 0.0874, 0.1266, 0.2964, 0.0391, 0.0574, 0.5576, 0.1224], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0143, 0.0158, 0.0124, 0.0143, 0.0114, 0.0144, 0.0115], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 16:23:51,359 INFO [train.py:894] (3/4) Epoch 21, batch 700, loss[loss=0.1587, simple_loss=0.2423, pruned_loss=0.03751, over 18481.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2642, pruned_loss=0.04414, over 3602697.86 frames. ], batch size: 43, lr: 5.51e-03, grad_scale: 16.0 2022-12-23 16:23:59,409 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp1.1 from training. Duration: 24.67275 2022-12-23 16:24:16,537 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.979e+02 3.538e+02 4.225e+02 5.093e+02 8.209e+02, threshold=8.450e+02, percent-clipped=1.0 2022-12-23 16:24:29,969 WARNING [train.py:1060] (3/4) Exclude cut with ID 3082-165428-0081-50734-0_sp0.9 from training. Duration: 21.8055625 2022-12-23 16:25:06,065 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0_sp0.9 from training. Duration: 22.6666875 2022-12-23 16:25:07,412 INFO [train.py:894] (3/4) Epoch 21, batch 750, loss[loss=0.1814, simple_loss=0.2726, pruned_loss=0.04507, over 18522.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2649, pruned_loss=0.04423, over 3627191.35 frames. ], batch size: 77, lr: 5.51e-03, grad_scale: 16.0 2022-12-23 16:25:26,412 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4571, 1.2326, 1.8452, 2.6803, 2.0068, 2.3319, 0.6836, 1.9134], device='cuda:3'), covar=tensor([0.1795, 0.1572, 0.1274, 0.0610, 0.1010, 0.0962, 0.2085, 0.1108], device='cuda:3'), in_proj_covar=tensor([0.0101, 0.0116, 0.0133, 0.0146, 0.0105, 0.0138, 0.0129, 0.0113], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 16:26:08,000 WARNING [train.py:1060] (3/4) Exclude cut with ID 2411-132532-0017-82279-0_sp1.1 from training. Duration: 0.9681875 2022-12-23 16:26:12,807 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70916.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:26:23,413 INFO [train.py:894] (3/4) Epoch 21, batch 800, loss[loss=0.177, simple_loss=0.2703, pruned_loss=0.04189, over 18471.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.265, pruned_loss=0.04407, over 3647240.15 frames. ], batch size: 64, lr: 5.51e-03, grad_scale: 16.0 2022-12-23 16:26:32,234 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0 from training. Duration: 22.485 2022-12-23 16:26:48,153 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.388e+02 3.554e+02 4.225e+02 4.915e+02 1.041e+03, threshold=8.450e+02, percent-clipped=3.0 2022-12-23 16:27:03,908 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2022-12-23 16:27:09,867 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp1.1 from training. Duration: 23.82275 2022-12-23 16:27:23,638 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0 from training. Duration: 20.77 2022-12-23 16:27:25,181 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=70964.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:27:32,373 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0_sp0.9 from training. Duration: 24.088875 2022-12-23 16:27:38,675 INFO [train.py:894] (3/4) Epoch 21, batch 850, loss[loss=0.1862, simple_loss=0.2807, pruned_loss=0.04583, over 18590.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2654, pruned_loss=0.04417, over 3662879.79 frames. ], batch size: 56, lr: 5.50e-03, grad_scale: 16.0 2022-12-23 16:28:01,309 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp1.1 from training. Duration: 20.4409375 2022-12-23 16:28:07,714 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70992.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:28:20,692 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5954, 1.2457, 1.3060, 1.7164, 1.4637, 3.1839, 1.3473, 1.3805], device='cuda:3'), covar=tensor([0.1023, 0.2494, 0.1207, 0.1184, 0.1941, 0.0252, 0.1878, 0.2085], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0083, 0.0073, 0.0075, 0.0092, 0.0076, 0.0085, 0.0078], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 16:28:53,477 INFO [train.py:894] (3/4) Epoch 21, batch 900, loss[loss=0.1611, simple_loss=0.2478, pruned_loss=0.03719, over 18706.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2658, pruned_loss=0.04425, over 3674517.00 frames. ], batch size: 46, lr: 5.50e-03, grad_scale: 16.0 2022-12-23 16:29:05,273 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71030.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:29:18,054 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.311e+02 3.474e+02 4.263e+02 5.002e+02 9.201e+02, threshold=8.526e+02, percent-clipped=2.0 2022-12-23 16:29:18,107 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0_sp0.9 from training. Duration: 22.511125 2022-12-23 16:29:18,132 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0 from training. Duration: 20.675 2022-12-23 16:29:38,847 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71053.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:30:09,405 INFO [train.py:894] (3/4) Epoch 21, batch 950, loss[loss=0.1685, simple_loss=0.2555, pruned_loss=0.0407, over 18704.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2668, pruned_loss=0.04444, over 3683695.37 frames. ], batch size: 50, lr: 5.50e-03, grad_scale: 16.0 2022-12-23 16:30:36,338 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71091.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:30:56,396 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp0.9 from training. Duration: 24.9833125 2022-12-23 16:31:24,683 INFO [train.py:894] (3/4) Epoch 21, batch 1000, loss[loss=0.1808, simple_loss=0.2716, pruned_loss=0.04503, over 18605.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2662, pruned_loss=0.0439, over 3690439.61 frames. ], batch size: 51, lr: 5.50e-03, grad_scale: 16.0 2022-12-23 16:31:27,949 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0 from training. Duration: 27.14 2022-12-23 16:31:45,529 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0 from training. Duration: 22.44 2022-12-23 16:31:48,411 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.909e+02 3.318e+02 3.818e+02 4.635e+02 9.263e+02, threshold=7.636e+02, percent-clipped=2.0 2022-12-23 16:32:37,335 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4753, 1.1685, 1.8503, 2.6775, 1.9702, 2.4538, 0.8618, 1.9743], device='cuda:3'), covar=tensor([0.1837, 0.1714, 0.1290, 0.0660, 0.1080, 0.0941, 0.2061, 0.1126], device='cuda:3'), in_proj_covar=tensor([0.0101, 0.0115, 0.0133, 0.0147, 0.0105, 0.0139, 0.0129, 0.0112], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 16:32:40,054 INFO [train.py:894] (3/4) Epoch 21, batch 1050, loss[loss=0.1419, simple_loss=0.2326, pruned_loss=0.02553, over 18553.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2648, pruned_loss=0.04346, over 3695556.83 frames. ], batch size: 49, lr: 5.50e-03, grad_scale: 16.0 2022-12-23 16:33:03,647 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0060-62364-0_sp0.9 from training. Duration: 21.361125 2022-12-23 16:33:09,672 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp1.1 from training. Duration: 27.0318125 2022-12-23 16:33:09,936 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2311, 2.7908, 2.6976, 1.1659, 2.9855, 2.2386, 0.7964, 1.7471], device='cuda:3'), covar=tensor([0.2085, 0.1257, 0.1767, 0.3826, 0.1038, 0.1031, 0.4305, 0.1873], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0143, 0.0157, 0.0124, 0.0143, 0.0114, 0.0143, 0.0114], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 16:33:20,155 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp0.9 from training. Duration: 28.638875 2022-12-23 16:33:34,552 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0 from training. Duration: 20.4 2022-12-23 16:33:56,815 INFO [train.py:894] (3/4) Epoch 21, batch 1100, loss[loss=0.1838, simple_loss=0.2792, pruned_loss=0.04418, over 18538.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2646, pruned_loss=0.04335, over 3700350.17 frames. ], batch size: 55, lr: 5.49e-03, grad_scale: 16.0 2022-12-23 16:34:07,190 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0 from training. Duration: 20.025 2022-12-23 16:34:07,207 WARNING [train.py:1060] (3/4) Exclude cut with ID 2364-131735-0112-64612-0_sp0.9 from training. Duration: 20.488875 2022-12-23 16:34:14,452 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0 from training. Duration: 29.735 2022-12-23 16:34:20,018 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.240e+02 3.581e+02 4.240e+02 5.301e+02 9.691e+02, threshold=8.480e+02, percent-clipped=4.0 2022-12-23 16:35:00,328 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2022-12-23 16:35:11,397 INFO [train.py:894] (3/4) Epoch 21, batch 1150, loss[loss=0.1563, simple_loss=0.254, pruned_loss=0.02937, over 18544.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2645, pruned_loss=0.04336, over 3703704.17 frames. ], batch size: 55, lr: 5.49e-03, grad_scale: 16.0 2022-12-23 16:35:27,558 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6139, 1.5790, 1.6199, 1.5325, 1.2379, 3.5930, 1.4882, 2.0969], device='cuda:3'), covar=tensor([0.3164, 0.2119, 0.1961, 0.2120, 0.1481, 0.0174, 0.1730, 0.0826], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0117, 0.0124, 0.0121, 0.0104, 0.0096, 0.0091, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 16:35:33,445 WARNING [train.py:1060] (3/4) Exclude cut with ID 7276-92427-0014-12983-0_sp0.9 from training. Duration: 21.3055625 2022-12-23 16:35:35,046 WARNING [train.py:1060] (3/4) Exclude cut with ID 1025-75365-0008-79168-0_sp0.9 from training. Duration: 22.0666875 2022-12-23 16:36:26,489 INFO [train.py:894] (3/4) Epoch 21, batch 1200, loss[loss=0.1698, simple_loss=0.2442, pruned_loss=0.04769, over 18401.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2639, pruned_loss=0.04301, over 3706409.95 frames. ], batch size: 42, lr: 5.49e-03, grad_scale: 32.0 2022-12-23 16:36:48,759 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.215e+02 3.270e+02 4.020e+02 4.826e+02 8.816e+02, threshold=8.041e+02, percent-clipped=1.0 2022-12-23 16:37:03,412 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=71348.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:37:24,164 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0 from training. Duration: 20.26 2022-12-23 16:37:38,415 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0030-9324-0_sp0.9 from training. Duration: 21.3444375 2022-12-23 16:37:41,233 INFO [train.py:894] (3/4) Epoch 21, batch 1250, loss[loss=0.1539, simple_loss=0.2423, pruned_loss=0.0328, over 18534.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2645, pruned_loss=0.04314, over 3707325.54 frames. ], batch size: 47, lr: 5.49e-03, grad_scale: 32.0 2022-12-23 16:38:00,379 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=71386.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:38:33,451 WARNING [train.py:1060] (3/4) Exclude cut with ID 497-129325-0061-62254-0_sp1.1 from training. Duration: 0.97725 2022-12-23 16:38:56,756 INFO [train.py:894] (3/4) Epoch 21, batch 1300, loss[loss=0.1775, simple_loss=0.2674, pruned_loss=0.04384, over 18599.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2641, pruned_loss=0.0429, over 3708512.61 frames. ], batch size: 51, lr: 5.49e-03, grad_scale: 32.0 2022-12-23 16:39:16,412 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2022-12-23 16:39:18,586 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0_sp0.9 from training. Duration: 22.97225 2022-12-23 16:39:19,918 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.912e+02 3.273e+02 4.091e+02 5.101e+02 9.665e+02, threshold=8.182e+02, percent-clipped=2.0 2022-12-23 16:39:48,956 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0047-39922-0_sp0.9 from training. Duration: 21.97775 2022-12-23 16:40:04,221 WARNING [train.py:1060] (3/4) Exclude cut with ID 1112-1043-0006-89194-0_sp0.9 from training. Duration: 21.8333125 2022-12-23 16:40:13,528 INFO [train.py:894] (3/4) Epoch 21, batch 1350, loss[loss=0.1932, simple_loss=0.2846, pruned_loss=0.05094, over 18447.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2644, pruned_loss=0.04329, over 3709737.05 frames. ], batch size: 64, lr: 5.48e-03, grad_scale: 32.0 2022-12-23 16:40:13,595 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0 from training. Duration: 20.47 2022-12-23 16:41:19,050 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0037-39912-0_sp0.9 from training. Duration: 20.67225 2022-12-23 16:41:27,516 INFO [train.py:894] (3/4) Epoch 21, batch 1400, loss[loss=0.187, simple_loss=0.2806, pruned_loss=0.04666, over 18577.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2637, pruned_loss=0.04289, over 3710822.96 frames. ], batch size: 57, lr: 5.48e-03, grad_scale: 32.0 2022-12-23 16:41:38,874 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp0.9 from training. Duration: 25.2444375 2022-12-23 16:41:51,595 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.246e+02 3.506e+02 4.118e+02 5.338e+02 9.015e+02, threshold=8.236e+02, percent-clipped=4.0 2022-12-23 16:42:00,605 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0021-76797-0_sp0.9 from training. Duration: 21.1445 2022-12-23 16:42:40,868 INFO [train.py:894] (3/4) Epoch 21, batch 1450, loss[loss=0.1437, simple_loss=0.228, pruned_loss=0.02967, over 18485.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.263, pruned_loss=0.04246, over 3711184.99 frames. ], batch size: 43, lr: 5.48e-03, grad_scale: 16.0 2022-12-23 16:43:15,962 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp0.9 from training. Duration: 33.038875 2022-12-23 16:43:16,173 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([5.5301, 4.9346, 5.2642, 5.2659, 4.6644, 4.6992, 5.6803, 1.7105], device='cuda:3'), covar=tensor([0.0955, 0.0999, 0.0790, 0.1690, 0.2249, 0.1978, 0.0788, 0.7439], device='cuda:3'), in_proj_covar=tensor([0.0332, 0.0218, 0.0228, 0.0258, 0.0314, 0.0258, 0.0279, 0.0276], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 16:43:46,340 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8807, 1.3045, 0.7881, 1.3894, 2.0792, 1.2023, 1.5073, 1.6293], device='cuda:3'), covar=tensor([0.1458, 0.2088, 0.2347, 0.1472, 0.1751, 0.1769, 0.1502, 0.1656], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0097, 0.0115, 0.0095, 0.0116, 0.0090, 0.0096, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 16:43:46,434 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71616.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:43:51,804 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64291-0000-16059-0_sp0.9 from training. Duration: 20.0944375 2022-12-23 16:43:56,088 INFO [train.py:894] (3/4) Epoch 21, batch 1500, loss[loss=0.1697, simple_loss=0.2445, pruned_loss=0.04741, over 18569.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2626, pruned_loss=0.0422, over 3712306.85 frames. ], batch size: 41, lr: 5.48e-03, grad_scale: 16.0 2022-12-23 16:44:07,036 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp1.1 from training. Duration: 20.4 2022-12-23 16:44:15,918 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0 from training. Duration: 20.085 2022-12-23 16:44:20,034 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1132, 2.1906, 1.7151, 2.4896, 2.1784, 2.0249, 2.6072, 2.2086], device='cuda:3'), covar=tensor([0.0849, 0.1570, 0.2340, 0.1581, 0.1637, 0.0881, 0.1046, 0.1222], device='cuda:3'), in_proj_covar=tensor([0.0177, 0.0208, 0.0251, 0.0288, 0.0235, 0.0190, 0.0206, 0.0202], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 16:44:22,491 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.435e+02 3.355e+02 3.963e+02 5.477e+02 1.143e+03, threshold=7.926e+02, percent-clipped=5.0 2022-12-23 16:44:26,761 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0_sp0.9 from training. Duration: 23.07775 2022-12-23 16:44:34,263 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=71648.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:45:11,907 INFO [train.py:894] (3/4) Epoch 21, batch 1550, loss[loss=0.1749, simple_loss=0.2733, pruned_loss=0.03823, over 18700.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2617, pruned_loss=0.04186, over 3712231.16 frames. ], batch size: 60, lr: 5.48e-03, grad_scale: 16.0 2022-12-23 16:45:14,896 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp0.9 from training. Duration: 24.9333125 2022-12-23 16:45:18,098 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71677.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:45:31,037 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=71686.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:45:45,731 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=71696.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:45:57,916 WARNING [train.py:1060] (3/4) Exclude cut with ID 5118-111612-0016-124680-0_sp0.9 from training. Duration: 20.388875 2022-12-23 16:46:04,396 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp1.1 from training. Duration: 20.3590625 2022-12-23 16:46:09,111 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([6.1188, 5.1302, 5.4176, 6.0847, 5.6687, 5.4604, 6.0819, 1.8607], device='cuda:3'), covar=tensor([0.0484, 0.0641, 0.0468, 0.0560, 0.1123, 0.0919, 0.0402, 0.5507], device='cuda:3'), in_proj_covar=tensor([0.0333, 0.0219, 0.0228, 0.0260, 0.0316, 0.0259, 0.0280, 0.0276], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 16:46:25,803 INFO [train.py:894] (3/4) Epoch 21, batch 1600, loss[loss=0.1856, simple_loss=0.2827, pruned_loss=0.04423, over 18515.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2625, pruned_loss=0.04216, over 3712676.40 frames. ], batch size: 55, lr: 5.47e-03, grad_scale: 16.0 2022-12-23 16:46:41,430 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=71734.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:46:50,331 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.261e+02 3.265e+02 3.938e+02 4.745e+02 1.111e+03, threshold=7.876e+02, percent-clipped=1.0 2022-12-23 16:47:13,738 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0_sp1.1 from training. Duration: 0.836375 2022-12-23 16:47:40,616 INFO [train.py:894] (3/4) Epoch 21, batch 1650, loss[loss=0.1982, simple_loss=0.2825, pruned_loss=0.05696, over 18429.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2637, pruned_loss=0.04359, over 3712564.18 frames. ], batch size: 48, lr: 5.47e-03, grad_scale: 16.0 2022-12-23 16:47:57,357 WARNING [train.py:1060] (3/4) Exclude cut with ID 8565-290391-0049-67394-0_sp0.9 from training. Duration: 21.3166875 2022-12-23 16:48:28,844 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0029-104863-0_sp0.9 from training. Duration: 22.1055625 2022-12-23 16:48:38,966 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp1.1 from training. Duration: 21.77725 2022-12-23 16:48:43,923 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6667, 2.6373, 2.3189, 2.1400, 3.0764, 3.0643, 2.6252, 2.2763], device='cuda:3'), covar=tensor([0.0382, 0.0351, 0.0460, 0.0556, 0.0243, 0.0248, 0.0412, 0.0674], device='cuda:3'), in_proj_covar=tensor([0.0125, 0.0129, 0.0130, 0.0120, 0.0100, 0.0123, 0.0137, 0.0159], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 16:48:56,726 INFO [train.py:894] (3/4) Epoch 21, batch 1700, loss[loss=0.1681, simple_loss=0.2476, pruned_loss=0.04427, over 18429.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2652, pruned_loss=0.04529, over 3713453.64 frames. ], batch size: 48, lr: 5.47e-03, grad_scale: 16.0 2022-12-23 16:48:59,650 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp0.9 from training. Duration: 27.8166875 2022-12-23 16:49:21,067 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp1.1 from training. Duration: 22.5090625 2022-12-23 16:49:22,521 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.287e+02 4.126e+02 5.487e+02 6.780e+02 1.150e+03, threshold=1.097e+03, percent-clipped=15.0 2022-12-23 16:49:27,559 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0 from training. Duration: 25.035 2022-12-23 16:49:47,161 WARNING [train.py:1060] (3/4) Exclude cut with ID 774-127930-0014-10412-0_sp1.1 from training. Duration: 0.95 2022-12-23 16:50:04,671 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp0.9 from training. Duration: 0.92225 2022-12-23 16:50:13,435 INFO [train.py:894] (3/4) Epoch 21, batch 1750, loss[loss=0.1607, simple_loss=0.2345, pruned_loss=0.04349, over 18415.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2653, pruned_loss=0.04649, over 3711528.04 frames. ], batch size: 42, lr: 5.47e-03, grad_scale: 16.0 2022-12-23 16:50:32,196 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0 from training. Duration: 21.97 2022-12-23 16:50:52,948 WARNING [train.py:1060] (3/4) Exclude cut with ID 7492-105653-0055-62765-0_sp0.9 from training. Duration: 21.97225 2022-12-23 16:50:54,404 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp0.9 from training. Duration: 25.3333125 2022-12-23 16:51:02,752 WARNING [train.py:1060] (3/4) Exclude cut with ID 5172-29468-0015-19128-0_sp0.9 from training. Duration: 21.5055625 2022-12-23 16:51:12,504 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2022-12-23 16:51:13,466 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp1.1 from training. Duration: 20.72725 2022-12-23 16:51:23,956 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71919.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:51:29,044 INFO [train.py:894] (3/4) Epoch 21, batch 1800, loss[loss=0.1858, simple_loss=0.2703, pruned_loss=0.05062, over 18580.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2664, pruned_loss=0.04748, over 3713104.07 frames. ], batch size: 69, lr: 5.47e-03, grad_scale: 16.0 2022-12-23 16:51:45,391 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp0.9 from training. Duration: 26.32775 2022-12-23 16:51:54,253 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.190e+02 4.570e+02 5.531e+02 6.951e+02 1.422e+03, threshold=1.106e+03, percent-clipped=2.0 2022-12-23 16:52:19,586 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0 from training. Duration: 20.025 2022-12-23 16:52:24,058 WARNING [train.py:1060] (3/4) Exclude cut with ID 6709-74022-0004-86860-0_sp1.1 from training. Duration: 0.9409375 2022-12-23 16:52:24,071 WARNING [train.py:1060] (3/4) Exclude cut with ID 4757-1811-0023-62229-0_sp0.9 from training. Duration: 21.37775 2022-12-23 16:52:41,872 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=71972.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:52:43,123 INFO [train.py:894] (3/4) Epoch 21, batch 1850, loss[loss=0.1912, simple_loss=0.2742, pruned_loss=0.05413, over 18722.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2674, pruned_loss=0.04853, over 3713844.74 frames. ], batch size: 52, lr: 5.47e-03, grad_scale: 16.0 2022-12-23 16:52:46,289 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0004-25974-0_sp0.9 from training. Duration: 21.17225 2022-12-23 16:52:46,308 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp0.9 from training. Duration: 27.511125 2022-12-23 16:52:54,587 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71980.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:53:07,911 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.8411, 2.4789, 2.0469, 0.9889, 1.9617, 2.1892, 1.8047, 2.1851], device='cuda:3'), covar=tensor([0.0553, 0.0553, 0.1254, 0.1600, 0.1301, 0.1330, 0.1524, 0.0828], device='cuda:3'), in_proj_covar=tensor([0.0169, 0.0182, 0.0204, 0.0189, 0.0206, 0.0197, 0.0211, 0.0200], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 16:53:20,247 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0 from training. Duration: 22.8 2022-12-23 16:53:23,301 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0 from training. Duration: 22.585 2022-12-23 16:53:58,052 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0001-146967-0_sp0.9 from training. Duration: 22.0166875 2022-12-23 16:54:02,864 INFO [train.py:894] (3/4) Epoch 21, batch 1900, loss[loss=0.1849, simple_loss=0.2744, pruned_loss=0.0477, over 18728.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2676, pruned_loss=0.04954, over 3714037.76 frames. ], batch size: 54, lr: 5.46e-03, grad_scale: 16.0 2022-12-23 16:54:14,601 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp1.1 from training. Duration: 24.395375 2022-12-23 16:54:19,189 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp0.9 from training. Duration: 27.47775 2022-12-23 16:54:23,902 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp0.9 from training. Duration: 24.8833125 2022-12-23 16:54:27,539 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0 from training. Duration: 23.39 2022-12-23 16:54:28,687 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.768e+02 4.620e+02 5.475e+02 6.791e+02 1.324e+03, threshold=1.095e+03, percent-clipped=3.0 2022-12-23 16:54:32,044 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp0.9 from training. Duration: 28.72225 2022-12-23 16:54:42,253 WARNING [train.py:1060] (3/4) Exclude cut with ID 585-294811-0110-133686-0_sp0.9 from training. Duration: 20.8944375 2022-12-23 16:54:57,097 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0_sp0.9 from training. Duration: 23.8444375 2022-12-23 16:55:18,378 INFO [train.py:894] (3/4) Epoch 21, batch 1950, loss[loss=0.1758, simple_loss=0.266, pruned_loss=0.04278, over 18727.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2664, pruned_loss=0.04933, over 3713893.13 frames. ], batch size: 54, lr: 5.46e-03, grad_scale: 16.0 2022-12-23 16:55:22,725 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0 from training. Duration: 25.85 2022-12-23 16:55:23,972 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0 from training. Duration: 21.39 2022-12-23 16:55:34,504 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0 from training. Duration: 27.92 2022-12-23 16:56:01,768 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0039-130165-0_sp0.9 from training. Duration: 20.661125 2022-12-23 16:56:24,262 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0043-15874-0_sp0.9 from training. Duration: 20.07225 2022-12-23 16:56:31,912 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0 from training. Duration: 21.01 2022-12-23 16:56:33,407 INFO [train.py:894] (3/4) Epoch 21, batch 2000, loss[loss=0.1545, simple_loss=0.2316, pruned_loss=0.03869, over 18555.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2663, pruned_loss=0.0502, over 3713636.25 frames. ], batch size: 44, lr: 5.46e-03, grad_scale: 16.0 2022-12-23 16:56:42,990 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72129.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:57:01,405 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.763e+02 4.237e+02 5.051e+02 6.066e+02 1.490e+03, threshold=1.010e+03, percent-clipped=3.0 2022-12-23 16:57:04,906 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6111, 2.3738, 2.0855, 1.3307, 2.9737, 2.7217, 2.4422, 1.7575], device='cuda:3'), covar=tensor([0.0376, 0.0461, 0.0500, 0.0790, 0.0245, 0.0352, 0.0471, 0.1001], device='cuda:3'), in_proj_covar=tensor([0.0124, 0.0127, 0.0129, 0.0119, 0.0099, 0.0123, 0.0135, 0.0157], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 16:57:38,855 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0 from training. Duration: 20.65 2022-12-23 16:57:47,628 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0 from training. Duration: 21.46 2022-12-23 16:57:50,428 INFO [train.py:894] (3/4) Epoch 21, batch 2050, loss[loss=0.1827, simple_loss=0.2685, pruned_loss=0.04851, over 18675.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2674, pruned_loss=0.05163, over 3714347.13 frames. ], batch size: 62, lr: 5.46e-03, grad_scale: 16.0 2022-12-23 16:58:09,481 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5033, 2.5279, 3.2329, 0.9053, 2.7201, 3.3958, 2.5356, 2.5791], device='cuda:3'), covar=tensor([0.0794, 0.0481, 0.0278, 0.0598, 0.0383, 0.0496, 0.0396, 0.0765], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0170, 0.0128, 0.0140, 0.0148, 0.0142, 0.0164, 0.0173], device='cuda:3'), out_proj_covar=tensor([1.1300e-04, 1.3026e-04, 9.6539e-05, 1.0425e-04, 1.1047e-04, 1.0861e-04, 1.2604e-04, 1.3217e-04], device='cuda:3') 2022-12-23 16:58:16,416 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=72190.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 16:58:33,045 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0 from training. Duration: 0.92 2022-12-23 16:58:39,078 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0_sp0.9 from training. Duration: 23.7666875 2022-12-23 16:59:00,837 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5227, 1.6734, 1.9000, 1.0643, 1.7874, 1.8253, 1.3873, 2.1753], device='cuda:3'), covar=tensor([0.1149, 0.1841, 0.1097, 0.1621, 0.0741, 0.1018, 0.2458, 0.0506], device='cuda:3'), in_proj_covar=tensor([0.0197, 0.0214, 0.0207, 0.0193, 0.0176, 0.0217, 0.0215, 0.0192], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 16:59:04,646 INFO [train.py:894] (3/4) Epoch 21, batch 2100, loss[loss=0.1931, simple_loss=0.2803, pruned_loss=0.0529, over 18643.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2676, pruned_loss=0.05191, over 3714759.30 frames. ], batch size: 53, lr: 5.46e-03, grad_scale: 16.0 2022-12-23 16:59:16,123 WARNING [train.py:1060] (3/4) Exclude cut with ID 8544-281189-0060-101339-0_sp0.9 from training. Duration: 20.861125 2022-12-23 16:59:26,204 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0_sp0.9 from training. Duration: 22.711125 2022-12-23 16:59:30,368 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.195e+02 4.359e+02 5.318e+02 6.615e+02 1.286e+03, threshold=1.064e+03, percent-clipped=4.0 2022-12-23 16:59:45,708 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.01 vs. limit=5.0 2022-12-23 16:59:54,592 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2022-12-23 17:00:07,784 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp1.1 from training. Duration: 22.986375 2022-12-23 17:00:11,423 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.58 vs. limit=2.0 2022-12-23 17:00:18,321 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72272.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:00:19,409 INFO [train.py:894] (3/4) Epoch 21, batch 2150, loss[loss=0.1994, simple_loss=0.2695, pruned_loss=0.06469, over 18524.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2671, pruned_loss=0.05178, over 3714327.96 frames. ], batch size: 47, lr: 5.45e-03, grad_scale: 16.0 2022-12-23 17:00:23,102 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72275.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:00:24,580 WARNING [train.py:1060] (3/4) Exclude cut with ID 8040-260924-0003-80960-0_sp0.9 from training. Duration: 22.07225 2022-12-23 17:00:30,578 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0045-26330-0_sp0.9 from training. Duration: 20.3055625 2022-12-23 17:00:32,725 WARNING [train.py:1060] (3/4) Exclude cut with ID 6356-271890-0060-94317-0_sp0.9 from training. Duration: 20.72225 2022-12-23 17:00:50,025 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp1.1 from training. Duration: 22.4818125 2022-12-23 17:01:14,397 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp0.9 from training. Duration: 25.0944375 2022-12-23 17:01:17,545 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0 from training. Duration: 21.515 2022-12-23 17:01:22,595 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp0.9 from training. Duration: 27.02225 2022-12-23 17:01:27,250 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0010-62480-0_sp0.9 from training. Duration: 22.22225 2022-12-23 17:01:33,674 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=72320.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:01:37,998 INFO [train.py:894] (3/4) Epoch 21, batch 2200, loss[loss=0.1589, simple_loss=0.2372, pruned_loss=0.04027, over 18426.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2671, pruned_loss=0.05189, over 3714172.16 frames. ], batch size: 42, lr: 5.45e-03, grad_scale: 16.0 2022-12-23 17:01:38,007 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0085-44554-0_sp0.9 from training. Duration: 20.85 2022-12-23 17:01:43,174 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72326.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:02:04,134 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.723e+02 4.275e+02 5.311e+02 6.247e+02 1.700e+03, threshold=1.062e+03, percent-clipped=3.0 2022-12-23 17:02:10,095 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0 from training. Duration: 21.54 2022-12-23 17:02:13,261 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp1.1 from training. Duration: 20.5318125 2022-12-23 17:02:22,026 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0012-134311-0_sp0.9 from training. Duration: 21.9333125 2022-12-23 17:02:33,484 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.12 vs. limit=5.0 2022-12-23 17:02:54,653 INFO [train.py:894] (3/4) Epoch 21, batch 2250, loss[loss=0.1601, simple_loss=0.2527, pruned_loss=0.03374, over 18416.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2665, pruned_loss=0.05169, over 3713866.35 frames. ], batch size: 48, lr: 5.45e-03, grad_scale: 16.0 2022-12-23 17:03:13,016 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0025-130151-0_sp0.9 from training. Duration: 21.7944375 2022-12-23 17:03:16,195 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=72387.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:03:26,961 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0_sp0.9 from training. Duration: 22.4666875 2022-12-23 17:03:34,304 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0 from training. Duration: 21.635 2022-12-23 17:03:40,413 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0_sp0.9 from training. Duration: 24.038875 2022-12-23 17:03:50,541 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2022-12-23 17:03:53,244 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2521, 2.8513, 2.7984, 1.1548, 2.8526, 2.0545, 0.4391, 1.8073], device='cuda:3'), covar=tensor([0.2059, 0.1509, 0.1607, 0.3637, 0.1297, 0.1275, 0.4657, 0.1685], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0145, 0.0160, 0.0125, 0.0147, 0.0115, 0.0144, 0.0114], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 17:04:10,560 INFO [train.py:894] (3/4) Epoch 21, batch 2300, loss[loss=0.2129, simple_loss=0.2877, pruned_loss=0.06908, over 18559.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2658, pruned_loss=0.05144, over 3712873.42 frames. ], batch size: 188, lr: 5.45e-03, grad_scale: 16.0 2022-12-23 17:04:23,724 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp1.1 from training. Duration: 21.786375 2022-12-23 17:04:35,980 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.221e+02 4.085e+02 4.933e+02 6.235e+02 1.106e+03, threshold=9.865e+02, percent-clipped=2.0 2022-12-23 17:04:36,047 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0 from training. Duration: 20.22 2022-12-23 17:05:25,823 INFO [train.py:894] (3/4) Epoch 21, batch 2350, loss[loss=0.2075, simple_loss=0.284, pruned_loss=0.06552, over 18604.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2664, pruned_loss=0.05176, over 3712825.36 frames. ], batch size: 182, lr: 5.45e-03, grad_scale: 16.0 2022-12-23 17:05:43,995 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72485.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:05:44,261 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8655, 1.7845, 1.5454, 1.5375, 1.8619, 2.1067, 2.1174, 1.3812], device='cuda:3'), covar=tensor([0.0283, 0.0300, 0.0430, 0.0246, 0.0204, 0.0344, 0.0242, 0.0329], device='cuda:3'), in_proj_covar=tensor([0.0092, 0.0124, 0.0147, 0.0121, 0.0114, 0.0116, 0.0094, 0.0123], device='cuda:3'), out_proj_covar=tensor([7.3442e-05, 9.8337e-05, 1.2134e-04, 9.6427e-05, 9.1914e-05, 8.9378e-05, 7.3671e-05, 9.7356e-05], device='cuda:3') 2022-12-23 17:06:35,843 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0 from training. Duration: 25.285 2022-12-23 17:06:41,685 INFO [train.py:894] (3/4) Epoch 21, batch 2400, loss[loss=0.1925, simple_loss=0.2749, pruned_loss=0.05508, over 18509.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2677, pruned_loss=0.05223, over 3712584.88 frames. ], batch size: 52, lr: 5.44e-03, grad_scale: 8.0 2022-12-23 17:07:03,447 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1667, 2.1338, 1.6284, 1.6609, 2.2055, 2.6571, 2.4770, 1.9081], device='cuda:3'), covar=tensor([0.0319, 0.0339, 0.0448, 0.0300, 0.0259, 0.0342, 0.0283, 0.0313], device='cuda:3'), in_proj_covar=tensor([0.0092, 0.0124, 0.0147, 0.0121, 0.0114, 0.0116, 0.0094, 0.0124], device='cuda:3'), out_proj_covar=tensor([7.3712e-05, 9.8699e-05, 1.2173e-04, 9.6368e-05, 9.2018e-05, 8.9502e-05, 7.3715e-05, 9.7676e-05], device='cuda:3') 2022-12-23 17:07:08,714 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.768e+02 4.305e+02 5.500e+02 6.750e+02 1.440e+03, threshold=1.100e+03, percent-clipped=3.0 2022-12-23 17:07:43,865 WARNING [train.py:1060] (3/4) Exclude cut with ID 811-130148-0001-63453-0_sp0.9 from training. Duration: 20.861125 2022-12-23 17:07:50,560 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.83 vs. limit=5.0 2022-12-23 17:07:58,613 INFO [train.py:894] (3/4) Epoch 21, batch 2450, loss[loss=0.2036, simple_loss=0.2877, pruned_loss=0.05974, over 18590.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2674, pruned_loss=0.05212, over 3713200.18 frames. ], batch size: 57, lr: 5.44e-03, grad_scale: 8.0 2022-12-23 17:08:02,412 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72575.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:08:05,006 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0 from training. Duration: 20.88 2022-12-23 17:08:36,201 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0_sp0.9 from training. Duration: 23.4166875 2022-12-23 17:09:15,644 INFO [train.py:894] (3/4) Epoch 21, batch 2500, loss[loss=0.198, simple_loss=0.2717, pruned_loss=0.06214, over 18563.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2677, pruned_loss=0.05235, over 3714021.82 frames. ], batch size: 49, lr: 5.44e-03, grad_scale: 8.0 2022-12-23 17:09:15,811 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=72623.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:09:42,410 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.111e+02 4.322e+02 5.104e+02 6.362e+02 1.624e+03, threshold=1.021e+03, percent-clipped=3.0 2022-12-23 17:09:53,814 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0 from training. Duration: 21.24 2022-12-23 17:09:53,830 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0_sp0.9 from training. Duration: 23.9055625 2022-12-23 17:10:27,879 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.7883, 1.5858, 1.6569, 1.0662, 1.3322, 1.6992, 1.6648, 1.4326], device='cuda:3'), covar=tensor([0.0729, 0.0342, 0.0277, 0.0380, 0.0327, 0.0433, 0.0232, 0.0637], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0172, 0.0127, 0.0141, 0.0147, 0.0143, 0.0164, 0.0174], device='cuda:3'), out_proj_covar=tensor([1.1345e-04, 1.3128e-04, 9.5576e-05, 1.0487e-04, 1.0953e-04, 1.0893e-04, 1.2596e-04, 1.3220e-04], device='cuda:3') 2022-12-23 17:10:29,145 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp0.9 from training. Duration: 25.988875 2022-12-23 17:10:30,274 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2022-12-23 17:10:32,088 INFO [train.py:894] (3/4) Epoch 21, batch 2550, loss[loss=0.1454, simple_loss=0.2305, pruned_loss=0.0301, over 18693.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2673, pruned_loss=0.05205, over 3714896.83 frames. ], batch size: 46, lr: 5.44e-03, grad_scale: 8.0 2022-12-23 17:10:37,591 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0001-134300-0_sp0.9 from training. Duration: 20.67225 2022-12-23 17:10:45,082 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72682.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:11:19,269 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.2747, 1.2860, 1.5132, 0.9381, 1.4514, 1.4054, 1.1828, 1.6764], device='cuda:3'), covar=tensor([0.1028, 0.1898, 0.1029, 0.1337, 0.0697, 0.1022, 0.2356, 0.0520], device='cuda:3'), in_proj_covar=tensor([0.0198, 0.0212, 0.0206, 0.0192, 0.0176, 0.0217, 0.0214, 0.0191], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 17:11:26,823 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0 from training. Duration: 20.34 2022-12-23 17:11:48,023 INFO [train.py:894] (3/4) Epoch 21, batch 2600, loss[loss=0.1703, simple_loss=0.2429, pruned_loss=0.04881, over 18465.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.268, pruned_loss=0.05208, over 3714918.46 frames. ], batch size: 41, lr: 5.44e-03, grad_scale: 8.0 2022-12-23 17:12:16,331 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.887e+02 4.261e+02 5.084e+02 5.993e+02 1.354e+03, threshold=1.017e+03, percent-clipped=3.0 2022-12-23 17:12:36,311 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp0.9 from training. Duration: 25.061125 2022-12-23 17:12:48,907 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0 from training. Duration: 0.83 2022-12-23 17:12:51,305 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2022-12-23 17:13:04,940 INFO [train.py:894] (3/4) Epoch 21, batch 2650, loss[loss=0.1692, simple_loss=0.2585, pruned_loss=0.03996, over 18723.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2675, pruned_loss=0.05176, over 3714961.68 frames. ], batch size: 52, lr: 5.44e-03, grad_scale: 8.0 2022-12-23 17:13:13,739 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0 from training. Duration: 24.73 2022-12-23 17:13:18,890 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3749, 1.0083, 1.6108, 2.6269, 1.8627, 2.1830, 0.7921, 1.9032], device='cuda:3'), covar=tensor([0.1900, 0.1834, 0.1436, 0.0769, 0.1139, 0.1088, 0.2094, 0.1158], device='cuda:3'), in_proj_covar=tensor([0.0102, 0.0117, 0.0134, 0.0148, 0.0106, 0.0140, 0.0129, 0.0112], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 17:13:23,543 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72785.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:13:28,279 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0 from training. Duration: 23.965 2022-12-23 17:13:35,771 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0030-146996-0_sp0.9 from training. Duration: 22.088875 2022-12-23 17:13:51,140 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0_sp0.9 from training. Duration: 23.6 2022-12-23 17:14:20,780 INFO [train.py:894] (3/4) Epoch 21, batch 2700, loss[loss=0.1664, simple_loss=0.2442, pruned_loss=0.04429, over 18692.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2674, pruned_loss=0.05181, over 3715187.77 frames. ], batch size: 46, lr: 5.43e-03, grad_scale: 8.0 2022-12-23 17:14:37,559 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=72833.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:14:49,643 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.034e+02 4.615e+02 5.963e+02 7.996e+02 1.786e+03, threshold=1.193e+03, percent-clipped=8.0 2022-12-23 17:15:35,335 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0 from training. Duration: 23.795 2022-12-23 17:15:38,028 INFO [train.py:894] (3/4) Epoch 21, batch 2750, loss[loss=0.2092, simple_loss=0.2924, pruned_loss=0.06302, over 18711.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2665, pruned_loss=0.05198, over 3715453.07 frames. ], batch size: 60, lr: 5.43e-03, grad_scale: 8.0 2022-12-23 17:15:51,699 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp1.1 from training. Duration: 21.5409375 2022-12-23 17:15:55,270 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp0.9 from training. Duration: 24.97775 2022-12-23 17:16:06,863 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0_sp0.9 from training. Duration: 23.3444375 2022-12-23 17:16:10,745 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72894.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:16:20,489 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6775, 3.6769, 3.6126, 1.4625, 3.7564, 2.7168, 0.6929, 2.4806], device='cuda:3'), covar=tensor([0.1896, 0.1328, 0.1395, 0.3789, 0.1063, 0.1051, 0.5006, 0.1543], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0145, 0.0161, 0.0125, 0.0147, 0.0115, 0.0146, 0.0115], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 17:16:35,158 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0_sp0.9 from training. Duration: 23.2 2022-12-23 17:16:39,743 WARNING [train.py:1060] (3/4) Exclude cut with ID 5653-46179-0060-117930-0_sp0.9 from training. Duration: 21.17225 2022-12-23 17:16:53,401 INFO [train.py:894] (3/4) Epoch 21, batch 2800, loss[loss=0.1875, simple_loss=0.2515, pruned_loss=0.06171, over 18578.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2655, pruned_loss=0.05137, over 3715096.66 frames. ], batch size: 41, lr: 5.43e-03, grad_scale: 8.0 2022-12-23 17:16:57,045 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.3758, 2.4346, 1.9976, 3.1068, 2.2598, 2.3853, 2.4018, 3.4101], device='cuda:3'), covar=tensor([0.1711, 0.3367, 0.1953, 0.2858, 0.3836, 0.1060, 0.3432, 0.0789], device='cuda:3'), in_proj_covar=tensor([0.0294, 0.0291, 0.0246, 0.0350, 0.0273, 0.0228, 0.0289, 0.0215], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 17:17:04,256 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp0.9 from training. Duration: 24.6555625 2022-12-23 17:17:21,263 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.700e+02 4.162e+02 4.894e+02 6.036e+02 1.354e+03, threshold=9.787e+02, percent-clipped=2.0 2022-12-23 17:17:41,941 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=72955.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:17:55,158 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72964.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 17:17:59,460 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0 from training. Duration: 20.44 2022-12-23 17:18:05,783 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5803, 2.1780, 1.7410, 2.3073, 1.9376, 2.0805, 1.9657, 2.4204], device='cuda:3'), covar=tensor([0.1957, 0.2881, 0.2041, 0.2449, 0.3352, 0.1095, 0.2861, 0.0952], device='cuda:3'), in_proj_covar=tensor([0.0294, 0.0291, 0.0246, 0.0350, 0.0272, 0.0228, 0.0289, 0.0215], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 17:18:07,925 INFO [train.py:894] (3/4) Epoch 21, batch 2850, loss[loss=0.1538, simple_loss=0.2429, pruned_loss=0.03235, over 18543.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2649, pruned_loss=0.0512, over 3715154.26 frames. ], batch size: 49, lr: 5.43e-03, grad_scale: 8.0 2022-12-23 17:18:14,418 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0_sp0.9 from training. Duration: 23.45 2022-12-23 17:18:23,028 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72982.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:18:43,806 WARNING [train.py:1060] (3/4) Exclude cut with ID 6945-60535-0076-12784-0_sp0.9 from training. Duration: 20.52225 2022-12-23 17:18:51,589 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0 from training. Duration: 22.19 2022-12-23 17:19:01,913 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp1.1 from training. Duration: 25.3818125 2022-12-23 17:19:17,396 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp0.9 from training. Duration: 28.0944375 2022-12-23 17:19:23,492 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0_sp0.9 from training. Duration: 22.9444375 2022-12-23 17:19:24,836 INFO [train.py:894] (3/4) Epoch 21, batch 2900, loss[loss=0.2065, simple_loss=0.2825, pruned_loss=0.06529, over 18526.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2643, pruned_loss=0.05101, over 3714736.92 frames. ], batch size: 58, lr: 5.43e-03, grad_scale: 8.0 2022-12-23 17:19:29,416 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73025.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 17:19:33,520 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp1.1 from training. Duration: 21.6318125 2022-12-23 17:19:36,586 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=73030.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:19:50,195 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0 from training. Duration: 23.695 2022-12-23 17:19:53,305 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.702e+02 4.963e+02 5.828e+02 7.417e+02 1.430e+03, threshold=1.166e+03, percent-clipped=7.0 2022-12-23 17:20:14,006 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0001, 1.8633, 1.5840, 1.6564, 1.7321, 1.8959, 1.7103, 1.7787], device='cuda:3'), covar=tensor([0.2306, 0.3165, 0.2052, 0.2630, 0.3364, 0.1146, 0.2977, 0.1125], device='cuda:3'), in_proj_covar=tensor([0.0293, 0.0290, 0.0245, 0.0348, 0.0271, 0.0227, 0.0288, 0.0214], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 17:20:14,992 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0 from training. Duration: 23.955 2022-12-23 17:20:41,617 INFO [train.py:894] (3/4) Epoch 21, batch 2950, loss[loss=0.1894, simple_loss=0.275, pruned_loss=0.05193, over 18666.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2637, pruned_loss=0.05045, over 3713556.60 frames. ], batch size: 60, lr: 5.42e-03, grad_scale: 8.0 2022-12-23 17:20:50,685 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp0.9 from training. Duration: 26.438875 2022-12-23 17:21:24,610 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2022-12-23 17:21:34,830 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0021-26306-0_sp0.9 from training. Duration: 21.2444375 2022-12-23 17:21:34,864 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp0.9 from training. Duration: 31.02225 2022-12-23 17:21:44,515 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0 from training. Duration: 22.395 2022-12-23 17:21:57,239 INFO [train.py:894] (3/4) Epoch 21, batch 3000, loss[loss=0.1865, simple_loss=0.2778, pruned_loss=0.04758, over 18480.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2649, pruned_loss=0.05117, over 3714463.97 frames. ], batch size: 64, lr: 5.42e-03, grad_scale: 8.0 2022-12-23 17:21:57,240 INFO [train.py:919] (3/4) Computing validation loss 2022-12-23 17:22:08,277 INFO [train.py:928] (3/4) Epoch 21, validation: loss=0.1664, simple_loss=0.2634, pruned_loss=0.03471, over 944034.00 frames. 2022-12-23 17:22:08,278 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-23 17:22:12,871 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0 from training. Duration: 21.075 2022-12-23 17:22:19,038 WARNING [train.py:1060] (3/4) Exclude cut with ID 6482-98857-0025-147532-0_sp0.9 from training. Duration: 20.0055625 2022-12-23 17:22:19,047 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0037-132304-0_sp0.9 from training. Duration: 22.05 2022-12-23 17:22:19,062 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0 from training. Duration: 26.8349375 2022-12-23 17:22:23,758 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp1.1 from training. Duration: 22.1090625 2022-12-23 17:22:29,490 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp0.9 from training. Duration: 26.6166875 2022-12-23 17:22:35,635 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.005e+02 4.313e+02 5.237e+02 6.690e+02 1.463e+03, threshold=1.047e+03, percent-clipped=1.0 2022-12-23 17:22:45,991 WARNING [train.py:1060] (3/4) Exclude cut with ID 2046-178027-0000-53705-0_sp0.9 from training. Duration: 20.3055625 2022-12-23 17:23:10,790 WARNING [train.py:1060] (3/4) Exclude cut with ID 7205-50138-0008-5373-0_sp0.9 from training. Duration: 20.7 2022-12-23 17:23:24,046 INFO [train.py:894] (3/4) Epoch 21, batch 3050, loss[loss=0.1704, simple_loss=0.2413, pruned_loss=0.04971, over 18510.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2653, pruned_loss=0.05102, over 3714292.15 frames. ], batch size: 44, lr: 5.42e-03, grad_scale: 8.0 2022-12-23 17:23:47,199 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.8416, 3.3226, 3.3017, 3.7763, 3.4542, 3.3542, 3.9945, 1.2163], device='cuda:3'), covar=tensor([0.0841, 0.0801, 0.0763, 0.0872, 0.1574, 0.1342, 0.0682, 0.5009], device='cuda:3'), in_proj_covar=tensor([0.0343, 0.0224, 0.0234, 0.0269, 0.0325, 0.0269, 0.0288, 0.0283], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 17:23:51,631 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0 from training. Duration: 22.48 2022-12-23 17:24:07,593 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp0.9 from training. Duration: 29.816625 2022-12-23 17:24:15,982 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.8091, 2.5224, 1.9931, 1.0292, 1.9705, 2.3336, 1.9970, 2.2378], device='cuda:3'), covar=tensor([0.0696, 0.0584, 0.1478, 0.1730, 0.1388, 0.1318, 0.1533, 0.0893], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0188, 0.0209, 0.0194, 0.0213, 0.0204, 0.0218, 0.0204], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 17:24:27,421 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp1.1 from training. Duration: 22.7590625 2022-12-23 17:24:33,303 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0 from training. Duration: 22.555 2022-12-23 17:24:41,061 INFO [train.py:894] (3/4) Epoch 21, batch 3100, loss[loss=0.1577, simple_loss=0.2422, pruned_loss=0.03657, over 18388.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2651, pruned_loss=0.05095, over 3714665.30 frames. ], batch size: 46, lr: 5.42e-03, grad_scale: 8.0 2022-12-23 17:24:53,231 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0005-25975-0_sp0.9 from training. Duration: 21.688875 2022-12-23 17:25:08,304 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.440e+02 3.995e+02 5.110e+02 7.002e+02 1.344e+03, threshold=1.022e+03, percent-clipped=3.0 2022-12-23 17:25:22,316 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73250.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:25:26,725 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0_sp0.9 from training. Duration: 22.6 2022-12-23 17:25:55,602 INFO [train.py:894] (3/4) Epoch 21, batch 3150, loss[loss=0.1868, simple_loss=0.2751, pruned_loss=0.04927, over 18672.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2652, pruned_loss=0.05094, over 3713683.20 frames. ], batch size: 60, lr: 5.42e-03, grad_scale: 8.0 2022-12-23 17:26:00,730 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2022-12-23 17:26:07,974 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0 from training. Duration: 24.32 2022-12-23 17:26:27,011 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73293.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:27:04,346 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-276745-0093-13116-0_sp0.9 from training. Duration: 21.061125 2022-12-23 17:27:07,388 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73320.0, num_to_drop=1, layers_to_drop={3} 2022-12-23 17:27:11,714 INFO [train.py:894] (3/4) Epoch 21, batch 3200, loss[loss=0.1876, simple_loss=0.2753, pruned_loss=0.04996, over 18727.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2649, pruned_loss=0.05045, over 3714100.85 frames. ], batch size: 54, lr: 5.41e-03, grad_scale: 8.0 2022-12-23 17:27:17,799 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0024-15855-0_sp0.9 from training. Duration: 20.32225 2022-12-23 17:27:30,192 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp1.1 from training. Duration: 0.7545625 2022-12-23 17:27:39,054 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.630e+02 4.267e+02 5.062e+02 6.321e+02 9.305e+02, threshold=1.012e+03, percent-clipped=0.0 2022-12-23 17:27:41,356 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2022-12-23 17:27:44,965 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0_sp0.9 from training. Duration: 23.9333125 2022-12-23 17:27:59,052 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73354.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:28:19,123 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp1.1 from training. Duration: 20.17275 2022-12-23 17:28:24,006 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp1.1 from training. Duration: 20.436375 2022-12-23 17:28:27,515 INFO [train.py:894] (3/4) Epoch 21, batch 3250, loss[loss=0.1749, simple_loss=0.2517, pruned_loss=0.04909, over 18698.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2644, pruned_loss=0.05017, over 3715001.99 frames. ], batch size: 50, lr: 5.41e-03, grad_scale: 8.0 2022-12-23 17:29:42,409 INFO [train.py:894] (3/4) Epoch 21, batch 3300, loss[loss=0.1848, simple_loss=0.2699, pruned_loss=0.04982, over 18643.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2647, pruned_loss=0.04998, over 3715413.02 frames. ], batch size: 98, lr: 5.41e-03, grad_scale: 8.0 2022-12-23 17:29:42,442 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0_sp0.9 from training. Duration: 23.1055625 2022-12-23 17:29:44,099 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp1.1 from training. Duration: 23.5 2022-12-23 17:29:56,994 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp0.9 from training. Duration: 26.62775 2022-12-23 17:30:10,520 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.729e+02 4.288e+02 5.181e+02 6.375e+02 1.469e+03, threshold=1.036e+03, percent-clipped=5.0 2022-12-23 17:30:10,588 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0 from training. Duration: 21.105 2022-12-23 17:30:14,525 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0_sp0.9 from training. Duration: 24.411125 2022-12-23 17:30:38,691 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp1.1 from training. Duration: 21.263625 2022-12-23 17:30:58,758 INFO [train.py:894] (3/4) Epoch 21, batch 3350, loss[loss=0.1867, simple_loss=0.2736, pruned_loss=0.04993, over 18456.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2641, pruned_loss=0.04949, over 3714262.47 frames. ], batch size: 54, lr: 5.41e-03, grad_scale: 8.0 2022-12-23 17:31:00,333 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2022-12-23 17:31:13,965 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0 from training. Duration: 20.795 2022-12-23 17:31:22,526 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0 from training. Duration: 24.76 2022-12-23 17:31:22,549 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0_sp0.9 from training. Duration: 22.25 2022-12-23 17:31:47,985 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp1.1 from training. Duration: 20.5045625 2022-12-23 17:32:17,087 INFO [train.py:894] (3/4) Epoch 21, batch 3400, loss[loss=0.2033, simple_loss=0.2803, pruned_loss=0.0632, over 18633.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2655, pruned_loss=0.05054, over 3714366.94 frames. ], batch size: 53, lr: 5.41e-03, grad_scale: 8.0 2022-12-23 17:32:43,249 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.477e+02 4.102e+02 5.173e+02 6.286e+02 1.561e+03, threshold=1.035e+03, percent-clipped=1.0 2022-12-23 17:32:56,718 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73550.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:33:29,502 INFO [train.py:894] (3/4) Epoch 21, batch 3450, loss[loss=0.1879, simple_loss=0.2733, pruned_loss=0.05129, over 18630.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2659, pruned_loss=0.0508, over 3714712.37 frames. ], batch size: 53, lr: 5.41e-03, grad_scale: 8.0 2022-12-23 17:33:53,940 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73590.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:34:05,256 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=73598.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:34:14,393 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9997, 1.7209, 2.2090, 1.2440, 2.1677, 2.0667, 1.4287, 2.5007], device='cuda:3'), covar=tensor([0.1224, 0.1981, 0.1198, 0.2042, 0.0789, 0.1290, 0.2493, 0.0579], device='cuda:3'), in_proj_covar=tensor([0.0197, 0.0211, 0.0207, 0.0194, 0.0176, 0.0218, 0.0215, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 17:34:37,728 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73620.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 17:34:42,262 INFO [train.py:894] (3/4) Epoch 21, batch 3500, loss[loss=0.2087, simple_loss=0.2875, pruned_loss=0.06491, over 18662.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.266, pruned_loss=0.0507, over 3715261.44 frames. ], batch size: 177, lr: 5.40e-03, grad_scale: 8.0 2022-12-23 17:35:05,199 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp1.1 from training. Duration: 22.2954375 2022-12-23 17:35:14,809 INFO [train.py:894] (3/4) Epoch 22, batch 0, loss[loss=0.1807, simple_loss=0.2747, pruned_loss=0.04337, over 18619.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2747, pruned_loss=0.04337, over 18619.00 frames. ], batch size: 53, lr: 5.28e-03, grad_scale: 8.0 2022-12-23 17:35:14,809 INFO [train.py:919] (3/4) Computing validation loss 2022-12-23 17:35:26,081 INFO [train.py:928] (3/4) Epoch 22, validation: loss=0.1635, simple_loss=0.2611, pruned_loss=0.03298, over 944034.00 frames. 2022-12-23 17:35:26,082 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-23 17:35:44,484 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.826e+02 4.412e+02 5.413e+02 6.975e+02 1.595e+03, threshold=1.083e+03, percent-clipped=3.0 2022-12-23 17:35:57,163 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73649.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:36:00,367 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73651.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:36:18,902 WARNING [train.py:1060] (3/4) Exclude cut with ID 298-126791-0067-24026-0_sp0.9 from training. Duration: 21.438875 2022-12-23 17:36:23,429 WARNING [train.py:1060] (3/4) Exclude cut with ID 5652-39938-0025-23684-0_sp0.9 from training. Duration: 22.2055625 2022-12-23 17:36:26,664 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=73668.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 17:36:43,283 INFO [train.py:894] (3/4) Epoch 22, batch 50, loss[loss=0.1509, simple_loss=0.2342, pruned_loss=0.03383, over 18587.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2627, pruned_loss=0.04368, over 837973.82 frames. ], batch size: 45, lr: 5.28e-03, grad_scale: 8.0 2022-12-23 17:37:33,408 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0162, 1.3969, 2.3314, 4.3991, 3.1377, 2.8447, 1.1218, 3.1511], device='cuda:3'), covar=tensor([0.1802, 0.1716, 0.1547, 0.0423, 0.0906, 0.1129, 0.2000, 0.0856], device='cuda:3'), in_proj_covar=tensor([0.0102, 0.0117, 0.0135, 0.0148, 0.0106, 0.0140, 0.0129, 0.0112], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 17:37:59,109 INFO [train.py:894] (3/4) Epoch 22, batch 100, loss[loss=0.1847, simple_loss=0.2808, pruned_loss=0.0443, over 18683.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2618, pruned_loss=0.04294, over 1475863.93 frames. ], batch size: 78, lr: 5.27e-03, grad_scale: 8.0 2022-12-23 17:38:03,893 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73732.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:38:16,525 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.153e+02 3.091e+02 3.725e+02 4.885e+02 1.063e+03, threshold=7.451e+02, percent-clipped=0.0 2022-12-23 17:39:15,705 INFO [train.py:894] (3/4) Epoch 22, batch 150, loss[loss=0.1826, simple_loss=0.2763, pruned_loss=0.0444, over 18486.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2613, pruned_loss=0.04289, over 1971846.97 frames. ], batch size: 54, lr: 5.27e-03, grad_scale: 8.0 2022-12-23 17:39:24,341 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0 from training. Duration: 24.525 2022-12-23 17:39:37,202 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73793.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:40:00,566 WARNING [train.py:1060] (3/4) Exclude cut with ID 3699-47246-0007-3408-0_sp0.9 from training. Duration: 20.26675 2022-12-23 17:40:14,064 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp0.9 from training. Duration: 27.25 2022-12-23 17:40:32,539 INFO [train.py:894] (3/4) Epoch 22, batch 200, loss[loss=0.1817, simple_loss=0.2686, pruned_loss=0.04739, over 18511.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2607, pruned_loss=0.04237, over 2357798.38 frames. ], batch size: 52, lr: 5.27e-03, grad_scale: 8.0 2022-12-23 17:40:49,759 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.045e+02 3.217e+02 3.912e+02 4.713e+02 9.147e+02, threshold=7.824e+02, percent-clipped=3.0 2022-12-23 17:41:20,415 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8603, 1.2714, 0.8096, 1.3417, 2.0536, 1.2526, 1.4574, 1.6044], device='cuda:3'), covar=tensor([0.1498, 0.2036, 0.2175, 0.1520, 0.1802, 0.1703, 0.1509, 0.1712], device='cuda:3'), in_proj_covar=tensor([0.0093, 0.0097, 0.0115, 0.0095, 0.0116, 0.0090, 0.0097, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 17:41:27,616 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0 from training. Duration: 21.68 2022-12-23 17:41:40,360 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0 from training. Duration: 21.6300625 2022-12-23 17:41:48,830 INFO [train.py:894] (3/4) Epoch 22, batch 250, loss[loss=0.1686, simple_loss=0.2552, pruned_loss=0.04093, over 18381.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2608, pruned_loss=0.04245, over 2659422.79 frames. ], batch size: 46, lr: 5.27e-03, grad_scale: 8.0 2022-12-23 17:42:05,179 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0_sp0.9 from training. Duration: 24.033375 2022-12-23 17:42:22,444 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2022-12-23 17:42:50,809 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2022-12-23 17:42:55,228 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4519, 1.3796, 1.4068, 1.3781, 0.8305, 2.2935, 0.8336, 1.3605], device='cuda:3'), covar=tensor([0.3312, 0.2309, 0.2197, 0.2177, 0.1600, 0.0326, 0.1829, 0.0959], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0118, 0.0125, 0.0121, 0.0104, 0.0097, 0.0091, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 17:43:02,019 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0 from training. Duration: 22.905 2022-12-23 17:43:03,554 INFO [train.py:894] (3/4) Epoch 22, batch 300, loss[loss=0.1683, simple_loss=0.2527, pruned_loss=0.04194, over 18433.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.261, pruned_loss=0.04244, over 2893598.92 frames. ], batch size: 42, lr: 5.27e-03, grad_scale: 8.0 2022-12-23 17:43:03,644 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp1.1 from training. Duration: 23.4318125 2022-12-23 17:43:22,884 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.285e+02 3.196e+02 4.010e+02 5.060e+02 9.017e+02, threshold=8.020e+02, percent-clipped=3.0 2022-12-23 17:43:30,107 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73946.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:43:35,551 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73949.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:43:35,756 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8020, 2.2296, 1.7276, 2.4596, 3.1255, 1.7874, 2.1974, 1.4267], device='cuda:3'), covar=tensor([0.1810, 0.1695, 0.1507, 0.0955, 0.1160, 0.1051, 0.1672, 0.1491], device='cuda:3'), in_proj_covar=tensor([0.0244, 0.0223, 0.0214, 0.0197, 0.0256, 0.0195, 0.0223, 0.0198], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 17:43:43,533 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9125, 1.2590, 0.7661, 1.3351, 2.3527, 1.2205, 1.5636, 1.6086], device='cuda:3'), covar=tensor([0.1493, 0.1963, 0.2238, 0.1530, 0.1566, 0.1770, 0.1465, 0.1762], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0098, 0.0115, 0.0096, 0.0116, 0.0091, 0.0098, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 17:44:20,093 INFO [train.py:894] (3/4) Epoch 22, batch 350, loss[loss=0.1852, simple_loss=0.2802, pruned_loss=0.04505, over 18526.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2614, pruned_loss=0.04241, over 3074786.67 frames. ], batch size: 55, lr: 5.26e-03, grad_scale: 8.0 2022-12-23 17:44:48,108 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=73997.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:45:10,528 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp1.1 from training. Duration: 20.82275 2022-12-23 17:45:11,864 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp0.9 from training. Duration: 25.45 2022-12-23 17:45:15,229 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8063, 2.3012, 1.6835, 2.5412, 2.8633, 1.7803, 1.9395, 1.4715], device='cuda:3'), covar=tensor([0.1843, 0.1647, 0.1547, 0.0930, 0.1405, 0.1050, 0.1940, 0.1509], device='cuda:3'), in_proj_covar=tensor([0.0245, 0.0224, 0.0214, 0.0198, 0.0256, 0.0195, 0.0223, 0.0199], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 17:45:40,512 INFO [train.py:894] (3/4) Epoch 22, batch 400, loss[loss=0.166, simple_loss=0.2639, pruned_loss=0.03403, over 18632.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2623, pruned_loss=0.04258, over 3216690.11 frames. ], batch size: 53, lr: 5.26e-03, grad_scale: 8.0 2022-12-23 17:45:57,386 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.190e+02 3.712e+02 4.382e+02 4.971e+02 1.685e+03, threshold=8.765e+02, percent-clipped=4.0 2022-12-23 17:46:11,645 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0 from training. Duration: 25.775 2022-12-23 17:46:32,054 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0_sp0.9 from training. Duration: 22.25 2022-12-23 17:46:32,455 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74064.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:46:52,523 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.21 vs. limit=5.0 2022-12-23 17:46:53,843 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.6193, 1.8777, 2.1123, 1.1384, 1.3097, 2.2771, 2.0771, 1.7390], device='cuda:3'), covar=tensor([0.0740, 0.0311, 0.0345, 0.0385, 0.0424, 0.0448, 0.0274, 0.0661], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0171, 0.0127, 0.0141, 0.0150, 0.0143, 0.0165, 0.0173], device='cuda:3'), out_proj_covar=tensor([1.1311e-04, 1.3081e-04, 9.5502e-05, 1.0485e-04, 1.1187e-04, 1.0930e-04, 1.2637e-04, 1.3150e-04], device='cuda:3') 2022-12-23 17:46:54,830 INFO [train.py:894] (3/4) Epoch 22, batch 450, loss[loss=0.1789, simple_loss=0.2729, pruned_loss=0.0424, over 18581.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2625, pruned_loss=0.04296, over 3327334.96 frames. ], batch size: 56, lr: 5.26e-03, grad_scale: 8.0 2022-12-23 17:46:59,053 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0 from training. Duration: 26.205 2022-12-23 17:47:08,178 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74088.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:47:15,380 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp0.9 from training. Duration: 30.1555625 2022-12-23 17:47:20,986 WARNING [train.py:1060] (3/4) Exclude cut with ID 1265-135635-0050-6781-0_sp0.9 from training. Duration: 21.8333125 2022-12-23 17:47:30,990 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp1.1 from training. Duration: 20.6545625 2022-12-23 17:48:04,683 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74125.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:48:09,818 INFO [train.py:894] (3/4) Epoch 22, batch 500, loss[loss=0.1948, simple_loss=0.284, pruned_loss=0.05276, over 18522.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2619, pruned_loss=0.04265, over 3412410.16 frames. ], batch size: 58, lr: 5.26e-03, grad_scale: 8.0 2022-12-23 17:48:12,528 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0045-39920-0_sp0.9 from training. Duration: 20.52225 2022-12-23 17:48:27,461 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.598e+02 3.456e+02 4.144e+02 5.225e+02 1.116e+03, threshold=8.288e+02, percent-clipped=3.0 2022-12-23 17:48:29,296 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7991, 0.7407, 1.6428, 1.4401, 1.8644, 1.9082, 1.5162, 1.6871], device='cuda:3'), covar=tensor([0.2240, 0.3376, 0.2629, 0.2726, 0.2081, 0.0978, 0.3170, 0.1327], device='cuda:3'), in_proj_covar=tensor([0.0269, 0.0300, 0.0278, 0.0316, 0.0309, 0.0253, 0.0343, 0.0242], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 17:48:31,678 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp0.9 from training. Duration: 29.1166875 2022-12-23 17:48:48,317 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.2538, 1.2443, 0.9523, 1.2320, 1.3239, 1.1652, 1.4844, 1.3389], device='cuda:3'), covar=tensor([0.0654, 0.1159, 0.1950, 0.1245, 0.1217, 0.0694, 0.0742, 0.0900], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0209, 0.0251, 0.0287, 0.0235, 0.0190, 0.0205, 0.0203], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 17:49:06,737 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9690, 1.8557, 1.9948, 1.1426, 2.0710, 2.1526, 1.5469, 2.4425], device='cuda:3'), covar=tensor([0.1065, 0.1825, 0.1207, 0.1806, 0.0718, 0.1095, 0.2253, 0.0559], device='cuda:3'), in_proj_covar=tensor([0.0196, 0.0210, 0.0205, 0.0191, 0.0173, 0.0215, 0.0211, 0.0191], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 17:49:25,521 INFO [train.py:894] (3/4) Epoch 22, batch 550, loss[loss=0.1848, simple_loss=0.2754, pruned_loss=0.04705, over 18632.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2634, pruned_loss=0.04305, over 3478930.57 frames. ], batch size: 53, lr: 5.26e-03, grad_scale: 8.0 2022-12-23 17:49:33,250 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133211-0007-59831-0_sp0.9 from training. Duration: 21.388875 2022-12-23 17:50:08,130 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0 from training. Duration: 22.72 2022-12-23 17:50:09,660 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0_sp0.9 from training. Duration: 22.7444375 2022-12-23 17:50:43,004 INFO [train.py:894] (3/4) Epoch 22, batch 600, loss[loss=0.1928, simple_loss=0.281, pruned_loss=0.05226, over 18594.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2636, pruned_loss=0.04294, over 3530758.24 frames. ], batch size: 51, lr: 5.26e-03, grad_scale: 8.0 2022-12-23 17:50:53,990 WARNING [train.py:1060] (3/4) Exclude cut with ID 4133-6541-0027-40495-0_sp1.1 from training. Duration: 0.9681875 2022-12-23 17:50:56,981 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0_sp0.9 from training. Duration: 22.3166875 2022-12-23 17:51:00,213 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74240.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:51:01,181 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.128e+02 3.307e+02 4.178e+02 5.054e+02 1.095e+03, threshold=8.355e+02, percent-clipped=2.0 2022-12-23 17:51:03,292 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133212-0015-59917-0_sp0.9 from training. Duration: 21.8166875 2022-12-23 17:51:09,583 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74246.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:51:23,106 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.77 vs. limit=5.0 2022-12-23 17:51:59,079 INFO [train.py:894] (3/4) Epoch 22, batch 650, loss[loss=0.176, simple_loss=0.2752, pruned_loss=0.03841, over 18482.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2635, pruned_loss=0.04283, over 3571304.83 frames. ], batch size: 54, lr: 5.25e-03, grad_scale: 8.0 2022-12-23 17:52:23,066 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=74294.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:52:33,803 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74301.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:52:46,079 WARNING [train.py:1060] (3/4) Exclude cut with ID 4957-30119-0041-23990-0_sp0.9 from training. Duration: 20.22775 2022-12-23 17:53:14,479 INFO [train.py:894] (3/4) Epoch 22, batch 700, loss[loss=0.1988, simple_loss=0.2855, pruned_loss=0.05609, over 18684.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2642, pruned_loss=0.04327, over 3604014.21 frames. ], batch size: 62, lr: 5.25e-03, grad_scale: 8.0 2022-12-23 17:53:31,857 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp1.1 from training. Duration: 24.67275 2022-12-23 17:53:33,148 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.504e+02 3.513e+02 4.127e+02 5.298e+02 1.016e+03, threshold=8.254e+02, percent-clipped=2.0 2022-12-23 17:54:00,383 WARNING [train.py:1060] (3/4) Exclude cut with ID 3082-165428-0081-50734-0_sp0.9 from training. Duration: 21.8055625 2022-12-23 17:54:03,611 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2022-12-23 17:54:30,481 INFO [train.py:894] (3/4) Epoch 22, batch 750, loss[loss=0.1575, simple_loss=0.2417, pruned_loss=0.03666, over 18386.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2639, pruned_loss=0.04308, over 3628911.02 frames. ], batch size: 46, lr: 5.25e-03, grad_scale: 8.0 2022-12-23 17:54:37,746 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0_sp0.9 from training. Duration: 22.6666875 2022-12-23 17:54:44,591 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74388.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:54:49,163 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.2817, 2.4094, 1.8862, 3.1485, 2.2466, 2.5418, 2.4330, 3.3707], device='cuda:3'), covar=tensor([0.1823, 0.3577, 0.1996, 0.2895, 0.4049, 0.1057, 0.3373, 0.0801], device='cuda:3'), in_proj_covar=tensor([0.0292, 0.0290, 0.0244, 0.0346, 0.0270, 0.0227, 0.0286, 0.0215], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 17:55:32,270 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74420.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:55:42,091 WARNING [train.py:1060] (3/4) Exclude cut with ID 2411-132532-0017-82279-0_sp1.1 from training. Duration: 0.9681875 2022-12-23 17:55:44,842 INFO [train.py:894] (3/4) Epoch 22, batch 800, loss[loss=0.1608, simple_loss=0.2473, pruned_loss=0.03712, over 18419.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2643, pruned_loss=0.04327, over 3648271.90 frames. ], batch size: 48, lr: 5.25e-03, grad_scale: 8.0 2022-12-23 17:55:55,880 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=74436.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:56:02,822 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.088e+02 3.453e+02 4.133e+02 5.100e+02 8.129e+02, threshold=8.265e+02, percent-clipped=0.0 2022-12-23 17:56:06,312 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0 from training. Duration: 22.485 2022-12-23 17:56:29,567 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5003, 1.8894, 2.0618, 2.0604, 2.3884, 2.4083, 2.1887, 1.9352], device='cuda:3'), covar=tensor([0.2253, 0.3345, 0.2654, 0.3107, 0.2039, 0.0979, 0.3744, 0.1370], device='cuda:3'), in_proj_covar=tensor([0.0268, 0.0298, 0.0278, 0.0315, 0.0308, 0.0252, 0.0342, 0.0242], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 17:56:43,751 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp1.1 from training. Duration: 23.82275 2022-12-23 17:56:45,415 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1792, 2.0300, 2.0801, 2.1110, 2.1933, 5.2844, 2.2950, 2.7346], device='cuda:3'), covar=tensor([0.2805, 0.1862, 0.1756, 0.1870, 0.1082, 0.0095, 0.1324, 0.0747], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0118, 0.0126, 0.0121, 0.0104, 0.0097, 0.0092, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 17:56:58,673 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0 from training. Duration: 20.77 2022-12-23 17:56:59,898 INFO [train.py:894] (3/4) Epoch 22, batch 850, loss[loss=0.1638, simple_loss=0.2595, pruned_loss=0.03401, over 18517.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.264, pruned_loss=0.04323, over 3662751.39 frames. ], batch size: 58, lr: 5.25e-03, grad_scale: 16.0 2022-12-23 17:57:05,363 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0_sp0.9 from training. Duration: 24.088875 2022-12-23 17:57:09,733 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2022-12-23 17:57:36,038 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp1.1 from training. Duration: 20.4409375 2022-12-23 17:58:00,885 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2022-12-23 17:58:15,038 INFO [train.py:894] (3/4) Epoch 22, batch 900, loss[loss=0.182, simple_loss=0.2779, pruned_loss=0.04303, over 18624.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2635, pruned_loss=0.04296, over 3673443.66 frames. ], batch size: 99, lr: 5.25e-03, grad_scale: 16.0 2022-12-23 17:58:27,404 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2022-12-23 17:58:33,728 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.249e+02 3.429e+02 4.111e+02 5.297e+02 1.524e+03, threshold=8.223e+02, percent-clipped=3.0 2022-12-23 17:58:54,565 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0_sp0.9 from training. Duration: 22.511125 2022-12-23 17:58:54,593 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0 from training. Duration: 20.675 2022-12-23 17:59:30,701 INFO [train.py:894] (3/4) Epoch 22, batch 950, loss[loss=0.206, simple_loss=0.2879, pruned_loss=0.06199, over 18730.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2641, pruned_loss=0.04321, over 3683351.17 frames. ], batch size: 52, lr: 5.24e-03, grad_scale: 16.0 2022-12-23 17:59:56,778 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74595.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 17:59:58,040 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74596.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 18:00:31,577 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp0.9 from training. Duration: 24.9833125 2022-12-23 18:00:48,656 INFO [train.py:894] (3/4) Epoch 22, batch 1000, loss[loss=0.1515, simple_loss=0.2394, pruned_loss=0.03183, over 18567.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2645, pruned_loss=0.04339, over 3690311.61 frames. ], batch size: 49, lr: 5.24e-03, grad_scale: 16.0 2022-12-23 18:01:03,980 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0 from training. Duration: 27.14 2022-12-23 18:01:06,413 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2022-12-23 18:01:06,918 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.184e+02 3.436e+02 4.038e+02 4.889e+02 8.968e+02, threshold=8.077e+02, percent-clipped=3.0 2022-12-23 18:01:19,148 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0 from training. Duration: 22.44 2022-12-23 18:01:24,467 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3979, 1.5406, 0.5101, 1.8726, 2.3840, 1.5916, 1.9377, 2.1853], device='cuda:3'), covar=tensor([0.1379, 0.2002, 0.2691, 0.1400, 0.1703, 0.1895, 0.1387, 0.1556], device='cuda:3'), in_proj_covar=tensor([0.0093, 0.0097, 0.0115, 0.0096, 0.0116, 0.0091, 0.0097, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 18:01:30,482 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74656.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 18:02:05,200 INFO [train.py:894] (3/4) Epoch 22, batch 1050, loss[loss=0.1734, simple_loss=0.2537, pruned_loss=0.04652, over 18621.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2642, pruned_loss=0.04315, over 3695473.70 frames. ], batch size: 45, lr: 5.24e-03, grad_scale: 16.0 2022-12-23 18:02:37,334 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0060-62364-0_sp0.9 from training. Duration: 21.361125 2022-12-23 18:02:43,489 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp1.1 from training. Duration: 27.0318125 2022-12-23 18:02:53,571 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp0.9 from training. Duration: 28.638875 2022-12-23 18:03:07,359 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74720.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 18:03:08,748 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0 from training. Duration: 20.4 2022-12-23 18:03:20,744 INFO [train.py:894] (3/4) Epoch 22, batch 1100, loss[loss=0.1578, simple_loss=0.238, pruned_loss=0.03881, over 18549.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2634, pruned_loss=0.04269, over 3699053.92 frames. ], batch size: 44, lr: 5.24e-03, grad_scale: 16.0 2022-12-23 18:03:33,943 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2022-12-23 18:03:39,047 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.975e+02 3.285e+02 3.886e+02 4.761e+02 1.198e+03, threshold=7.773e+02, percent-clipped=4.0 2022-12-23 18:03:40,629 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0 from training. Duration: 20.025 2022-12-23 18:03:40,641 WARNING [train.py:1060] (3/4) Exclude cut with ID 2364-131735-0112-64612-0_sp0.9 from training. Duration: 20.488875 2022-12-23 18:03:45,608 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0 from training. Duration: 29.735 2022-12-23 18:04:20,903 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=74768.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 18:04:37,412 INFO [train.py:894] (3/4) Epoch 22, batch 1150, loss[loss=0.1849, simple_loss=0.2756, pruned_loss=0.04714, over 18498.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2632, pruned_loss=0.04265, over 3701946.36 frames. ], batch size: 52, lr: 5.24e-03, grad_scale: 16.0 2022-12-23 18:05:06,948 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4269, 1.1058, 1.6530, 2.6435, 1.9357, 2.4325, 0.7483, 1.8808], device='cuda:3'), covar=tensor([0.1876, 0.1808, 0.1471, 0.0711, 0.1077, 0.0981, 0.2196, 0.1272], device='cuda:3'), in_proj_covar=tensor([0.0102, 0.0117, 0.0136, 0.0149, 0.0104, 0.0141, 0.0129, 0.0113], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 18:05:08,157 WARNING [train.py:1060] (3/4) Exclude cut with ID 7276-92427-0014-12983-0_sp0.9 from training. Duration: 21.3055625 2022-12-23 18:05:09,613 WARNING [train.py:1060] (3/4) Exclude cut with ID 1025-75365-0008-79168-0_sp0.9 from training. Duration: 22.0666875 2022-12-23 18:05:53,486 INFO [train.py:894] (3/4) Epoch 22, batch 1200, loss[loss=0.1745, simple_loss=0.2701, pruned_loss=0.03938, over 18698.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2622, pruned_loss=0.04213, over 3703318.02 frames. ], batch size: 65, lr: 5.23e-03, grad_scale: 16.0 2022-12-23 18:06:10,681 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.171e+02 3.301e+02 3.924e+02 4.920e+02 7.003e+02, threshold=7.847e+02, percent-clipped=0.0 2022-12-23 18:06:43,758 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2022-12-23 18:06:56,716 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0 from training. Duration: 20.26 2022-12-23 18:07:07,534 INFO [train.py:894] (3/4) Epoch 22, batch 1250, loss[loss=0.1799, simple_loss=0.266, pruned_loss=0.04692, over 18669.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2629, pruned_loss=0.04209, over 3705572.11 frames. ], batch size: 46, lr: 5.23e-03, grad_scale: 16.0 2022-12-23 18:07:10,307 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0030-9324-0_sp0.9 from training. Duration: 21.3444375 2022-12-23 18:07:33,648 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74896.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 18:07:47,079 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2492, 2.0011, 2.3050, 1.3891, 2.3533, 2.3652, 1.5668, 2.6122], device='cuda:3'), covar=tensor([0.1135, 0.1868, 0.1370, 0.2031, 0.0776, 0.1226, 0.2444, 0.0597], device='cuda:3'), in_proj_covar=tensor([0.0194, 0.0208, 0.0204, 0.0190, 0.0171, 0.0213, 0.0211, 0.0191], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 18:07:53,263 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74909.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 18:08:04,019 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2022-12-23 18:08:06,614 WARNING [train.py:1060] (3/4) Exclude cut with ID 497-129325-0061-62254-0_sp1.1 from training. Duration: 0.97725 2022-12-23 18:08:22,524 INFO [train.py:894] (3/4) Epoch 22, batch 1300, loss[loss=0.1657, simple_loss=0.2502, pruned_loss=0.04062, over 18433.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2634, pruned_loss=0.04223, over 3708424.25 frames. ], batch size: 48, lr: 5.23e-03, grad_scale: 16.0 2022-12-23 18:08:40,175 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.220e+02 3.437e+02 4.231e+02 5.337e+02 9.980e+02, threshold=8.462e+02, percent-clipped=5.0 2022-12-23 18:08:44,641 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=74944.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 18:08:50,289 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0_sp0.9 from training. Duration: 22.97225 2022-12-23 18:08:55,008 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74951.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 18:09:14,749 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74964.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 18:09:22,558 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0047-39922-0_sp0.9 from training. Duration: 21.97775 2022-12-23 18:09:24,312 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74970.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 18:09:37,113 INFO [train.py:894] (3/4) Epoch 22, batch 1350, loss[loss=0.1904, simple_loss=0.2826, pruned_loss=0.04905, over 18486.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2634, pruned_loss=0.04227, over 3709433.72 frames. ], batch size: 64, lr: 5.23e-03, grad_scale: 16.0 2022-12-23 18:09:37,167 WARNING [train.py:1060] (3/4) Exclude cut with ID 1112-1043-0006-89194-0_sp0.9 from training. Duration: 21.8333125 2022-12-23 18:09:46,859 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6488, 1.3579, 1.4231, 1.9086, 1.7718, 3.2428, 1.2385, 1.5273], device='cuda:3'), covar=tensor([0.0785, 0.1855, 0.1074, 0.0875, 0.1334, 0.0233, 0.1482, 0.1497], device='cuda:3'), in_proj_covar=tensor([0.0072, 0.0082, 0.0071, 0.0074, 0.0090, 0.0075, 0.0085, 0.0077], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 18:09:49,477 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0 from training. Duration: 20.47 2022-12-23 18:10:05,294 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.0767, 0.9348, 1.1554, 0.6160, 0.6472, 1.2223, 1.2074, 1.1707], device='cuda:3'), covar=tensor([0.0829, 0.0364, 0.0345, 0.0413, 0.0505, 0.0513, 0.0320, 0.0631], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0173, 0.0128, 0.0142, 0.0150, 0.0144, 0.0168, 0.0175], device='cuda:3'), out_proj_covar=tensor([1.1460e-04, 1.3224e-04, 9.5747e-05, 1.0514e-04, 1.1176e-04, 1.0952e-04, 1.2868e-04, 1.3275e-04], device='cuda:3') 2022-12-23 18:10:27,657 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5531, 2.3794, 1.8599, 1.2638, 2.7629, 2.6638, 2.3305, 1.7753], device='cuda:3'), covar=tensor([0.0381, 0.0411, 0.0571, 0.0834, 0.0308, 0.0352, 0.0468, 0.0901], device='cuda:3'), in_proj_covar=tensor([0.0125, 0.0127, 0.0129, 0.0118, 0.0101, 0.0123, 0.0133, 0.0158], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 18:10:41,714 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6900, 1.3456, 1.1167, 0.2355, 1.1369, 1.5327, 1.3317, 1.4801], device='cuda:3'), covar=tensor([0.0725, 0.0652, 0.1138, 0.1830, 0.1279, 0.1774, 0.1906, 0.0723], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0184, 0.0205, 0.0190, 0.0209, 0.0199, 0.0214, 0.0199], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 18:10:47,802 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75025.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 18:10:53,382 INFO [train.py:894] (3/4) Epoch 22, batch 1400, loss[loss=0.1692, simple_loss=0.2567, pruned_loss=0.04085, over 18429.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2641, pruned_loss=0.04249, over 3709785.23 frames. ], batch size: 48, lr: 5.23e-03, grad_scale: 16.0 2022-12-23 18:10:53,478 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0037-39912-0_sp0.9 from training. Duration: 20.67225 2022-12-23 18:11:11,918 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.292e+02 3.487e+02 4.292e+02 5.613e+02 1.451e+03, threshold=8.583e+02, percent-clipped=4.0 2022-12-23 18:11:11,986 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp0.9 from training. Duration: 25.2444375 2022-12-23 18:11:34,006 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0021-76797-0_sp0.9 from training. Duration: 21.1445 2022-12-23 18:11:46,706 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2659, 1.6653, 1.9269, 1.9010, 2.2027, 2.2037, 2.0569, 1.8485], device='cuda:3'), covar=tensor([0.2259, 0.3350, 0.2614, 0.2958, 0.2148, 0.1004, 0.3420, 0.1398], device='cuda:3'), in_proj_covar=tensor([0.0268, 0.0299, 0.0279, 0.0318, 0.0309, 0.0253, 0.0345, 0.0243], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 18:12:07,253 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8612, 2.4020, 1.6800, 2.4355, 3.0844, 1.7560, 1.9839, 1.4283], device='cuda:3'), covar=tensor([0.1833, 0.1575, 0.1588, 0.0993, 0.1231, 0.1086, 0.1870, 0.1560], device='cuda:3'), in_proj_covar=tensor([0.0246, 0.0226, 0.0216, 0.0199, 0.0259, 0.0196, 0.0224, 0.0199], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 18:12:08,229 INFO [train.py:894] (3/4) Epoch 22, batch 1450, loss[loss=0.1797, simple_loss=0.2702, pruned_loss=0.04458, over 18554.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2641, pruned_loss=0.04258, over 3711137.02 frames. ], batch size: 69, lr: 5.23e-03, grad_scale: 16.0 2022-12-23 18:12:48,722 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp0.9 from training. Duration: 33.038875 2022-12-23 18:12:55,177 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3580, 2.5117, 1.9681, 1.7616, 2.5349, 2.9619, 2.7643, 2.1285], device='cuda:3'), covar=tensor([0.0360, 0.0260, 0.0410, 0.0270, 0.0209, 0.0340, 0.0301, 0.0300], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0125, 0.0151, 0.0123, 0.0115, 0.0119, 0.0098, 0.0126], device='cuda:3'), out_proj_covar=tensor([7.4453e-05, 9.9088e-05, 1.2420e-04, 9.7759e-05, 9.2422e-05, 9.1741e-05, 7.6317e-05, 9.9647e-05], device='cuda:3') 2022-12-23 18:13:25,051 INFO [train.py:894] (3/4) Epoch 22, batch 1500, loss[loss=0.1468, simple_loss=0.2286, pruned_loss=0.03248, over 18405.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2638, pruned_loss=0.0425, over 3710911.87 frames. ], batch size: 42, lr: 5.22e-03, grad_scale: 16.0 2022-12-23 18:13:25,102 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64291-0000-16059-0_sp0.9 from training. Duration: 20.0944375 2022-12-23 18:13:31,304 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2022-12-23 18:13:39,633 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp1.1 from training. Duration: 20.4 2022-12-23 18:13:42,479 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.040e+02 3.361e+02 3.944e+02 4.833e+02 7.905e+02, threshold=7.889e+02, percent-clipped=0.0 2022-12-23 18:13:48,205 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0 from training. Duration: 20.085 2022-12-23 18:13:53,073 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6437, 1.6822, 1.8468, 1.1141, 1.7914, 1.9067, 1.4424, 2.1247], device='cuda:3'), covar=tensor([0.1137, 0.1858, 0.1058, 0.1744, 0.0734, 0.1119, 0.2317, 0.0573], device='cuda:3'), in_proj_covar=tensor([0.0193, 0.0208, 0.0203, 0.0191, 0.0170, 0.0212, 0.0208, 0.0190], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 18:14:00,982 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0_sp0.9 from training. Duration: 23.07775 2022-12-23 18:14:11,957 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3099, 2.3044, 1.7676, 2.6819, 2.5643, 2.2634, 3.1951, 2.3280], device='cuda:3'), covar=tensor([0.0789, 0.1763, 0.2668, 0.1667, 0.1640, 0.0829, 0.0792, 0.1194], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0208, 0.0250, 0.0287, 0.0235, 0.0190, 0.0203, 0.0203], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 18:14:19,019 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5535, 4.1048, 3.9349, 1.7331, 4.2092, 3.2111, 0.7890, 2.6073], device='cuda:3'), covar=tensor([0.2167, 0.1264, 0.1224, 0.3540, 0.0753, 0.0889, 0.5172, 0.1658], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0140, 0.0156, 0.0124, 0.0144, 0.0114, 0.0143, 0.0114], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 18:14:26,137 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.1991, 1.4405, 1.7116, 0.9064, 1.0718, 1.7812, 1.7184, 1.5538], device='cuda:3'), covar=tensor([0.0791, 0.0354, 0.0311, 0.0371, 0.0440, 0.0490, 0.0256, 0.0711], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0172, 0.0128, 0.0140, 0.0149, 0.0143, 0.0166, 0.0175], device='cuda:3'), out_proj_covar=tensor([1.1436e-04, 1.3149e-04, 9.5502e-05, 1.0421e-04, 1.1123e-04, 1.0903e-04, 1.2735e-04, 1.3292e-04], device='cuda:3') 2022-12-23 18:14:40,482 INFO [train.py:894] (3/4) Epoch 22, batch 1550, loss[loss=0.1947, simple_loss=0.284, pruned_loss=0.05264, over 18454.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2625, pruned_loss=0.04197, over 3710912.17 frames. ], batch size: 64, lr: 5.22e-03, grad_scale: 16.0 2022-12-23 18:14:46,248 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp0.9 from training. Duration: 24.9333125 2022-12-23 18:15:27,812 WARNING [train.py:1060] (3/4) Exclude cut with ID 5118-111612-0016-124680-0_sp0.9 from training. Duration: 20.388875 2022-12-23 18:15:32,159 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp1.1 from training. Duration: 20.3590625 2022-12-23 18:15:55,815 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=75228.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 18:15:56,904 INFO [train.py:894] (3/4) Epoch 22, batch 1600, loss[loss=0.1901, simple_loss=0.277, pruned_loss=0.05162, over 18600.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2627, pruned_loss=0.04176, over 3711907.75 frames. ], batch size: 57, lr: 5.22e-03, grad_scale: 16.0 2022-12-23 18:16:14,678 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.171e+02 3.261e+02 3.774e+02 4.790e+02 7.917e+02, threshold=7.548e+02, percent-clipped=1.0 2022-12-23 18:16:30,120 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75251.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 18:16:42,660 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0_sp1.1 from training. Duration: 0.836375 2022-12-23 18:16:51,741 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75265.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 18:16:59,877 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=75270.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 18:17:13,158 INFO [train.py:894] (3/4) Epoch 22, batch 1650, loss[loss=0.1483, simple_loss=0.2246, pruned_loss=0.03603, over 18496.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2634, pruned_loss=0.04242, over 3712242.29 frames. ], batch size: 43, lr: 5.22e-03, grad_scale: 16.0 2022-12-23 18:17:15,151 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8205, 2.6128, 1.8312, 2.5732, 3.2336, 1.9252, 2.0754, 1.4711], device='cuda:3'), covar=tensor([0.1815, 0.1550, 0.1491, 0.0955, 0.1174, 0.1042, 0.1733, 0.1499], device='cuda:3'), in_proj_covar=tensor([0.0247, 0.0227, 0.0216, 0.0201, 0.0261, 0.0197, 0.0226, 0.0200], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 18:17:25,089 WARNING [train.py:1060] (3/4) Exclude cut with ID 8565-290391-0049-67394-0_sp0.9 from training. Duration: 21.3166875 2022-12-23 18:17:28,467 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75289.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 18:17:43,249 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=75299.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 18:17:46,050 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.59 vs. limit=2.0 2022-12-23 18:17:55,541 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0029-104863-0_sp0.9 from training. Duration: 22.1055625 2022-12-23 18:18:06,226 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp1.1 from training. Duration: 21.77725 2022-12-23 18:18:07,351 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2022-12-23 18:18:15,438 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75320.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 18:18:25,491 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp0.9 from training. Duration: 27.8166875 2022-12-23 18:18:28,720 INFO [train.py:894] (3/4) Epoch 22, batch 1700, loss[loss=0.1527, simple_loss=0.2397, pruned_loss=0.03285, over 18382.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2641, pruned_loss=0.0438, over 3712891.55 frames. ], batch size: 46, lr: 5.22e-03, grad_scale: 16.0 2022-12-23 18:18:32,209 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75331.0, num_to_drop=1, layers_to_drop={3} 2022-12-23 18:18:46,422 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.494e+02 3.900e+02 4.543e+02 5.903e+02 1.187e+03, threshold=9.086e+02, percent-clipped=10.0 2022-12-23 18:18:51,133 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp1.1 from training. Duration: 22.5090625 2022-12-23 18:18:56,348 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0 from training. Duration: 25.035 2022-12-23 18:19:08,869 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.47 vs. limit=5.0 2022-12-23 18:19:14,632 WARNING [train.py:1060] (3/4) Exclude cut with ID 774-127930-0014-10412-0_sp1.1 from training. Duration: 0.95 2022-12-23 18:19:27,405 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.58 vs. limit=2.0 2022-12-23 18:19:33,786 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp0.9 from training. Duration: 0.92225 2022-12-23 18:19:44,171 INFO [train.py:894] (3/4) Epoch 22, batch 1750, loss[loss=0.158, simple_loss=0.2382, pruned_loss=0.0389, over 18468.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2651, pruned_loss=0.04533, over 3713213.93 frames. ], batch size: 43, lr: 5.22e-03, grad_scale: 16.0 2022-12-23 18:20:00,928 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0 from training. Duration: 21.97 2022-12-23 18:20:19,489 WARNING [train.py:1060] (3/4) Exclude cut with ID 7492-105653-0055-62765-0_sp0.9 from training. Duration: 21.97225 2022-12-23 18:20:19,524 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp0.9 from training. Duration: 25.3333125 2022-12-23 18:20:25,130 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.1275, 0.9816, 0.9745, 1.1490, 1.2243, 1.1615, 1.1199, 0.9989], device='cuda:3'), covar=tensor([0.0263, 0.0259, 0.0548, 0.0208, 0.0226, 0.0375, 0.0267, 0.0305], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0127, 0.0153, 0.0124, 0.0117, 0.0121, 0.0099, 0.0128], device='cuda:3'), out_proj_covar=tensor([7.4852e-05, 1.0068e-04, 1.2607e-04, 9.8926e-05, 9.4238e-05, 9.2783e-05, 7.6910e-05, 1.0078e-04], device='cuda:3') 2022-12-23 18:20:30,562 WARNING [train.py:1060] (3/4) Exclude cut with ID 5172-29468-0015-19128-0_sp0.9 from training. Duration: 21.5055625 2022-12-23 18:20:39,783 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp1.1 from training. Duration: 20.72725 2022-12-23 18:21:00,858 INFO [train.py:894] (3/4) Epoch 22, batch 1800, loss[loss=0.1842, simple_loss=0.2647, pruned_loss=0.05179, over 18662.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.266, pruned_loss=0.04696, over 3713830.79 frames. ], batch size: 48, lr: 5.21e-03, grad_scale: 16.0 2022-12-23 18:21:10,292 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp0.9 from training. Duration: 26.32775 2022-12-23 18:21:19,341 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.815e+02 4.099e+02 5.163e+02 6.501e+02 1.367e+03, threshold=1.033e+03, percent-clipped=5.0 2022-12-23 18:21:41,008 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0 from training. Duration: 20.025 2022-12-23 18:21:46,253 WARNING [train.py:1060] (3/4) Exclude cut with ID 6709-74022-0004-86860-0_sp1.1 from training. Duration: 0.9409375 2022-12-23 18:21:47,550 WARNING [train.py:1060] (3/4) Exclude cut with ID 4757-1811-0023-62229-0_sp0.9 from training. Duration: 21.37775 2022-12-23 18:22:07,042 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0004-25974-0_sp0.9 from training. Duration: 21.17225 2022-12-23 18:22:08,328 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp0.9 from training. Duration: 27.511125 2022-12-23 18:22:17,419 INFO [train.py:894] (3/4) Epoch 22, batch 1850, loss[loss=0.1756, simple_loss=0.2596, pruned_loss=0.04584, over 18433.00 frames. ], tot_loss[loss=0.183, simple_loss=0.268, pruned_loss=0.04899, over 3713637.60 frames. ], batch size: 48, lr: 5.21e-03, grad_scale: 16.0 2022-12-23 18:22:40,786 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0 from training. Duration: 22.8 2022-12-23 18:22:45,989 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0 from training. Duration: 22.585 2022-12-23 18:22:46,232 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.7037, 3.2749, 3.2264, 3.6019, 3.3516, 3.2864, 3.8114, 1.4044], device='cuda:3'), covar=tensor([0.0826, 0.0652, 0.0713, 0.0940, 0.1416, 0.1149, 0.0771, 0.4528], device='cuda:3'), in_proj_covar=tensor([0.0339, 0.0223, 0.0233, 0.0268, 0.0322, 0.0264, 0.0286, 0.0282], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 18:23:15,786 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.68 vs. limit=2.0 2022-12-23 18:23:17,351 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0001-146967-0_sp0.9 from training. Duration: 22.0166875 2022-12-23 18:23:32,330 INFO [train.py:894] (3/4) Epoch 22, batch 1900, loss[loss=0.1658, simple_loss=0.2528, pruned_loss=0.03944, over 18390.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.268, pruned_loss=0.0496, over 3713717.11 frames. ], batch size: 53, lr: 5.21e-03, grad_scale: 16.0 2022-12-23 18:23:33,783 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp1.1 from training. Duration: 24.395375 2022-12-23 18:23:41,119 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp0.9 from training. Duration: 27.47775 2022-12-23 18:23:45,186 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp0.9 from training. Duration: 24.8833125 2022-12-23 18:23:47,956 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0 from training. Duration: 23.39 2022-12-23 18:23:48,425 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.1491, 2.4188, 1.8964, 2.7756, 2.2985, 2.5206, 2.5194, 3.0523], device='cuda:3'), covar=tensor([0.1847, 0.3065, 0.1967, 0.2996, 0.3573, 0.0995, 0.2977, 0.0920], device='cuda:3'), in_proj_covar=tensor([0.0292, 0.0289, 0.0243, 0.0345, 0.0270, 0.0226, 0.0285, 0.0214], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 18:23:49,338 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.656e+02 4.293e+02 5.268e+02 6.591e+02 1.165e+03, threshold=1.054e+03, percent-clipped=3.0 2022-12-23 18:23:53,674 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp0.9 from training. Duration: 28.72225 2022-12-23 18:24:03,165 WARNING [train.py:1060] (3/4) Exclude cut with ID 585-294811-0110-133686-0_sp0.9 from training. Duration: 20.8944375 2022-12-23 18:24:20,435 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0_sp0.9 from training. Duration: 23.8444375 2022-12-23 18:24:26,566 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75565.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 18:24:43,019 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0 from training. Duration: 25.85 2022-12-23 18:24:43,027 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0 from training. Duration: 21.39 2022-12-23 18:24:47,508 INFO [train.py:894] (3/4) Epoch 22, batch 1950, loss[loss=0.2017, simple_loss=0.2849, pruned_loss=0.05923, over 18667.00 frames. ], tot_loss[loss=0.184, simple_loss=0.268, pruned_loss=0.04998, over 3713837.45 frames. ], batch size: 62, lr: 5.21e-03, grad_scale: 16.0 2022-12-23 18:24:53,759 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0 from training. Duration: 27.92 2022-12-23 18:24:55,329 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75584.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 18:25:23,657 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0039-130165-0_sp0.9 from training. Duration: 20.661125 2022-12-23 18:25:40,003 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=75613.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 18:25:45,907 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0043-15874-0_sp0.9 from training. Duration: 20.07225 2022-12-23 18:25:50,807 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75620.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 18:25:53,467 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0 from training. Duration: 21.01 2022-12-23 18:26:00,013 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75626.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 18:26:04,785 INFO [train.py:894] (3/4) Epoch 22, batch 2000, loss[loss=0.1934, simple_loss=0.2714, pruned_loss=0.05769, over 18567.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2676, pruned_loss=0.05054, over 3714037.70 frames. ], batch size: 57, lr: 5.21e-03, grad_scale: 16.0 2022-12-23 18:26:23,206 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.177e+02 4.476e+02 5.612e+02 7.026e+02 1.347e+03, threshold=1.122e+03, percent-clipped=1.0 2022-12-23 18:27:03,401 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0 from training. Duration: 20.65 2022-12-23 18:27:04,176 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=75668.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 18:27:11,607 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0 from training. Duration: 21.46 2022-12-23 18:27:20,877 INFO [train.py:894] (3/4) Epoch 22, batch 2050, loss[loss=0.166, simple_loss=0.2401, pruned_loss=0.04589, over 18417.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2665, pruned_loss=0.05045, over 3713304.79 frames. ], batch size: 42, lr: 5.21e-03, grad_scale: 16.0 2022-12-23 18:27:26,932 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([5.6493, 4.7894, 4.9455, 5.6225, 5.1948, 5.0146, 5.7290, 1.7531], device='cuda:3'), covar=tensor([0.0625, 0.0704, 0.0577, 0.0684, 0.1357, 0.1100, 0.0458, 0.5144], device='cuda:3'), in_proj_covar=tensor([0.0345, 0.0227, 0.0237, 0.0273, 0.0327, 0.0269, 0.0290, 0.0285], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 18:27:56,744 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0 from training. Duration: 0.92 2022-12-23 18:28:03,722 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0_sp0.9 from training. Duration: 23.7666875 2022-12-23 18:28:34,391 INFO [train.py:894] (3/4) Epoch 22, batch 2100, loss[loss=0.1728, simple_loss=0.2534, pruned_loss=0.04611, over 18527.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2672, pruned_loss=0.05109, over 3714011.73 frames. ], batch size: 47, lr: 5.20e-03, grad_scale: 16.0 2022-12-23 18:28:40,316 WARNING [train.py:1060] (3/4) Exclude cut with ID 8544-281189-0060-101339-0_sp0.9 from training. Duration: 20.861125 2022-12-23 18:28:53,129 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.853e+02 4.059e+02 5.294e+02 7.011e+02 1.474e+03, threshold=1.059e+03, percent-clipped=5.0 2022-12-23 18:28:53,185 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0_sp0.9 from training. Duration: 22.711125 2022-12-23 18:28:59,590 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4766, 1.4121, 1.3072, 1.4562, 1.7397, 1.5770, 1.5676, 1.1740], device='cuda:3'), covar=tensor([0.0272, 0.0220, 0.0489, 0.0178, 0.0184, 0.0376, 0.0250, 0.0293], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0127, 0.0153, 0.0124, 0.0116, 0.0121, 0.0098, 0.0127], device='cuda:3'), out_proj_covar=tensor([7.4724e-05, 1.0044e-04, 1.2563e-04, 9.8832e-05, 9.3765e-05, 9.3314e-05, 7.6560e-05, 1.0054e-04], device='cuda:3') 2022-12-23 18:29:34,875 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp1.1 from training. Duration: 22.986375 2022-12-23 18:29:50,874 INFO [train.py:894] (3/4) Epoch 22, batch 2150, loss[loss=0.1816, simple_loss=0.2608, pruned_loss=0.05118, over 18677.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2662, pruned_loss=0.05066, over 3714365.79 frames. ], batch size: 50, lr: 5.20e-03, grad_scale: 8.0 2022-12-23 18:29:52,313 WARNING [train.py:1060] (3/4) Exclude cut with ID 8040-260924-0003-80960-0_sp0.9 from training. Duration: 22.07225 2022-12-23 18:29:55,876 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.86 vs. limit=5.0 2022-12-23 18:29:56,509 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0045-26330-0_sp0.9 from training. Duration: 20.3055625 2022-12-23 18:29:57,848 WARNING [train.py:1060] (3/4) Exclude cut with ID 6356-271890-0060-94317-0_sp0.9 from training. Duration: 20.72225 2022-12-23 18:30:18,101 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp1.1 from training. Duration: 22.4818125 2022-12-23 18:30:45,444 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp0.9 from training. Duration: 25.0944375 2022-12-23 18:30:49,961 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0 from training. Duration: 21.515 2022-12-23 18:30:56,355 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp0.9 from training. Duration: 27.02225 2022-12-23 18:31:00,804 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0010-62480-0_sp0.9 from training. Duration: 22.22225 2022-12-23 18:31:06,951 INFO [train.py:894] (3/4) Epoch 22, batch 2200, loss[loss=0.2067, simple_loss=0.2862, pruned_loss=0.06362, over 18582.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2658, pruned_loss=0.05096, over 3713896.66 frames. ], batch size: 98, lr: 5.20e-03, grad_scale: 8.0 2022-12-23 18:31:08,385 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0085-44554-0_sp0.9 from training. Duration: 20.85 2022-12-23 18:31:27,607 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.539e+02 4.148e+02 5.070e+02 6.126e+02 9.714e+02, threshold=1.014e+03, percent-clipped=0.0 2022-12-23 18:31:41,022 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0 from training. Duration: 21.54 2022-12-23 18:31:44,189 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp1.1 from training. Duration: 20.5318125 2022-12-23 18:31:54,949 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0012-134311-0_sp0.9 from training. Duration: 21.9333125 2022-12-23 18:32:22,615 INFO [train.py:894] (3/4) Epoch 22, batch 2250, loss[loss=0.2045, simple_loss=0.2825, pruned_loss=0.06324, over 18700.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2661, pruned_loss=0.05118, over 3713715.94 frames. ], batch size: 97, lr: 5.20e-03, grad_scale: 8.0 2022-12-23 18:32:31,521 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75884.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 18:32:40,680 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8892, 0.7590, 1.6743, 1.4738, 1.8950, 1.8997, 1.5328, 1.6842], device='cuda:3'), covar=tensor([0.2173, 0.3327, 0.2621, 0.2699, 0.1962, 0.0990, 0.2946, 0.1344], device='cuda:3'), in_proj_covar=tensor([0.0269, 0.0299, 0.0280, 0.0318, 0.0308, 0.0252, 0.0344, 0.0242], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 18:32:46,313 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0025-130151-0_sp0.9 from training. Duration: 21.7944375 2022-12-23 18:32:58,616 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0_sp0.9 from training. Duration: 22.4666875 2022-12-23 18:33:06,169 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0 from training. Duration: 21.635 2022-12-23 18:33:12,494 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0_sp0.9 from training. Duration: 24.038875 2022-12-23 18:33:36,560 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75926.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 18:33:41,667 INFO [train.py:894] (3/4) Epoch 22, batch 2300, loss[loss=0.1991, simple_loss=0.2754, pruned_loss=0.06137, over 18578.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2673, pruned_loss=0.05195, over 3714626.37 frames. ], batch size: 176, lr: 5.20e-03, grad_scale: 8.0 2022-12-23 18:33:46,492 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=75932.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 18:33:48,697 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.47 vs. limit=5.0 2022-12-23 18:33:55,239 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp1.1 from training. Duration: 21.786375 2022-12-23 18:33:55,747 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5158, 2.0964, 1.6705, 2.2489, 1.9018, 2.0075, 1.9090, 2.4867], device='cuda:3'), covar=tensor([0.2081, 0.3216, 0.1957, 0.2841, 0.3582, 0.1119, 0.3257, 0.0929], device='cuda:3'), in_proj_covar=tensor([0.0293, 0.0290, 0.0243, 0.0346, 0.0270, 0.0227, 0.0287, 0.0215], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 18:34:01,093 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.457e+02 4.223e+02 4.994e+02 5.710e+02 1.631e+03, threshold=9.988e+02, percent-clipped=1.0 2022-12-23 18:34:05,468 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0 from training. Duration: 20.22 2022-12-23 18:34:22,807 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6944, 1.4110, 1.4577, 1.8418, 1.7941, 3.1903, 1.4630, 1.4333], device='cuda:3'), covar=tensor([0.0839, 0.1849, 0.1153, 0.0920, 0.1448, 0.0301, 0.1413, 0.1642], device='cuda:3'), in_proj_covar=tensor([0.0072, 0.0082, 0.0072, 0.0074, 0.0092, 0.0076, 0.0085, 0.0078], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 18:34:50,788 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=75974.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 18:34:57,500 INFO [train.py:894] (3/4) Epoch 22, batch 2350, loss[loss=0.1555, simple_loss=0.2363, pruned_loss=0.03737, over 18552.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2669, pruned_loss=0.05159, over 3714711.72 frames. ], batch size: 44, lr: 5.20e-03, grad_scale: 8.0 2022-12-23 18:36:03,274 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2022-12-23 18:36:10,261 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0 from training. Duration: 25.285 2022-12-23 18:36:17,374 INFO [train.py:894] (3/4) Epoch 22, batch 2400, loss[loss=0.1982, simple_loss=0.2736, pruned_loss=0.0614, over 18512.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2668, pruned_loss=0.05153, over 3714457.29 frames. ], batch size: 52, lr: 5.19e-03, grad_scale: 8.0 2022-12-23 18:36:36,132 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.436e+02 4.375e+02 5.426e+02 6.772e+02 1.323e+03, threshold=1.085e+03, percent-clipped=1.0 2022-12-23 18:37:13,570 WARNING [train.py:1060] (3/4) Exclude cut with ID 811-130148-0001-63453-0_sp0.9 from training. Duration: 20.861125 2022-12-23 18:37:32,744 INFO [train.py:894] (3/4) Epoch 22, batch 2450, loss[loss=0.1966, simple_loss=0.2808, pruned_loss=0.05617, over 18682.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2664, pruned_loss=0.05139, over 3714594.52 frames. ], batch size: 60, lr: 5.19e-03, grad_scale: 8.0 2022-12-23 18:37:37,594 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0 from training. Duration: 20.88 2022-12-23 18:38:10,792 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0_sp0.9 from training. Duration: 23.4166875 2022-12-23 18:38:47,655 INFO [train.py:894] (3/4) Epoch 22, batch 2500, loss[loss=0.1918, simple_loss=0.2816, pruned_loss=0.05095, over 18606.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2657, pruned_loss=0.05098, over 3714371.76 frames. ], batch size: 51, lr: 5.19e-03, grad_scale: 8.0 2022-12-23 18:38:48,137 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7633, 1.7229, 1.6324, 1.6141, 1.9316, 1.9965, 1.9726, 1.3740], device='cuda:3'), covar=tensor([0.0327, 0.0227, 0.0445, 0.0207, 0.0190, 0.0316, 0.0253, 0.0293], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0126, 0.0152, 0.0123, 0.0117, 0.0121, 0.0099, 0.0127], device='cuda:3'), out_proj_covar=tensor([7.5127e-05, 1.0018e-04, 1.2485e-04, 9.8088e-05, 9.4042e-05, 9.2891e-05, 7.6972e-05, 1.0044e-04], device='cuda:3') 2022-12-23 18:39:05,876 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.797e+02 4.308e+02 5.296e+02 6.059e+02 1.304e+03, threshold=1.059e+03, percent-clipped=2.0 2022-12-23 18:39:30,971 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0 from training. Duration: 21.24 2022-12-23 18:39:30,988 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0_sp0.9 from training. Duration: 23.9055625 2022-12-23 18:40:03,983 INFO [train.py:894] (3/4) Epoch 22, batch 2550, loss[loss=0.188, simple_loss=0.28, pruned_loss=0.04796, over 18560.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2659, pruned_loss=0.05111, over 3713520.91 frames. ], batch size: 57, lr: 5.19e-03, grad_scale: 8.0 2022-12-23 18:40:04,071 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp0.9 from training. Duration: 25.988875 2022-12-23 18:40:12,757 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0001-134300-0_sp0.9 from training. Duration: 20.67225 2022-12-23 18:40:17,482 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2040, 2.0752, 1.9464, 1.3186, 2.4472, 2.3116, 2.0300, 1.6767], device='cuda:3'), covar=tensor([0.0393, 0.0426, 0.0460, 0.0758, 0.0308, 0.0340, 0.0434, 0.0883], device='cuda:3'), in_proj_covar=tensor([0.0125, 0.0126, 0.0130, 0.0119, 0.0101, 0.0124, 0.0133, 0.0159], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 18:41:00,539 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0 from training. Duration: 20.34 2022-12-23 18:41:12,795 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.1827, 3.6389, 3.5712, 4.1008, 3.7981, 3.6821, 4.3078, 1.2173], device='cuda:3'), covar=tensor([0.0755, 0.0719, 0.0757, 0.0847, 0.1460, 0.1243, 0.0641, 0.5272], device='cuda:3'), in_proj_covar=tensor([0.0346, 0.0227, 0.0237, 0.0273, 0.0328, 0.0270, 0.0292, 0.0285], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 18:41:19,988 INFO [train.py:894] (3/4) Epoch 22, batch 2600, loss[loss=0.1638, simple_loss=0.2496, pruned_loss=0.03899, over 18658.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.265, pruned_loss=0.05079, over 3712838.92 frames. ], batch size: 48, lr: 5.19e-03, grad_scale: 8.0 2022-12-23 18:41:40,555 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.086e+02 4.163e+02 5.127e+02 6.183e+02 1.258e+03, threshold=1.025e+03, percent-clipped=3.0 2022-12-23 18:42:16,447 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp0.9 from training. Duration: 25.061125 2022-12-23 18:42:21,115 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76269.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 18:42:26,671 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0 from training. Duration: 0.83 2022-12-23 18:42:36,181 INFO [train.py:894] (3/4) Epoch 22, batch 2650, loss[loss=0.19, simple_loss=0.2742, pruned_loss=0.05283, over 18508.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2649, pruned_loss=0.05099, over 3712443.95 frames. ], batch size: 58, lr: 5.18e-03, grad_scale: 8.0 2022-12-23 18:42:51,409 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0 from training. Duration: 24.73 2022-12-23 18:43:00,269 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76294.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 18:43:04,009 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0 from training. Duration: 23.965 2022-12-23 18:43:12,762 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0030-146996-0_sp0.9 from training. Duration: 22.088875 2022-12-23 18:43:26,126 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8810, 1.8248, 2.2046, 1.3226, 2.2080, 2.3239, 1.6150, 2.6046], device='cuda:3'), covar=tensor([0.1212, 0.1955, 0.1245, 0.1946, 0.0727, 0.1054, 0.2322, 0.0528], device='cuda:3'), in_proj_covar=tensor([0.0200, 0.0214, 0.0209, 0.0197, 0.0176, 0.0219, 0.0217, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 18:43:28,549 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0_sp0.9 from training. Duration: 23.6 2022-12-23 18:43:44,033 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76323.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 18:43:52,452 INFO [train.py:894] (3/4) Epoch 22, batch 2700, loss[loss=0.1674, simple_loss=0.2564, pruned_loss=0.03919, over 18456.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2654, pruned_loss=0.05119, over 3713134.35 frames. ], batch size: 54, lr: 5.18e-03, grad_scale: 8.0 2022-12-23 18:43:54,855 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76330.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 18:44:13,386 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.173e+02 4.323e+02 5.325e+02 6.564e+02 1.409e+03, threshold=1.065e+03, percent-clipped=3.0 2022-12-23 18:44:34,742 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76355.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 18:45:10,779 INFO [train.py:894] (3/4) Epoch 22, batch 2750, loss[loss=0.1659, simple_loss=0.2488, pruned_loss=0.04154, over 18683.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2649, pruned_loss=0.05095, over 3713481.78 frames. ], batch size: 41, lr: 5.18e-03, grad_scale: 8.0 2022-12-23 18:45:10,811 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0 from training. Duration: 23.795 2022-12-23 18:45:19,013 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76384.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 18:45:27,246 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp1.1 from training. Duration: 21.5409375 2022-12-23 18:45:30,205 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp0.9 from training. Duration: 24.97775 2022-12-23 18:45:43,363 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0_sp0.9 from training. Duration: 23.3444375 2022-12-23 18:45:57,447 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([5.9161, 4.9864, 5.1806, 5.8763, 5.4782, 5.1792, 5.9612, 1.6760], device='cuda:3'), covar=tensor([0.0538, 0.0638, 0.0554, 0.0668, 0.1131, 0.1178, 0.0453, 0.5313], device='cuda:3'), in_proj_covar=tensor([0.0347, 0.0227, 0.0238, 0.0273, 0.0327, 0.0271, 0.0293, 0.0286], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 18:46:06,743 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5515, 1.4755, 1.5308, 1.5453, 1.0269, 3.0103, 1.1793, 1.7296], device='cuda:3'), covar=tensor([0.3172, 0.2126, 0.2034, 0.2101, 0.1557, 0.0252, 0.1734, 0.0884], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0118, 0.0125, 0.0121, 0.0104, 0.0096, 0.0091, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 18:46:09,316 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0_sp0.9 from training. Duration: 23.2 2022-12-23 18:46:15,897 WARNING [train.py:1060] (3/4) Exclude cut with ID 5653-46179-0060-117930-0_sp0.9 from training. Duration: 21.17225 2022-12-23 18:46:27,838 INFO [train.py:894] (3/4) Epoch 22, batch 2800, loss[loss=0.1888, simple_loss=0.2748, pruned_loss=0.05137, over 18490.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2654, pruned_loss=0.05112, over 3713347.82 frames. ], batch size: 52, lr: 5.18e-03, grad_scale: 8.0 2022-12-23 18:46:36,514 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp0.9 from training. Duration: 24.6555625 2022-12-23 18:46:47,869 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.459e+02 4.216e+02 4.683e+02 5.845e+02 1.389e+03, threshold=9.366e+02, percent-clipped=1.0 2022-12-23 18:47:31,199 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0 from training. Duration: 20.44 2022-12-23 18:47:43,343 INFO [train.py:894] (3/4) Epoch 22, batch 2850, loss[loss=0.2155, simple_loss=0.2907, pruned_loss=0.07018, over 18445.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2651, pruned_loss=0.05103, over 3713128.50 frames. ], batch size: 54, lr: 5.18e-03, grad_scale: 8.0 2022-12-23 18:47:47,927 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0_sp0.9 from training. Duration: 23.45 2022-12-23 18:48:19,988 WARNING [train.py:1060] (3/4) Exclude cut with ID 6945-60535-0076-12784-0_sp0.9 from training. Duration: 20.52225 2022-12-23 18:48:26,982 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0 from training. Duration: 22.19 2022-12-23 18:48:37,763 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp1.1 from training. Duration: 25.3818125 2022-12-23 18:48:41,285 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0964, 2.0443, 2.3208, 1.7151, 2.2709, 2.3589, 1.8940, 2.4698], device='cuda:3'), covar=tensor([0.0928, 0.1326, 0.1226, 0.1421, 0.0613, 0.0832, 0.1628, 0.0474], device='cuda:3'), in_proj_covar=tensor([0.0198, 0.0212, 0.0207, 0.0195, 0.0175, 0.0216, 0.0214, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 18:48:53,375 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp0.9 from training. Duration: 28.0944375 2022-12-23 18:48:59,559 INFO [train.py:894] (3/4) Epoch 22, batch 2900, loss[loss=0.2185, simple_loss=0.2912, pruned_loss=0.0729, over 18543.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2656, pruned_loss=0.05167, over 3714500.39 frames. ], batch size: 177, lr: 5.18e-03, grad_scale: 8.0 2022-12-23 18:48:59,670 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0_sp0.9 from training. Duration: 22.9444375 2022-12-23 18:49:08,599 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp1.1 from training. Duration: 21.6318125 2022-12-23 18:49:20,302 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.306e+02 4.171e+02 5.104e+02 6.624e+02 1.059e+03, threshold=1.021e+03, percent-clipped=3.0 2022-12-23 18:49:28,802 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0 from training. Duration: 23.695 2022-12-23 18:49:54,599 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0 from training. Duration: 23.955 2022-12-23 18:49:58,427 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5064, 1.3780, 1.4537, 1.3733, 0.7854, 2.2596, 0.9043, 1.3860], device='cuda:3'), covar=tensor([0.3180, 0.2192, 0.2027, 0.2124, 0.1618, 0.0340, 0.1729, 0.0913], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0118, 0.0125, 0.0121, 0.0104, 0.0096, 0.0091, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 18:50:11,707 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5025, 1.4274, 1.3450, 1.6401, 1.7214, 2.7445, 1.5033, 1.5611], device='cuda:3'), covar=tensor([0.0857, 0.1715, 0.1164, 0.0880, 0.1313, 0.0348, 0.1244, 0.1410], device='cuda:3'), in_proj_covar=tensor([0.0072, 0.0082, 0.0072, 0.0074, 0.0091, 0.0075, 0.0085, 0.0078], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 18:50:15,596 INFO [train.py:894] (3/4) Epoch 22, batch 2950, loss[loss=0.1677, simple_loss=0.2501, pruned_loss=0.04269, over 18681.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2653, pruned_loss=0.05127, over 3714108.59 frames. ], batch size: 48, lr: 5.17e-03, grad_scale: 8.0 2022-12-23 18:50:28,620 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp0.9 from training. Duration: 26.438875 2022-12-23 18:51:12,832 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0021-26306-0_sp0.9 from training. Duration: 21.2444375 2022-12-23 18:51:12,874 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp0.9 from training. Duration: 31.02225 2022-12-23 18:51:23,327 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0 from training. Duration: 22.395 2022-12-23 18:51:25,404 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.56 vs. limit=2.0 2022-12-23 18:51:27,905 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76625.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 18:51:34,304 INFO [train.py:894] (3/4) Epoch 22, batch 3000, loss[loss=0.1948, simple_loss=0.275, pruned_loss=0.05732, over 18576.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.265, pruned_loss=0.05087, over 3714303.50 frames. ], batch size: 49, lr: 5.17e-03, grad_scale: 8.0 2022-12-23 18:51:34,305 INFO [train.py:919] (3/4) Computing validation loss 2022-12-23 18:51:45,374 INFO [train.py:928] (3/4) Epoch 22, validation: loss=0.1633, simple_loss=0.2609, pruned_loss=0.03279, over 944034.00 frames. 2022-12-23 18:51:45,375 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-23 18:51:51,308 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0 from training. Duration: 21.075 2022-12-23 18:51:57,360 WARNING [train.py:1060] (3/4) Exclude cut with ID 6482-98857-0025-147532-0_sp0.9 from training. Duration: 20.0055625 2022-12-23 18:51:57,372 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0037-132304-0_sp0.9 from training. Duration: 22.05 2022-12-23 18:51:57,386 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0 from training. Duration: 26.8349375 2022-12-23 18:52:01,015 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp1.1 from training. Duration: 22.1090625 2022-12-23 18:52:05,253 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.086e+02 4.293e+02 5.248e+02 6.560e+02 1.696e+03, threshold=1.050e+03, percent-clipped=5.0 2022-12-23 18:52:08,362 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp0.9 from training. Duration: 26.6166875 2022-12-23 18:52:18,134 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76650.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 18:52:26,827 WARNING [train.py:1060] (3/4) Exclude cut with ID 2046-178027-0000-53705-0_sp0.9 from training. Duration: 20.3055625 2022-12-23 18:52:43,994 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1220, 2.1799, 1.7675, 2.5047, 2.2685, 2.0062, 2.6514, 2.2184], device='cuda:3'), covar=tensor([0.0766, 0.1468, 0.2122, 0.1374, 0.1411, 0.0785, 0.0910, 0.1028], device='cuda:3'), in_proj_covar=tensor([0.0179, 0.0211, 0.0254, 0.0291, 0.0238, 0.0193, 0.0209, 0.0204], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 18:52:47,271 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.05 vs. limit=5.0 2022-12-23 18:52:48,308 WARNING [train.py:1060] (3/4) Exclude cut with ID 7205-50138-0008-5373-0_sp0.9 from training. Duration: 20.7 2022-12-23 18:53:02,170 INFO [train.py:894] (3/4) Epoch 22, batch 3050, loss[loss=0.1817, simple_loss=0.2673, pruned_loss=0.04804, over 18519.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2647, pruned_loss=0.0507, over 3714063.58 frames. ], batch size: 98, lr: 5.17e-03, grad_scale: 8.0 2022-12-23 18:53:02,354 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76679.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 18:53:28,728 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0 from training. Duration: 22.48 2022-12-23 18:53:46,459 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp0.9 from training. Duration: 29.816625 2022-12-23 18:54:04,750 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp1.1 from training. Duration: 22.7590625 2022-12-23 18:54:08,950 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0 from training. Duration: 22.555 2022-12-23 18:54:19,693 INFO [train.py:894] (3/4) Epoch 22, batch 3100, loss[loss=0.1616, simple_loss=0.2331, pruned_loss=0.04508, over 18490.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2651, pruned_loss=0.05111, over 3714339.89 frames. ], batch size: 43, lr: 5.17e-03, grad_scale: 8.0 2022-12-23 18:54:31,505 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0005-25975-0_sp0.9 from training. Duration: 21.688875 2022-12-23 18:54:39,574 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.526e+02 4.110e+02 5.190e+02 6.058e+02 1.141e+03, threshold=1.038e+03, percent-clipped=2.0 2022-12-23 18:55:07,025 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0_sp0.9 from training. Duration: 22.6 2022-12-23 18:55:27,255 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2022-12-23 18:55:35,648 INFO [train.py:894] (3/4) Epoch 22, batch 3150, loss[loss=0.1821, simple_loss=0.2678, pruned_loss=0.04822, over 18544.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2643, pruned_loss=0.05044, over 3715920.57 frames. ], batch size: 55, lr: 5.17e-03, grad_scale: 8.0 2022-12-23 18:55:46,401 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0 from training. Duration: 24.32 2022-12-23 18:56:44,990 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-276745-0093-13116-0_sp0.9 from training. Duration: 21.061125 2022-12-23 18:56:50,985 INFO [train.py:894] (3/4) Epoch 22, batch 3200, loss[loss=0.169, simple_loss=0.2536, pruned_loss=0.04226, over 18579.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2643, pruned_loss=0.05034, over 3715369.44 frames. ], batch size: 57, lr: 5.17e-03, grad_scale: 8.0 2022-12-23 18:56:58,987 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0024-15855-0_sp0.9 from training. Duration: 20.32225 2022-12-23 18:57:06,660 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76839.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 18:57:11,039 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.867e+02 4.150e+02 5.402e+02 6.828e+02 2.211e+03, threshold=1.080e+03, percent-clipped=5.0 2022-12-23 18:57:11,097 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp1.1 from training. Duration: 0.7545625 2022-12-23 18:57:25,232 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0_sp0.9 from training. Duration: 23.9333125 2022-12-23 18:57:36,013 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4828, 1.2040, 0.7390, 1.2647, 1.9373, 0.7723, 1.2591, 1.4354], device='cuda:3'), covar=tensor([0.1755, 0.2086, 0.1947, 0.1506, 0.1909, 0.1770, 0.1547, 0.1728], device='cuda:3'), in_proj_covar=tensor([0.0095, 0.0098, 0.0117, 0.0097, 0.0119, 0.0092, 0.0098, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 18:57:58,572 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp1.1 from training. Duration: 20.17275 2022-12-23 18:58:04,304 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp1.1 from training. Duration: 20.436375 2022-12-23 18:58:07,838 INFO [train.py:894] (3/4) Epoch 22, batch 3250, loss[loss=0.185, simple_loss=0.2686, pruned_loss=0.05071, over 18733.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2644, pruned_loss=0.05029, over 3714096.14 frames. ], batch size: 52, lr: 5.16e-03, grad_scale: 8.0 2022-12-23 18:58:40,893 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76900.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 18:58:42,293 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5542, 1.1389, 1.9514, 2.9298, 2.1268, 2.5012, 0.8888, 2.1081], device='cuda:3'), covar=tensor([0.1915, 0.1774, 0.1403, 0.0728, 0.1087, 0.1141, 0.2208, 0.1067], device='cuda:3'), in_proj_covar=tensor([0.0103, 0.0118, 0.0137, 0.0153, 0.0106, 0.0143, 0.0130, 0.0115], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2022-12-23 18:59:19,254 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76925.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 18:59:24,928 INFO [train.py:894] (3/4) Epoch 22, batch 3300, loss[loss=0.164, simple_loss=0.2461, pruned_loss=0.04101, over 18682.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2648, pruned_loss=0.05006, over 3714266.87 frames. ], batch size: 48, lr: 5.16e-03, grad_scale: 8.0 2022-12-23 18:59:26,505 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0_sp0.9 from training. Duration: 23.1055625 2022-12-23 18:59:28,087 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp1.1 from training. Duration: 23.5 2022-12-23 18:59:38,793 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp0.9 from training. Duration: 26.62775 2022-12-23 18:59:44,629 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.429e+02 4.022e+02 4.936e+02 5.959e+02 1.401e+03, threshold=9.873e+02, percent-clipped=1.0 2022-12-23 18:59:52,357 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0 from training. Duration: 21.105 2022-12-23 18:59:55,428 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0_sp0.9 from training. Duration: 24.411125 2022-12-23 18:59:56,973 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76950.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:00:23,307 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp1.1 from training. Duration: 21.263625 2022-12-23 19:00:31,537 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=76973.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 19:00:40,034 INFO [train.py:894] (3/4) Epoch 22, batch 3350, loss[loss=0.1694, simple_loss=0.2465, pruned_loss=0.04613, over 18535.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2649, pruned_loss=0.05004, over 3714289.00 frames. ], batch size: 44, lr: 5.16e-03, grad_scale: 8.0 2022-12-23 19:00:40,297 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76979.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 19:00:51,276 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3261, 1.4814, 1.2483, 1.7700, 1.7508, 1.4808, 1.0426, 1.2315], device='cuda:3'), covar=tensor([0.1918, 0.1778, 0.1674, 0.1042, 0.1146, 0.1058, 0.2165, 0.1517], device='cuda:3'), in_proj_covar=tensor([0.0247, 0.0227, 0.0215, 0.0199, 0.0259, 0.0196, 0.0225, 0.0200], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 19:00:58,078 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0 from training. Duration: 20.795 2022-12-23 19:01:07,292 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0 from training. Duration: 24.76 2022-12-23 19:01:07,314 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0_sp0.9 from training. Duration: 22.25 2022-12-23 19:01:08,563 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=76998.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:01:32,822 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp1.1 from training. Duration: 20.5045625 2022-12-23 19:01:53,154 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=77027.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 19:01:55,609 INFO [train.py:894] (3/4) Epoch 22, batch 3400, loss[loss=0.2102, simple_loss=0.2876, pruned_loss=0.06645, over 18612.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.264, pruned_loss=0.04976, over 3714633.63 frames. ], batch size: 78, lr: 5.16e-03, grad_scale: 8.0 2022-12-23 19:02:16,429 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.658e+02 4.239e+02 5.278e+02 6.844e+02 1.140e+03, threshold=1.056e+03, percent-clipped=4.0 2022-12-23 19:02:38,250 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5901, 1.5513, 1.6594, 1.6287, 1.2009, 3.8091, 1.5750, 2.0283], device='cuda:3'), covar=tensor([0.3344, 0.2267, 0.2043, 0.2215, 0.1637, 0.0206, 0.1637, 0.0926], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0117, 0.0125, 0.0120, 0.0104, 0.0096, 0.0091, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 19:03:08,435 INFO [train.py:894] (3/4) Epoch 22, batch 3450, loss[loss=0.1993, simple_loss=0.2784, pruned_loss=0.06011, over 18603.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2622, pruned_loss=0.04887, over 3714487.98 frames. ], batch size: 51, lr: 5.16e-03, grad_scale: 8.0 2022-12-23 19:03:32,300 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77095.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:04:08,128 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2022-12-23 19:04:23,452 INFO [train.py:894] (3/4) Epoch 22, batch 3500, loss[loss=0.2257, simple_loss=0.2934, pruned_loss=0.07905, over 18649.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2622, pruned_loss=0.04933, over 3714492.63 frames. ], batch size: 178, lr: 5.16e-03, grad_scale: 8.0 2022-12-23 19:04:45,127 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp1.1 from training. Duration: 22.2954375 2022-12-23 19:04:56,375 INFO [train.py:894] (3/4) Epoch 23, batch 0, loss[loss=0.2186, simple_loss=0.2966, pruned_loss=0.0703, over 18604.00 frames. ], tot_loss[loss=0.2186, simple_loss=0.2966, pruned_loss=0.0703, over 18604.00 frames. ], batch size: 173, lr: 5.04e-03, grad_scale: 8.0 2022-12-23 19:04:56,375 INFO [train.py:919] (3/4) Computing validation loss 2022-12-23 19:05:07,270 INFO [train.py:928] (3/4) Epoch 23, validation: loss=0.1628, simple_loss=0.2608, pruned_loss=0.03242, over 944034.00 frames. 2022-12-23 19:05:07,271 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-23 19:05:17,281 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.568e+02 4.089e+02 5.198e+02 6.379e+02 2.209e+03, threshold=1.040e+03, percent-clipped=4.0 2022-12-23 19:05:38,581 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77156.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 19:05:55,699 WARNING [train.py:1060] (3/4) Exclude cut with ID 298-126791-0067-24026-0_sp0.9 from training. Duration: 21.438875 2022-12-23 19:06:01,465 WARNING [train.py:1060] (3/4) Exclude cut with ID 5652-39938-0025-23684-0_sp0.9 from training. Duration: 22.2055625 2022-12-23 19:06:21,065 INFO [train.py:894] (3/4) Epoch 23, batch 50, loss[loss=0.17, simple_loss=0.2586, pruned_loss=0.04067, over 18454.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2644, pruned_loss=0.04307, over 838234.28 frames. ], batch size: 50, lr: 5.04e-03, grad_scale: 8.0 2022-12-23 19:06:24,832 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2022-12-23 19:06:35,669 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77195.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:07:01,049 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77211.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:07:35,095 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3401, 2.0852, 2.5887, 2.7163, 2.5272, 3.7780, 2.1719, 2.1305], device='cuda:3'), covar=tensor([0.0733, 0.1435, 0.1003, 0.0728, 0.1096, 0.0265, 0.1098, 0.1282], device='cuda:3'), in_proj_covar=tensor([0.0072, 0.0081, 0.0071, 0.0074, 0.0090, 0.0075, 0.0084, 0.0077], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 19:07:37,913 INFO [train.py:894] (3/4) Epoch 23, batch 100, loss[loss=0.149, simple_loss=0.2276, pruned_loss=0.03525, over 18602.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2605, pruned_loss=0.04167, over 1476895.43 frames. ], batch size: 41, lr: 5.04e-03, grad_scale: 8.0 2022-12-23 19:07:48,677 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.031e+02 3.278e+02 3.906e+02 4.585e+02 8.522e+02, threshold=7.812e+02, percent-clipped=0.0 2022-12-23 19:07:59,000 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0799, 1.2982, 1.8021, 1.6948, 2.1136, 2.1281, 1.8315, 1.7569], device='cuda:3'), covar=tensor([0.2393, 0.3591, 0.2743, 0.2945, 0.2086, 0.0986, 0.3522, 0.1409], device='cuda:3'), in_proj_covar=tensor([0.0270, 0.0301, 0.0282, 0.0320, 0.0312, 0.0255, 0.0346, 0.0244], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 19:08:20,298 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.97 vs. limit=5.0 2022-12-23 19:08:34,877 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77272.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:08:36,925 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77273.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:08:42,366 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77277.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:08:53,297 INFO [train.py:894] (3/4) Epoch 23, batch 150, loss[loss=0.2203, simple_loss=0.3053, pruned_loss=0.06762, over 18582.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2615, pruned_loss=0.04246, over 1972257.46 frames. ], batch size: 56, lr: 5.04e-03, grad_scale: 8.0 2022-12-23 19:09:04,349 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0 from training. Duration: 24.525 2022-12-23 19:09:23,703 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2022-12-23 19:09:37,444 WARNING [train.py:1060] (3/4) Exclude cut with ID 3699-47246-0007-3408-0_sp0.9 from training. Duration: 20.26675 2022-12-23 19:09:50,847 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp0.9 from training. Duration: 27.25 2022-12-23 19:10:08,181 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77334.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:10:09,218 INFO [train.py:894] (3/4) Epoch 23, batch 200, loss[loss=0.1914, simple_loss=0.2806, pruned_loss=0.05111, over 18587.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2599, pruned_loss=0.04169, over 2358883.09 frames. ], batch size: 57, lr: 5.03e-03, grad_scale: 8.0 2022-12-23 19:10:14,907 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77338.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:10:21,079 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.110e+02 3.066e+02 3.714e+02 4.513e+02 7.950e+02, threshold=7.428e+02, percent-clipped=1.0 2022-12-23 19:10:25,759 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5296, 2.5510, 1.7673, 3.2672, 2.8773, 2.3814, 3.7529, 2.5055], device='cuda:3'), covar=tensor([0.0775, 0.1876, 0.2778, 0.1629, 0.1693, 0.0892, 0.0797, 0.1282], device='cuda:3'), in_proj_covar=tensor([0.0177, 0.0210, 0.0252, 0.0290, 0.0238, 0.0192, 0.0207, 0.0205], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 19:10:39,240 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3644, 2.4214, 2.9764, 1.4595, 2.8114, 2.8338, 1.8119, 3.1551], device='cuda:3'), covar=tensor([0.1482, 0.1982, 0.1234, 0.2302, 0.0963, 0.1494, 0.2317, 0.0670], device='cuda:3'), in_proj_covar=tensor([0.0199, 0.0214, 0.0209, 0.0196, 0.0177, 0.0217, 0.0216, 0.0194], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 19:11:06,931 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0 from training. Duration: 21.68 2022-12-23 19:11:18,510 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0 from training. Duration: 21.6300625 2022-12-23 19:11:20,918 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2022-12-23 19:11:24,636 INFO [train.py:894] (3/4) Epoch 23, batch 250, loss[loss=0.1657, simple_loss=0.2586, pruned_loss=0.0364, over 18592.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2579, pruned_loss=0.04051, over 2659106.93 frames. ], batch size: 51, lr: 5.03e-03, grad_scale: 8.0 2022-12-23 19:11:41,185 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0_sp0.9 from training. Duration: 24.033375 2022-12-23 19:12:39,046 INFO [train.py:894] (3/4) Epoch 23, batch 300, loss[loss=0.1445, simple_loss=0.2222, pruned_loss=0.03345, over 18471.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2572, pruned_loss=0.03987, over 2893536.24 frames. ], batch size: 43, lr: 5.03e-03, grad_scale: 8.0 2022-12-23 19:12:41,911 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0 from training. Duration: 22.905 2022-12-23 19:12:43,913 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp1.1 from training. Duration: 23.4318125 2022-12-23 19:12:49,726 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.384e+02 3.220e+02 3.955e+02 4.566e+02 1.223e+03, threshold=7.910e+02, percent-clipped=4.0 2022-12-23 19:13:02,828 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77451.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 19:13:09,121 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2022-12-23 19:13:53,143 INFO [train.py:894] (3/4) Epoch 23, batch 350, loss[loss=0.169, simple_loss=0.2655, pruned_loss=0.03629, over 18607.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2579, pruned_loss=0.04027, over 3075640.57 frames. ], batch size: 78, lr: 5.03e-03, grad_scale: 8.0 2022-12-23 19:14:09,203 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77495.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:14:36,026 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp1.1 from training. Duration: 20.82275 2022-12-23 19:14:36,072 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp0.9 from training. Duration: 25.45 2022-12-23 19:14:52,556 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2022-12-23 19:15:09,539 INFO [train.py:894] (3/4) Epoch 23, batch 400, loss[loss=0.1895, simple_loss=0.2752, pruned_loss=0.05188, over 18565.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2579, pruned_loss=0.04058, over 3217560.95 frames. ], batch size: 57, lr: 5.03e-03, grad_scale: 8.0 2022-12-23 19:15:20,430 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.174e+02 3.387e+02 4.125e+02 5.058e+02 9.204e+02, threshold=8.250e+02, percent-clipped=5.0 2022-12-23 19:15:22,082 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=77543.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:15:32,243 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0 from training. Duration: 25.775 2022-12-23 19:15:45,752 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.94 vs. limit=5.0 2022-12-23 19:15:54,480 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0_sp0.9 from training. Duration: 22.25 2022-12-23 19:15:57,092 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1588, 1.2803, 1.8527, 1.7422, 2.1785, 2.1539, 1.9185, 1.8170], device='cuda:3'), covar=tensor([0.2239, 0.3348, 0.2574, 0.2922, 0.1976, 0.0939, 0.3125, 0.1316], device='cuda:3'), in_proj_covar=tensor([0.0266, 0.0297, 0.0279, 0.0317, 0.0309, 0.0252, 0.0342, 0.0241], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 19:15:58,308 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77567.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:16:03,725 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2022-12-23 19:16:24,177 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0 from training. Duration: 26.205 2022-12-23 19:16:25,499 INFO [train.py:894] (3/4) Epoch 23, batch 450, loss[loss=0.1758, simple_loss=0.2573, pruned_loss=0.04713, over 18519.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2591, pruned_loss=0.04084, over 3327562.90 frames. ], batch size: 44, lr: 5.03e-03, grad_scale: 8.0 2022-12-23 19:16:38,997 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp0.9 from training. Duration: 30.1555625 2022-12-23 19:16:39,455 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1206, 2.2298, 1.4904, 2.7359, 2.5006, 2.1353, 3.1182, 2.1480], device='cuda:3'), covar=tensor([0.0793, 0.1690, 0.2809, 0.1590, 0.1571, 0.0861, 0.0801, 0.1250], device='cuda:3'), in_proj_covar=tensor([0.0177, 0.0209, 0.0251, 0.0288, 0.0237, 0.0191, 0.0206, 0.0204], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 19:16:44,918 WARNING [train.py:1060] (3/4) Exclude cut with ID 1265-135635-0050-6781-0_sp0.9 from training. Duration: 21.8333125 2022-12-23 19:16:53,752 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp1.1 from training. Duration: 20.6545625 2022-12-23 19:16:57,254 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6442, 4.1440, 3.9054, 1.8785, 4.1535, 3.2467, 0.8747, 2.7471], device='cuda:3'), covar=tensor([0.2118, 0.0986, 0.1270, 0.3352, 0.0793, 0.0851, 0.4946, 0.1427], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0142, 0.0160, 0.0125, 0.0146, 0.0116, 0.0146, 0.0116], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 19:17:12,059 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77616.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:17:31,630 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77629.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:17:35,995 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2022-12-23 19:17:37,738 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0045-39920-0_sp0.9 from training. Duration: 20.52225 2022-12-23 19:17:37,912 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77633.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:17:40,694 INFO [train.py:894] (3/4) Epoch 23, batch 500, loss[loss=0.177, simple_loss=0.264, pruned_loss=0.04498, over 18537.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2604, pruned_loss=0.04147, over 3414462.74 frames. ], batch size: 47, lr: 5.02e-03, grad_scale: 8.0 2022-12-23 19:17:48,827 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2022-12-23 19:17:50,549 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.063e+02 3.373e+02 4.171e+02 5.386e+02 1.486e+03, threshold=8.341e+02, percent-clipped=6.0 2022-12-23 19:17:55,429 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp0.9 from training. Duration: 29.1166875 2022-12-23 19:18:25,228 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2022-12-23 19:18:43,928 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77677.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:18:44,034 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77677.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:18:55,112 INFO [train.py:894] (3/4) Epoch 23, batch 550, loss[loss=0.1751, simple_loss=0.2656, pruned_loss=0.04229, over 18520.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2607, pruned_loss=0.04153, over 3480463.07 frames. ], batch size: 55, lr: 5.02e-03, grad_scale: 8.0 2022-12-23 19:18:55,182 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133211-0007-59831-0_sp0.9 from training. Duration: 21.388875 2022-12-23 19:19:31,896 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0 from training. Duration: 22.72 2022-12-23 19:19:31,940 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0_sp0.9 from training. Duration: 22.7444375 2022-12-23 19:20:09,201 INFO [train.py:894] (3/4) Epoch 23, batch 600, loss[loss=0.1656, simple_loss=0.261, pruned_loss=0.03514, over 18542.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2607, pruned_loss=0.04169, over 3532445.79 frames. ], batch size: 55, lr: 5.02e-03, grad_scale: 8.0 2022-12-23 19:20:14,117 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77738.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:20:16,306 WARNING [train.py:1060] (3/4) Exclude cut with ID 4133-6541-0027-40495-0_sp1.1 from training. Duration: 0.9681875 2022-12-23 19:20:19,169 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.410e+02 3.680e+02 4.304e+02 5.162e+02 1.245e+03, threshold=8.609e+02, percent-clipped=2.0 2022-12-23 19:20:19,232 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0_sp0.9 from training. Duration: 22.3166875 2022-12-23 19:20:24,751 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133212-0015-59917-0_sp0.9 from training. Duration: 21.8166875 2022-12-23 19:20:33,027 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77751.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 19:21:24,611 INFO [train.py:894] (3/4) Epoch 23, batch 650, loss[loss=0.1554, simple_loss=0.2357, pruned_loss=0.03755, over 18593.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2614, pruned_loss=0.04219, over 3572513.89 frames. ], batch size: 45, lr: 5.02e-03, grad_scale: 16.0 2022-12-23 19:21:45,782 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=77799.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:22:07,365 WARNING [train.py:1060] (3/4) Exclude cut with ID 4957-30119-0041-23990-0_sp0.9 from training. Duration: 20.22775 2022-12-23 19:22:40,276 INFO [train.py:894] (3/4) Epoch 23, batch 700, loss[loss=0.185, simple_loss=0.2721, pruned_loss=0.04892, over 18517.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2626, pruned_loss=0.04255, over 3604339.53 frames. ], batch size: 98, lr: 5.02e-03, grad_scale: 16.0 2022-12-23 19:22:50,403 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.138e+02 3.371e+02 3.749e+02 4.634e+02 8.391e+02, threshold=7.499e+02, percent-clipped=0.0 2022-12-23 19:22:50,483 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp1.1 from training. Duration: 24.67275 2022-12-23 19:23:19,160 WARNING [train.py:1060] (3/4) Exclude cut with ID 3082-165428-0081-50734-0_sp0.9 from training. Duration: 21.8055625 2022-12-23 19:23:27,881 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77867.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:23:42,353 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2022-12-23 19:23:54,213 INFO [train.py:894] (3/4) Epoch 23, batch 750, loss[loss=0.1656, simple_loss=0.265, pruned_loss=0.03305, over 18503.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2628, pruned_loss=0.04257, over 3629494.13 frames. ], batch size: 52, lr: 5.02e-03, grad_scale: 16.0 2022-12-23 19:23:55,753 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0_sp0.9 from training. Duration: 22.6666875 2022-12-23 19:24:39,712 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=77915.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:24:59,345 WARNING [train.py:1060] (3/4) Exclude cut with ID 2411-132532-0017-82279-0_sp1.1 from training. Duration: 0.9681875 2022-12-23 19:25:01,047 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77929.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:25:06,791 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77933.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:25:09,354 INFO [train.py:894] (3/4) Epoch 23, batch 800, loss[loss=0.1734, simple_loss=0.2717, pruned_loss=0.03758, over 18693.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2627, pruned_loss=0.04264, over 3647531.87 frames. ], batch size: 60, lr: 5.01e-03, grad_scale: 16.0 2022-12-23 19:25:19,140 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.500e+02 3.367e+02 3.962e+02 4.709e+02 1.041e+03, threshold=7.923e+02, percent-clipped=5.0 2022-12-23 19:25:25,817 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0 from training. Duration: 22.485 2022-12-23 19:25:59,326 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2022-12-23 19:26:01,089 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2022-12-23 19:26:05,079 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp1.1 from training. Duration: 23.82275 2022-12-23 19:26:05,203 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77972.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:26:12,811 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=77977.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:26:15,835 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7507, 5.3945, 4.7348, 3.0375, 5.4337, 4.2328, 1.0370, 3.4675], device='cuda:3'), covar=tensor([0.2143, 0.0835, 0.1259, 0.2587, 0.0566, 0.0683, 0.4802, 0.1251], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0141, 0.0158, 0.0123, 0.0144, 0.0114, 0.0143, 0.0114], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 19:26:18,636 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=77981.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:26:19,938 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0 from training. Duration: 20.77 2022-12-23 19:26:24,430 INFO [train.py:894] (3/4) Epoch 23, batch 850, loss[loss=0.2047, simple_loss=0.2849, pruned_loss=0.06219, over 18516.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2632, pruned_loss=0.04289, over 3662201.01 frames. ], batch size: 58, lr: 5.01e-03, grad_scale: 16.0 2022-12-23 19:26:27,361 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0_sp0.9 from training. Duration: 24.088875 2022-12-23 19:27:00,202 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp1.1 from training. Duration: 20.4409375 2022-12-23 19:27:39,988 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78033.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:27:42,668 INFO [train.py:894] (3/4) Epoch 23, batch 900, loss[loss=0.1763, simple_loss=0.2668, pruned_loss=0.04291, over 18595.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2638, pruned_loss=0.04267, over 3673264.35 frames. ], batch size: 51, lr: 5.01e-03, grad_scale: 16.0 2022-12-23 19:27:52,432 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.942e+02 3.202e+02 3.974e+02 4.965e+02 8.252e+02, threshold=7.949e+02, percent-clipped=1.0 2022-12-23 19:28:18,024 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0_sp0.9 from training. Duration: 22.511125 2022-12-23 19:28:18,048 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0 from training. Duration: 20.675 2022-12-23 19:28:39,408 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4517, 1.5480, 1.2878, 1.3666, 1.6449, 1.5501, 1.5684, 1.2952], device='cuda:3'), covar=tensor([0.0335, 0.0283, 0.0429, 0.0240, 0.0225, 0.0406, 0.0234, 0.0309], device='cuda:3'), in_proj_covar=tensor([0.0095, 0.0127, 0.0152, 0.0124, 0.0118, 0.0123, 0.0099, 0.0127], device='cuda:3'), out_proj_covar=tensor([7.5410e-05, 1.0083e-04, 1.2525e-04, 9.8568e-05, 9.4736e-05, 9.4295e-05, 7.7354e-05, 1.0046e-04], device='cuda:3') 2022-12-23 19:28:58,721 INFO [train.py:894] (3/4) Epoch 23, batch 950, loss[loss=0.1603, simple_loss=0.2432, pruned_loss=0.03867, over 18404.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2634, pruned_loss=0.04231, over 3682767.32 frames. ], batch size: 42, lr: 5.01e-03, grad_scale: 16.0 2022-12-23 19:29:57,245 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp0.9 from training. Duration: 24.9833125 2022-12-23 19:30:03,415 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0196, 1.9036, 1.5161, 1.6979, 1.7008, 1.8717, 1.7304, 1.7551], device='cuda:3'), covar=tensor([0.2285, 0.3004, 0.1982, 0.2406, 0.3432, 0.1081, 0.2843, 0.1088], device='cuda:3'), in_proj_covar=tensor([0.0296, 0.0294, 0.0248, 0.0348, 0.0275, 0.0230, 0.0290, 0.0216], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 19:30:13,047 INFO [train.py:894] (3/4) Epoch 23, batch 1000, loss[loss=0.1803, simple_loss=0.2795, pruned_loss=0.04058, over 18493.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2637, pruned_loss=0.04232, over 3689303.03 frames. ], batch size: 77, lr: 5.01e-03, grad_scale: 16.0 2022-12-23 19:30:23,694 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.248e+02 3.311e+02 3.940e+02 4.673e+02 6.833e+02, threshold=7.881e+02, percent-clipped=0.0 2022-12-23 19:30:28,176 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0 from training. Duration: 27.14 2022-12-23 19:30:44,251 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0 from training. Duration: 22.44 2022-12-23 19:31:27,920 INFO [train.py:894] (3/4) Epoch 23, batch 1050, loss[loss=0.1933, simple_loss=0.2818, pruned_loss=0.05241, over 18521.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2638, pruned_loss=0.0423, over 3695851.23 frames. ], batch size: 58, lr: 5.01e-03, grad_scale: 8.0 2022-12-23 19:31:34,790 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5565, 2.6677, 2.8415, 1.2689, 2.5717, 3.0890, 2.6701, 2.4769], device='cuda:3'), covar=tensor([0.0794, 0.0349, 0.0321, 0.0533, 0.0385, 0.0509, 0.0341, 0.0701], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0172, 0.0129, 0.0140, 0.0149, 0.0144, 0.0166, 0.0175], device='cuda:3'), out_proj_covar=tensor([1.1302e-04, 1.3053e-04, 9.5806e-05, 1.0341e-04, 1.1007e-04, 1.0934e-04, 1.2690e-04, 1.3254e-04], device='cuda:3') 2022-12-23 19:32:03,623 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0060-62364-0_sp0.9 from training. Duration: 21.361125 2022-12-23 19:32:10,764 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp1.1 from training. Duration: 27.0318125 2022-12-23 19:32:20,805 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp0.9 from training. Duration: 28.638875 2022-12-23 19:32:34,587 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0 from training. Duration: 20.4 2022-12-23 19:32:41,293 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78232.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:32:45,134 INFO [train.py:894] (3/4) Epoch 23, batch 1100, loss[loss=0.1858, simple_loss=0.283, pruned_loss=0.04437, over 18634.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2624, pruned_loss=0.04192, over 3700056.68 frames. ], batch size: 53, lr: 5.01e-03, grad_scale: 8.0 2022-12-23 19:32:56,856 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.120e+02 3.350e+02 4.234e+02 4.822e+02 9.215e+02, threshold=8.468e+02, percent-clipped=3.0 2022-12-23 19:33:04,633 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0 from training. Duration: 20.025 2022-12-23 19:33:05,957 WARNING [train.py:1060] (3/4) Exclude cut with ID 2364-131735-0112-64612-0_sp0.9 from training. Duration: 20.488875 2022-12-23 19:33:12,703 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0 from training. Duration: 29.735 2022-12-23 19:33:18,504 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.1559, 5.4141, 4.7643, 2.8210, 5.4875, 4.2844, 1.3528, 3.5875], device='cuda:3'), covar=tensor([0.1760, 0.0745, 0.1220, 0.2837, 0.0525, 0.0718, 0.4504, 0.1182], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0142, 0.0159, 0.0123, 0.0145, 0.0115, 0.0143, 0.0115], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 19:33:41,650 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78272.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:34:00,732 INFO [train.py:894] (3/4) Epoch 23, batch 1150, loss[loss=0.1873, simple_loss=0.2695, pruned_loss=0.05253, over 18512.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2615, pruned_loss=0.04171, over 3702705.02 frames. ], batch size: 41, lr: 5.00e-03, grad_scale: 8.0 2022-12-23 19:34:01,565 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2022-12-23 19:34:12,683 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78293.0, num_to_drop=1, layers_to_drop={3} 2022-12-23 19:34:25,308 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.55 vs. limit=2.0 2022-12-23 19:34:34,593 WARNING [train.py:1060] (3/4) Exclude cut with ID 7276-92427-0014-12983-0_sp0.9 from training. Duration: 21.3055625 2022-12-23 19:34:35,970 WARNING [train.py:1060] (3/4) Exclude cut with ID 1025-75365-0008-79168-0_sp0.9 from training. Duration: 22.0666875 2022-12-23 19:34:53,262 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=78320.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:35:12,653 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78333.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:35:15,188 INFO [train.py:894] (3/4) Epoch 23, batch 1200, loss[loss=0.2113, simple_loss=0.2941, pruned_loss=0.06423, over 18539.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2606, pruned_loss=0.04121, over 3705100.83 frames. ], batch size: 77, lr: 5.00e-03, grad_scale: 8.0 2022-12-23 19:35:15,964 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2022-12-23 19:35:27,165 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.472e+02 3.425e+02 4.075e+02 4.983e+02 9.037e+02, threshold=8.150e+02, percent-clipped=1.0 2022-12-23 19:35:53,552 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8111, 1.4022, 1.6511, 1.9493, 1.6611, 3.6504, 1.2381, 1.4472], device='cuda:3'), covar=tensor([0.0816, 0.1852, 0.1117, 0.1009, 0.1522, 0.0216, 0.1502, 0.1565], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0082, 0.0073, 0.0076, 0.0091, 0.0076, 0.0085, 0.0078], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2022-12-23 19:36:24,124 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=78381.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:36:26,706 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0 from training. Duration: 20.26 2022-12-23 19:36:29,570 INFO [train.py:894] (3/4) Epoch 23, batch 1250, loss[loss=0.1621, simple_loss=0.2545, pruned_loss=0.0348, over 18731.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2599, pruned_loss=0.04095, over 3707542.25 frames. ], batch size: 54, lr: 5.00e-03, grad_scale: 8.0 2022-12-23 19:36:38,534 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0030-9324-0_sp0.9 from training. Duration: 21.3444375 2022-12-23 19:37:37,488 WARNING [train.py:1060] (3/4) Exclude cut with ID 497-129325-0061-62254-0_sp1.1 from training. Duration: 0.97725 2022-12-23 19:37:45,283 INFO [train.py:894] (3/4) Epoch 23, batch 1300, loss[loss=0.1508, simple_loss=0.2404, pruned_loss=0.03066, over 18696.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2586, pruned_loss=0.0405, over 3708686.42 frames. ], batch size: 50, lr: 5.00e-03, grad_scale: 8.0 2022-12-23 19:37:57,536 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.004e+02 3.125e+02 3.830e+02 4.659e+02 1.286e+03, threshold=7.660e+02, percent-clipped=4.0 2022-12-23 19:38:21,904 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0_sp0.9 from training. Duration: 22.97225 2022-12-23 19:38:23,340 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78460.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:38:50,727 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0047-39922-0_sp0.9 from training. Duration: 21.97775 2022-12-23 19:38:53,154 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2679, 2.0155, 1.6448, 1.9435, 1.8019, 1.9491, 1.9178, 2.1845], device='cuda:3'), covar=tensor([0.2291, 0.3199, 0.2027, 0.2616, 0.3585, 0.1188, 0.2849, 0.1012], device='cuda:3'), in_proj_covar=tensor([0.0294, 0.0293, 0.0246, 0.0344, 0.0273, 0.0228, 0.0288, 0.0214], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 19:38:59,646 INFO [train.py:894] (3/4) Epoch 23, batch 1350, loss[loss=0.161, simple_loss=0.2414, pruned_loss=0.04033, over 18429.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2589, pruned_loss=0.04056, over 3709026.61 frames. ], batch size: 48, lr: 5.00e-03, grad_scale: 8.0 2022-12-23 19:39:02,915 WARNING [train.py:1060] (3/4) Exclude cut with ID 1112-1043-0006-89194-0_sp0.9 from training. Duration: 21.8333125 2022-12-23 19:39:12,351 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0 from training. Duration: 20.47 2022-12-23 19:39:55,058 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78521.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:40:13,622 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.1972, 1.2238, 0.9547, 1.3102, 1.4205, 2.4210, 1.2414, 1.3843], device='cuda:3'), covar=tensor([0.0873, 0.1827, 0.1086, 0.0944, 0.1532, 0.0323, 0.1415, 0.1555], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0082, 0.0072, 0.0075, 0.0090, 0.0076, 0.0085, 0.0077], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 19:40:14,601 INFO [train.py:894] (3/4) Epoch 23, batch 1400, loss[loss=0.1934, simple_loss=0.2815, pruned_loss=0.05262, over 18549.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2597, pruned_loss=0.04054, over 3709669.81 frames. ], batch size: 57, lr: 5.00e-03, grad_scale: 8.0 2022-12-23 19:40:18,928 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0037-39912-0_sp0.9 from training. Duration: 20.67225 2022-12-23 19:40:26,387 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.565e+02 3.335e+02 3.901e+02 5.226e+02 1.409e+03, threshold=7.802e+02, percent-clipped=5.0 2022-12-23 19:40:30,256 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2022-12-23 19:40:34,405 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5645, 2.1757, 2.2556, 2.2381, 2.4450, 2.5411, 2.4015, 2.0935], device='cuda:3'), covar=tensor([0.2012, 0.3036, 0.2287, 0.2782, 0.1965, 0.0879, 0.3375, 0.1224], device='cuda:3'), in_proj_covar=tensor([0.0269, 0.0299, 0.0281, 0.0319, 0.0310, 0.0253, 0.0345, 0.0242], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 19:40:36,790 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp0.9 from training. Duration: 25.2444375 2022-12-23 19:40:46,203 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78556.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:40:59,492 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0021-76797-0_sp0.9 from training. Duration: 21.1445 2022-12-23 19:41:08,920 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3199, 1.4813, 1.2595, 1.8110, 1.7277, 1.4323, 0.9906, 1.2152], device='cuda:3'), covar=tensor([0.2033, 0.1911, 0.1789, 0.1170, 0.1249, 0.1155, 0.2480, 0.1613], device='cuda:3'), in_proj_covar=tensor([0.0247, 0.0224, 0.0215, 0.0198, 0.0257, 0.0196, 0.0223, 0.0199], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 19:41:29,578 INFO [train.py:894] (3/4) Epoch 23, batch 1450, loss[loss=0.1865, simple_loss=0.2805, pruned_loss=0.04621, over 18714.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2604, pruned_loss=0.04077, over 3711194.65 frames. ], batch size: 65, lr: 4.99e-03, grad_scale: 8.0 2022-12-23 19:41:33,953 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78588.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 19:41:41,797 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.77 vs. limit=5.0 2022-12-23 19:42:15,456 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp0.9 from training. Duration: 33.038875 2022-12-23 19:42:18,555 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78617.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:42:40,143 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4969, 2.2654, 1.7529, 0.7394, 1.6145, 1.9849, 1.7095, 1.9551], device='cuda:3'), covar=tensor([0.0648, 0.0510, 0.1173, 0.1776, 0.1381, 0.1621, 0.1749, 0.0765], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0185, 0.0205, 0.0188, 0.0209, 0.0199, 0.0213, 0.0198], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 19:42:45,868 INFO [train.py:894] (3/4) Epoch 23, batch 1500, loss[loss=0.1717, simple_loss=0.269, pruned_loss=0.03724, over 18519.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2608, pruned_loss=0.04092, over 3711161.02 frames. ], batch size: 69, lr: 4.99e-03, grad_scale: 8.0 2022-12-23 19:42:53,028 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64291-0000-16059-0_sp0.9 from training. Duration: 20.0944375 2022-12-23 19:42:57,189 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.151e+02 3.273e+02 3.888e+02 4.739e+02 1.159e+03, threshold=7.776e+02, percent-clipped=4.0 2022-12-23 19:42:58,095 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2022-12-23 19:43:08,957 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp1.1 from training. Duration: 20.4 2022-12-23 19:43:16,096 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0 from training. Duration: 20.085 2022-12-23 19:43:27,900 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0_sp0.9 from training. Duration: 23.07775 2022-12-23 19:44:00,106 INFO [train.py:894] (3/4) Epoch 23, batch 1550, loss[loss=0.178, simple_loss=0.2694, pruned_loss=0.0433, over 18537.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2613, pruned_loss=0.04127, over 3711479.57 frames. ], batch size: 97, lr: 4.99e-03, grad_scale: 8.0 2022-12-23 19:44:06,598 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.45 vs. limit=2.0 2022-12-23 19:44:12,085 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp0.9 from training. Duration: 24.9333125 2022-12-23 19:44:57,461 WARNING [train.py:1060] (3/4) Exclude cut with ID 5118-111612-0016-124680-0_sp0.9 from training. Duration: 20.388875 2022-12-23 19:45:03,230 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp1.1 from training. Duration: 20.3590625 2022-12-23 19:45:14,753 INFO [train.py:894] (3/4) Epoch 23, batch 1600, loss[loss=0.1906, simple_loss=0.2708, pruned_loss=0.0552, over 18559.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2602, pruned_loss=0.04108, over 3711727.14 frames. ], batch size: 49, lr: 4.99e-03, grad_scale: 8.0 2022-12-23 19:45:27,223 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.292e+02 3.292e+02 3.867e+02 4.586e+02 1.296e+03, threshold=7.734e+02, percent-clipped=3.0 2022-12-23 19:45:38,707 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2022-12-23 19:46:12,309 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0_sp1.1 from training. Duration: 0.836375 2022-12-23 19:46:31,062 INFO [train.py:894] (3/4) Epoch 23, batch 1650, loss[loss=0.1713, simple_loss=0.261, pruned_loss=0.04083, over 18511.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2613, pruned_loss=0.04201, over 3711329.36 frames. ], batch size: 52, lr: 4.99e-03, grad_scale: 8.0 2022-12-23 19:46:42,446 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7606, 1.4620, 1.6162, 2.0220, 1.8648, 3.3454, 1.3975, 1.5238], device='cuda:3'), covar=tensor([0.0816, 0.1768, 0.1080, 0.0929, 0.1350, 0.0262, 0.1446, 0.1523], device='cuda:3'), in_proj_covar=tensor([0.0072, 0.0082, 0.0072, 0.0074, 0.0090, 0.0075, 0.0084, 0.0077], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 19:46:55,284 WARNING [train.py:1060] (3/4) Exclude cut with ID 8565-290391-0049-67394-0_sp0.9 from training. Duration: 21.3166875 2022-12-23 19:47:11,971 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2022-12-23 19:47:15,602 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.15 vs. limit=5.0 2022-12-23 19:47:19,301 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78816.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:47:26,349 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0029-104863-0_sp0.9 from training. Duration: 22.1055625 2022-12-23 19:47:36,877 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp1.1 from training. Duration: 21.77725 2022-12-23 19:47:46,979 INFO [train.py:894] (3/4) Epoch 23, batch 1700, loss[loss=0.1573, simple_loss=0.2348, pruned_loss=0.03991, over 18544.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2619, pruned_loss=0.04346, over 3712624.59 frames. ], batch size: 44, lr: 4.99e-03, grad_scale: 8.0 2022-12-23 19:47:55,464 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp0.9 from training. Duration: 27.8166875 2022-12-23 19:48:00,053 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.331e+02 3.650e+02 4.431e+02 5.330e+02 1.177e+03, threshold=8.861e+02, percent-clipped=5.0 2022-12-23 19:48:02,509 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7040, 1.3382, 0.9520, 0.2401, 1.0829, 1.5160, 1.1539, 1.2396], device='cuda:3'), covar=tensor([0.0733, 0.0758, 0.1350, 0.1902, 0.1271, 0.1826, 0.2233, 0.0942], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0188, 0.0208, 0.0191, 0.0213, 0.0202, 0.0217, 0.0201], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 19:48:08,511 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7636, 1.6227, 1.9415, 1.0832, 1.8124, 1.8962, 1.4884, 2.2024], device='cuda:3'), covar=tensor([0.0977, 0.1858, 0.0976, 0.1530, 0.0706, 0.0974, 0.2234, 0.0523], device='cuda:3'), in_proj_covar=tensor([0.0196, 0.0212, 0.0207, 0.0194, 0.0174, 0.0215, 0.0211, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 19:48:14,678 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2022-12-23 19:48:21,246 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp1.1 from training. Duration: 22.5090625 2022-12-23 19:48:27,367 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0 from training. Duration: 25.035 2022-12-23 19:48:44,716 WARNING [train.py:1060] (3/4) Exclude cut with ID 774-127930-0014-10412-0_sp1.1 from training. Duration: 0.95 2022-12-23 19:49:02,882 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp0.9 from training. Duration: 0.92225 2022-12-23 19:49:04,290 INFO [train.py:894] (3/4) Epoch 23, batch 1750, loss[loss=0.2099, simple_loss=0.3003, pruned_loss=0.05971, over 18668.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2634, pruned_loss=0.04528, over 3713404.39 frames. ], batch size: 62, lr: 4.98e-03, grad_scale: 8.0 2022-12-23 19:49:08,913 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78888.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:49:31,088 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0 from training. Duration: 21.97 2022-12-23 19:49:44,861 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78912.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:49:49,564 WARNING [train.py:1060] (3/4) Exclude cut with ID 7492-105653-0055-62765-0_sp0.9 from training. Duration: 21.97225 2022-12-23 19:49:51,128 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp0.9 from training. Duration: 25.3333125 2022-12-23 19:50:01,622 WARNING [train.py:1060] (3/4) Exclude cut with ID 5172-29468-0015-19128-0_sp0.9 from training. Duration: 21.5055625 2022-12-23 19:50:10,183 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp1.1 from training. Duration: 20.72725 2022-12-23 19:50:19,180 INFO [train.py:894] (3/4) Epoch 23, batch 1800, loss[loss=0.1853, simple_loss=0.2722, pruned_loss=0.04919, over 18668.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2644, pruned_loss=0.0466, over 3713439.05 frames. ], batch size: 60, lr: 4.98e-03, grad_scale: 8.0 2022-12-23 19:50:20,900 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=78936.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:50:32,367 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.020e+02 4.798e+02 5.662e+02 6.863e+02 1.253e+03, threshold=1.132e+03, percent-clipped=4.0 2022-12-23 19:50:43,745 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp0.9 from training. Duration: 26.32775 2022-12-23 19:50:56,583 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4103, 1.7309, 2.0726, 2.0217, 2.3369, 2.3342, 2.1922, 1.9805], device='cuda:3'), covar=tensor([0.2114, 0.2990, 0.2445, 0.2789, 0.1910, 0.0950, 0.2992, 0.1311], device='cuda:3'), in_proj_covar=tensor([0.0268, 0.0298, 0.0280, 0.0318, 0.0308, 0.0252, 0.0344, 0.0241], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 19:51:04,751 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78965.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:51:13,318 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6505, 2.2205, 1.7611, 2.5439, 2.0177, 2.1394, 2.1435, 2.6388], device='cuda:3'), covar=tensor([0.2047, 0.3064, 0.1929, 0.2581, 0.3628, 0.1068, 0.2971, 0.0889], device='cuda:3'), in_proj_covar=tensor([0.0297, 0.0295, 0.0249, 0.0349, 0.0275, 0.0230, 0.0291, 0.0217], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 19:51:15,713 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0 from training. Duration: 20.025 2022-12-23 19:51:18,385 WARNING [train.py:1060] (3/4) Exclude cut with ID 6709-74022-0004-86860-0_sp1.1 from training. Duration: 0.9409375 2022-12-23 19:51:18,397 WARNING [train.py:1060] (3/4) Exclude cut with ID 4757-1811-0023-62229-0_sp0.9 from training. Duration: 21.37775 2022-12-23 19:51:34,064 INFO [train.py:894] (3/4) Epoch 23, batch 1850, loss[loss=0.1679, simple_loss=0.2493, pruned_loss=0.04319, over 18656.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2644, pruned_loss=0.04759, over 3712913.77 frames. ], batch size: 48, lr: 4.98e-03, grad_scale: 8.0 2022-12-23 19:51:41,111 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0004-25974-0_sp0.9 from training. Duration: 21.17225 2022-12-23 19:51:41,121 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp0.9 from training. Duration: 27.511125 2022-12-23 19:51:53,677 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.25 vs. limit=5.0 2022-12-23 19:52:12,882 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0 from training. Duration: 22.8 2022-12-23 19:52:17,510 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0 from training. Duration: 22.585 2022-12-23 19:52:17,932 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.1170, 0.9977, 1.1636, 0.5036, 0.6449, 1.2025, 1.2423, 1.2080], device='cuda:3'), covar=tensor([0.0787, 0.0369, 0.0360, 0.0413, 0.0454, 0.0517, 0.0313, 0.0639], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0172, 0.0130, 0.0141, 0.0149, 0.0144, 0.0168, 0.0177], device='cuda:3'), out_proj_covar=tensor([1.1428e-04, 1.3062e-04, 9.6547e-05, 1.0458e-04, 1.1045e-04, 1.0946e-04, 1.2822e-04, 1.3398e-04], device='cuda:3') 2022-12-23 19:52:38,030 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79026.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:52:46,946 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0001-146967-0_sp0.9 from training. Duration: 22.0166875 2022-12-23 19:52:51,170 INFO [train.py:894] (3/4) Epoch 23, batch 1900, loss[loss=0.1808, simple_loss=0.2585, pruned_loss=0.0515, over 18431.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2646, pruned_loss=0.04844, over 3713662.14 frames. ], batch size: 48, lr: 4.98e-03, grad_scale: 8.0 2022-12-23 19:53:01,424 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp1.1 from training. Duration: 24.395375 2022-12-23 19:53:04,451 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.829e+02 4.333e+02 5.435e+02 7.142e+02 1.314e+03, threshold=1.087e+03, percent-clipped=4.0 2022-12-23 19:53:07,771 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp0.9 from training. Duration: 27.47775 2022-12-23 19:53:14,086 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp0.9 from training. Duration: 24.8833125 2022-12-23 19:53:16,958 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0 from training. Duration: 23.39 2022-12-23 19:53:21,465 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp0.9 from training. Duration: 28.72225 2022-12-23 19:53:29,888 WARNING [train.py:1060] (3/4) Exclude cut with ID 585-294811-0110-133686-0_sp0.9 from training. Duration: 20.8944375 2022-12-23 19:53:44,486 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0_sp0.9 from training. Duration: 23.8444375 2022-12-23 19:54:07,793 INFO [train.py:894] (3/4) Epoch 23, batch 1950, loss[loss=0.2134, simple_loss=0.2963, pruned_loss=0.06532, over 18714.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2656, pruned_loss=0.04939, over 3713735.38 frames. ], batch size: 54, lr: 4.98e-03, grad_scale: 8.0 2022-12-23 19:54:10,821 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0 from training. Duration: 25.85 2022-12-23 19:54:10,833 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0 from training. Duration: 21.39 2022-12-23 19:54:13,963 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0111, 1.2397, 0.8137, 1.4514, 2.1668, 1.3358, 1.4956, 1.7183], device='cuda:3'), covar=tensor([0.1477, 0.2139, 0.2226, 0.1491, 0.1740, 0.1789, 0.1462, 0.1707], device='cuda:3'), in_proj_covar=tensor([0.0093, 0.0098, 0.0116, 0.0097, 0.0117, 0.0091, 0.0099, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 19:54:20,178 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6457, 2.1371, 1.7054, 2.3820, 2.7080, 1.6490, 1.7023, 1.2788], device='cuda:3'), covar=tensor([0.1901, 0.1600, 0.1509, 0.0958, 0.1182, 0.1087, 0.1978, 0.1580], device='cuda:3'), in_proj_covar=tensor([0.0250, 0.0226, 0.0217, 0.0200, 0.0260, 0.0197, 0.0226, 0.0201], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 19:54:22,574 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0 from training. Duration: 27.92 2022-12-23 19:54:49,929 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0039-130165-0_sp0.9 from training. Duration: 20.661125 2022-12-23 19:54:54,493 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79116.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:55:11,752 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0043-15874-0_sp0.9 from training. Duration: 20.07225 2022-12-23 19:55:17,931 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0 from training. Duration: 21.01 2022-12-23 19:55:23,804 INFO [train.py:894] (3/4) Epoch 23, batch 2000, loss[loss=0.2359, simple_loss=0.3086, pruned_loss=0.08159, over 18665.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2654, pruned_loss=0.04956, over 3714434.24 frames. ], batch size: 184, lr: 4.98e-03, grad_scale: 8.0 2022-12-23 19:55:36,310 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.867e+02 4.288e+02 5.070e+02 5.717e+02 1.182e+03, threshold=1.014e+03, percent-clipped=2.0 2022-12-23 19:55:49,087 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2022-12-23 19:56:08,477 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=79164.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:56:25,642 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0 from training. Duration: 20.65 2022-12-23 19:56:34,302 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0 from training. Duration: 21.46 2022-12-23 19:56:37,521 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7473, 1.4104, 1.7202, 1.9865, 1.6997, 3.5186, 1.4667, 1.4679], device='cuda:3'), covar=tensor([0.0845, 0.1870, 0.1107, 0.0976, 0.1510, 0.0300, 0.1457, 0.1622], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0082, 0.0072, 0.0075, 0.0091, 0.0076, 0.0084, 0.0077], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 19:56:39,815 INFO [train.py:894] (3/4) Epoch 23, batch 2050, loss[loss=0.2042, simple_loss=0.282, pruned_loss=0.06322, over 18600.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2652, pruned_loss=0.04992, over 3714170.41 frames. ], batch size: 69, lr: 4.98e-03, grad_scale: 8.0 2022-12-23 19:57:02,170 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2926, 2.3645, 1.7346, 2.6977, 2.5413, 2.1710, 3.1593, 2.3701], device='cuda:3'), covar=tensor([0.0809, 0.1547, 0.2556, 0.1603, 0.1557, 0.0812, 0.0816, 0.1110], device='cuda:3'), in_proj_covar=tensor([0.0178, 0.0210, 0.0250, 0.0289, 0.0238, 0.0190, 0.0205, 0.0204], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 19:57:11,105 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5930, 1.4805, 1.4920, 0.8523, 1.6437, 1.5682, 1.5973, 1.3289], device='cuda:3'), covar=tensor([0.0370, 0.0541, 0.0456, 0.0761, 0.0425, 0.0414, 0.0428, 0.1005], device='cuda:3'), in_proj_covar=tensor([0.0126, 0.0130, 0.0131, 0.0120, 0.0102, 0.0124, 0.0134, 0.0161], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 19:57:19,539 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0 from training. Duration: 0.92 2022-12-23 19:57:19,683 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79212.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:57:27,493 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0_sp0.9 from training. Duration: 23.7666875 2022-12-23 19:57:54,619 INFO [train.py:894] (3/4) Epoch 23, batch 2100, loss[loss=0.1755, simple_loss=0.2513, pruned_loss=0.04987, over 18404.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2651, pruned_loss=0.04996, over 3713109.52 frames. ], batch size: 46, lr: 4.97e-03, grad_scale: 8.0 2022-12-23 19:58:02,591 WARNING [train.py:1060] (3/4) Exclude cut with ID 8544-281189-0060-101339-0_sp0.9 from training. Duration: 20.861125 2022-12-23 19:58:06,584 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.746e+02 4.115e+02 5.224e+02 6.588e+02 1.506e+03, threshold=1.045e+03, percent-clipped=3.0 2022-12-23 19:58:11,492 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0_sp0.9 from training. Duration: 22.711125 2022-12-23 19:58:26,727 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7538, 2.2129, 1.7585, 2.4551, 2.9542, 1.6648, 1.9727, 1.3236], device='cuda:3'), covar=tensor([0.2051, 0.1803, 0.1634, 0.1056, 0.1419, 0.1198, 0.1993, 0.1707], device='cuda:3'), in_proj_covar=tensor([0.0251, 0.0228, 0.0218, 0.0201, 0.0262, 0.0198, 0.0227, 0.0202], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 19:58:32,781 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=79260.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 19:58:55,630 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp1.1 from training. Duration: 22.986375 2022-12-23 19:59:10,426 INFO [train.py:894] (3/4) Epoch 23, batch 2150, loss[loss=0.1846, simple_loss=0.2644, pruned_loss=0.0524, over 18421.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2654, pruned_loss=0.05012, over 3713796.21 frames. ], batch size: 48, lr: 4.97e-03, grad_scale: 8.0 2022-12-23 19:59:10,485 WARNING [train.py:1060] (3/4) Exclude cut with ID 8040-260924-0003-80960-0_sp0.9 from training. Duration: 22.07225 2022-12-23 19:59:10,875 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79285.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 19:59:15,250 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0045-26330-0_sp0.9 from training. Duration: 20.3055625 2022-12-23 19:59:19,086 WARNING [train.py:1060] (3/4) Exclude cut with ID 6356-271890-0060-94317-0_sp0.9 from training. Duration: 20.72225 2022-12-23 19:59:36,690 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp1.1 from training. Duration: 22.4818125 2022-12-23 19:59:50,544 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5033, 3.3275, 3.2279, 1.2134, 3.4370, 2.5971, 0.5749, 2.0345], device='cuda:3'), covar=tensor([0.2066, 0.1364, 0.1562, 0.3935, 0.1023, 0.1072, 0.5044, 0.1882], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0144, 0.0160, 0.0125, 0.0147, 0.0116, 0.0146, 0.0117], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 20:00:03,935 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp0.9 from training. Duration: 25.0944375 2022-12-23 20:00:04,059 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79321.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:00:08,065 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0 from training. Duration: 21.515 2022-12-23 20:00:13,990 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp0.9 from training. Duration: 27.02225 2022-12-23 20:00:18,786 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0010-62480-0_sp0.9 from training. Duration: 22.22225 2022-12-23 20:00:21,910 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79333.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:00:24,471 INFO [train.py:894] (3/4) Epoch 23, batch 2200, loss[loss=0.1702, simple_loss=0.2623, pruned_loss=0.03905, over 18642.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2649, pruned_loss=0.04985, over 3714043.97 frames. ], batch size: 100, lr: 4.97e-03, grad_scale: 8.0 2022-12-23 20:00:24,568 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0085-44554-0_sp0.9 from training. Duration: 20.85 2022-12-23 20:00:36,434 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.368e+02 4.433e+02 5.321e+02 6.846e+02 1.189e+03, threshold=1.064e+03, percent-clipped=4.0 2022-12-23 20:00:41,261 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79346.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 20:01:00,248 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0 from training. Duration: 21.54 2022-12-23 20:01:03,853 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp1.1 from training. Duration: 20.5318125 2022-12-23 20:01:08,149 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79364.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:01:13,991 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0012-134311-0_sp0.9 from training. Duration: 21.9333125 2022-12-23 20:01:24,575 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79375.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 20:01:39,048 INFO [train.py:894] (3/4) Epoch 23, batch 2250, loss[loss=0.1731, simple_loss=0.2581, pruned_loss=0.04411, over 18676.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2667, pruned_loss=0.05085, over 3714289.56 frames. ], batch size: 62, lr: 4.97e-03, grad_scale: 8.0 2022-12-23 20:01:52,957 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79394.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:02:01,491 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0025-130151-0_sp0.9 from training. Duration: 21.7944375 2022-12-23 20:02:15,885 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0_sp0.9 from training. Duration: 22.4666875 2022-12-23 20:02:21,801 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0 from training. Duration: 21.635 2022-12-23 20:02:28,830 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0_sp0.9 from training. Duration: 24.038875 2022-12-23 20:02:39,822 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79425.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:02:53,957 INFO [train.py:894] (3/4) Epoch 23, batch 2300, loss[loss=0.2035, simple_loss=0.2917, pruned_loss=0.05766, over 18537.00 frames. ], tot_loss[loss=0.184, simple_loss=0.266, pruned_loss=0.05097, over 3714015.15 frames. ], batch size: 55, lr: 4.97e-03, grad_scale: 8.0 2022-12-23 20:02:55,749 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79436.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 20:03:05,585 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.680e+02 4.602e+02 5.590e+02 7.460e+02 2.406e+03, threshold=1.118e+03, percent-clipped=8.0 2022-12-23 20:03:09,817 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp1.1 from training. Duration: 21.786375 2022-12-23 20:03:18,788 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 2022-12-23 20:03:24,231 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0 from training. Duration: 20.22 2022-12-23 20:04:09,401 INFO [train.py:894] (3/4) Epoch 23, batch 2350, loss[loss=0.1638, simple_loss=0.2522, pruned_loss=0.03773, over 18503.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2655, pruned_loss=0.05063, over 3714334.16 frames. ], batch size: 52, lr: 4.97e-03, grad_scale: 8.0 2022-12-23 20:05:18,909 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.1802, 3.1515, 2.1471, 1.6372, 3.7310, 3.5131, 2.9791, 2.5190], device='cuda:3'), covar=tensor([0.0375, 0.0385, 0.0571, 0.0766, 0.0208, 0.0337, 0.0437, 0.0776], device='cuda:3'), in_proj_covar=tensor([0.0125, 0.0129, 0.0130, 0.0119, 0.0102, 0.0123, 0.0133, 0.0159], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 20:05:21,577 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0 from training. Duration: 25.285 2022-12-23 20:05:26,459 INFO [train.py:894] (3/4) Epoch 23, batch 2400, loss[loss=0.2127, simple_loss=0.2922, pruned_loss=0.06657, over 18657.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2649, pruned_loss=0.05041, over 3715036.25 frames. ], batch size: 182, lr: 4.96e-03, grad_scale: 8.0 2022-12-23 20:05:38,580 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.551e+02 4.138e+02 4.947e+02 5.790e+02 8.686e+02, threshold=9.894e+02, percent-clipped=0.0 2022-12-23 20:05:39,429 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2022-12-23 20:06:27,748 WARNING [train.py:1060] (3/4) Exclude cut with ID 811-130148-0001-63453-0_sp0.9 from training. Duration: 20.861125 2022-12-23 20:06:29,424 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3699, 1.8283, 1.1559, 2.0547, 2.7310, 2.1661, 2.0967, 2.2169], device='cuda:3'), covar=tensor([0.1387, 0.1984, 0.2239, 0.1324, 0.1433, 0.1449, 0.1334, 0.1530], device='cuda:3'), in_proj_covar=tensor([0.0093, 0.0097, 0.0115, 0.0095, 0.0117, 0.0091, 0.0098, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 20:06:42,123 INFO [train.py:894] (3/4) Epoch 23, batch 2450, loss[loss=0.1558, simple_loss=0.2397, pruned_loss=0.03598, over 18376.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2643, pruned_loss=0.05033, over 3715101.64 frames. ], batch size: 46, lr: 4.96e-03, grad_scale: 8.0 2022-12-23 20:06:42,537 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79585.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:06:51,665 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0 from training. Duration: 20.88 2022-12-23 20:07:22,972 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0_sp0.9 from training. Duration: 23.4166875 2022-12-23 20:07:37,238 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79621.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:07:42,299 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.56 vs. limit=2.0 2022-12-23 20:07:47,850 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5467, 1.2198, 1.8028, 2.9371, 2.1613, 2.2691, 0.8251, 2.0426], device='cuda:3'), covar=tensor([0.1799, 0.1649, 0.1366, 0.0680, 0.0985, 0.1228, 0.2113, 0.1076], device='cuda:3'), in_proj_covar=tensor([0.0104, 0.0118, 0.0137, 0.0153, 0.0107, 0.0144, 0.0130, 0.0114], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2022-12-23 20:07:58,607 INFO [train.py:894] (3/4) Epoch 23, batch 2500, loss[loss=0.1617, simple_loss=0.2412, pruned_loss=0.04115, over 18528.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2652, pruned_loss=0.05079, over 3715929.22 frames. ], batch size: 44, lr: 4.96e-03, grad_scale: 8.0 2022-12-23 20:08:08,458 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79641.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 20:08:11,485 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.617e+02 4.368e+02 5.588e+02 6.657e+02 1.417e+03, threshold=1.118e+03, percent-clipped=5.0 2022-12-23 20:08:16,338 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79646.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:08:32,508 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79657.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:08:35,885 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.04 vs. limit=5.0 2022-12-23 20:08:36,483 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0 from training. Duration: 21.24 2022-12-23 20:08:36,500 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0_sp0.9 from training. Duration: 23.9055625 2022-12-23 20:08:49,543 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=79669.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:09:10,150 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp0.9 from training. Duration: 25.988875 2022-12-23 20:09:12,913 INFO [train.py:894] (3/4) Epoch 23, batch 2550, loss[loss=0.1806, simple_loss=0.2741, pruned_loss=0.04359, over 18523.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2647, pruned_loss=0.04994, over 3714534.18 frames. ], batch size: 55, lr: 4.96e-03, grad_scale: 8.0 2022-12-23 20:09:19,526 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0001-134300-0_sp0.9 from training. Duration: 20.67225 2022-12-23 20:09:19,687 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79689.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:09:46,123 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.51 vs. limit=2.0 2022-12-23 20:10:04,286 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79718.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:10:06,866 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79720.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:10:08,194 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0 from training. Duration: 20.34 2022-12-23 20:10:23,150 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79731.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 20:10:29,807 INFO [train.py:894] (3/4) Epoch 23, batch 2600, loss[loss=0.1854, simple_loss=0.2715, pruned_loss=0.04962, over 18629.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2647, pruned_loss=0.05009, over 3713114.67 frames. ], batch size: 98, lr: 4.96e-03, grad_scale: 8.0 2022-12-23 20:10:35,371 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2022-12-23 20:10:41,465 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.544e+02 4.012e+02 4.948e+02 6.016e+02 1.385e+03, threshold=9.896e+02, percent-clipped=1.0 2022-12-23 20:11:04,955 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2970, 1.5957, 1.9975, 1.8826, 2.2222, 2.2298, 2.0357, 1.8852], device='cuda:3'), covar=tensor([0.2212, 0.3283, 0.2470, 0.2780, 0.2027, 0.0991, 0.3031, 0.1331], device='cuda:3'), in_proj_covar=tensor([0.0272, 0.0302, 0.0284, 0.0321, 0.0313, 0.0256, 0.0348, 0.0245], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 20:11:10,460 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79762.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:11:22,898 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp0.9 from training. Duration: 25.061125 2022-12-23 20:11:33,698 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0 from training. Duration: 0.83 2022-12-23 20:11:35,913 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7759, 2.3908, 1.6896, 2.5680, 3.0313, 1.7247, 2.0033, 1.3998], device='cuda:3'), covar=tensor([0.1987, 0.1610, 0.1575, 0.1006, 0.1286, 0.1141, 0.1887, 0.1585], device='cuda:3'), in_proj_covar=tensor([0.0251, 0.0230, 0.0218, 0.0203, 0.0264, 0.0198, 0.0228, 0.0202], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 20:11:44,509 INFO [train.py:894] (3/4) Epoch 23, batch 2650, loss[loss=0.1551, simple_loss=0.2304, pruned_loss=0.03987, over 18524.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2644, pruned_loss=0.05022, over 3713950.88 frames. ], batch size: 44, lr: 4.96e-03, grad_scale: 8.0 2022-12-23 20:11:58,559 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0 from training. Duration: 24.73 2022-12-23 20:12:09,655 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6097, 1.5529, 1.4577, 1.5819, 1.8454, 1.7395, 1.7324, 1.2570], device='cuda:3'), covar=tensor([0.0312, 0.0241, 0.0468, 0.0200, 0.0186, 0.0379, 0.0287, 0.0325], device='cuda:3'), in_proj_covar=tensor([0.0095, 0.0127, 0.0153, 0.0125, 0.0117, 0.0121, 0.0099, 0.0128], device='cuda:3'), out_proj_covar=tensor([7.5408e-05, 1.0043e-04, 1.2557e-04, 9.9276e-05, 9.4686e-05, 9.3282e-05, 7.7016e-05, 1.0055e-04], device='cuda:3') 2022-12-23 20:12:12,080 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0 from training. Duration: 23.965 2022-12-23 20:12:13,839 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6910, 1.4335, 1.1054, 0.2722, 1.1506, 1.5634, 1.3406, 1.4679], device='cuda:3'), covar=tensor([0.0709, 0.0602, 0.1071, 0.1686, 0.1178, 0.1642, 0.1839, 0.0712], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0189, 0.0209, 0.0191, 0.0212, 0.0203, 0.0219, 0.0202], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 20:12:20,582 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0030-146996-0_sp0.9 from training. Duration: 22.088875 2022-12-23 20:12:36,659 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0_sp0.9 from training. Duration: 23.6 2022-12-23 20:12:41,200 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79823.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:13:00,067 INFO [train.py:894] (3/4) Epoch 23, batch 2700, loss[loss=0.2107, simple_loss=0.2974, pruned_loss=0.06201, over 18620.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2653, pruned_loss=0.05046, over 3714038.55 frames. ], batch size: 78, lr: 4.95e-03, grad_scale: 8.0 2022-12-23 20:13:12,606 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.225e+02 4.087e+02 5.258e+02 6.716e+02 1.165e+03, threshold=1.052e+03, percent-clipped=3.0 2022-12-23 20:14:16,438 INFO [train.py:894] (3/4) Epoch 23, batch 2750, loss[loss=0.1687, simple_loss=0.2642, pruned_loss=0.03664, over 18711.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2649, pruned_loss=0.05024, over 3713918.34 frames. ], batch size: 78, lr: 4.95e-03, grad_scale: 8.0 2022-12-23 20:14:16,474 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0 from training. Duration: 23.795 2022-12-23 20:14:33,595 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp1.1 from training. Duration: 21.5409375 2022-12-23 20:14:34,024 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5520, 1.9756, 1.5600, 2.1893, 2.2966, 1.6387, 1.5595, 1.3497], device='cuda:3'), covar=tensor([0.2139, 0.1736, 0.1771, 0.1073, 0.1367, 0.1228, 0.2420, 0.1676], device='cuda:3'), in_proj_covar=tensor([0.0251, 0.0230, 0.0218, 0.0203, 0.0264, 0.0198, 0.0228, 0.0202], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 20:14:36,528 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp0.9 from training. Duration: 24.97775 2022-12-23 20:14:46,723 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0_sp0.9 from training. Duration: 23.3444375 2022-12-23 20:15:16,345 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0_sp0.9 from training. Duration: 23.2 2022-12-23 20:15:22,237 WARNING [train.py:1060] (3/4) Exclude cut with ID 5653-46179-0060-117930-0_sp0.9 from training. Duration: 21.17225 2022-12-23 20:15:32,915 INFO [train.py:894] (3/4) Epoch 23, batch 2800, loss[loss=0.173, simple_loss=0.2516, pruned_loss=0.04726, over 18512.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2634, pruned_loss=0.04954, over 3714667.34 frames. ], batch size: 47, lr: 4.95e-03, grad_scale: 8.0 2022-12-23 20:15:40,673 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp0.9 from training. Duration: 24.6555625 2022-12-23 20:15:42,345 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79941.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:15:42,455 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79941.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 20:15:45,136 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.473e+02 3.775e+02 4.575e+02 5.641e+02 1.095e+03, threshold=9.150e+02, percent-clipped=1.0 2022-12-23 20:15:52,446 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.94 vs. limit=5.0 2022-12-23 20:15:55,980 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7981, 1.7839, 1.8441, 1.8069, 1.2995, 3.4730, 1.5900, 2.1845], device='cuda:3'), covar=tensor([0.3020, 0.2007, 0.1747, 0.1985, 0.1494, 0.0227, 0.1743, 0.0860], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0118, 0.0124, 0.0120, 0.0105, 0.0097, 0.0091, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 20:16:15,875 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4510, 1.8993, 1.4657, 2.1689, 2.2086, 1.5828, 1.3570, 1.2753], device='cuda:3'), covar=tensor([0.2090, 0.1763, 0.1724, 0.1029, 0.1363, 0.1165, 0.2422, 0.1667], device='cuda:3'), in_proj_covar=tensor([0.0253, 0.0232, 0.0220, 0.0204, 0.0266, 0.0200, 0.0230, 0.0204], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 20:16:36,526 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0 from training. Duration: 20.44 2022-12-23 20:16:49,083 INFO [train.py:894] (3/4) Epoch 23, batch 2850, loss[loss=0.1995, simple_loss=0.2852, pruned_loss=0.05688, over 18679.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2638, pruned_loss=0.04956, over 3714532.75 frames. ], batch size: 62, lr: 4.95e-03, grad_scale: 8.0 2022-12-23 20:16:53,624 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0_sp0.9 from training. Duration: 23.45 2022-12-23 20:16:55,243 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=79989.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 20:16:55,264 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79989.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:17:00,178 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.57 vs. limit=5.0 2022-12-23 20:17:04,603 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.71 vs. limit=5.0 2022-12-23 20:17:24,491 WARNING [train.py:1060] (3/4) Exclude cut with ID 6945-60535-0076-12784-0_sp0.9 from training. Duration: 20.52225 2022-12-23 20:17:33,588 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0 from training. Duration: 22.19 2022-12-23 20:17:35,281 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80013.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:17:43,672 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp1.1 from training. Duration: 25.3818125 2022-12-23 20:17:45,321 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80020.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:18:00,164 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp0.9 from training. Duration: 28.0944375 2022-12-23 20:18:02,027 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80031.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 20:18:02,134 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6247, 1.6401, 1.3498, 1.5065, 1.8133, 1.7635, 1.7296, 1.3117], device='cuda:3'), covar=tensor([0.0341, 0.0270, 0.0517, 0.0247, 0.0217, 0.0432, 0.0320, 0.0366], device='cuda:3'), in_proj_covar=tensor([0.0095, 0.0127, 0.0152, 0.0125, 0.0117, 0.0121, 0.0099, 0.0128], device='cuda:3'), out_proj_covar=tensor([7.5274e-05, 1.0020e-04, 1.2499e-04, 9.9143e-05, 9.4482e-05, 9.3109e-05, 7.7290e-05, 1.0051e-04], device='cuda:3') 2022-12-23 20:18:05,938 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0_sp0.9 from training. Duration: 22.9444375 2022-12-23 20:18:07,429 INFO [train.py:894] (3/4) Epoch 23, batch 2900, loss[loss=0.1845, simple_loss=0.2735, pruned_loss=0.04775, over 18644.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.264, pruned_loss=0.04945, over 3714591.48 frames. ], batch size: 97, lr: 4.95e-03, grad_scale: 8.0 2022-12-23 20:18:10,375 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=80037.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:18:13,173 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp1.1 from training. Duration: 21.6318125 2022-12-23 20:18:18,879 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.810e+02 4.423e+02 5.143e+02 6.532e+02 1.388e+03, threshold=1.029e+03, percent-clipped=7.0 2022-12-23 20:18:29,734 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0 from training. Duration: 23.695 2022-12-23 20:18:56,109 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0 from training. Duration: 23.955 2022-12-23 20:18:57,664 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=80068.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:19:13,686 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=80079.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 20:19:22,081 INFO [train.py:894] (3/4) Epoch 23, batch 2950, loss[loss=0.1968, simple_loss=0.2784, pruned_loss=0.0576, over 18718.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2642, pruned_loss=0.0499, over 3715162.00 frames. ], batch size: 65, lr: 4.95e-03, grad_scale: 8.0 2022-12-23 20:19:26,462 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp0.9 from training. Duration: 26.438875 2022-12-23 20:19:31,732 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4752, 2.8074, 3.1879, 1.0541, 2.6884, 3.5657, 2.5219, 2.7933], device='cuda:3'), covar=tensor([0.0836, 0.0369, 0.0257, 0.0526, 0.0360, 0.0342, 0.0366, 0.0617], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0170, 0.0129, 0.0140, 0.0148, 0.0143, 0.0167, 0.0175], device='cuda:3'), out_proj_covar=tensor([1.1255e-04, 1.2881e-04, 9.6272e-05, 1.0389e-04, 1.0931e-04, 1.0906e-04, 1.2742e-04, 1.3245e-04], device='cuda:3') 2022-12-23 20:19:49,696 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2022-12-23 20:20:10,743 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0021-26306-0_sp0.9 from training. Duration: 21.2444375 2022-12-23 20:20:12,073 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp0.9 from training. Duration: 31.02225 2022-12-23 20:20:12,197 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80118.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:20:19,168 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0 from training. Duration: 22.395 2022-12-23 20:20:36,788 INFO [train.py:894] (3/4) Epoch 23, batch 3000, loss[loss=0.1392, simple_loss=0.2318, pruned_loss=0.02331, over 18580.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2639, pruned_loss=0.04953, over 3714935.12 frames. ], batch size: 51, lr: 4.95e-03, grad_scale: 8.0 2022-12-23 20:20:36,788 INFO [train.py:919] (3/4) Computing validation loss 2022-12-23 20:20:47,841 INFO [train.py:928] (3/4) Epoch 23, validation: loss=0.1641, simple_loss=0.261, pruned_loss=0.03364, over 944034.00 frames. 2022-12-23 20:20:47,842 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-23 20:20:47,863 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0 from training. Duration: 21.075 2022-12-23 20:20:52,106 WARNING [train.py:1060] (3/4) Exclude cut with ID 6482-98857-0025-147532-0_sp0.9 from training. Duration: 20.0055625 2022-12-23 20:20:52,113 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0037-132304-0_sp0.9 from training. Duration: 22.05 2022-12-23 20:20:52,124 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0 from training. Duration: 26.8349375 2022-12-23 20:20:55,163 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp1.1 from training. Duration: 22.1090625 2022-12-23 20:21:00,210 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.015e+02 4.784e+02 5.536e+02 6.733e+02 1.386e+03, threshold=1.107e+03, percent-clipped=4.0 2022-12-23 20:21:03,078 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp0.9 from training. Duration: 26.6166875 2022-12-23 20:21:16,869 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2022-12-23 20:21:21,894 WARNING [train.py:1060] (3/4) Exclude cut with ID 2046-178027-0000-53705-0_sp0.9 from training. Duration: 20.3055625 2022-12-23 20:21:42,214 WARNING [train.py:1060] (3/4) Exclude cut with ID 7205-50138-0008-5373-0_sp0.9 from training. Duration: 20.7 2022-12-23 20:22:02,878 INFO [train.py:894] (3/4) Epoch 23, batch 3050, loss[loss=0.1638, simple_loss=0.2513, pruned_loss=0.03815, over 18721.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2644, pruned_loss=0.04989, over 3715584.12 frames. ], batch size: 52, lr: 4.94e-03, grad_scale: 16.0 2022-12-23 20:22:28,948 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0 from training. Duration: 22.48 2022-12-23 20:22:43,524 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp0.9 from training. Duration: 29.816625 2022-12-23 20:22:56,416 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2022-12-23 20:23:04,905 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp1.1 from training. Duration: 22.7590625 2022-12-23 20:23:10,805 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0 from training. Duration: 22.555 2022-12-23 20:23:15,119 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.2823, 4.1872, 4.0503, 2.6947, 4.2208, 3.3625, 1.6808, 3.1030], device='cuda:3'), covar=tensor([0.1973, 0.1266, 0.1156, 0.2463, 0.0916, 0.0781, 0.3663, 0.1286], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0146, 0.0160, 0.0126, 0.0148, 0.0117, 0.0146, 0.0117], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 20:23:19,540 INFO [train.py:894] (3/4) Epoch 23, batch 3100, loss[loss=0.1994, simple_loss=0.2802, pruned_loss=0.0593, over 18455.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2643, pruned_loss=0.04975, over 3715246.94 frames. ], batch size: 64, lr: 4.94e-03, grad_scale: 16.0 2022-12-23 20:23:29,226 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80241.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:23:32,363 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.802e+02 4.142e+02 4.904e+02 7.040e+02 2.138e+03, threshold=9.809e+02, percent-clipped=3.0 2022-12-23 20:23:32,382 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0005-25975-0_sp0.9 from training. Duration: 21.688875 2022-12-23 20:24:04,786 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0_sp0.9 from training. Duration: 22.6 2022-12-23 20:24:32,734 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80283.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:24:35,760 INFO [train.py:894] (3/4) Epoch 23, batch 3150, loss[loss=0.1887, simple_loss=0.2793, pruned_loss=0.04903, over 18678.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2643, pruned_loss=0.04957, over 3716394.59 frames. ], batch size: 60, lr: 4.94e-03, grad_scale: 16.0 2022-12-23 20:24:41,667 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=80289.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:24:43,322 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0 from training. Duration: 24.32 2022-12-23 20:25:17,160 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80313.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:25:23,111 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.28 vs. limit=5.0 2022-12-23 20:25:40,352 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-276745-0093-13116-0_sp0.9 from training. Duration: 21.061125 2022-12-23 20:25:51,251 INFO [train.py:894] (3/4) Epoch 23, batch 3200, loss[loss=0.1588, simple_loss=0.2369, pruned_loss=0.04037, over 18691.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2634, pruned_loss=0.04944, over 3715640.64 frames. ], batch size: 41, lr: 4.94e-03, grad_scale: 16.0 2022-12-23 20:25:54,349 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0024-15855-0_sp0.9 from training. Duration: 20.32225 2022-12-23 20:26:03,086 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.560e+02 4.270e+02 5.295e+02 6.467e+02 1.258e+03, threshold=1.059e+03, percent-clipped=3.0 2022-12-23 20:26:04,955 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80344.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:26:06,059 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp1.1 from training. Duration: 0.7545625 2022-12-23 20:26:20,788 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0_sp0.9 from training. Duration: 23.9333125 2022-12-23 20:26:29,590 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=80361.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:26:38,154 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6373, 3.8309, 3.6668, 1.5630, 3.9174, 2.9244, 0.6319, 2.5054], device='cuda:3'), covar=tensor([0.2206, 0.1463, 0.1478, 0.3913, 0.0979, 0.0979, 0.5610, 0.1660], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0147, 0.0163, 0.0127, 0.0150, 0.0118, 0.0147, 0.0118], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 20:26:57,755 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp1.1 from training. Duration: 20.17275 2022-12-23 20:27:02,356 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp1.1 from training. Duration: 20.436375 2022-12-23 20:27:06,852 INFO [train.py:894] (3/4) Epoch 23, batch 3250, loss[loss=0.1825, simple_loss=0.2703, pruned_loss=0.04739, over 18672.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2638, pruned_loss=0.04947, over 3715020.19 frames. ], batch size: 60, lr: 4.94e-03, grad_scale: 16.0 2022-12-23 20:27:19,115 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80393.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:27:57,562 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80418.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:28:21,759 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0_sp0.9 from training. Duration: 23.1055625 2022-12-23 20:28:23,435 INFO [train.py:894] (3/4) Epoch 23, batch 3300, loss[loss=0.1727, simple_loss=0.2505, pruned_loss=0.04745, over 18647.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2633, pruned_loss=0.04907, over 3715326.47 frames. ], batch size: 41, lr: 4.94e-03, grad_scale: 16.0 2022-12-23 20:28:24,874 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp1.1 from training. Duration: 23.5 2022-12-23 20:28:35,115 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.089e+02 4.153e+02 5.096e+02 6.144e+02 1.740e+03, threshold=1.019e+03, percent-clipped=6.0 2022-12-23 20:28:36,516 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp0.9 from training. Duration: 26.62775 2022-12-23 20:28:38,371 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5434, 2.6993, 3.2631, 1.6195, 2.9912, 3.1221, 2.1609, 3.6267], device='cuda:3'), covar=tensor([0.1326, 0.1769, 0.1242, 0.2271, 0.0808, 0.1264, 0.2148, 0.0550], device='cuda:3'), in_proj_covar=tensor([0.0196, 0.0213, 0.0207, 0.0194, 0.0174, 0.0216, 0.0214, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 20:28:49,867 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0 from training. Duration: 21.105 2022-12-23 20:28:51,417 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80454.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:28:52,486 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0_sp0.9 from training. Duration: 24.411125 2022-12-23 20:29:04,899 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80463.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:29:08,678 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=80466.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:29:20,189 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp1.1 from training. Duration: 21.263625 2022-12-23 20:29:37,277 INFO [train.py:894] (3/4) Epoch 23, batch 3350, loss[loss=0.198, simple_loss=0.2792, pruned_loss=0.05836, over 18586.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2626, pruned_loss=0.04872, over 3714652.62 frames. ], batch size: 51, lr: 4.94e-03, grad_scale: 16.0 2022-12-23 20:29:50,827 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0 from training. Duration: 20.795 2022-12-23 20:29:51,197 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80494.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:29:55,889 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6761, 1.6430, 1.6283, 1.7147, 1.4292, 3.8888, 1.6978, 2.1444], device='cuda:3'), covar=tensor([0.3684, 0.2255, 0.2163, 0.2274, 0.1555, 0.0209, 0.1641, 0.0942], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0118, 0.0125, 0.0121, 0.0105, 0.0097, 0.0091, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 20:30:01,122 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0 from training. Duration: 24.76 2022-12-23 20:30:01,137 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0_sp0.9 from training. Duration: 22.25 2022-12-23 20:30:28,425 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp1.1 from training. Duration: 20.5045625 2022-12-23 20:30:36,503 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.2580, 3.1755, 2.1023, 1.6671, 3.6191, 3.5489, 3.1163, 2.6170], device='cuda:3'), covar=tensor([0.0357, 0.0356, 0.0617, 0.0774, 0.0248, 0.0356, 0.0436, 0.0727], device='cuda:3'), in_proj_covar=tensor([0.0124, 0.0129, 0.0129, 0.0120, 0.0102, 0.0125, 0.0135, 0.0160], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 20:30:36,560 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80524.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:30:53,195 INFO [train.py:894] (3/4) Epoch 23, batch 3400, loss[loss=0.1603, simple_loss=0.2375, pruned_loss=0.04156, over 18532.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2625, pruned_loss=0.04851, over 3714523.48 frames. ], batch size: 44, lr: 4.93e-03, grad_scale: 8.0 2022-12-23 20:30:54,834 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([5.9506, 5.0612, 5.2709, 5.8565, 5.4777, 5.2563, 6.0251, 1.5750], device='cuda:3'), covar=tensor([0.0559, 0.0697, 0.0518, 0.0736, 0.1253, 0.1099, 0.0430, 0.5345], device='cuda:3'), in_proj_covar=tensor([0.0350, 0.0231, 0.0243, 0.0278, 0.0332, 0.0273, 0.0295, 0.0288], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 20:31:06,193 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.766e+02 4.348e+02 5.300e+02 6.386e+02 1.214e+03, threshold=1.060e+03, percent-clipped=1.0 2022-12-23 20:31:23,017 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80555.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:32:04,414 INFO [train.py:894] (3/4) Epoch 23, batch 3450, loss[loss=0.1754, simple_loss=0.2658, pruned_loss=0.04252, over 18722.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2617, pruned_loss=0.04809, over 3714507.03 frames. ], batch size: 52, lr: 4.93e-03, grad_scale: 8.0 2022-12-23 20:32:21,547 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2022-12-23 20:33:19,550 INFO [train.py:894] (3/4) Epoch 23, batch 3500, loss[loss=0.2077, simple_loss=0.2851, pruned_loss=0.06512, over 18671.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2625, pruned_loss=0.04853, over 3714545.84 frames. ], batch size: 187, lr: 4.93e-03, grad_scale: 8.0 2022-12-23 20:33:26,348 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80639.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:33:41,175 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp1.1 from training. Duration: 22.2954375 2022-12-23 20:33:51,323 INFO [train.py:894] (3/4) Epoch 24, batch 0, loss[loss=0.1603, simple_loss=0.2414, pruned_loss=0.03958, over 18396.00 frames. ], tot_loss[loss=0.1603, simple_loss=0.2414, pruned_loss=0.03958, over 18396.00 frames. ], batch size: 42, lr: 4.82e-03, grad_scale: 8.0 2022-12-23 20:33:51,324 INFO [train.py:919] (3/4) Computing validation loss 2022-12-23 20:34:02,130 INFO [train.py:928] (3/4) Epoch 24, validation: loss=0.1635, simple_loss=0.2609, pruned_loss=0.03309, over 944034.00 frames. 2022-12-23 20:34:02,131 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-23 20:34:06,644 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.712e+02 4.449e+02 5.372e+02 6.832e+02 1.472e+03, threshold=1.074e+03, percent-clipped=3.0 2022-12-23 20:34:30,446 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7170, 1.6842, 1.6043, 1.4416, 1.9403, 1.9403, 1.9543, 1.3547], device='cuda:3'), covar=tensor([0.0405, 0.0306, 0.0519, 0.0290, 0.0225, 0.0434, 0.0308, 0.0390], device='cuda:3'), in_proj_covar=tensor([0.0097, 0.0130, 0.0156, 0.0128, 0.0119, 0.0124, 0.0101, 0.0130], device='cuda:3'), out_proj_covar=tensor([7.7245e-05, 1.0308e-04, 1.2772e-04, 1.0175e-04, 9.5951e-05, 9.5445e-05, 7.8557e-05, 1.0269e-04], device='cuda:3') 2022-12-23 20:34:51,927 WARNING [train.py:1060] (3/4) Exclude cut with ID 298-126791-0067-24026-0_sp0.9 from training. Duration: 21.438875 2022-12-23 20:34:57,060 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([5.7829, 4.9358, 5.0449, 5.7915, 5.3304, 5.1208, 5.8522, 1.8412], device='cuda:3'), covar=tensor([0.0552, 0.0655, 0.0590, 0.0612, 0.1326, 0.1079, 0.0419, 0.5230], device='cuda:3'), in_proj_covar=tensor([0.0352, 0.0232, 0.0244, 0.0280, 0.0335, 0.0275, 0.0297, 0.0291], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 20:34:58,222 WARNING [train.py:1060] (3/4) Exclude cut with ID 5652-39938-0025-23684-0_sp0.9 from training. Duration: 22.2055625 2022-12-23 20:35:18,395 INFO [train.py:894] (3/4) Epoch 24, batch 50, loss[loss=0.1824, simple_loss=0.268, pruned_loss=0.04837, over 18457.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.263, pruned_loss=0.04187, over 838273.62 frames. ], batch size: 50, lr: 4.82e-03, grad_scale: 8.0 2022-12-23 20:36:14,478 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0160, 1.8994, 1.4657, 2.0222, 2.1765, 1.8919, 2.6353, 2.0725], device='cuda:3'), covar=tensor([0.0883, 0.1773, 0.2843, 0.1807, 0.1809, 0.0946, 0.0919, 0.1290], device='cuda:3'), in_proj_covar=tensor([0.0179, 0.0214, 0.0253, 0.0293, 0.0240, 0.0194, 0.0210, 0.0207], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 20:36:27,779 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80736.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:36:30,743 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.2684, 2.8830, 2.8485, 3.2494, 2.9590, 2.8337, 3.3725, 1.0070], device='cuda:3'), covar=tensor([0.1013, 0.0913, 0.0903, 0.0938, 0.1733, 0.1508, 0.0891, 0.5082], device='cuda:3'), in_proj_covar=tensor([0.0351, 0.0233, 0.0244, 0.0280, 0.0335, 0.0274, 0.0297, 0.0291], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 20:36:34,714 INFO [train.py:894] (3/4) Epoch 24, batch 100, loss[loss=0.1517, simple_loss=0.2408, pruned_loss=0.03127, over 18400.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2617, pruned_loss=0.04251, over 1476008.28 frames. ], batch size: 46, lr: 4.82e-03, grad_scale: 8.0 2022-12-23 20:36:39,362 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.348e+02 3.139e+02 3.771e+02 4.682e+02 1.046e+03, threshold=7.543e+02, percent-clipped=0.0 2022-12-23 20:36:47,531 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80749.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:37:22,392 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80773.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:37:49,413 INFO [train.py:894] (3/4) Epoch 24, batch 150, loss[loss=0.1625, simple_loss=0.2458, pruned_loss=0.03955, over 18434.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2595, pruned_loss=0.04133, over 1970893.37 frames. ], batch size: 48, lr: 4.82e-03, grad_scale: 8.0 2022-12-23 20:37:56,505 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0 from training. Duration: 24.525 2022-12-23 20:37:59,653 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80797.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:38:31,038 WARNING [train.py:1060] (3/4) Exclude cut with ID 3699-47246-0007-3408-0_sp0.9 from training. Duration: 20.26675 2022-12-23 20:38:31,148 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80819.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:38:37,484 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.3810, 2.5804, 2.0667, 3.0296, 2.4579, 2.4798, 2.5519, 3.3604], device='cuda:3'), covar=tensor([0.1838, 0.3224, 0.1915, 0.2884, 0.3857, 0.1034, 0.3142, 0.0799], device='cuda:3'), in_proj_covar=tensor([0.0303, 0.0299, 0.0251, 0.0356, 0.0279, 0.0234, 0.0295, 0.0222], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 20:38:44,190 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp0.9 from training. Duration: 27.25 2022-12-23 20:38:53,840 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80834.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:39:03,940 INFO [train.py:894] (3/4) Epoch 24, batch 200, loss[loss=0.1555, simple_loss=0.2417, pruned_loss=0.03466, over 18552.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2584, pruned_loss=0.04095, over 2357338.22 frames. ], batch size: 44, lr: 4.82e-03, grad_scale: 8.0 2022-12-23 20:39:09,446 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.872e+02 2.949e+02 3.615e+02 4.514e+02 7.125e+02, threshold=7.230e+02, percent-clipped=0.0 2022-12-23 20:39:18,430 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80850.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:40:01,269 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0 from training. Duration: 21.68 2022-12-23 20:40:11,436 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0 from training. Duration: 21.6300625 2022-12-23 20:40:19,789 INFO [train.py:894] (3/4) Epoch 24, batch 250, loss[loss=0.1762, simple_loss=0.266, pruned_loss=0.04323, over 18624.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2569, pruned_loss=0.04083, over 2657924.97 frames. ], batch size: 171, lr: 4.82e-03, grad_scale: 8.0 2022-12-23 20:40:34,926 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0_sp0.9 from training. Duration: 24.033375 2022-12-23 20:41:16,444 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2022-12-23 20:41:30,832 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80939.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:41:32,051 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0 from training. Duration: 22.905 2022-12-23 20:41:34,075 INFO [train.py:894] (3/4) Epoch 24, batch 300, loss[loss=0.1676, simple_loss=0.2612, pruned_loss=0.03699, over 18547.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2565, pruned_loss=0.04044, over 2891278.45 frames. ], batch size: 55, lr: 4.82e-03, grad_scale: 8.0 2022-12-23 20:41:34,105 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp1.1 from training. Duration: 23.4318125 2022-12-23 20:41:38,662 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.257e+02 3.429e+02 4.110e+02 5.616e+02 1.556e+03, threshold=8.221e+02, percent-clipped=7.0 2022-12-23 20:42:13,887 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2022-12-23 20:42:43,282 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=80987.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:42:49,391 INFO [train.py:894] (3/4) Epoch 24, batch 350, loss[loss=0.1501, simple_loss=0.2363, pruned_loss=0.03199, over 18714.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2569, pruned_loss=0.04054, over 3074554.03 frames. ], batch size: 46, lr: 4.81e-03, grad_scale: 8.0 2022-12-23 20:42:50,130 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.38 vs. limit=5.0 2022-12-23 20:42:51,777 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.93 vs. limit=5.0 2022-12-23 20:43:31,015 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp1.1 from training. Duration: 20.82275 2022-12-23 20:43:31,518 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6796, 2.2319, 1.8160, 2.4599, 2.0094, 2.2068, 2.1206, 2.6522], device='cuda:3'), covar=tensor([0.1901, 0.3223, 0.1890, 0.2623, 0.3659, 0.1055, 0.3108, 0.0880], device='cuda:3'), in_proj_covar=tensor([0.0302, 0.0299, 0.0252, 0.0353, 0.0280, 0.0235, 0.0296, 0.0222], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 20:43:32,484 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp0.9 from training. Duration: 25.45 2022-12-23 20:44:07,086 INFO [train.py:894] (3/4) Epoch 24, batch 400, loss[loss=0.1688, simple_loss=0.2598, pruned_loss=0.03887, over 18673.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.258, pruned_loss=0.0409, over 3215597.93 frames. ], batch size: 60, lr: 4.81e-03, grad_scale: 8.0 2022-12-23 20:44:11,335 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.000e+02 3.338e+02 3.884e+02 4.805e+02 8.135e+02, threshold=7.768e+02, percent-clipped=0.0 2022-12-23 20:44:19,469 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81049.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:44:32,705 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0 from training. Duration: 25.775 2022-12-23 20:44:54,583 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0_sp0.9 from training. Duration: 22.25 2022-12-23 20:45:06,917 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3517, 2.3457, 2.9476, 1.7617, 2.8638, 2.8290, 2.0704, 3.0050], device='cuda:3'), covar=tensor([0.1359, 0.1925, 0.1431, 0.2275, 0.0751, 0.1298, 0.2156, 0.0606], device='cuda:3'), in_proj_covar=tensor([0.0197, 0.0215, 0.0207, 0.0194, 0.0173, 0.0216, 0.0214, 0.0194], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 20:45:21,127 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0 from training. Duration: 26.205 2022-12-23 20:45:22,653 INFO [train.py:894] (3/4) Epoch 24, batch 450, loss[loss=0.1506, simple_loss=0.2307, pruned_loss=0.03524, over 18548.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2585, pruned_loss=0.04092, over 3325612.30 frames. ], batch size: 41, lr: 4.81e-03, grad_scale: 8.0 2022-12-23 20:45:24,954 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81092.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:45:32,206 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=81097.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:45:37,961 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp0.9 from training. Duration: 30.1555625 2022-12-23 20:45:43,629 WARNING [train.py:1060] (3/4) Exclude cut with ID 1265-135635-0050-6781-0_sp0.9 from training. Duration: 21.8333125 2022-12-23 20:45:51,983 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp1.1 from training. Duration: 20.6545625 2022-12-23 20:45:59,978 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81116.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:46:04,115 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81119.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:46:18,914 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81129.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:46:32,826 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0045-39920-0_sp0.9 from training. Duration: 20.52225 2022-12-23 20:46:37,968 INFO [train.py:894] (3/4) Epoch 24, batch 500, loss[loss=0.1532, simple_loss=0.2423, pruned_loss=0.032, over 18409.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2603, pruned_loss=0.04135, over 3411868.39 frames. ], batch size: 48, lr: 4.81e-03, grad_scale: 8.0 2022-12-23 20:46:42,459 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.626e+02 3.730e+02 4.462e+02 5.472e+02 2.665e+03, threshold=8.924e+02, percent-clipped=9.0 2022-12-23 20:46:51,210 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81150.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:46:54,093 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp0.9 from training. Duration: 29.1166875 2022-12-23 20:47:16,091 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=81167.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:47:31,547 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81177.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:47:53,254 INFO [train.py:894] (3/4) Epoch 24, batch 550, loss[loss=0.1933, simple_loss=0.2831, pruned_loss=0.05178, over 18664.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2607, pruned_loss=0.0413, over 3477831.24 frames. ], batch size: 60, lr: 4.81e-03, grad_scale: 8.0 2022-12-23 20:47:53,329 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133211-0007-59831-0_sp0.9 from training. Duration: 21.388875 2022-12-23 20:48:04,028 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=81198.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:48:11,390 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4852, 4.1007, 3.8536, 1.6275, 4.2267, 3.1344, 0.7541, 2.5618], device='cuda:3'), covar=tensor([0.2380, 0.1140, 0.1497, 0.3637, 0.0831, 0.0882, 0.5088, 0.1567], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0144, 0.0159, 0.0125, 0.0147, 0.0115, 0.0144, 0.0116], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 20:48:28,986 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0 from training. Duration: 22.72 2022-12-23 20:48:30,525 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0_sp0.9 from training. Duration: 22.7444375 2022-12-23 20:49:09,400 INFO [train.py:894] (3/4) Epoch 24, batch 600, loss[loss=0.1405, simple_loss=0.2246, pruned_loss=0.02822, over 18603.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2607, pruned_loss=0.04127, over 3529643.48 frames. ], batch size: 45, lr: 4.81e-03, grad_scale: 8.0 2022-12-23 20:49:14,225 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.016e+02 3.241e+02 3.841e+02 4.795e+02 1.272e+03, threshold=7.682e+02, percent-clipped=3.0 2022-12-23 20:49:15,709 WARNING [train.py:1060] (3/4) Exclude cut with ID 4133-6541-0027-40495-0_sp1.1 from training. Duration: 0.9681875 2022-12-23 20:49:18,647 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0_sp0.9 from training. Duration: 22.3166875 2022-12-23 20:49:24,212 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133212-0015-59917-0_sp0.9 from training. Duration: 21.8166875 2022-12-23 20:50:04,752 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.7400, 4.0653, 3.9738, 4.6687, 4.3274, 4.0733, 4.8874, 1.2589], device='cuda:3'), covar=tensor([0.0591, 0.0676, 0.0722, 0.0733, 0.1297, 0.1225, 0.0507, 0.5686], device='cuda:3'), in_proj_covar=tensor([0.0347, 0.0230, 0.0242, 0.0277, 0.0330, 0.0270, 0.0296, 0.0289], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 20:50:24,740 INFO [train.py:894] (3/4) Epoch 24, batch 650, loss[loss=0.188, simple_loss=0.2815, pruned_loss=0.04723, over 18733.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2605, pruned_loss=0.04099, over 3570154.63 frames. ], batch size: 60, lr: 4.81e-03, grad_scale: 8.0 2022-12-23 20:50:36,997 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81299.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:50:39,752 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81301.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:51:05,777 WARNING [train.py:1060] (3/4) Exclude cut with ID 4957-30119-0041-23990-0_sp0.9 from training. Duration: 20.22775 2022-12-23 20:51:39,382 INFO [train.py:894] (3/4) Epoch 24, batch 700, loss[loss=0.2077, simple_loss=0.3014, pruned_loss=0.05704, over 18526.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2619, pruned_loss=0.04138, over 3602208.33 frames. ], batch size: 58, lr: 4.80e-03, grad_scale: 8.0 2022-12-23 20:51:43,774 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.135e+02 3.349e+02 3.927e+02 5.051e+02 1.281e+03, threshold=7.854e+02, percent-clipped=5.0 2022-12-23 20:51:44,489 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.96 vs. limit=5.0 2022-12-23 20:51:48,229 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp1.1 from training. Duration: 24.67275 2022-12-23 20:52:07,697 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81360.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:52:10,690 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81362.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:52:17,830 WARNING [train.py:1060] (3/4) Exclude cut with ID 3082-165428-0081-50734-0_sp0.9 from training. Duration: 21.8055625 2022-12-23 20:52:55,133 INFO [train.py:894] (3/4) Epoch 24, batch 750, loss[loss=0.1699, simple_loss=0.2689, pruned_loss=0.03541, over 18501.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2621, pruned_loss=0.04127, over 3627475.68 frames. ], batch size: 52, lr: 4.80e-03, grad_scale: 8.0 2022-12-23 20:52:57,128 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81392.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:52:58,228 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0_sp0.9 from training. Duration: 22.6666875 2022-12-23 20:53:52,771 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81429.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:54:00,256 WARNING [train.py:1060] (3/4) Exclude cut with ID 2411-132532-0017-82279-0_sp1.1 from training. Duration: 0.9681875 2022-12-23 20:54:04,053 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2022-12-23 20:54:09,057 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=81440.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:54:09,321 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6626, 1.4430, 1.0485, 0.2997, 1.0528, 1.5921, 1.3837, 1.4074], device='cuda:3'), covar=tensor([0.0710, 0.0655, 0.1175, 0.1867, 0.1275, 0.1776, 0.1952, 0.0789], device='cuda:3'), in_proj_covar=tensor([0.0173, 0.0189, 0.0208, 0.0192, 0.0213, 0.0203, 0.0219, 0.0204], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 20:54:10,204 INFO [train.py:894] (3/4) Epoch 24, batch 800, loss[loss=0.181, simple_loss=0.2714, pruned_loss=0.04535, over 18694.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2616, pruned_loss=0.04129, over 3646661.01 frames. ], batch size: 50, lr: 4.80e-03, grad_scale: 8.0 2022-12-23 20:54:14,888 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.956e+02 3.275e+02 3.803e+02 4.513e+02 1.046e+03, threshold=7.605e+02, percent-clipped=6.0 2022-12-23 20:54:25,525 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0 from training. Duration: 22.485 2022-12-23 20:54:57,623 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81472.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:55:04,607 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp1.1 from training. Duration: 23.82275 2022-12-23 20:55:05,341 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=81477.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:55:19,140 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0 from training. Duration: 20.77 2022-12-23 20:55:22,878 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.89 vs. limit=5.0 2022-12-23 20:55:25,358 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0_sp0.9 from training. Duration: 24.088875 2022-12-23 20:55:26,777 INFO [train.py:894] (3/4) Epoch 24, batch 850, loss[loss=0.1785, simple_loss=0.2697, pruned_loss=0.04364, over 18522.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2611, pruned_loss=0.04108, over 3662009.38 frames. ], batch size: 98, lr: 4.80e-03, grad_scale: 8.0 2022-12-23 20:55:54,256 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp1.1 from training. Duration: 20.4409375 2022-12-23 20:55:58,997 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7726, 2.4284, 2.1178, 0.8415, 2.0297, 2.1812, 1.9606, 2.2452], device='cuda:3'), covar=tensor([0.0605, 0.0576, 0.1335, 0.1863, 0.1402, 0.1462, 0.1492, 0.0845], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0186, 0.0205, 0.0189, 0.0209, 0.0201, 0.0215, 0.0201], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 20:56:41,813 INFO [train.py:894] (3/4) Epoch 24, batch 900, loss[loss=0.1903, simple_loss=0.2771, pruned_loss=0.0517, over 18706.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2607, pruned_loss=0.04107, over 3673501.35 frames. ], batch size: 78, lr: 4.80e-03, grad_scale: 8.0 2022-12-23 20:56:46,272 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.385e+02 3.273e+02 3.910e+02 4.576e+02 1.457e+03, threshold=7.821e+02, percent-clipped=4.0 2022-12-23 20:56:51,446 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4653, 1.3699, 1.1259, 1.6149, 1.6589, 2.9780, 1.3366, 1.4774], device='cuda:3'), covar=tensor([0.0884, 0.1851, 0.1219, 0.0972, 0.1532, 0.0262, 0.1463, 0.1614], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0082, 0.0072, 0.0075, 0.0091, 0.0076, 0.0085, 0.0077], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2022-12-23 20:57:06,940 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0_sp0.9 from training. Duration: 22.511125 2022-12-23 20:57:08,406 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0 from training. Duration: 20.675 2022-12-23 20:57:21,598 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8300, 1.6399, 1.3628, 1.4701, 1.5039, 1.6428, 1.5231, 1.6529], device='cuda:3'), covar=tensor([0.2082, 0.2885, 0.1871, 0.2404, 0.3119, 0.1071, 0.2675, 0.1039], device='cuda:3'), in_proj_covar=tensor([0.0297, 0.0296, 0.0248, 0.0347, 0.0276, 0.0232, 0.0291, 0.0219], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 20:57:50,493 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4787, 2.2888, 1.9537, 1.4182, 2.8135, 2.5995, 2.5719, 1.8494], device='cuda:3'), covar=tensor([0.0357, 0.0409, 0.0513, 0.0783, 0.0269, 0.0359, 0.0349, 0.0847], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0128, 0.0128, 0.0119, 0.0102, 0.0123, 0.0134, 0.0159], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 20:57:57,312 INFO [train.py:894] (3/4) Epoch 24, batch 950, loss[loss=0.1799, simple_loss=0.2783, pruned_loss=0.04078, over 18561.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.262, pruned_loss=0.04132, over 3682484.26 frames. ], batch size: 56, lr: 4.80e-03, grad_scale: 8.0 2022-12-23 20:58:15,681 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4391, 1.0828, 1.7945, 2.8169, 2.0969, 2.6661, 0.7155, 2.2225], device='cuda:3'), covar=tensor([0.2290, 0.2485, 0.1741, 0.0913, 0.1255, 0.1049, 0.2682, 0.1312], device='cuda:3'), in_proj_covar=tensor([0.0102, 0.0117, 0.0134, 0.0150, 0.0105, 0.0141, 0.0128, 0.0112], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 20:58:51,313 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp0.9 from training. Duration: 24.9833125 2022-12-23 20:59:07,237 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([5.9205, 5.0243, 5.1627, 5.8552, 5.4584, 5.1783, 6.0105, 1.6753], device='cuda:3'), covar=tensor([0.0582, 0.0725, 0.0646, 0.0727, 0.1246, 0.1155, 0.0433, 0.5462], device='cuda:3'), in_proj_covar=tensor([0.0345, 0.0229, 0.0240, 0.0275, 0.0327, 0.0268, 0.0293, 0.0287], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 20:59:13,228 INFO [train.py:894] (3/4) Epoch 24, batch 1000, loss[loss=0.1668, simple_loss=0.2636, pruned_loss=0.03498, over 18696.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2613, pruned_loss=0.04122, over 3690166.53 frames. ], batch size: 60, lr: 4.79e-03, grad_scale: 8.0 2022-12-23 20:59:17,609 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.354e+02 3.373e+02 4.012e+02 4.702e+02 7.910e+02, threshold=8.023e+02, percent-clipped=1.0 2022-12-23 20:59:21,908 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0 from training. Duration: 27.14 2022-12-23 20:59:22,417 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.2869, 1.3769, 1.1577, 1.6148, 1.6785, 1.4023, 1.0508, 1.1526], device='cuda:3'), covar=tensor([0.2297, 0.2285, 0.2214, 0.1456, 0.1342, 0.1460, 0.2625, 0.1886], device='cuda:3'), in_proj_covar=tensor([0.0248, 0.0227, 0.0217, 0.0201, 0.0261, 0.0196, 0.0225, 0.0201], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 20:59:33,980 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81655.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:59:35,222 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0 from training. Duration: 22.44 2022-12-23 20:59:36,764 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81657.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 20:59:56,971 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2022-12-23 20:59:58,354 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81670.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:00:24,160 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.31 vs. limit=5.0 2022-12-23 21:00:29,155 INFO [train.py:894] (3/4) Epoch 24, batch 1050, loss[loss=0.1567, simple_loss=0.2319, pruned_loss=0.0408, over 18507.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2607, pruned_loss=0.04114, over 3695316.64 frames. ], batch size: 43, lr: 4.79e-03, grad_scale: 8.0 2022-12-23 21:00:54,274 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0060-62364-0_sp0.9 from training. Duration: 21.361125 2022-12-23 21:01:02,219 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp1.1 from training. Duration: 27.0318125 2022-12-23 21:01:11,293 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp0.9 from training. Duration: 28.638875 2022-12-23 21:01:26,341 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0 from training. Duration: 20.4 2022-12-23 21:01:29,844 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81731.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 21:01:43,893 INFO [train.py:894] (3/4) Epoch 24, batch 1100, loss[loss=0.1888, simple_loss=0.2768, pruned_loss=0.05036, over 18505.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2605, pruned_loss=0.04086, over 3698640.30 frames. ], batch size: 58, lr: 4.79e-03, grad_scale: 8.0 2022-12-23 21:01:47,090 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.0549, 5.5361, 5.0537, 2.8409, 5.5952, 4.2250, 0.6772, 3.5211], device='cuda:3'), covar=tensor([0.1715, 0.0672, 0.1093, 0.2551, 0.0442, 0.0669, 0.5026, 0.1232], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0144, 0.0158, 0.0124, 0.0145, 0.0114, 0.0144, 0.0115], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 21:01:48,279 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.333e+02 3.535e+02 4.177e+02 5.058e+02 9.147e+02, threshold=8.355e+02, percent-clipped=1.0 2022-12-23 21:01:58,859 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0 from training. Duration: 20.025 2022-12-23 21:01:58,871 WARNING [train.py:1060] (3/4) Exclude cut with ID 2364-131735-0112-64612-0_sp0.9 from training. Duration: 20.488875 2022-12-23 21:02:04,937 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0 from training. Duration: 29.735 2022-12-23 21:02:31,814 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81772.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:02:59,302 INFO [train.py:894] (3/4) Epoch 24, batch 1150, loss[loss=0.1599, simple_loss=0.2557, pruned_loss=0.03208, over 18515.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2604, pruned_loss=0.04095, over 3701188.59 frames. ], batch size: 52, lr: 4.79e-03, grad_scale: 8.0 2022-12-23 21:03:15,277 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2022-12-23 21:03:25,714 WARNING [train.py:1060] (3/4) Exclude cut with ID 7276-92427-0014-12983-0_sp0.9 from training. Duration: 21.3055625 2022-12-23 21:03:27,211 WARNING [train.py:1060] (3/4) Exclude cut with ID 1025-75365-0008-79168-0_sp0.9 from training. Duration: 22.0666875 2022-12-23 21:03:45,361 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=81820.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:04:11,723 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.1502, 3.5254, 3.5337, 4.0671, 3.7924, 3.6357, 4.2680, 1.3405], device='cuda:3'), covar=tensor([0.0901, 0.0970, 0.0823, 0.0961, 0.1627, 0.1475, 0.0836, 0.5619], device='cuda:3'), in_proj_covar=tensor([0.0344, 0.0228, 0.0239, 0.0274, 0.0326, 0.0267, 0.0293, 0.0287], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 21:04:15,786 INFO [train.py:894] (3/4) Epoch 24, batch 1200, loss[loss=0.1753, simple_loss=0.2652, pruned_loss=0.04267, over 18584.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2605, pruned_loss=0.04093, over 3703654.59 frames. ], batch size: 57, lr: 4.79e-03, grad_scale: 8.0 2022-12-23 21:04:19,890 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.558e+02 3.274e+02 3.968e+02 4.873e+02 1.280e+03, threshold=7.935e+02, percent-clipped=3.0 2022-12-23 21:04:23,723 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.97 vs. limit=5.0 2022-12-23 21:05:14,411 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.10 vs. limit=5.0 2022-12-23 21:05:19,323 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0 from training. Duration: 20.26 2022-12-23 21:05:30,578 INFO [train.py:894] (3/4) Epoch 24, batch 1250, loss[loss=0.1954, simple_loss=0.2854, pruned_loss=0.05274, over 18694.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2605, pruned_loss=0.04082, over 3705428.76 frames. ], batch size: 60, lr: 4.79e-03, grad_scale: 8.0 2022-12-23 21:05:32,107 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0030-9324-0_sp0.9 from training. Duration: 21.3444375 2022-12-23 21:05:39,733 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9281, 1.8673, 2.0931, 1.1970, 2.1362, 2.1870, 1.5194, 2.5364], device='cuda:3'), covar=tensor([0.1219, 0.1883, 0.1370, 0.2061, 0.0813, 0.1227, 0.2443, 0.0556], device='cuda:3'), in_proj_covar=tensor([0.0195, 0.0212, 0.0206, 0.0192, 0.0173, 0.0215, 0.0212, 0.0192], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 21:06:21,943 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2022-12-23 21:06:28,833 WARNING [train.py:1060] (3/4) Exclude cut with ID 497-129325-0061-62254-0_sp1.1 from training. Duration: 0.97725 2022-12-23 21:06:46,114 INFO [train.py:894] (3/4) Epoch 24, batch 1300, loss[loss=0.1552, simple_loss=0.2413, pruned_loss=0.03458, over 18526.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2605, pruned_loss=0.04055, over 3706300.25 frames. ], batch size: 44, lr: 4.79e-03, grad_scale: 8.0 2022-12-23 21:06:50,342 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.062e+02 3.110e+02 3.685e+02 4.502e+02 1.281e+03, threshold=7.371e+02, percent-clipped=3.0 2022-12-23 21:07:07,420 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81955.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:07:10,902 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0_sp0.9 from training. Duration: 22.97225 2022-12-23 21:07:11,135 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81957.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:07:31,789 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0698, 1.0625, 1.7394, 1.6171, 1.9890, 2.0327, 1.7384, 1.7578], device='cuda:3'), covar=tensor([0.2163, 0.3538, 0.2632, 0.2808, 0.2159, 0.1022, 0.3253, 0.1364], device='cuda:3'), in_proj_covar=tensor([0.0269, 0.0299, 0.0282, 0.0320, 0.0312, 0.0255, 0.0347, 0.0243], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 21:07:40,757 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5538, 1.4764, 1.5131, 1.4304, 1.1258, 3.0196, 1.2226, 1.7109], device='cuda:3'), covar=tensor([0.3136, 0.2138, 0.2012, 0.2209, 0.1514, 0.0211, 0.1751, 0.0913], device='cuda:3'), in_proj_covar=tensor([0.0131, 0.0117, 0.0123, 0.0120, 0.0104, 0.0096, 0.0090, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 21:07:41,945 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0047-39922-0_sp0.9 from training. Duration: 21.97775 2022-12-23 21:07:55,395 WARNING [train.py:1060] (3/4) Exclude cut with ID 1112-1043-0006-89194-0_sp0.9 from training. Duration: 21.8333125 2022-12-23 21:07:57,303 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81987.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:08:02,963 INFO [train.py:894] (3/4) Epoch 24, batch 1350, loss[loss=0.1598, simple_loss=0.2504, pruned_loss=0.03464, over 18585.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2602, pruned_loss=0.04064, over 3707335.40 frames. ], batch size: 51, lr: 4.78e-03, grad_scale: 8.0 2022-12-23 21:08:05,929 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0 from training. Duration: 20.47 2022-12-23 21:08:25,308 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=82003.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:08:28,686 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=82005.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:09:01,124 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82026.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 21:09:16,349 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0037-39912-0_sp0.9 from training. Duration: 20.67225 2022-12-23 21:09:19,812 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.1773, 2.5265, 1.8850, 2.8718, 2.4109, 2.4429, 2.4424, 3.2267], device='cuda:3'), covar=tensor([0.1870, 0.3262, 0.1915, 0.2930, 0.3639, 0.1036, 0.3220, 0.0819], device='cuda:3'), in_proj_covar=tensor([0.0296, 0.0294, 0.0247, 0.0343, 0.0274, 0.0230, 0.0289, 0.0218], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 21:09:22,128 INFO [train.py:894] (3/4) Epoch 24, batch 1400, loss[loss=0.1745, simple_loss=0.2659, pruned_loss=0.04156, over 18596.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2595, pruned_loss=0.0401, over 3708394.98 frames. ], batch size: 51, lr: 4.78e-03, grad_scale: 8.0 2022-12-23 21:09:26,688 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.199e+02 3.133e+02 3.786e+02 4.539e+02 8.263e+02, threshold=7.572e+02, percent-clipped=1.0 2022-12-23 21:09:33,604 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82048.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:09:34,737 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp0.9 from training. Duration: 25.2444375 2022-12-23 21:09:45,240 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82055.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:09:48,962 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2022-12-23 21:10:00,186 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0021-76797-0_sp0.9 from training. Duration: 21.1445 2022-12-23 21:10:37,964 INFO [train.py:894] (3/4) Epoch 24, batch 1450, loss[loss=0.1413, simple_loss=0.2231, pruned_loss=0.02977, over 18544.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2593, pruned_loss=0.04021, over 3709089.56 frames. ], batch size: 41, lr: 4.78e-03, grad_scale: 8.0 2022-12-23 21:11:15,421 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp0.9 from training. Duration: 33.038875 2022-12-23 21:11:17,128 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82116.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:11:52,247 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64291-0000-16059-0_sp0.9 from training. Duration: 20.0944375 2022-12-23 21:11:53,451 INFO [train.py:894] (3/4) Epoch 24, batch 1500, loss[loss=0.1574, simple_loss=0.2537, pruned_loss=0.03049, over 18624.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2594, pruned_loss=0.04019, over 3710351.46 frames. ], batch size: 53, lr: 4.78e-03, grad_scale: 8.0 2022-12-23 21:11:57,690 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.080e+02 3.225e+02 3.712e+02 4.583e+02 1.375e+03, threshold=7.423e+02, percent-clipped=2.0 2022-12-23 21:12:07,775 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp1.1 from training. Duration: 20.4 2022-12-23 21:12:15,644 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0 from training. Duration: 20.085 2022-12-23 21:12:26,509 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0_sp0.9 from training. Duration: 23.07775 2022-12-23 21:13:07,890 INFO [train.py:894] (3/4) Epoch 24, batch 1550, loss[loss=0.1485, simple_loss=0.24, pruned_loss=0.02844, over 18437.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2592, pruned_loss=0.04004, over 3710655.97 frames. ], batch size: 48, lr: 4.78e-03, grad_scale: 8.0 2022-12-23 21:13:12,595 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2112, 2.1851, 1.5559, 2.5775, 2.3552, 2.0604, 2.9629, 2.2363], device='cuda:3'), covar=tensor([0.0835, 0.1749, 0.2792, 0.1667, 0.1743, 0.0908, 0.0891, 0.1228], device='cuda:3'), in_proj_covar=tensor([0.0177, 0.0213, 0.0252, 0.0288, 0.0239, 0.0192, 0.0206, 0.0206], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 21:13:13,571 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp0.9 from training. Duration: 24.9333125 2022-12-23 21:13:23,289 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0800, 2.0427, 1.5239, 2.3532, 2.2713, 1.9815, 2.8579, 2.1791], device='cuda:3'), covar=tensor([0.0856, 0.1703, 0.2682, 0.1658, 0.1668, 0.0866, 0.0804, 0.1229], device='cuda:3'), in_proj_covar=tensor([0.0177, 0.0214, 0.0253, 0.0289, 0.0239, 0.0192, 0.0206, 0.0206], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 21:13:59,124 WARNING [train.py:1060] (3/4) Exclude cut with ID 5118-111612-0016-124680-0_sp0.9 from training. Duration: 20.388875 2022-12-23 21:14:05,126 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp1.1 from training. Duration: 20.3590625 2022-12-23 21:14:23,210 INFO [train.py:894] (3/4) Epoch 24, batch 1600, loss[loss=0.1904, simple_loss=0.2784, pruned_loss=0.05126, over 18583.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2606, pruned_loss=0.04037, over 3712017.97 frames. ], batch size: 57, lr: 4.78e-03, grad_scale: 8.0 2022-12-23 21:14:25,103 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.5235, 2.9194, 2.7486, 1.5355, 2.7221, 2.4311, 2.0892, 2.7609], device='cuda:3'), covar=tensor([0.0603, 0.0590, 0.1339, 0.1703, 0.1417, 0.1397, 0.1675, 0.0912], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0187, 0.0207, 0.0190, 0.0211, 0.0203, 0.0218, 0.0204], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 21:14:27,465 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.380e+02 3.366e+02 4.050e+02 5.021e+02 8.278e+02, threshold=8.100e+02, percent-clipped=2.0 2022-12-23 21:14:30,936 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4457, 1.3655, 1.2192, 1.6646, 1.6747, 3.0386, 1.3173, 1.5246], device='cuda:3'), covar=tensor([0.0840, 0.1823, 0.1057, 0.0876, 0.1419, 0.0258, 0.1438, 0.1545], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0082, 0.0072, 0.0075, 0.0091, 0.0076, 0.0084, 0.0077], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2022-12-23 21:14:49,328 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4857, 3.3664, 3.3011, 1.2811, 3.4753, 2.6571, 0.7794, 2.2593], device='cuda:3'), covar=tensor([0.2110, 0.1319, 0.1449, 0.4113, 0.0937, 0.0950, 0.4730, 0.1686], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0144, 0.0158, 0.0125, 0.0145, 0.0115, 0.0145, 0.0114], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 21:15:15,684 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0_sp1.1 from training. Duration: 0.836375 2022-12-23 21:15:20,729 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5100, 3.7081, 3.5589, 1.1792, 3.7991, 2.8398, 0.7656, 2.3681], device='cuda:3'), covar=tensor([0.2198, 0.1143, 0.1424, 0.4006, 0.0806, 0.0945, 0.4818, 0.1524], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0144, 0.0158, 0.0125, 0.0145, 0.0115, 0.0145, 0.0114], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 21:15:38,369 INFO [train.py:894] (3/4) Epoch 24, batch 1650, loss[loss=0.1441, simple_loss=0.2256, pruned_loss=0.03129, over 18543.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2604, pruned_loss=0.0406, over 3712288.36 frames. ], batch size: 44, lr: 4.78e-03, grad_scale: 8.0 2022-12-23 21:16:01,339 WARNING [train.py:1060] (3/4) Exclude cut with ID 8565-290391-0049-67394-0_sp0.9 from training. Duration: 21.3166875 2022-12-23 21:16:05,761 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5086, 3.8972, 3.7112, 1.4564, 3.9661, 2.9784, 0.9011, 2.5678], device='cuda:3'), covar=tensor([0.2294, 0.1222, 0.1381, 0.3905, 0.0807, 0.0996, 0.4661, 0.1569], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0144, 0.0158, 0.0124, 0.0144, 0.0115, 0.0144, 0.0114], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 21:16:31,553 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82326.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:16:32,797 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0029-104863-0_sp0.9 from training. Duration: 22.1055625 2022-12-23 21:16:43,033 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp1.1 from training. Duration: 21.77725 2022-12-23 21:16:53,962 INFO [train.py:894] (3/4) Epoch 24, batch 1700, loss[loss=0.1645, simple_loss=0.252, pruned_loss=0.03851, over 18720.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2606, pruned_loss=0.04135, over 3712392.41 frames. ], batch size: 52, lr: 4.77e-03, grad_scale: 8.0 2022-12-23 21:16:57,082 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82343.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:16:58,410 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.394e+02 3.814e+02 4.522e+02 5.754e+02 2.798e+03, threshold=9.045e+02, percent-clipped=8.0 2022-12-23 21:17:01,700 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82346.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:17:04,235 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp0.9 from training. Duration: 27.8166875 2022-12-23 21:17:32,342 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp1.1 from training. Duration: 22.5090625 2022-12-23 21:17:38,415 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0 from training. Duration: 25.035 2022-12-23 21:17:44,252 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=82374.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:17:57,040 WARNING [train.py:1060] (3/4) Exclude cut with ID 774-127930-0014-10412-0_sp1.1 from training. Duration: 0.95 2022-12-23 21:18:09,468 INFO [train.py:894] (3/4) Epoch 24, batch 1750, loss[loss=0.1748, simple_loss=0.2601, pruned_loss=0.04476, over 18502.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2608, pruned_loss=0.04231, over 3712949.56 frames. ], batch size: 52, lr: 4.77e-03, grad_scale: 8.0 2022-12-23 21:18:15,943 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9916, 1.8684, 1.8432, 1.0856, 2.2723, 2.0955, 1.9988, 1.5295], device='cuda:3'), covar=tensor([0.0412, 0.0516, 0.0466, 0.0828, 0.0332, 0.0384, 0.0423, 0.0999], device='cuda:3'), in_proj_covar=tensor([0.0122, 0.0126, 0.0127, 0.0119, 0.0101, 0.0123, 0.0132, 0.0157], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 21:18:17,032 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp0.9 from training. Duration: 0.92225 2022-12-23 21:18:25,367 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2302, 2.7998, 2.7166, 1.1607, 2.9661, 2.1697, 0.5730, 1.7241], device='cuda:3'), covar=tensor([0.2192, 0.1567, 0.1759, 0.3988, 0.1238, 0.1152, 0.4822, 0.1842], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0145, 0.0158, 0.0125, 0.0145, 0.0116, 0.0144, 0.0114], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 21:18:34,182 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82407.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:18:39,717 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82411.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:18:44,512 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0 from training. Duration: 21.97 2022-12-23 21:19:01,051 WARNING [train.py:1060] (3/4) Exclude cut with ID 7492-105653-0055-62765-0_sp0.9 from training. Duration: 21.97225 2022-12-23 21:19:01,097 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp0.9 from training. Duration: 25.3333125 2022-12-23 21:19:12,079 WARNING [train.py:1060] (3/4) Exclude cut with ID 5172-29468-0015-19128-0_sp0.9 from training. Duration: 21.5055625 2022-12-23 21:19:22,263 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp1.1 from training. Duration: 20.72725 2022-12-23 21:19:24,922 INFO [train.py:894] (3/4) Epoch 24, batch 1800, loss[loss=0.186, simple_loss=0.2736, pruned_loss=0.04916, over 18485.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2613, pruned_loss=0.04349, over 3711805.54 frames. ], batch size: 64, lr: 4.77e-03, grad_scale: 8.0 2022-12-23 21:19:29,071 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.173e+02 4.124e+02 5.123e+02 6.216e+02 1.031e+03, threshold=1.025e+03, percent-clipped=5.0 2022-12-23 21:19:52,653 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp0.9 from training. Duration: 26.32775 2022-12-23 21:20:23,949 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0 from training. Duration: 20.025 2022-12-23 21:20:28,396 WARNING [train.py:1060] (3/4) Exclude cut with ID 6709-74022-0004-86860-0_sp1.1 from training. Duration: 0.9409375 2022-12-23 21:20:28,402 WARNING [train.py:1060] (3/4) Exclude cut with ID 4757-1811-0023-62229-0_sp0.9 from training. Duration: 21.37775 2022-12-23 21:20:40,156 INFO [train.py:894] (3/4) Epoch 24, batch 1850, loss[loss=0.1716, simple_loss=0.2527, pruned_loss=0.04522, over 18544.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2623, pruned_loss=0.04537, over 3713240.26 frames. ], batch size: 47, lr: 4.77e-03, grad_scale: 8.0 2022-12-23 21:20:43,245 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82493.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 21:20:49,406 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0004-25974-0_sp0.9 from training. Duration: 21.17225 2022-12-23 21:20:49,420 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp0.9 from training. Duration: 27.511125 2022-12-23 21:21:22,557 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0 from training. Duration: 22.8 2022-12-23 21:21:26,869 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0 from training. Duration: 22.585 2022-12-23 21:21:53,437 INFO [train.py:894] (3/4) Epoch 24, batch 1900, loss[loss=0.1456, simple_loss=0.2287, pruned_loss=0.03127, over 18529.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2628, pruned_loss=0.0466, over 3712973.07 frames. ], batch size: 47, lr: 4.77e-03, grad_scale: 16.0 2022-12-23 21:21:58,950 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.610e+02 4.545e+02 5.793e+02 7.153e+02 1.257e+03, threshold=1.159e+03, percent-clipped=6.0 2022-12-23 21:21:59,003 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0001-146967-0_sp0.9 from training. Duration: 22.0166875 2022-12-23 21:22:14,564 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82554.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 21:22:17,065 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp1.1 from training. Duration: 24.395375 2022-12-23 21:22:22,998 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp0.9 from training. Duration: 27.47775 2022-12-23 21:22:28,778 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp0.9 from training. Duration: 24.8833125 2022-12-23 21:22:30,258 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0 from training. Duration: 23.39 2022-12-23 21:22:36,493 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82569.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:22:37,625 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp0.9 from training. Duration: 28.72225 2022-12-23 21:22:48,497 WARNING [train.py:1060] (3/4) Exclude cut with ID 585-294811-0110-133686-0_sp0.9 from training. Duration: 20.8944375 2022-12-23 21:23:04,096 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0_sp0.9 from training. Duration: 23.8444375 2022-12-23 21:23:05,913 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5356, 1.3906, 1.4531, 1.3053, 0.8658, 2.2785, 0.7642, 1.4164], device='cuda:3'), covar=tensor([0.3256, 0.2263, 0.2142, 0.2351, 0.1724, 0.0363, 0.1886, 0.0931], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0118, 0.0125, 0.0122, 0.0105, 0.0097, 0.0091, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 21:23:08,471 INFO [train.py:894] (3/4) Epoch 24, batch 1950, loss[loss=0.1836, simple_loss=0.2697, pruned_loss=0.04872, over 18589.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2622, pruned_loss=0.04682, over 3712117.94 frames. ], batch size: 56, lr: 4.77e-03, grad_scale: 16.0 2022-12-23 21:23:27,358 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0 from training. Duration: 25.85 2022-12-23 21:23:27,369 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0 from training. Duration: 21.39 2022-12-23 21:23:37,508 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0 from training. Duration: 27.92 2022-12-23 21:24:06,475 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0039-130165-0_sp0.9 from training. Duration: 20.661125 2022-12-23 21:24:09,583 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82630.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:24:26,716 INFO [train.py:894] (3/4) Epoch 24, batch 2000, loss[loss=0.2034, simple_loss=0.2835, pruned_loss=0.06169, over 18581.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.263, pruned_loss=0.04784, over 3712849.31 frames. ], batch size: 187, lr: 4.77e-03, grad_scale: 16.0 2022-12-23 21:24:29,935 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82643.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:24:30,846 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.140e+02 4.292e+02 5.088e+02 6.391e+02 1.331e+03, threshold=1.018e+03, percent-clipped=1.0 2022-12-23 21:24:30,871 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0043-15874-0_sp0.9 from training. Duration: 20.07225 2022-12-23 21:24:37,222 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0 from training. Duration: 21.01 2022-12-23 21:25:42,843 INFO [train.py:894] (3/4) Epoch 24, batch 2050, loss[loss=0.1695, simple_loss=0.2432, pruned_loss=0.04789, over 18492.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2638, pruned_loss=0.04861, over 3713034.68 frames. ], batch size: 43, lr: 4.76e-03, grad_scale: 16.0 2022-12-23 21:25:43,047 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=82691.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:25:47,392 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0 from training. Duration: 20.65 2022-12-23 21:25:53,898 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0 from training. Duration: 21.46 2022-12-23 21:25:59,844 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82702.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:26:01,651 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.0407, 1.0057, 0.9848, 1.1658, 1.2568, 1.1536, 1.1204, 0.9514], device='cuda:3'), covar=tensor([0.0325, 0.0254, 0.0575, 0.0222, 0.0261, 0.0415, 0.0268, 0.0322], device='cuda:3'), in_proj_covar=tensor([0.0096, 0.0126, 0.0152, 0.0124, 0.0116, 0.0121, 0.0100, 0.0126], device='cuda:3'), out_proj_covar=tensor([7.6206e-05, 9.9736e-05, 1.2490e-04, 9.8819e-05, 9.3411e-05, 9.3153e-05, 7.7659e-05, 9.9693e-05], device='cuda:3') 2022-12-23 21:26:12,832 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82711.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:26:40,179 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0 from training. Duration: 0.92 2022-12-23 21:26:48,030 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0_sp0.9 from training. Duration: 23.7666875 2022-12-23 21:26:57,936 INFO [train.py:894] (3/4) Epoch 24, batch 2100, loss[loss=0.1773, simple_loss=0.2616, pruned_loss=0.0465, over 18553.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2637, pruned_loss=0.04846, over 3712653.38 frames. ], batch size: 99, lr: 4.76e-03, grad_scale: 16.0 2022-12-23 21:27:03,148 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.401e+02 4.641e+02 5.481e+02 6.606e+02 1.267e+03, threshold=1.096e+03, percent-clipped=3.0 2022-12-23 21:27:22,186 WARNING [train.py:1060] (3/4) Exclude cut with ID 8544-281189-0060-101339-0_sp0.9 from training. Duration: 20.861125 2022-12-23 21:27:25,210 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=82759.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:27:33,536 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0_sp0.9 from training. Duration: 22.711125 2022-12-23 21:28:10,056 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8908, 1.8874, 2.1203, 1.1790, 2.1848, 2.2194, 1.6137, 2.5727], device='cuda:3'), covar=tensor([0.1053, 0.1755, 0.1279, 0.1978, 0.0669, 0.1121, 0.2223, 0.0517], device='cuda:3'), in_proj_covar=tensor([0.0199, 0.0216, 0.0208, 0.0195, 0.0176, 0.0217, 0.0216, 0.0196], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 21:28:14,012 INFO [train.py:894] (3/4) Epoch 24, batch 2150, loss[loss=0.1857, simple_loss=0.2701, pruned_loss=0.0507, over 18702.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.264, pruned_loss=0.04877, over 3713083.97 frames. ], batch size: 78, lr: 4.76e-03, grad_scale: 16.0 2022-12-23 21:28:16,084 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp1.1 from training. Duration: 22.986375 2022-12-23 21:28:26,905 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2022-12-23 21:28:32,495 WARNING [train.py:1060] (3/4) Exclude cut with ID 8040-260924-0003-80960-0_sp0.9 from training. Duration: 22.07225 2022-12-23 21:28:36,511 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0045-26330-0_sp0.9 from training. Duration: 20.3055625 2022-12-23 21:28:39,174 WARNING [train.py:1060] (3/4) Exclude cut with ID 6356-271890-0060-94317-0_sp0.9 from training. Duration: 20.72225 2022-12-23 21:28:56,004 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp1.1 from training. Duration: 22.4818125 2022-12-23 21:29:00,261 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2022-12-23 21:29:22,902 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp0.9 from training. Duration: 25.0944375 2022-12-23 21:29:26,241 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0 from training. Duration: 21.515 2022-12-23 21:29:30,883 INFO [train.py:894] (3/4) Epoch 24, batch 2200, loss[loss=0.1689, simple_loss=0.2575, pruned_loss=0.04014, over 18653.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2638, pruned_loss=0.04909, over 3713877.98 frames. ], batch size: 97, lr: 4.76e-03, grad_scale: 16.0 2022-12-23 21:29:33,925 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp0.9 from training. Duration: 27.02225 2022-12-23 21:29:35,207 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.795e+02 4.049e+02 5.149e+02 6.445e+02 1.343e+03, threshold=1.030e+03, percent-clipped=3.0 2022-12-23 21:29:38,115 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0010-62480-0_sp0.9 from training. Duration: 22.22225 2022-12-23 21:29:42,535 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82849.0, num_to_drop=1, layers_to_drop={3} 2022-12-23 21:29:45,340 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0085-44554-0_sp0.9 from training. Duration: 20.85 2022-12-23 21:30:17,867 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0 from training. Duration: 21.54 2022-12-23 21:30:21,319 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp1.1 from training. Duration: 20.5318125 2022-12-23 21:30:30,139 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0012-134311-0_sp0.9 from training. Duration: 21.9333125 2022-12-23 21:30:48,587 INFO [train.py:894] (3/4) Epoch 24, batch 2250, loss[loss=0.1472, simple_loss=0.2326, pruned_loss=0.03094, over 18676.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2646, pruned_loss=0.05006, over 3712958.46 frames. ], batch size: 48, lr: 4.76e-03, grad_scale: 16.0 2022-12-23 21:31:02,226 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2022-12-23 21:31:18,796 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0025-130151-0_sp0.9 from training. Duration: 21.7944375 2022-12-23 21:31:22,726 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5356, 1.9857, 2.1422, 2.2299, 2.4089, 2.3774, 2.3340, 1.9805], device='cuda:3'), covar=tensor([0.2162, 0.3309, 0.2546, 0.2903, 0.2017, 0.0998, 0.3518, 0.1323], device='cuda:3'), in_proj_covar=tensor([0.0269, 0.0299, 0.0283, 0.0320, 0.0311, 0.0257, 0.0348, 0.0244], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 21:31:30,012 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0_sp0.9 from training. Duration: 22.4666875 2022-12-23 21:31:35,885 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0 from training. Duration: 21.635 2022-12-23 21:31:40,546 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0_sp0.9 from training. Duration: 24.038875 2022-12-23 21:31:40,680 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82925.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:32:06,056 INFO [train.py:894] (3/4) Epoch 24, batch 2300, loss[loss=0.1719, simple_loss=0.2473, pruned_loss=0.04824, over 18527.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2647, pruned_loss=0.04978, over 3713733.68 frames. ], batch size: 44, lr: 4.76e-03, grad_scale: 16.0 2022-12-23 21:32:10,496 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.645e+02 4.227e+02 4.886e+02 6.108e+02 1.315e+03, threshold=9.771e+02, percent-clipped=2.0 2022-12-23 21:32:23,989 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp1.1 from training. Duration: 21.786375 2022-12-23 21:32:36,999 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0 from training. Duration: 20.22 2022-12-23 21:32:40,223 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82963.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:33:21,974 INFO [train.py:894] (3/4) Epoch 24, batch 2350, loss[loss=0.1472, simple_loss=0.2241, pruned_loss=0.03516, over 18654.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2643, pruned_loss=0.04949, over 3714154.95 frames. ], batch size: 41, lr: 4.76e-03, grad_scale: 16.0 2022-12-23 21:33:36,768 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2022-12-23 21:33:39,469 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83002.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:34:12,393 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83024.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:34:37,748 INFO [train.py:894] (3/4) Epoch 24, batch 2400, loss[loss=0.184, simple_loss=0.2727, pruned_loss=0.04767, over 18579.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2639, pruned_loss=0.04905, over 3714206.66 frames. ], batch size: 57, lr: 4.75e-03, grad_scale: 16.0 2022-12-23 21:34:39,321 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0 from training. Duration: 25.285 2022-12-23 21:34:41,926 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.252e+02 3.943e+02 4.892e+02 6.290e+02 1.556e+03, threshold=9.783e+02, percent-clipped=4.0 2022-12-23 21:34:51,793 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=83050.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:35:28,449 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2022-12-23 21:35:31,209 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1751, 1.4063, 1.8896, 1.8401, 2.1888, 2.2095, 1.9612, 1.8232], device='cuda:3'), covar=tensor([0.2287, 0.3394, 0.2655, 0.2839, 0.2055, 0.0996, 0.3146, 0.1339], device='cuda:3'), in_proj_covar=tensor([0.0269, 0.0300, 0.0284, 0.0321, 0.0313, 0.0257, 0.0349, 0.0244], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 21:35:45,767 WARNING [train.py:1060] (3/4) Exclude cut with ID 811-130148-0001-63453-0_sp0.9 from training. Duration: 20.861125 2022-12-23 21:35:54,371 INFO [train.py:894] (3/4) Epoch 24, batch 2450, loss[loss=0.1776, simple_loss=0.2635, pruned_loss=0.04581, over 18412.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2637, pruned_loss=0.04946, over 3714125.65 frames. ], batch size: 48, lr: 4.75e-03, grad_scale: 16.0 2022-12-23 21:35:56,847 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2022-12-23 21:36:03,231 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2022-12-23 21:36:06,761 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0 from training. Duration: 20.88 2022-12-23 21:36:07,223 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.9481, 2.2387, 1.6698, 2.9251, 2.1812, 2.0714, 2.2239, 3.0054], device='cuda:3'), covar=tensor([0.2288, 0.3495, 0.2379, 0.3198, 0.4196, 0.1333, 0.3736, 0.0911], device='cuda:3'), in_proj_covar=tensor([0.0299, 0.0296, 0.0249, 0.0348, 0.0276, 0.0232, 0.0291, 0.0219], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 21:36:34,756 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.0622, 0.9899, 1.1891, 0.6017, 0.5686, 1.1959, 1.1897, 1.1843], device='cuda:3'), covar=tensor([0.0739, 0.0373, 0.0348, 0.0395, 0.0465, 0.0510, 0.0294, 0.0597], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0174, 0.0131, 0.0143, 0.0150, 0.0145, 0.0167, 0.0178], device='cuda:3'), out_proj_covar=tensor([1.1419e-04, 1.3175e-04, 9.7098e-05, 1.0537e-04, 1.1060e-04, 1.0981e-04, 1.2708e-04, 1.3425e-04], device='cuda:3') 2022-12-23 21:36:40,293 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0_sp0.9 from training. Duration: 23.4166875 2022-12-23 21:37:10,375 INFO [train.py:894] (3/4) Epoch 24, batch 2500, loss[loss=0.1761, simple_loss=0.2566, pruned_loss=0.04782, over 18682.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2636, pruned_loss=0.04913, over 3715551.51 frames. ], batch size: 50, lr: 4.75e-03, grad_scale: 16.0 2022-12-23 21:37:15,185 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.926e+02 4.415e+02 5.249e+02 6.665e+02 1.083e+03, threshold=1.050e+03, percent-clipped=2.0 2022-12-23 21:37:22,665 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83149.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 21:37:57,655 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0 from training. Duration: 21.24 2022-12-23 21:37:57,670 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0_sp0.9 from training. Duration: 23.9055625 2022-12-23 21:38:05,869 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2022-12-23 21:38:24,788 INFO [train.py:894] (3/4) Epoch 24, batch 2550, loss[loss=0.1676, simple_loss=0.2537, pruned_loss=0.04077, over 18521.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2633, pruned_loss=0.04916, over 3716530.12 frames. ], batch size: 58, lr: 4.75e-03, grad_scale: 16.0 2022-12-23 21:38:31,597 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp0.9 from training. Duration: 25.988875 2022-12-23 21:38:33,467 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83196.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:38:34,641 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=83197.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 21:38:41,005 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0001-134300-0_sp0.9 from training. Duration: 20.67225 2022-12-23 21:38:43,255 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2022-12-23 21:39:17,923 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83225.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:39:23,526 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2032, 2.2035, 1.7553, 2.3120, 2.4404, 2.0845, 2.8550, 2.2978], device='cuda:3'), covar=tensor([0.0810, 0.1498, 0.2534, 0.1604, 0.1578, 0.0824, 0.0891, 0.1160], device='cuda:3'), in_proj_covar=tensor([0.0179, 0.0213, 0.0254, 0.0290, 0.0239, 0.0193, 0.0206, 0.0206], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 21:39:27,567 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0 from training. Duration: 20.34 2022-12-23 21:39:41,996 INFO [train.py:894] (3/4) Epoch 24, batch 2600, loss[loss=0.1733, simple_loss=0.2504, pruned_loss=0.04813, over 18521.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2634, pruned_loss=0.04961, over 3715158.47 frames. ], batch size: 47, lr: 4.75e-03, grad_scale: 8.0 2022-12-23 21:39:47,858 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.557e+02 4.021e+02 4.692e+02 5.692e+02 9.573e+02, threshold=9.385e+02, percent-clipped=0.0 2022-12-23 21:40:07,049 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83257.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:40:26,927 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2022-12-23 21:40:34,320 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=83273.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:40:34,598 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.6203, 3.0304, 2.7377, 1.4413, 2.5237, 2.4407, 2.2785, 2.7931], device='cuda:3'), covar=tensor([0.0580, 0.0627, 0.1323, 0.1759, 0.1556, 0.1399, 0.1493, 0.0925], device='cuda:3'), in_proj_covar=tensor([0.0173, 0.0187, 0.0205, 0.0189, 0.0209, 0.0202, 0.0216, 0.0201], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 21:40:42,873 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp0.9 from training. Duration: 25.061125 2022-12-23 21:40:51,180 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83284.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:40:55,173 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0 from training. Duration: 0.83 2022-12-23 21:41:01,169 INFO [train.py:894] (3/4) Epoch 24, batch 2650, loss[loss=0.1626, simple_loss=0.2398, pruned_loss=0.04271, over 18459.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2637, pruned_loss=0.04933, over 3713339.57 frames. ], batch size: 43, lr: 4.75e-03, grad_scale: 8.0 2022-12-23 21:41:19,913 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0 from training. Duration: 24.73 2022-12-23 21:41:31,835 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0 from training. Duration: 23.965 2022-12-23 21:41:40,147 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0030-146996-0_sp0.9 from training. Duration: 22.088875 2022-12-23 21:41:44,592 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6270, 4.0607, 3.8701, 1.6355, 4.1606, 3.0304, 0.5946, 2.7201], device='cuda:3'), covar=tensor([0.2002, 0.0950, 0.1377, 0.3386, 0.0869, 0.0938, 0.5073, 0.1408], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0146, 0.0161, 0.0126, 0.0148, 0.0116, 0.0145, 0.0115], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 21:41:44,595 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83319.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:41:57,047 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0_sp0.9 from training. Duration: 23.6 2022-12-23 21:42:18,622 INFO [train.py:894] (3/4) Epoch 24, batch 2700, loss[loss=0.1556, simple_loss=0.2382, pruned_loss=0.03646, over 18530.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2635, pruned_loss=0.04966, over 3714218.24 frames. ], batch size: 47, lr: 4.75e-03, grad_scale: 8.0 2022-12-23 21:42:21,464 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2022-12-23 21:42:24,631 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.744e+02 4.043e+02 4.781e+02 6.098e+02 9.863e+02, threshold=9.563e+02, percent-clipped=3.0 2022-12-23 21:42:25,187 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83345.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:42:45,290 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.0014, 5.5878, 5.1174, 2.5339, 5.5638, 4.2792, 1.1333, 4.0838], device='cuda:3'), covar=tensor([0.1940, 0.0932, 0.1355, 0.3208, 0.0769, 0.0720, 0.4935, 0.1109], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0147, 0.0161, 0.0126, 0.0149, 0.0116, 0.0146, 0.0115], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 21:43:35,054 INFO [train.py:894] (3/4) Epoch 24, batch 2750, loss[loss=0.1747, simple_loss=0.254, pruned_loss=0.04765, over 18580.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2641, pruned_loss=0.04986, over 3714824.17 frames. ], batch size: 51, lr: 4.74e-03, grad_scale: 8.0 2022-12-23 21:43:35,119 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0 from training. Duration: 23.795 2022-12-23 21:43:50,848 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp1.1 from training. Duration: 21.5409375 2022-12-23 21:43:54,035 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp0.9 from training. Duration: 24.97775 2022-12-23 21:44:04,748 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0_sp0.9 from training. Duration: 23.3444375 2022-12-23 21:44:33,506 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0_sp0.9 from training. Duration: 23.2 2022-12-23 21:44:38,920 WARNING [train.py:1060] (3/4) Exclude cut with ID 5653-46179-0060-117930-0_sp0.9 from training. Duration: 21.17225 2022-12-23 21:44:51,834 INFO [train.py:894] (3/4) Epoch 24, batch 2800, loss[loss=0.162, simple_loss=0.2379, pruned_loss=0.04306, over 18718.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2637, pruned_loss=0.04934, over 3715692.81 frames. ], batch size: 46, lr: 4.74e-03, grad_scale: 8.0 2022-12-23 21:44:57,540 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.010e+02 4.230e+02 5.439e+02 6.718e+02 1.478e+03, threshold=1.088e+03, percent-clipped=3.0 2022-12-23 21:44:58,916 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp0.9 from training. Duration: 24.6555625 2022-12-23 21:45:29,973 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9481, 0.7136, 1.7977, 1.5147, 1.9662, 2.0033, 1.5977, 1.7904], device='cuda:3'), covar=tensor([0.2161, 0.3163, 0.2461, 0.2598, 0.1974, 0.0979, 0.2930, 0.1284], device='cuda:3'), in_proj_covar=tensor([0.0271, 0.0299, 0.0283, 0.0319, 0.0312, 0.0257, 0.0349, 0.0244], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 21:45:53,873 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0 from training. Duration: 20.44 2022-12-23 21:46:08,183 INFO [train.py:894] (3/4) Epoch 24, batch 2850, loss[loss=0.1782, simple_loss=0.2737, pruned_loss=0.04133, over 18730.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2637, pruned_loss=0.0493, over 3716074.94 frames. ], batch size: 52, lr: 4.74e-03, grad_scale: 8.0 2022-12-23 21:46:09,834 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0_sp0.9 from training. Duration: 23.45 2022-12-23 21:46:18,367 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6103, 2.1358, 2.1793, 2.2095, 2.4216, 2.4501, 2.3763, 2.1092], device='cuda:3'), covar=tensor([0.2100, 0.3346, 0.2614, 0.3097, 0.2181, 0.1053, 0.3657, 0.1371], device='cuda:3'), in_proj_covar=tensor([0.0271, 0.0301, 0.0284, 0.0321, 0.0313, 0.0258, 0.0350, 0.0244], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 21:46:35,538 WARNING [train.py:1060] (3/4) Exclude cut with ID 6945-60535-0076-12784-0_sp0.9 from training. Duration: 20.52225 2022-12-23 21:46:43,301 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0 from training. Duration: 22.19 2022-12-23 21:46:52,204 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp1.1 from training. Duration: 25.3818125 2022-12-23 21:46:55,645 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7359, 2.2339, 1.7853, 0.8918, 1.8244, 2.2011, 1.7651, 2.0827], device='cuda:3'), covar=tensor([0.0605, 0.0671, 0.1279, 0.1652, 0.1239, 0.1337, 0.1700, 0.0848], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0186, 0.0204, 0.0188, 0.0207, 0.0201, 0.0214, 0.0200], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 21:47:07,790 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp0.9 from training. Duration: 28.0944375 2022-12-23 21:47:13,876 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0_sp0.9 from training. Duration: 22.9444375 2022-12-23 21:47:23,318 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp1.1 from training. Duration: 21.6318125 2022-12-23 21:47:26,369 INFO [train.py:894] (3/4) Epoch 24, batch 2900, loss[loss=0.1864, simple_loss=0.2735, pruned_loss=0.04969, over 18688.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2639, pruned_loss=0.04978, over 3715559.55 frames. ], batch size: 99, lr: 4.74e-03, grad_scale: 8.0 2022-12-23 21:47:31,835 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.246e+02 4.184e+02 5.440e+02 6.811e+02 1.934e+03, threshold=1.088e+03, percent-clipped=6.0 2022-12-23 21:47:39,870 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0 from training. Duration: 23.695 2022-12-23 21:47:43,545 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83552.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:48:05,025 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2022-12-23 21:48:05,419 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0 from training. Duration: 23.955 2022-12-23 21:48:17,536 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5456, 1.9583, 2.1549, 2.2039, 2.4602, 2.3962, 2.3501, 1.9673], device='cuda:3'), covar=tensor([0.2159, 0.3317, 0.2534, 0.2908, 0.2067, 0.0978, 0.3477, 0.1368], device='cuda:3'), in_proj_covar=tensor([0.0272, 0.0301, 0.0284, 0.0321, 0.0313, 0.0258, 0.0350, 0.0245], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 21:48:40,615 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp0.9 from training. Duration: 26.438875 2022-12-23 21:48:42,068 INFO [train.py:894] (3/4) Epoch 24, batch 2950, loss[loss=0.163, simple_loss=0.2428, pruned_loss=0.04165, over 18673.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2635, pruned_loss=0.04929, over 3715212.95 frames. ], batch size: 46, lr: 4.74e-03, grad_scale: 8.0 2022-12-23 21:49:25,363 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83619.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:49:26,586 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0021-26306-0_sp0.9 from training. Duration: 21.2444375 2022-12-23 21:49:27,964 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp0.9 from training. Duration: 31.02225 2022-12-23 21:49:29,670 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4777, 1.2066, 1.8582, 3.0783, 2.1392, 2.3570, 0.7388, 2.3070], device='cuda:3'), covar=tensor([0.2074, 0.2054, 0.1714, 0.0894, 0.1195, 0.1430, 0.2685, 0.1237], device='cuda:3'), in_proj_covar=tensor([0.0103, 0.0117, 0.0136, 0.0152, 0.0105, 0.0144, 0.0129, 0.0113], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 21:49:33,887 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8297, 1.5992, 1.7580, 1.6519, 1.2263, 3.6022, 1.5473, 1.9733], device='cuda:3'), covar=tensor([0.2994, 0.2055, 0.1936, 0.2138, 0.1500, 0.0205, 0.1586, 0.0849], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0118, 0.0125, 0.0121, 0.0105, 0.0097, 0.0091, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 21:49:39,488 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0 from training. Duration: 22.395 2022-12-23 21:49:57,035 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0 from training. Duration: 21.075 2022-12-23 21:49:57,205 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83640.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:49:58,473 INFO [train.py:894] (3/4) Epoch 24, batch 3000, loss[loss=0.1523, simple_loss=0.2312, pruned_loss=0.03666, over 18527.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2634, pruned_loss=0.04905, over 3714901.31 frames. ], batch size: 44, lr: 4.74e-03, grad_scale: 8.0 2022-12-23 21:49:58,473 INFO [train.py:919] (3/4) Computing validation loss 2022-12-23 21:50:09,481 INFO [train.py:928] (3/4) Epoch 24, validation: loss=0.1636, simple_loss=0.2603, pruned_loss=0.03339, over 944034.00 frames. 2022-12-23 21:50:09,482 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-23 21:50:13,150 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2022-12-23 21:50:13,780 WARNING [train.py:1060] (3/4) Exclude cut with ID 6482-98857-0025-147532-0_sp0.9 from training. Duration: 20.0055625 2022-12-23 21:50:13,787 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0037-132304-0_sp0.9 from training. Duration: 22.05 2022-12-23 21:50:13,802 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0 from training. Duration: 26.8349375 2022-12-23 21:50:15,121 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.874e+02 4.185e+02 5.061e+02 6.325e+02 1.554e+03, threshold=1.012e+03, percent-clipped=3.0 2022-12-23 21:50:18,099 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp1.1 from training. Duration: 22.1090625 2022-12-23 21:50:23,872 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp0.9 from training. Duration: 26.6166875 2022-12-23 21:50:43,414 WARNING [train.py:1060] (3/4) Exclude cut with ID 2046-178027-0000-53705-0_sp0.9 from training. Duration: 20.3055625 2022-12-23 21:50:48,990 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=83667.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:51:03,977 WARNING [train.py:1060] (3/4) Exclude cut with ID 7205-50138-0008-5373-0_sp0.9 from training. Duration: 20.7 2022-12-23 21:51:24,966 INFO [train.py:894] (3/4) Epoch 24, batch 3050, loss[loss=0.157, simple_loss=0.2442, pruned_loss=0.03483, over 18428.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.263, pruned_loss=0.04877, over 3713598.06 frames. ], batch size: 48, lr: 4.74e-03, grad_scale: 8.0 2022-12-23 21:51:29,938 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0373, 1.7013, 1.9990, 2.5419, 2.1843, 4.5842, 1.7397, 1.6763], device='cuda:3'), covar=tensor([0.0773, 0.1747, 0.0998, 0.0836, 0.1339, 0.0217, 0.1359, 0.1577], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0082, 0.0072, 0.0074, 0.0091, 0.0077, 0.0084, 0.0076], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2022-12-23 21:51:31,638 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.9924, 2.4891, 2.5065, 1.8232, 2.5822, 2.3012, 2.0428, 2.4888], device='cuda:3'), covar=tensor([0.0840, 0.0852, 0.1642, 0.2010, 0.1547, 0.1381, 0.1782, 0.1087], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0190, 0.0208, 0.0192, 0.0211, 0.0203, 0.0218, 0.0203], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 21:51:48,734 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0 from training. Duration: 22.48 2022-12-23 21:52:05,713 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp0.9 from training. Duration: 29.816625 2022-12-23 21:52:08,195 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2022-12-23 21:52:24,369 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp1.1 from training. Duration: 22.7590625 2022-12-23 21:52:28,547 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0 from training. Duration: 22.555 2022-12-23 21:52:40,211 INFO [train.py:894] (3/4) Epoch 24, batch 3100, loss[loss=0.1626, simple_loss=0.2483, pruned_loss=0.03838, over 18666.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2632, pruned_loss=0.04893, over 3714117.46 frames. ], batch size: 48, lr: 4.73e-03, grad_scale: 8.0 2022-12-23 21:52:44,005 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.24 vs. limit=5.0 2022-12-23 21:52:46,094 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.916e+02 4.123e+02 4.986e+02 6.362e+02 1.111e+03, threshold=9.972e+02, percent-clipped=1.0 2022-12-23 21:52:50,244 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0005-25975-0_sp0.9 from training. Duration: 21.688875 2022-12-23 21:53:09,672 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83760.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:53:28,008 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0_sp0.9 from training. Duration: 22.6 2022-12-23 21:53:56,001 INFO [train.py:894] (3/4) Epoch 24, batch 3150, loss[loss=0.2005, simple_loss=0.2823, pruned_loss=0.05935, over 18480.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2627, pruned_loss=0.0486, over 3714205.61 frames. ], batch size: 77, lr: 4.73e-03, grad_scale: 8.0 2022-12-23 21:53:57,973 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.9245, 2.6790, 2.1645, 0.9845, 2.0754, 2.3115, 1.9087, 2.3609], device='cuda:3'), covar=tensor([0.0619, 0.0505, 0.1291, 0.1812, 0.1315, 0.1375, 0.1625, 0.0789], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0189, 0.0208, 0.0192, 0.0211, 0.0204, 0.0218, 0.0203], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 21:54:04,912 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0 from training. Duration: 24.32 2022-12-23 21:54:12,695 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.5844, 1.4241, 1.4647, 1.0377, 1.0793, 1.5034, 1.5396, 1.3146], device='cuda:3'), covar=tensor([0.0653, 0.0282, 0.0271, 0.0343, 0.0328, 0.0433, 0.0217, 0.0575], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0172, 0.0129, 0.0141, 0.0148, 0.0143, 0.0165, 0.0177], device='cuda:3'), out_proj_covar=tensor([1.1274e-04, 1.3041e-04, 9.5528e-05, 1.0437e-04, 1.0894e-04, 1.0857e-04, 1.2558e-04, 1.3344e-04], device='cuda:3') 2022-12-23 21:54:43,384 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83821.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:55:01,071 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-276745-0093-13116-0_sp0.9 from training. Duration: 21.061125 2022-12-23 21:55:13,155 INFO [train.py:894] (3/4) Epoch 24, batch 3200, loss[loss=0.1748, simple_loss=0.254, pruned_loss=0.04785, over 18562.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2619, pruned_loss=0.04828, over 3714472.39 frames. ], batch size: 49, lr: 4.73e-03, grad_scale: 8.0 2022-12-23 21:55:13,278 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0024-15855-0_sp0.9 from training. Duration: 20.32225 2022-12-23 21:55:19,613 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.688e+02 4.275e+02 5.159e+02 6.611e+02 1.406e+03, threshold=1.032e+03, percent-clipped=3.0 2022-12-23 21:55:20,755 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.3323, 1.6441, 1.8651, 1.2586, 1.2818, 1.9758, 1.8804, 1.5613], device='cuda:3'), covar=tensor([0.0751, 0.0327, 0.0327, 0.0329, 0.0394, 0.0517, 0.0235, 0.0688], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0173, 0.0130, 0.0142, 0.0149, 0.0144, 0.0166, 0.0178], device='cuda:3'), out_proj_covar=tensor([1.1325e-04, 1.3110e-04, 9.6006e-05, 1.0496e-04, 1.0958e-04, 1.0917e-04, 1.2635e-04, 1.3418e-04], device='cuda:3') 2022-12-23 21:55:25,879 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp1.1 from training. Duration: 0.7545625 2022-12-23 21:55:29,429 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9605, 1.5798, 1.9157, 1.8800, 1.6057, 5.0993, 1.9873, 2.6685], device='cuda:3'), covar=tensor([0.2916, 0.2048, 0.1891, 0.2043, 0.1342, 0.0103, 0.1433, 0.0744], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0117, 0.0125, 0.0121, 0.0105, 0.0097, 0.0091, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 21:55:30,679 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83852.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:55:40,066 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0_sp0.9 from training. Duration: 23.9333125 2022-12-23 21:56:12,151 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp1.1 from training. Duration: 20.17275 2022-12-23 21:56:15,064 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp1.1 from training. Duration: 20.436375 2022-12-23 21:56:31,574 INFO [train.py:894] (3/4) Epoch 24, batch 3250, loss[loss=0.1863, simple_loss=0.2747, pruned_loss=0.04893, over 18666.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2619, pruned_loss=0.04843, over 3714388.13 frames. ], batch size: 60, lr: 4.73e-03, grad_scale: 8.0 2022-12-23 21:56:45,370 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=83900.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:57:21,644 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.3113, 1.6200, 1.9373, 1.1200, 1.2419, 1.9551, 1.9394, 1.6037], device='cuda:3'), covar=tensor([0.0709, 0.0340, 0.0312, 0.0348, 0.0387, 0.0406, 0.0207, 0.0674], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0173, 0.0130, 0.0142, 0.0149, 0.0145, 0.0166, 0.0178], device='cuda:3'), out_proj_covar=tensor([1.1343e-04, 1.3132e-04, 9.6146e-05, 1.0501e-04, 1.0994e-04, 1.0950e-04, 1.2655e-04, 1.3414e-04], device='cuda:3') 2022-12-23 21:57:30,914 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.56 vs. limit=5.0 2022-12-23 21:57:39,301 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0_sp0.9 from training. Duration: 23.1055625 2022-12-23 21:57:40,790 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp1.1 from training. Duration: 23.5 2022-12-23 21:57:44,634 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83940.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:57:45,943 INFO [train.py:894] (3/4) Epoch 24, batch 3300, loss[loss=0.1612, simple_loss=0.2452, pruned_loss=0.03864, over 18701.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2622, pruned_loss=0.04851, over 3714196.43 frames. ], batch size: 46, lr: 4.73e-03, grad_scale: 8.0 2022-12-23 21:57:51,364 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.326e+02 4.315e+02 5.085e+02 6.619e+02 1.531e+03, threshold=1.017e+03, percent-clipped=2.0 2022-12-23 21:57:52,705 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp0.9 from training. Duration: 26.62775 2022-12-23 21:57:58,183 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83949.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:58:07,498 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0 from training. Duration: 21.105 2022-12-23 21:58:10,496 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0_sp0.9 from training. Duration: 24.411125 2022-12-23 21:58:28,882 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5335, 1.3515, 0.9738, 0.2938, 1.0050, 1.4151, 1.2488, 1.3190], device='cuda:3'), covar=tensor([0.0669, 0.0546, 0.1017, 0.1568, 0.1037, 0.1480, 0.1587, 0.0687], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0190, 0.0210, 0.0193, 0.0212, 0.0206, 0.0219, 0.0205], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 21:58:38,815 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp1.1 from training. Duration: 21.263625 2022-12-23 21:58:54,237 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83985.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:58:58,260 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=83988.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 21:59:02,661 INFO [train.py:894] (3/4) Epoch 24, batch 3350, loss[loss=0.1841, simple_loss=0.2732, pruned_loss=0.04753, over 18521.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2616, pruned_loss=0.04808, over 3713784.91 frames. ], batch size: 58, lr: 4.73e-03, grad_scale: 8.0 2022-12-23 21:59:10,453 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0 from training. Duration: 20.795 2022-12-23 21:59:25,141 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0 from training. Duration: 24.76 2022-12-23 21:59:25,160 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0_sp0.9 from training. Duration: 22.25 2022-12-23 21:59:35,986 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84010.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 21:59:49,894 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp1.1 from training. Duration: 20.5045625 2022-12-23 22:00:21,132 INFO [train.py:894] (3/4) Epoch 24, batch 3400, loss[loss=0.2021, simple_loss=0.2806, pruned_loss=0.06182, over 18628.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2616, pruned_loss=0.04807, over 3714065.38 frames. ], batch size: 53, lr: 4.73e-03, grad_scale: 8.0 2022-12-23 22:00:27,538 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.020e+02 4.214e+02 5.258e+02 6.086e+02 1.654e+03, threshold=1.052e+03, percent-clipped=1.0 2022-12-23 22:00:29,298 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84046.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:01:33,007 INFO [train.py:894] (3/4) Epoch 24, batch 3450, loss[loss=0.1563, simple_loss=0.231, pruned_loss=0.04085, over 18486.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2626, pruned_loss=0.0487, over 3714918.22 frames. ], batch size: 43, lr: 4.72e-03, grad_scale: 8.0 2022-12-23 22:01:41,772 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84097.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:01:53,779 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5700, 4.1048, 3.8946, 1.7018, 4.2283, 3.1858, 0.8425, 2.7832], device='cuda:3'), covar=tensor([0.2218, 0.1298, 0.1405, 0.3577, 0.0883, 0.0862, 0.4932, 0.1411], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0149, 0.0164, 0.0126, 0.0151, 0.0116, 0.0147, 0.0117], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 22:02:09,386 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84116.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:02:45,419 INFO [train.py:894] (3/4) Epoch 24, batch 3500, loss[loss=0.21, simple_loss=0.2884, pruned_loss=0.06585, over 18624.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2633, pruned_loss=0.04891, over 3714226.55 frames. ], batch size: 175, lr: 4.72e-03, grad_scale: 8.0 2022-12-23 22:02:52,081 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.085e+02 4.483e+02 5.516e+02 7.129e+02 1.524e+03, threshold=1.103e+03, percent-clipped=5.0 2022-12-23 22:03:07,402 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp1.1 from training. Duration: 22.2954375 2022-12-23 22:03:16,195 INFO [train.py:894] (3/4) Epoch 25, batch 0, loss[loss=0.1654, simple_loss=0.2544, pruned_loss=0.03818, over 18714.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2544, pruned_loss=0.03818, over 18714.00 frames. ], batch size: 50, lr: 4.63e-03, grad_scale: 8.0 2022-12-23 22:03:16,195 INFO [train.py:919] (3/4) Computing validation loss 2022-12-23 22:03:27,095 INFO [train.py:928] (3/4) Epoch 25, validation: loss=0.1626, simple_loss=0.2599, pruned_loss=0.03263, over 944034.00 frames. 2022-12-23 22:03:27,096 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-23 22:03:44,426 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84158.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:03:50,749 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.17 vs. limit=5.0 2022-12-23 22:04:20,782 WARNING [train.py:1060] (3/4) Exclude cut with ID 298-126791-0067-24026-0_sp0.9 from training. Duration: 21.438875 2022-12-23 22:04:25,163 WARNING [train.py:1060] (3/4) Exclude cut with ID 5652-39938-0025-23684-0_sp0.9 from training. Duration: 22.2055625 2022-12-23 22:04:41,761 INFO [train.py:894] (3/4) Epoch 25, batch 50, loss[loss=0.1764, simple_loss=0.2721, pruned_loss=0.04033, over 18507.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2598, pruned_loss=0.04074, over 837942.90 frames. ], batch size: 77, lr: 4.62e-03, grad_scale: 8.0 2022-12-23 22:04:45,773 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84199.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:05:40,598 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84236.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:05:54,057 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.027e+02 3.187e+02 3.567e+02 4.493e+02 9.384e+02, threshold=7.134e+02, percent-clipped=0.0 2022-12-23 22:05:56,769 INFO [train.py:894] (3/4) Epoch 25, batch 100, loss[loss=0.1467, simple_loss=0.2408, pruned_loss=0.0263, over 18397.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2594, pruned_loss=0.04102, over 1474395.89 frames. ], batch size: 46, lr: 4.62e-03, grad_scale: 8.0 2022-12-23 22:06:10,192 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2022-12-23 22:06:16,930 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84260.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:07:10,954 INFO [train.py:894] (3/4) Epoch 25, batch 150, loss[loss=0.1829, simple_loss=0.2706, pruned_loss=0.04761, over 18457.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2592, pruned_loss=0.041, over 1970505.62 frames. ], batch size: 64, lr: 4.62e-03, grad_scale: 8.0 2022-12-23 22:07:11,457 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84297.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:07:23,447 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84305.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 22:07:27,705 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0 from training. Duration: 24.525 2022-12-23 22:07:31,012 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7596, 1.2460, 0.8422, 1.3893, 2.0709, 1.2095, 1.4618, 1.6396], device='cuda:3'), covar=tensor([0.1686, 0.2256, 0.2276, 0.1568, 0.1895, 0.1887, 0.1532, 0.1817], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0098, 0.0117, 0.0096, 0.0119, 0.0092, 0.0098, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 22:08:02,200 WARNING [train.py:1060] (3/4) Exclude cut with ID 3699-47246-0007-3408-0_sp0.9 from training. Duration: 20.26675 2022-12-23 22:08:10,516 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84336.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:08:16,175 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp0.9 from training. Duration: 27.25 2022-12-23 22:08:17,453 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84341.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:08:23,124 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.898e+02 3.025e+02 3.482e+02 4.265e+02 1.190e+03, threshold=6.964e+02, percent-clipped=5.0 2022-12-23 22:08:26,661 INFO [train.py:894] (3/4) Epoch 25, batch 200, loss[loss=0.1557, simple_loss=0.2524, pruned_loss=0.02953, over 18580.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2585, pruned_loss=0.04041, over 2357597.87 frames. ], batch size: 51, lr: 4.62e-03, grad_scale: 8.0 2022-12-23 22:08:53,185 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1864, 1.5553, 1.8267, 1.9036, 2.1199, 2.1226, 2.0669, 1.7758], device='cuda:3'), covar=tensor([0.2326, 0.3719, 0.2917, 0.3089, 0.2234, 0.1138, 0.3387, 0.1450], device='cuda:3'), in_proj_covar=tensor([0.0271, 0.0301, 0.0285, 0.0323, 0.0315, 0.0258, 0.0350, 0.0246], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 22:09:26,908 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0 from training. Duration: 21.68 2022-12-23 22:09:38,752 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0 from training. Duration: 21.6300625 2022-12-23 22:09:41,362 INFO [train.py:894] (3/4) Epoch 25, batch 250, loss[loss=0.1819, simple_loss=0.2605, pruned_loss=0.05165, over 18573.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2571, pruned_loss=0.04, over 2658112.01 frames. ], batch size: 49, lr: 4.62e-03, grad_scale: 8.0 2022-12-23 22:09:41,838 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84397.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:10:02,545 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0_sp0.9 from training. Duration: 24.033375 2022-12-23 22:10:09,741 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84416.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:10:53,295 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.137e+02 3.339e+02 3.977e+02 4.779e+02 8.289e+02, threshold=7.955e+02, percent-clipped=6.0 2022-12-23 22:10:56,495 INFO [train.py:894] (3/4) Epoch 25, batch 300, loss[loss=0.1303, simple_loss=0.2159, pruned_loss=0.02238, over 18606.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2575, pruned_loss=0.04005, over 2893373.93 frames. ], batch size: 45, lr: 4.62e-03, grad_scale: 8.0 2022-12-23 22:10:57,996 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0 from training. Duration: 22.905 2022-12-23 22:10:59,451 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp1.1 from training. Duration: 23.4318125 2022-12-23 22:11:06,073 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84453.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:11:22,660 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=84464.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:12:12,050 INFO [train.py:894] (3/4) Epoch 25, batch 350, loss[loss=0.1784, simple_loss=0.2638, pruned_loss=0.04651, over 18664.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2574, pruned_loss=0.03967, over 3075177.85 frames. ], batch size: 48, lr: 4.62e-03, grad_scale: 8.0 2022-12-23 22:12:41,503 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2504, 1.4611, 1.9712, 1.8167, 2.2267, 2.2104, 2.0222, 1.8935], device='cuda:3'), covar=tensor([0.2253, 0.3474, 0.2790, 0.3148, 0.2136, 0.1056, 0.3238, 0.1358], device='cuda:3'), in_proj_covar=tensor([0.0270, 0.0300, 0.0284, 0.0322, 0.0314, 0.0257, 0.0349, 0.0244], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 22:12:47,343 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9987, 1.8772, 1.6742, 1.0737, 2.2499, 2.0570, 1.7912, 1.5883], device='cuda:3'), covar=tensor([0.0412, 0.0494, 0.0535, 0.0797, 0.0363, 0.0430, 0.0515, 0.0987], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0129, 0.0128, 0.0119, 0.0102, 0.0125, 0.0133, 0.0159], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 22:12:54,652 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84525.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:12:57,210 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp1.1 from training. Duration: 20.82275 2022-12-23 22:12:58,717 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp0.9 from training. Duration: 25.45 2022-12-23 22:13:24,703 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.178e+02 3.360e+02 4.039e+02 5.320e+02 1.136e+03, threshold=8.077e+02, percent-clipped=4.0 2022-12-23 22:13:27,740 INFO [train.py:894] (3/4) Epoch 25, batch 400, loss[loss=0.1998, simple_loss=0.288, pruned_loss=0.05582, over 18592.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.259, pruned_loss=0.04035, over 3216529.58 frames. ], batch size: 77, lr: 4.61e-03, grad_scale: 8.0 2022-12-23 22:13:40,357 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84555.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:13:58,017 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0 from training. Duration: 25.775 2022-12-23 22:14:06,372 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.50 vs. limit=5.0 2022-12-23 22:14:20,061 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0_sp0.9 from training. Duration: 22.25 2022-12-23 22:14:26,144 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84586.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:14:35,691 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84592.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:14:42,388 INFO [train.py:894] (3/4) Epoch 25, batch 450, loss[loss=0.1843, simple_loss=0.2754, pruned_loss=0.04661, over 18558.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2597, pruned_loss=0.04078, over 3326655.07 frames. ], batch size: 57, lr: 4.61e-03, grad_scale: 8.0 2022-12-23 22:14:46,835 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0 from training. Duration: 26.205 2022-12-23 22:14:54,956 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84605.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:15:05,408 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp0.9 from training. Duration: 30.1555625 2022-12-23 22:15:12,037 WARNING [train.py:1060] (3/4) Exclude cut with ID 1265-135635-0050-6781-0_sp0.9 from training. Duration: 21.8333125 2022-12-23 22:15:20,689 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp1.1 from training. Duration: 20.6545625 2022-12-23 22:15:22,508 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7724, 1.7733, 1.5267, 1.6588, 2.0590, 2.0096, 2.0885, 1.4007], device='cuda:3'), covar=tensor([0.0347, 0.0284, 0.0516, 0.0226, 0.0199, 0.0364, 0.0266, 0.0338], device='cuda:3'), in_proj_covar=tensor([0.0097, 0.0129, 0.0156, 0.0126, 0.0118, 0.0123, 0.0101, 0.0129], device='cuda:3'), out_proj_covar=tensor([7.7063e-05, 1.0196e-04, 1.2759e-04, 1.0020e-04, 9.4861e-05, 9.3991e-05, 7.8359e-05, 1.0186e-04], device='cuda:3') 2022-12-23 22:15:37,865 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84633.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:15:43,905 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4039, 1.3271, 1.3950, 1.3503, 1.1138, 2.3670, 1.1646, 1.6159], device='cuda:3'), covar=tensor([0.2538, 0.1835, 0.1647, 0.1802, 0.1251, 0.0255, 0.1876, 0.0759], device='cuda:3'), in_proj_covar=tensor([0.0131, 0.0117, 0.0125, 0.0120, 0.0105, 0.0097, 0.0090, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 22:15:45,303 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7261, 3.6822, 3.5331, 1.4912, 3.8478, 2.8935, 0.8908, 2.5523], device='cuda:3'), covar=tensor([0.2052, 0.1152, 0.1485, 0.3832, 0.0799, 0.0951, 0.4814, 0.1582], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0146, 0.0161, 0.0125, 0.0149, 0.0115, 0.0144, 0.0115], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 22:15:50,221 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84641.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:15:53,477 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.10 vs. limit=5.0 2022-12-23 22:15:55,472 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.349e+02 3.119e+02 3.751e+02 4.585e+02 8.162e+02, threshold=7.502e+02, percent-clipped=1.0 2022-12-23 22:15:58,513 INFO [train.py:894] (3/4) Epoch 25, batch 500, loss[loss=0.1535, simple_loss=0.2391, pruned_loss=0.03397, over 18516.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2589, pruned_loss=0.04046, over 3412899.16 frames. ], batch size: 44, lr: 4.61e-03, grad_scale: 8.0 2022-12-23 22:16:01,592 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0045-39920-0_sp0.9 from training. Duration: 20.52225 2022-12-23 22:16:05,286 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0471, 1.9729, 1.4681, 2.0971, 2.1157, 1.9902, 2.5752, 2.1303], device='cuda:3'), covar=tensor([0.0898, 0.1743, 0.2904, 0.1740, 0.1886, 0.0906, 0.1012, 0.1292], device='cuda:3'), in_proj_covar=tensor([0.0181, 0.0214, 0.0256, 0.0292, 0.0241, 0.0194, 0.0209, 0.0207], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 22:16:06,603 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.8032, 4.1205, 4.1553, 4.6825, 4.4739, 4.2301, 4.9205, 1.2007], device='cuda:3'), covar=tensor([0.0524, 0.0620, 0.0600, 0.0695, 0.0990, 0.0997, 0.0437, 0.5457], device='cuda:3'), in_proj_covar=tensor([0.0349, 0.0230, 0.0242, 0.0276, 0.0328, 0.0270, 0.0296, 0.0287], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 22:16:07,942 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=84653.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:16:21,241 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp0.9 from training. Duration: 29.1166875 2022-12-23 22:17:02,970 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=84689.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:17:07,082 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84692.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:17:10,264 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84694.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:17:14,082 INFO [train.py:894] (3/4) Epoch 25, batch 550, loss[loss=0.1607, simple_loss=0.2488, pruned_loss=0.03633, over 18371.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2584, pruned_loss=0.0401, over 3480020.75 frames. ], batch size: 46, lr: 4.61e-03, grad_scale: 8.0 2022-12-23 22:17:19,797 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133211-0007-59831-0_sp0.9 from training. Duration: 21.388875 2022-12-23 22:17:57,273 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0 from training. Duration: 22.72 2022-12-23 22:17:57,323 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0_sp0.9 from training. Duration: 22.7444375 2022-12-23 22:18:03,261 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.4061, 3.8226, 3.7841, 4.3378, 3.9665, 3.8456, 4.5371, 1.3363], device='cuda:3'), covar=tensor([0.0734, 0.0777, 0.0736, 0.0778, 0.1475, 0.1203, 0.0629, 0.5242], device='cuda:3'), in_proj_covar=tensor([0.0348, 0.0230, 0.0241, 0.0274, 0.0327, 0.0269, 0.0294, 0.0285], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 22:18:07,873 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7228, 1.4869, 1.0360, 0.3426, 1.1469, 1.5278, 1.2665, 1.4231], device='cuda:3'), covar=tensor([0.0789, 0.0779, 0.1436, 0.2176, 0.1469, 0.1912, 0.2339, 0.0919], device='cuda:3'), in_proj_covar=tensor([0.0173, 0.0187, 0.0207, 0.0190, 0.0210, 0.0202, 0.0216, 0.0202], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 22:18:25,920 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.064e+02 3.578e+02 4.188e+02 5.086e+02 7.637e+02, threshold=8.377e+02, percent-clipped=1.0 2022-12-23 22:18:29,241 INFO [train.py:894] (3/4) Epoch 25, batch 600, loss[loss=0.1559, simple_loss=0.2433, pruned_loss=0.03424, over 18678.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2599, pruned_loss=0.04062, over 3531862.06 frames. ], batch size: 48, lr: 4.61e-03, grad_scale: 8.0 2022-12-23 22:18:38,652 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84753.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:18:39,821 WARNING [train.py:1060] (3/4) Exclude cut with ID 4133-6541-0027-40495-0_sp1.1 from training. Duration: 0.9681875 2022-12-23 22:18:44,187 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0_sp0.9 from training. Duration: 22.3166875 2022-12-23 22:18:49,733 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2022-12-23 22:18:50,448 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133212-0015-59917-0_sp0.9 from training. Duration: 21.8166875 2022-12-23 22:19:44,490 INFO [train.py:894] (3/4) Epoch 25, batch 650, loss[loss=0.1574, simple_loss=0.241, pruned_loss=0.03688, over 18517.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2609, pruned_loss=0.04106, over 3573317.30 frames. ], batch size: 44, lr: 4.61e-03, grad_scale: 8.0 2022-12-23 22:19:49,844 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84800.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:19:51,039 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=84801.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:19:54,145 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7240, 1.5172, 1.7374, 1.6579, 1.1349, 3.8995, 1.6979, 2.1442], device='cuda:3'), covar=tensor([0.3073, 0.2182, 0.1952, 0.2108, 0.1570, 0.0153, 0.1548, 0.0847], device='cuda:3'), in_proj_covar=tensor([0.0131, 0.0117, 0.0124, 0.0120, 0.0105, 0.0096, 0.0090, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 22:20:14,348 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5340, 1.4804, 1.5975, 1.4293, 0.9181, 2.3091, 0.9925, 1.4872], device='cuda:3'), covar=tensor([0.3261, 0.2192, 0.1967, 0.2185, 0.1512, 0.0349, 0.1696, 0.0846], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0117, 0.0124, 0.0121, 0.0105, 0.0097, 0.0090, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 22:20:31,418 WARNING [train.py:1060] (3/4) Exclude cut with ID 4957-30119-0041-23990-0_sp0.9 from training. Duration: 20.22775 2022-12-23 22:20:54,923 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.139e+02 3.497e+02 4.132e+02 4.951e+02 1.041e+03, threshold=8.263e+02, percent-clipped=1.0 2022-12-23 22:20:57,875 INFO [train.py:894] (3/4) Epoch 25, batch 700, loss[loss=0.1653, simple_loss=0.2606, pruned_loss=0.03496, over 18638.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2608, pruned_loss=0.04078, over 3605639.94 frames. ], batch size: 53, lr: 4.61e-03, grad_scale: 8.0 2022-12-23 22:21:10,360 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84855.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:21:17,971 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp1.1 from training. Duration: 24.67275 2022-12-23 22:21:19,753 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84861.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:21:43,268 WARNING [train.py:1060] (3/4) Exclude cut with ID 3082-165428-0081-50734-0_sp0.9 from training. Duration: 21.8055625 2022-12-23 22:21:49,925 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84881.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:22:05,721 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84892.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:22:13,732 INFO [train.py:894] (3/4) Epoch 25, batch 750, loss[loss=0.1744, simple_loss=0.243, pruned_loss=0.05291, over 18549.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2605, pruned_loss=0.04054, over 3629749.14 frames. ], batch size: 41, lr: 4.61e-03, grad_scale: 8.0 2022-12-23 22:22:21,307 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0_sp0.9 from training. Duration: 22.6666875 2022-12-23 22:22:22,848 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=84903.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:23:20,265 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=84940.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:23:23,248 WARNING [train.py:1060] (3/4) Exclude cut with ID 2411-132532-0017-82279-0_sp1.1 from training. Duration: 0.9681875 2022-12-23 22:23:27,847 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.164e+02 3.238e+02 4.165e+02 4.876e+02 9.107e+02, threshold=8.330e+02, percent-clipped=1.0 2022-12-23 22:23:30,812 INFO [train.py:894] (3/4) Epoch 25, batch 800, loss[loss=0.185, simple_loss=0.2781, pruned_loss=0.04597, over 18631.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2611, pruned_loss=0.04029, over 3647892.93 frames. ], batch size: 99, lr: 4.60e-03, grad_scale: 8.0 2022-12-23 22:23:37,343 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84951.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:23:49,128 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0 from training. Duration: 22.485 2022-12-23 22:24:27,105 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp1.1 from training. Duration: 23.82275 2022-12-23 22:24:34,830 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84989.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:24:39,358 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84992.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:24:40,374 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0 from training. Duration: 20.77 2022-12-23 22:24:46,810 INFO [train.py:894] (3/4) Epoch 25, batch 850, loss[loss=0.1804, simple_loss=0.2758, pruned_loss=0.04248, over 18533.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2612, pruned_loss=0.04053, over 3663228.74 frames. ], batch size: 77, lr: 4.60e-03, grad_scale: 8.0 2022-12-23 22:24:48,328 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0_sp0.9 from training. Duration: 24.088875 2022-12-23 22:25:02,211 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2022-12-23 22:25:10,843 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85012.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:25:11,492 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2022-12-23 22:25:18,984 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp1.1 from training. Duration: 20.4409375 2022-12-23 22:25:51,510 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=85040.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:25:59,253 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.350e+02 3.402e+02 3.934e+02 4.822e+02 1.128e+03, threshold=7.867e+02, percent-clipped=2.0 2022-12-23 22:26:02,197 INFO [train.py:894] (3/4) Epoch 25, batch 900, loss[loss=0.1661, simple_loss=0.2589, pruned_loss=0.03665, over 18511.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2614, pruned_loss=0.04068, over 3674380.30 frames. ], batch size: 52, lr: 4.60e-03, grad_scale: 8.0 2022-12-23 22:26:34,955 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0_sp0.9 from training. Duration: 22.511125 2022-12-23 22:26:34,979 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0 from training. Duration: 20.675 2022-12-23 22:27:16,899 INFO [train.py:894] (3/4) Epoch 25, batch 950, loss[loss=0.1628, simple_loss=0.2566, pruned_loss=0.03451, over 18662.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2616, pruned_loss=0.04098, over 3682900.19 frames. ], batch size: 48, lr: 4.60e-03, grad_scale: 8.0 2022-12-23 22:27:19,309 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4877, 1.1914, 1.8982, 3.3919, 2.4557, 2.5237, 1.0133, 2.3423], device='cuda:3'), covar=tensor([0.2004, 0.1694, 0.1567, 0.0504, 0.0944, 0.1095, 0.2121, 0.1037], device='cuda:3'), in_proj_covar=tensor([0.0102, 0.0117, 0.0134, 0.0152, 0.0105, 0.0143, 0.0128, 0.0113], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 22:27:32,926 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85107.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:27:44,453 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85115.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:28:13,653 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp0.9 from training. Duration: 24.9833125 2022-12-23 22:28:29,077 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.943e+02 3.246e+02 3.932e+02 4.617e+02 8.441e+02, threshold=7.864e+02, percent-clipped=1.0 2022-12-23 22:28:32,095 INFO [train.py:894] (3/4) Epoch 25, batch 1000, loss[loss=0.1529, simple_loss=0.2341, pruned_loss=0.03582, over 18693.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2612, pruned_loss=0.0407, over 3690421.61 frames. ], batch size: 46, lr: 4.60e-03, grad_scale: 8.0 2022-12-23 22:28:46,690 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0 from training. Duration: 27.14 2022-12-23 22:28:46,857 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85156.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:29:00,953 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0 from training. Duration: 22.44 2022-12-23 22:29:04,399 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85168.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 22:29:16,371 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85176.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:29:23,978 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85181.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:29:48,017 INFO [train.py:894] (3/4) Epoch 25, batch 1050, loss[loss=0.1468, simple_loss=0.2308, pruned_loss=0.0314, over 18412.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2597, pruned_loss=0.04016, over 3695269.26 frames. ], batch size: 42, lr: 4.60e-03, grad_scale: 8.0 2022-12-23 22:30:21,413 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0060-62364-0_sp0.9 from training. Duration: 21.361125 2022-12-23 22:30:28,589 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp1.1 from training. Duration: 27.0318125 2022-12-23 22:30:31,923 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6580, 1.7842, 1.5480, 1.5584, 2.0187, 1.9571, 1.9385, 1.3192], device='cuda:3'), covar=tensor([0.0377, 0.0224, 0.0495, 0.0217, 0.0185, 0.0372, 0.0265, 0.0325], device='cuda:3'), in_proj_covar=tensor([0.0095, 0.0126, 0.0151, 0.0123, 0.0115, 0.0120, 0.0098, 0.0127], device='cuda:3'), out_proj_covar=tensor([7.5138e-05, 9.9362e-05, 1.2423e-04, 9.7547e-05, 9.2790e-05, 9.1528e-05, 7.6519e-05, 9.9638e-05], device='cuda:3') 2022-12-23 22:30:34,716 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8062, 1.3473, 0.8825, 1.3732, 2.1052, 1.1791, 1.5455, 1.6104], device='cuda:3'), covar=tensor([0.1635, 0.2087, 0.2262, 0.1528, 0.1809, 0.1896, 0.1503, 0.1792], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0098, 0.0117, 0.0096, 0.0119, 0.0092, 0.0099, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 22:30:37,454 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=85229.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:30:38,794 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp0.9 from training. Duration: 28.638875 2022-12-23 22:30:46,249 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.8639, 3.3282, 3.3300, 3.8392, 3.5195, 3.3791, 3.9787, 1.2266], device='cuda:3'), covar=tensor([0.0766, 0.0830, 0.0779, 0.0875, 0.1457, 0.1231, 0.0736, 0.5411], device='cuda:3'), in_proj_covar=tensor([0.0351, 0.0231, 0.0243, 0.0276, 0.0329, 0.0272, 0.0298, 0.0288], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 22:30:52,457 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0 from training. Duration: 20.4 2022-12-23 22:31:01,075 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.159e+02 3.176e+02 3.831e+02 4.678e+02 1.344e+03, threshold=7.662e+02, percent-clipped=2.0 2022-12-23 22:31:04,281 INFO [train.py:894] (3/4) Epoch 25, batch 1100, loss[loss=0.1758, simple_loss=0.2718, pruned_loss=0.0399, over 18508.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2606, pruned_loss=0.0403, over 3699294.45 frames. ], batch size: 58, lr: 4.60e-03, grad_scale: 16.0 2022-12-23 22:31:25,865 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0 from training. Duration: 20.025 2022-12-23 22:31:25,880 WARNING [train.py:1060] (3/4) Exclude cut with ID 2364-131735-0112-64612-0_sp0.9 from training. Duration: 20.488875 2022-12-23 22:31:30,304 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0 from training. Duration: 29.735 2022-12-23 22:32:08,352 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85289.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:32:19,893 INFO [train.py:894] (3/4) Epoch 25, batch 1150, loss[loss=0.1924, simple_loss=0.2839, pruned_loss=0.05039, over 18493.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2601, pruned_loss=0.0402, over 3702774.65 frames. ], batch size: 54, lr: 4.59e-03, grad_scale: 16.0 2022-12-23 22:32:35,461 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85307.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:32:54,235 WARNING [train.py:1060] (3/4) Exclude cut with ID 7276-92427-0014-12983-0_sp0.9 from training. Duration: 21.3055625 2022-12-23 22:32:55,699 WARNING [train.py:1060] (3/4) Exclude cut with ID 1025-75365-0008-79168-0_sp0.9 from training. Duration: 22.0666875 2022-12-23 22:33:19,336 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=85337.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:33:31,172 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.188e+02 3.174e+02 3.958e+02 5.147e+02 9.508e+02, threshold=7.916e+02, percent-clipped=3.0 2022-12-23 22:33:34,604 INFO [train.py:894] (3/4) Epoch 25, batch 1200, loss[loss=0.1744, simple_loss=0.2652, pruned_loss=0.04181, over 18588.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2598, pruned_loss=0.04032, over 3704763.37 frames. ], batch size: 57, lr: 4.59e-03, grad_scale: 16.0 2022-12-23 22:34:14,059 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2022-12-23 22:34:18,110 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2022-12-23 22:34:42,458 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0 from training. Duration: 20.26 2022-12-23 22:34:47,914 INFO [train.py:894] (3/4) Epoch 25, batch 1250, loss[loss=0.1629, simple_loss=0.2559, pruned_loss=0.035, over 18698.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.259, pruned_loss=0.03995, over 3707207.59 frames. ], batch size: 50, lr: 4.59e-03, grad_scale: 16.0 2022-12-23 22:34:57,989 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0030-9324-0_sp0.9 from training. Duration: 21.3444375 2022-12-23 22:35:03,241 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2022-12-23 22:35:34,918 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85428.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:35:53,831 WARNING [train.py:1060] (3/4) Exclude cut with ID 497-129325-0061-62254-0_sp1.1 from training. Duration: 0.97725 2022-12-23 22:36:00,830 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.956e+02 3.502e+02 4.076e+02 5.122e+02 9.671e+02, threshold=8.152e+02, percent-clipped=4.0 2022-12-23 22:36:03,834 INFO [train.py:894] (3/4) Epoch 25, batch 1300, loss[loss=0.1745, simple_loss=0.2698, pruned_loss=0.03966, over 18522.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2592, pruned_loss=0.04, over 3708783.01 frames. ], batch size: 58, lr: 4.59e-03, grad_scale: 16.0 2022-12-23 22:36:11,205 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.0664, 2.4933, 2.0040, 3.1228, 2.3240, 2.4679, 2.4296, 3.1830], device='cuda:3'), covar=tensor([0.1930, 0.3388, 0.1909, 0.2513, 0.3750, 0.1031, 0.3249, 0.0836], device='cuda:3'), in_proj_covar=tensor([0.0299, 0.0295, 0.0250, 0.0348, 0.0277, 0.0232, 0.0293, 0.0219], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 22:36:16,904 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85456.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:36:27,063 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85463.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 22:36:34,053 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0_sp0.9 from training. Duration: 22.97225 2022-12-23 22:36:38,370 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85471.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:36:56,344 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85483.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:37:04,622 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0047-39922-0_sp0.9 from training. Duration: 21.97775 2022-12-23 22:37:05,018 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85489.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:37:17,332 INFO [train.py:894] (3/4) Epoch 25, batch 1350, loss[loss=0.1434, simple_loss=0.2294, pruned_loss=0.02869, over 18429.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2594, pruned_loss=0.04007, over 3710083.97 frames. ], batch size: 42, lr: 4.59e-03, grad_scale: 16.0 2022-12-23 22:37:20,154 WARNING [train.py:1060] (3/4) Exclude cut with ID 1112-1043-0006-89194-0_sp0.9 from training. Duration: 21.8333125 2022-12-23 22:37:27,363 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=85504.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:37:28,672 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0 from training. Duration: 20.47 2022-12-23 22:37:35,042 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5209, 1.8736, 2.1348, 2.2863, 2.4646, 2.5641, 2.4274, 2.0205], device='cuda:3'), covar=tensor([0.2177, 0.3365, 0.2539, 0.2951, 0.2063, 0.0958, 0.3559, 0.1313], device='cuda:3'), in_proj_covar=tensor([0.0272, 0.0300, 0.0284, 0.0322, 0.0316, 0.0258, 0.0350, 0.0246], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 22:38:06,266 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85530.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:38:27,637 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85544.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:38:28,666 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.968e+02 3.372e+02 4.467e+02 5.937e+02 1.365e+03, threshold=8.935e+02, percent-clipped=9.0 2022-12-23 22:38:31,866 INFO [train.py:894] (3/4) Epoch 25, batch 1400, loss[loss=0.1961, simple_loss=0.2824, pruned_loss=0.05488, over 18601.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2603, pruned_loss=0.04036, over 3711303.96 frames. ], batch size: 51, lr: 4.59e-03, grad_scale: 16.0 2022-12-23 22:38:31,920 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0037-39912-0_sp0.9 from training. Duration: 20.67225 2022-12-23 22:38:49,041 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5853, 1.4283, 1.3949, 0.6334, 1.7325, 1.5691, 1.4586, 1.2875], device='cuda:3'), covar=tensor([0.0415, 0.0579, 0.0540, 0.0913, 0.0448, 0.0453, 0.0488, 0.1036], device='cuda:3'), in_proj_covar=tensor([0.0124, 0.0129, 0.0129, 0.0119, 0.0102, 0.0125, 0.0132, 0.0159], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 22:38:50,202 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp0.9 from training. Duration: 25.2444375 2022-12-23 22:39:13,358 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0021-76797-0_sp0.9 from training. Duration: 21.1445 2022-12-23 22:39:38,451 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85591.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:39:47,004 INFO [train.py:894] (3/4) Epoch 25, batch 1450, loss[loss=0.1506, simple_loss=0.2355, pruned_loss=0.0328, over 18614.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2603, pruned_loss=0.04041, over 3712380.67 frames. ], batch size: 45, lr: 4.59e-03, grad_scale: 16.0 2022-12-23 22:40:02,260 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85607.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:40:28,389 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp0.9 from training. Duration: 33.038875 2022-12-23 22:40:28,793 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85625.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:40:59,334 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.452e+02 3.397e+02 4.023e+02 4.690e+02 9.642e+02, threshold=8.046e+02, percent-clipped=1.0 2022-12-23 22:41:02,421 INFO [train.py:894] (3/4) Epoch 25, batch 1500, loss[loss=0.1702, simple_loss=0.2701, pruned_loss=0.03515, over 18387.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2596, pruned_loss=0.0401, over 3712250.25 frames. ], batch size: 53, lr: 4.59e-03, grad_scale: 16.0 2022-12-23 22:41:06,816 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64291-0000-16059-0_sp0.9 from training. Duration: 20.0944375 2022-12-23 22:41:14,996 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=85655.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:41:21,942 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp1.1 from training. Duration: 20.4 2022-12-23 22:41:30,402 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0 from training. Duration: 20.085 2022-12-23 22:41:42,396 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0_sp0.9 from training. Duration: 23.07775 2022-12-23 22:42:00,707 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85686.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 22:42:16,522 INFO [train.py:894] (3/4) Epoch 25, batch 1550, loss[loss=0.1477, simple_loss=0.2382, pruned_loss=0.02859, over 18671.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2596, pruned_loss=0.04004, over 3712764.96 frames. ], batch size: 46, lr: 4.58e-03, grad_scale: 16.0 2022-12-23 22:42:27,774 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp0.9 from training. Duration: 24.9333125 2022-12-23 22:43:12,641 WARNING [train.py:1060] (3/4) Exclude cut with ID 5118-111612-0016-124680-0_sp0.9 from training. Duration: 20.388875 2022-12-23 22:43:19,150 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp1.1 from training. Duration: 20.3590625 2022-12-23 22:43:29,018 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.994e+02 3.114e+02 3.818e+02 4.657e+02 1.147e+03, threshold=7.635e+02, percent-clipped=4.0 2022-12-23 22:43:31,796 INFO [train.py:894] (3/4) Epoch 25, batch 1600, loss[loss=0.1611, simple_loss=0.2372, pruned_loss=0.04247, over 18411.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2595, pruned_loss=0.04004, over 3713416.44 frames. ], batch size: 42, lr: 4.58e-03, grad_scale: 16.0 2022-12-23 22:43:55,489 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85763.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:44:07,582 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85771.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:44:26,671 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0_sp1.1 from training. Duration: 0.836375 2022-12-23 22:44:26,793 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85784.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:44:45,846 INFO [train.py:894] (3/4) Epoch 25, batch 1650, loss[loss=0.2085, simple_loss=0.2922, pruned_loss=0.0624, over 18601.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2606, pruned_loss=0.04093, over 3713568.32 frames. ], batch size: 69, lr: 4.58e-03, grad_scale: 16.0 2022-12-23 22:44:54,545 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7995, 1.6653, 1.6204, 1.6242, 1.6514, 4.8840, 2.1452, 2.8145], device='cuda:3'), covar=tensor([0.4295, 0.2818, 0.2671, 0.2989, 0.1548, 0.0217, 0.1478, 0.0817], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0118, 0.0125, 0.0121, 0.0105, 0.0096, 0.0090, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 22:45:06,473 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=85811.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:45:10,588 WARNING [train.py:1060] (3/4) Exclude cut with ID 8565-290391-0049-67394-0_sp0.9 from training. Duration: 21.3166875 2022-12-23 22:45:18,195 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=85819.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:45:41,570 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0029-104863-0_sp0.9 from training. Duration: 22.1055625 2022-12-23 22:45:47,985 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85839.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:45:53,775 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp1.1 from training. Duration: 21.77725 2022-12-23 22:45:56,761 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.405e+02 3.603e+02 4.272e+02 4.915e+02 8.898e+02, threshold=8.543e+02, percent-clipped=2.0 2022-12-23 22:45:59,561 INFO [train.py:894] (3/4) Epoch 25, batch 1700, loss[loss=0.2242, simple_loss=0.2971, pruned_loss=0.07568, over 18583.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2611, pruned_loss=0.04187, over 3714067.46 frames. ], batch size: 56, lr: 4.58e-03, grad_scale: 16.0 2022-12-23 22:46:11,848 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp0.9 from training. Duration: 27.8166875 2022-12-23 22:46:38,690 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp1.1 from training. Duration: 22.5090625 2022-12-23 22:46:42,050 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85875.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:46:45,541 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0 from training. Duration: 25.035 2022-12-23 22:46:58,518 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85886.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:47:02,117 WARNING [train.py:1060] (3/4) Exclude cut with ID 774-127930-0014-10412-0_sp1.1 from training. Duration: 0.95 2022-12-23 22:47:15,474 INFO [train.py:894] (3/4) Epoch 25, batch 1750, loss[loss=0.1876, simple_loss=0.2658, pruned_loss=0.05468, over 18446.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2619, pruned_loss=0.04333, over 3713437.93 frames. ], batch size: 50, lr: 4.58e-03, grad_scale: 16.0 2022-12-23 22:47:20,399 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp0.9 from training. Duration: 0.92225 2022-12-23 22:47:41,422 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1424, 1.5637, 2.5557, 4.4152, 3.1845, 2.8225, 0.8005, 3.2881], device='cuda:3'), covar=tensor([0.1651, 0.1607, 0.1362, 0.0535, 0.0900, 0.1241, 0.2154, 0.0792], device='cuda:3'), in_proj_covar=tensor([0.0103, 0.0118, 0.0136, 0.0154, 0.0106, 0.0144, 0.0129, 0.0114], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2022-12-23 22:47:44,367 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0 from training. Duration: 21.97 2022-12-23 22:48:02,284 WARNING [train.py:1060] (3/4) Exclude cut with ID 7492-105653-0055-62765-0_sp0.9 from training. Duration: 21.97225 2022-12-23 22:48:03,804 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp0.9 from training. Duration: 25.3333125 2022-12-23 22:48:05,605 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85930.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 22:48:14,179 WARNING [train.py:1060] (3/4) Exclude cut with ID 5172-29468-0015-19128-0_sp0.9 from training. Duration: 21.5055625 2022-12-23 22:48:14,748 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85936.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:48:17,891 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2857, 2.2272, 1.9365, 1.7258, 2.5326, 3.1228, 2.9085, 2.1495], device='cuda:3'), covar=tensor([0.0413, 0.0324, 0.0450, 0.0312, 0.0262, 0.0360, 0.0307, 0.0330], device='cuda:3'), in_proj_covar=tensor([0.0095, 0.0127, 0.0152, 0.0124, 0.0116, 0.0121, 0.0100, 0.0128], device='cuda:3'), out_proj_covar=tensor([7.5698e-05, 1.0004e-04, 1.2480e-04, 9.8198e-05, 9.3406e-05, 9.2671e-05, 7.7747e-05, 1.0073e-04], device='cuda:3') 2022-12-23 22:48:25,214 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp1.1 from training. Duration: 20.72725 2022-12-23 22:48:27,936 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.947e+02 4.240e+02 5.276e+02 6.362e+02 1.395e+03, threshold=1.055e+03, percent-clipped=9.0 2022-12-23 22:48:30,847 INFO [train.py:894] (3/4) Epoch 25, batch 1800, loss[loss=0.1668, simple_loss=0.2616, pruned_loss=0.03601, over 18393.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2625, pruned_loss=0.04464, over 3713616.43 frames. ], batch size: 53, lr: 4.58e-03, grad_scale: 16.0 2022-12-23 22:48:45,873 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85957.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:48:57,732 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp0.9 from training. Duration: 26.32775 2022-12-23 22:49:02,911 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9725, 2.0176, 2.3024, 1.3209, 2.3428, 2.3846, 1.6295, 2.7140], device='cuda:3'), covar=tensor([0.1201, 0.1646, 0.1258, 0.1914, 0.0763, 0.1082, 0.2247, 0.0538], device='cuda:3'), in_proj_covar=tensor([0.0195, 0.0212, 0.0206, 0.0191, 0.0172, 0.0213, 0.0213, 0.0191], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 22:49:06,610 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.0398, 2.5176, 2.3256, 3.0524, 2.5191, 2.5148, 2.4057, 3.1948], device='cuda:3'), covar=tensor([0.1741, 0.2875, 0.1503, 0.2296, 0.3217, 0.1009, 0.3093, 0.0754], device='cuda:3'), in_proj_covar=tensor([0.0299, 0.0297, 0.0251, 0.0350, 0.0278, 0.0233, 0.0295, 0.0219], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 22:49:21,100 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4597, 2.1199, 2.6627, 3.5032, 2.7923, 3.0398, 1.7564, 2.6689], device='cuda:3'), covar=tensor([0.1337, 0.1221, 0.1080, 0.0614, 0.0859, 0.1942, 0.1576, 0.0916], device='cuda:3'), in_proj_covar=tensor([0.0103, 0.0117, 0.0136, 0.0153, 0.0106, 0.0143, 0.0129, 0.0114], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 22:49:22,891 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85981.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 22:49:24,606 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.1262, 1.2738, 0.8857, 1.3313, 1.5204, 2.4178, 1.2916, 1.4740], device='cuda:3'), covar=tensor([0.0938, 0.1870, 0.1137, 0.0871, 0.1423, 0.0363, 0.1466, 0.1535], device='cuda:3'), in_proj_covar=tensor([0.0072, 0.0081, 0.0071, 0.0073, 0.0090, 0.0075, 0.0084, 0.0076], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 22:49:27,411 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0 from training. Duration: 20.025 2022-12-23 22:49:33,352 WARNING [train.py:1060] (3/4) Exclude cut with ID 6709-74022-0004-86860-0_sp1.1 from training. Duration: 0.9409375 2022-12-23 22:49:33,361 WARNING [train.py:1060] (3/4) Exclude cut with ID 4757-1811-0023-62229-0_sp0.9 from training. Duration: 21.37775 2022-12-23 22:49:34,314 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6116, 2.8833, 3.0028, 1.6013, 3.1878, 3.2395, 2.2069, 3.4720], device='cuda:3'), covar=tensor([0.1197, 0.1553, 0.1405, 0.2262, 0.0734, 0.1043, 0.2096, 0.0516], device='cuda:3'), in_proj_covar=tensor([0.0196, 0.0212, 0.0206, 0.0191, 0.0172, 0.0214, 0.0214, 0.0192], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 22:49:38,511 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85991.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 22:49:46,966 INFO [train.py:894] (3/4) Epoch 25, batch 1850, loss[loss=0.1649, simple_loss=0.2493, pruned_loss=0.04028, over 18678.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.263, pruned_loss=0.04595, over 3714324.34 frames. ], batch size: 48, lr: 4.58e-03, grad_scale: 16.0 2022-12-23 22:49:57,627 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0004-25974-0_sp0.9 from training. Duration: 21.17225 2022-12-23 22:49:57,641 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp0.9 from training. Duration: 27.511125 2022-12-23 22:49:57,966 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86002.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:50:21,058 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2022-12-23 22:50:23,218 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86018.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:50:31,428 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0 from training. Duration: 22.8 2022-12-23 22:50:36,341 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0 from training. Duration: 22.585 2022-12-23 22:51:02,098 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.089e+02 4.084e+02 4.781e+02 5.802e+02 1.070e+03, threshold=9.561e+02, percent-clipped=1.0 2022-12-23 22:51:04,972 INFO [train.py:894] (3/4) Epoch 25, batch 1900, loss[loss=0.2041, simple_loss=0.2815, pruned_loss=0.06334, over 18720.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2631, pruned_loss=0.04685, over 3714058.46 frames. ], batch size: 65, lr: 4.57e-03, grad_scale: 16.0 2022-12-23 22:51:08,125 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0001-146967-0_sp0.9 from training. Duration: 22.0166875 2022-12-23 22:51:12,622 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86052.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:51:24,343 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp1.1 from training. Duration: 24.395375 2022-12-23 22:51:29,997 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86063.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:51:31,062 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp0.9 from training. Duration: 27.47775 2022-12-23 22:51:36,762 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp0.9 from training. Duration: 24.8833125 2022-12-23 22:51:38,442 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86069.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:51:39,469 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0 from training. Duration: 23.39 2022-12-23 22:51:45,375 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp0.9 from training. Duration: 28.72225 2022-12-23 22:51:54,999 WARNING [train.py:1060] (3/4) Exclude cut with ID 585-294811-0110-133686-0_sp0.9 from training. Duration: 20.8944375 2022-12-23 22:52:00,887 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86084.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:52:07,195 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6014, 2.3283, 1.9839, 1.3397, 2.7269, 2.6681, 2.3581, 1.8395], device='cuda:3'), covar=tensor([0.0373, 0.0481, 0.0568, 0.0806, 0.0324, 0.0359, 0.0463, 0.0908], device='cuda:3'), in_proj_covar=tensor([0.0124, 0.0130, 0.0129, 0.0119, 0.0103, 0.0124, 0.0133, 0.0159], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2022-12-23 22:52:09,478 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0_sp0.9 from training. Duration: 23.8444375 2022-12-23 22:52:20,024 INFO [train.py:894] (3/4) Epoch 25, batch 1950, loss[loss=0.1861, simple_loss=0.2756, pruned_loss=0.04828, over 18683.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2638, pruned_loss=0.04725, over 3714410.10 frames. ], batch size: 60, lr: 4.57e-03, grad_scale: 16.0 2022-12-23 22:52:32,205 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0 from training. Duration: 25.85 2022-12-23 22:52:32,216 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0 from training. Duration: 21.39 2022-12-23 22:52:44,138 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0 from training. Duration: 27.92 2022-12-23 22:52:44,539 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86113.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:53:09,644 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86130.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:53:11,936 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0039-130165-0_sp0.9 from training. Duration: 20.661125 2022-12-23 22:53:12,037 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86132.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:53:23,000 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86139.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:53:31,647 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.321e+02 4.247e+02 5.001e+02 6.020e+02 1.322e+03, threshold=1.000e+03, percent-clipped=4.0 2022-12-23 22:53:34,768 INFO [train.py:894] (3/4) Epoch 25, batch 2000, loss[loss=0.1638, simple_loss=0.2436, pruned_loss=0.042, over 18709.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2632, pruned_loss=0.04736, over 3714145.09 frames. ], batch size: 46, lr: 4.57e-03, grad_scale: 16.0 2022-12-23 22:53:36,282 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0043-15874-0_sp0.9 from training. Duration: 20.07225 2022-12-23 22:53:44,154 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0 from training. Duration: 21.01 2022-12-23 22:54:13,321 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9444, 1.8543, 1.5263, 1.0591, 1.6185, 1.6841, 1.4459, 1.7283], device='cuda:3'), covar=tensor([0.0574, 0.0471, 0.1076, 0.1221, 0.0937, 0.1040, 0.1321, 0.0633], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0189, 0.0209, 0.0191, 0.0211, 0.0204, 0.0219, 0.0203], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 22:54:34,855 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86186.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:54:36,179 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86187.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:54:49,898 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0 from training. Duration: 20.65 2022-12-23 22:54:51,335 INFO [train.py:894] (3/4) Epoch 25, batch 2050, loss[loss=0.2084, simple_loss=0.2872, pruned_loss=0.0648, over 18653.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2639, pruned_loss=0.04832, over 3713977.56 frames. ], batch size: 62, lr: 4.57e-03, grad_scale: 16.0 2022-12-23 22:54:56,332 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7576, 2.3100, 1.8200, 2.4414, 2.0256, 2.1535, 2.1737, 2.6062], device='cuda:3'), covar=tensor([0.2071, 0.3100, 0.2010, 0.3019, 0.3895, 0.1111, 0.3147, 0.0992], device='cuda:3'), in_proj_covar=tensor([0.0298, 0.0296, 0.0251, 0.0349, 0.0278, 0.0233, 0.0294, 0.0219], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 22:54:57,558 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0 from training. Duration: 21.46 2022-12-23 22:55:18,333 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5150, 1.5511, 1.3644, 1.3161, 1.7116, 1.6825, 1.6295, 1.1901], device='cuda:3'), covar=tensor([0.0317, 0.0242, 0.0438, 0.0232, 0.0200, 0.0388, 0.0296, 0.0324], device='cuda:3'), in_proj_covar=tensor([0.0095, 0.0126, 0.0151, 0.0123, 0.0115, 0.0120, 0.0099, 0.0127], device='cuda:3'), out_proj_covar=tensor([7.5385e-05, 9.9207e-05, 1.2390e-04, 9.7457e-05, 9.2652e-05, 9.2137e-05, 7.7309e-05, 9.9821e-05], device='cuda:3') 2022-12-23 22:55:31,657 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2631, 1.6438, 1.9112, 1.9113, 2.2758, 2.2526, 2.1054, 1.9157], device='cuda:3'), covar=tensor([0.2324, 0.3335, 0.2565, 0.2986, 0.2017, 0.0967, 0.3245, 0.1328], device='cuda:3'), in_proj_covar=tensor([0.0272, 0.0302, 0.0285, 0.0324, 0.0316, 0.0258, 0.0351, 0.0247], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 22:55:37,762 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.9192, 5.2774, 4.8381, 2.8647, 5.3961, 4.0390, 0.9559, 3.6698], device='cuda:3'), covar=tensor([0.2198, 0.1191, 0.1332, 0.2891, 0.0699, 0.0749, 0.5015, 0.1256], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0147, 0.0161, 0.0124, 0.0149, 0.0114, 0.0144, 0.0115], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 22:55:42,237 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0 from training. Duration: 0.92 2022-12-23 22:55:43,875 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86231.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:55:48,086 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0_sp0.9 from training. Duration: 23.7666875 2022-12-23 22:55:48,211 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86234.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:55:48,520 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.5142, 1.9272, 2.2239, 1.0955, 1.4936, 2.3679, 2.1019, 1.7893], device='cuda:3'), covar=tensor([0.0778, 0.0392, 0.0286, 0.0422, 0.0400, 0.0425, 0.0256, 0.0731], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0174, 0.0130, 0.0142, 0.0150, 0.0146, 0.0167, 0.0178], device='cuda:3'), out_proj_covar=tensor([1.1448e-04, 1.3173e-04, 9.6554e-05, 1.0444e-04, 1.1020e-04, 1.1001e-04, 1.2669e-04, 1.3422e-04], device='cuda:3') 2022-12-23 22:56:04,973 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.056e+02 4.151e+02 5.152e+02 6.542e+02 1.596e+03, threshold=1.030e+03, percent-clipped=4.0 2022-12-23 22:56:07,211 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86246.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:56:08,501 INFO [train.py:894] (3/4) Epoch 25, batch 2100, loss[loss=0.1616, simple_loss=0.2533, pruned_loss=0.03496, over 18436.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2631, pruned_loss=0.0482, over 3714073.67 frames. ], batch size: 50, lr: 4.57e-03, grad_scale: 16.0 2022-12-23 22:56:22,371 WARNING [train.py:1060] (3/4) Exclude cut with ID 8544-281189-0060-101339-0_sp0.9 from training. Duration: 20.861125 2022-12-23 22:56:32,148 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0_sp0.9 from training. Duration: 22.711125 2022-12-23 22:56:36,847 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2747, 1.6080, 0.7360, 1.7652, 2.4719, 1.7155, 2.0407, 2.2545], device='cuda:3'), covar=tensor([0.1590, 0.2082, 0.2419, 0.1481, 0.1716, 0.1691, 0.1386, 0.1624], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0096, 0.0116, 0.0095, 0.0118, 0.0092, 0.0097, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 22:57:00,620 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.2735, 1.3649, 1.5228, 0.8586, 1.2979, 1.4235, 1.2309, 1.6530], device='cuda:3'), covar=tensor([0.1085, 0.1996, 0.1026, 0.1502, 0.0869, 0.1054, 0.2483, 0.0599], device='cuda:3'), in_proj_covar=tensor([0.0195, 0.0213, 0.0207, 0.0191, 0.0172, 0.0214, 0.0214, 0.0192], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 22:57:02,489 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86281.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:57:09,382 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86286.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 22:57:15,371 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp1.1 from training. Duration: 22.986375 2022-12-23 22:57:26,188 INFO [train.py:894] (3/4) Epoch 25, batch 2150, loss[loss=0.1578, simple_loss=0.2433, pruned_loss=0.03616, over 18429.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2631, pruned_loss=0.04801, over 3714460.17 frames. ], batch size: 48, lr: 4.57e-03, grad_scale: 16.0 2022-12-23 22:57:30,701 WARNING [train.py:1060] (3/4) Exclude cut with ID 8040-260924-0003-80960-0_sp0.9 from training. Duration: 22.07225 2022-12-23 22:57:35,151 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0045-26330-0_sp0.9 from training. Duration: 20.3055625 2022-12-23 22:57:38,003 WARNING [train.py:1060] (3/4) Exclude cut with ID 6356-271890-0060-94317-0_sp0.9 from training. Duration: 20.72225 2022-12-23 22:57:41,889 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86307.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:57:50,576 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86313.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:57:58,374 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp1.1 from training. Duration: 22.4818125 2022-12-23 22:58:14,498 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6731, 1.4435, 1.0133, 0.3027, 1.1192, 1.5139, 1.2212, 1.3600], device='cuda:3'), covar=tensor([0.0809, 0.0832, 0.1463, 0.2168, 0.1458, 0.2062, 0.2499, 0.0970], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0188, 0.0209, 0.0191, 0.0211, 0.0204, 0.0218, 0.0202], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 22:58:15,620 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86329.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:58:24,087 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp0.9 from training. Duration: 25.0944375 2022-12-23 22:58:28,496 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0 from training. Duration: 21.515 2022-12-23 22:58:34,256 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp0.9 from training. Duration: 27.02225 2022-12-23 22:58:39,251 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.878e+02 4.086e+02 4.973e+02 6.168e+02 9.928e+02, threshold=9.946e+02, percent-clipped=0.0 2022-12-23 22:58:39,385 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0010-62480-0_sp0.9 from training. Duration: 22.22225 2022-12-23 22:58:42,239 INFO [train.py:894] (3/4) Epoch 25, batch 2200, loss[loss=0.1736, simple_loss=0.2592, pruned_loss=0.04401, over 18459.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2614, pruned_loss=0.04727, over 3713843.96 frames. ], batch size: 50, lr: 4.57e-03, grad_scale: 16.0 2022-12-23 22:58:46,044 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.97 vs. limit=5.0 2022-12-23 22:58:46,769 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0085-44554-0_sp0.9 from training. Duration: 20.85 2022-12-23 22:58:59,756 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86358.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 22:59:20,009 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0 from training. Duration: 21.54 2022-12-23 22:59:24,511 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp1.1 from training. Duration: 20.5318125 2022-12-23 22:59:34,840 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0012-134311-0_sp0.9 from training. Duration: 21.9333125 2022-12-23 22:59:57,001 INFO [train.py:894] (3/4) Epoch 25, batch 2250, loss[loss=0.1849, simple_loss=0.2617, pruned_loss=0.05403, over 18384.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2613, pruned_loss=0.04734, over 3713801.29 frames. ], batch size: 46, lr: 4.57e-03, grad_scale: 16.0 2022-12-23 23:00:05,840 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5250, 2.1247, 1.6594, 0.7682, 1.7584, 2.0988, 1.5718, 1.9061], device='cuda:3'), covar=tensor([0.0753, 0.0770, 0.1708, 0.1939, 0.1383, 0.1551, 0.2208, 0.0970], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0188, 0.0209, 0.0191, 0.0212, 0.0205, 0.0219, 0.0203], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 23:00:11,565 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6519, 1.6582, 1.9039, 1.0962, 1.8175, 1.9079, 1.4823, 2.2283], device='cuda:3'), covar=tensor([0.1110, 0.1835, 0.1014, 0.1585, 0.0741, 0.1032, 0.2308, 0.0491], device='cuda:3'), in_proj_covar=tensor([0.0198, 0.0216, 0.0210, 0.0194, 0.0175, 0.0217, 0.0217, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 23:00:14,469 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86408.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:00:22,309 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0025-130151-0_sp0.9 from training. Duration: 21.7944375 2022-12-23 23:00:34,960 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0_sp0.9 from training. Duration: 22.4666875 2022-12-23 23:00:40,801 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86425.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:00:42,135 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0 from training. Duration: 21.635 2022-12-23 23:00:47,835 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0_sp0.9 from training. Duration: 24.038875 2022-12-23 23:00:59,350 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6703, 1.4230, 1.5538, 1.8358, 1.7006, 3.4125, 1.4585, 1.5947], device='cuda:3'), covar=tensor([0.0886, 0.1891, 0.1136, 0.0955, 0.1494, 0.0318, 0.1496, 0.1608], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0082, 0.0072, 0.0074, 0.0091, 0.0076, 0.0085, 0.0077], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2022-12-23 23:01:03,644 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86440.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:01:10,456 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.162e+02 3.881e+02 4.977e+02 5.869e+02 1.458e+03, threshold=9.954e+02, percent-clipped=4.0 2022-12-23 23:01:10,828 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86445.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:01:13,389 INFO [train.py:894] (3/4) Epoch 25, batch 2300, loss[loss=0.1614, simple_loss=0.2473, pruned_loss=0.03774, over 18410.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2618, pruned_loss=0.04814, over 3715055.10 frames. ], batch size: 48, lr: 4.56e-03, grad_scale: 16.0 2022-12-23 23:01:34,110 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp1.1 from training. Duration: 21.786375 2022-12-23 23:01:41,739 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7760, 1.8696, 2.1023, 1.1831, 2.0794, 2.1018, 1.6011, 2.4476], device='cuda:3'), covar=tensor([0.1199, 0.1780, 0.1122, 0.1804, 0.0713, 0.1151, 0.2376, 0.0506], device='cuda:3'), in_proj_covar=tensor([0.0198, 0.0215, 0.0208, 0.0193, 0.0174, 0.0216, 0.0216, 0.0194], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 23:01:44,917 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0 from training. Duration: 20.22 2022-12-23 23:01:48,735 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2022-12-23 23:02:30,732 INFO [train.py:894] (3/4) Epoch 25, batch 2350, loss[loss=0.187, simple_loss=0.2572, pruned_loss=0.05836, over 18692.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2622, pruned_loss=0.04837, over 3714852.92 frames. ], batch size: 48, lr: 4.56e-03, grad_scale: 16.0 2022-12-23 23:02:36,283 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 2022-12-23 23:02:37,170 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86501.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:02:38,557 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.1719, 3.6269, 3.5674, 4.1372, 3.7687, 3.6301, 4.3174, 1.3198], device='cuda:3'), covar=tensor([0.0762, 0.0772, 0.0774, 0.0901, 0.1419, 0.1300, 0.0616, 0.5168], device='cuda:3'), in_proj_covar=tensor([0.0352, 0.0232, 0.0244, 0.0275, 0.0332, 0.0275, 0.0297, 0.0289], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 23:02:44,762 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86506.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 23:03:22,411 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86531.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:03:43,755 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.777e+02 3.901e+02 5.093e+02 6.285e+02 1.254e+03, threshold=1.019e+03, percent-clipped=1.0 2022-12-23 23:03:43,789 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0 from training. Duration: 25.285 2022-12-23 23:03:46,473 INFO [train.py:894] (3/4) Epoch 25, batch 2400, loss[loss=0.1637, simple_loss=0.2443, pruned_loss=0.04158, over 18682.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2626, pruned_loss=0.04853, over 3716225.11 frames. ], batch size: 46, lr: 4.56e-03, grad_scale: 16.0 2022-12-23 23:04:32,864 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.4939, 1.8776, 2.1339, 1.0064, 1.3921, 2.2773, 2.1134, 1.8068], device='cuda:3'), covar=tensor([0.0805, 0.0342, 0.0318, 0.0422, 0.0399, 0.0440, 0.0230, 0.0682], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0173, 0.0130, 0.0141, 0.0149, 0.0145, 0.0166, 0.0177], device='cuda:3'), out_proj_covar=tensor([1.1425e-04, 1.3067e-04, 9.6190e-05, 1.0383e-04, 1.0923e-04, 1.0916e-04, 1.2611e-04, 1.3351e-04], device='cuda:3') 2022-12-23 23:04:36,258 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86579.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:04:37,997 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86580.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:04:47,225 WARNING [train.py:1060] (3/4) Exclude cut with ID 811-130148-0001-63453-0_sp0.9 from training. Duration: 20.861125 2022-12-23 23:04:47,453 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86586.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 23:05:03,675 INFO [train.py:894] (3/4) Epoch 25, batch 2450, loss[loss=0.1396, simple_loss=0.2221, pruned_loss=0.02849, over 18601.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.262, pruned_loss=0.04813, over 3715425.53 frames. ], batch size: 41, lr: 4.56e-03, grad_scale: 16.0 2022-12-23 23:05:08,243 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0 from training. Duration: 20.88 2022-12-23 23:05:12,160 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86602.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:05:28,309 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86613.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:05:41,687 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0_sp0.9 from training. Duration: 23.4166875 2022-12-23 23:06:00,838 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86634.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 23:06:11,094 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86641.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:06:16,262 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6650, 1.4865, 1.0070, 0.2286, 1.0700, 1.5993, 1.4242, 1.4312], device='cuda:3'), covar=tensor([0.0721, 0.0624, 0.1228, 0.1843, 0.1218, 0.1758, 0.1872, 0.0832], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0189, 0.0209, 0.0192, 0.0212, 0.0206, 0.0219, 0.0204], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 23:06:17,293 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.818e+02 4.531e+02 5.452e+02 6.835e+02 1.192e+03, threshold=1.090e+03, percent-clipped=5.0 2022-12-23 23:06:20,415 INFO [train.py:894] (3/4) Epoch 25, batch 2500, loss[loss=0.1647, simple_loss=0.2545, pruned_loss=0.03742, over 18386.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2626, pruned_loss=0.04832, over 3715106.19 frames. ], batch size: 53, lr: 4.56e-03, grad_scale: 16.0 2022-12-23 23:06:28,819 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6597, 1.6475, 1.7360, 1.5913, 1.3253, 3.7076, 1.6368, 2.1596], device='cuda:3'), covar=tensor([0.3216, 0.2123, 0.1948, 0.2134, 0.1496, 0.0190, 0.1649, 0.0851], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0118, 0.0125, 0.0122, 0.0105, 0.0097, 0.0091, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 23:06:37,406 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86658.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:06:41,798 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86661.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:06:58,271 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0 from training. Duration: 21.24 2022-12-23 23:06:58,281 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0_sp0.9 from training. Duration: 23.9055625 2022-12-23 23:07:31,396 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp0.9 from training. Duration: 25.988875 2022-12-23 23:07:35,997 INFO [train.py:894] (3/4) Epoch 25, batch 2550, loss[loss=0.1896, simple_loss=0.2751, pruned_loss=0.05203, over 18526.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2631, pruned_loss=0.04866, over 3715237.92 frames. ], batch size: 58, lr: 4.56e-03, grad_scale: 16.0 2022-12-23 23:07:40,384 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0001-134300-0_sp0.9 from training. Duration: 20.67225 2022-12-23 23:07:49,244 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86706.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:07:52,234 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86708.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:08:17,805 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86725.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:08:26,511 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0 from training. Duration: 20.34 2022-12-23 23:08:48,783 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.379e+02 3.888e+02 5.016e+02 6.192e+02 1.481e+03, threshold=1.003e+03, percent-clipped=3.0 2022-12-23 23:08:51,484 INFO [train.py:894] (3/4) Epoch 25, batch 2600, loss[loss=0.1577, simple_loss=0.2354, pruned_loss=0.04003, over 18616.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2631, pruned_loss=0.04876, over 3714907.56 frames. ], batch size: 45, lr: 4.56e-03, grad_scale: 16.0 2022-12-23 23:09:04,693 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86756.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:09:28,220 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.0777, 0.9234, 0.8951, 1.1888, 1.2962, 1.1966, 1.1513, 0.9897], device='cuda:3'), covar=tensor([0.0291, 0.0236, 0.0630, 0.0208, 0.0230, 0.0375, 0.0298, 0.0332], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0125, 0.0151, 0.0123, 0.0116, 0.0121, 0.0099, 0.0127], device='cuda:3'), out_proj_covar=tensor([7.4603e-05, 9.8995e-05, 1.2374e-04, 9.7344e-05, 9.2900e-05, 9.2500e-05, 7.7374e-05, 9.9652e-05], device='cuda:3') 2022-12-23 23:09:30,644 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86773.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:09:40,544 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp0.9 from training. Duration: 25.061125 2022-12-23 23:09:51,622 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0 from training. Duration: 0.83 2022-12-23 23:10:05,386 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86796.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:10:06,497 INFO [train.py:894] (3/4) Epoch 25, batch 2650, loss[loss=0.1929, simple_loss=0.2771, pruned_loss=0.05434, over 18519.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2626, pruned_loss=0.0482, over 3714317.37 frames. ], batch size: 52, lr: 4.55e-03, grad_scale: 16.0 2022-12-23 23:10:12,631 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86801.0, num_to_drop=1, layers_to_drop={3} 2022-12-23 23:10:18,195 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0 from training. Duration: 24.73 2022-12-23 23:10:30,285 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0 from training. Duration: 23.965 2022-12-23 23:10:32,002 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86814.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:10:39,999 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0030-146996-0_sp0.9 from training. Duration: 22.088875 2022-12-23 23:10:57,426 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0_sp0.9 from training. Duration: 23.6 2022-12-23 23:11:00,208 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2022-12-23 23:11:00,967 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.8834, 5.1275, 4.6839, 2.1210, 5.1369, 3.9032, 0.8244, 3.4369], device='cuda:3'), covar=tensor([0.2136, 0.1014, 0.1338, 0.3570, 0.0793, 0.0808, 0.5226, 0.1301], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0147, 0.0161, 0.0125, 0.0151, 0.0115, 0.0144, 0.0115], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 23:11:18,490 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.776e+02 4.332e+02 4.963e+02 5.753e+02 1.372e+03, threshold=9.925e+02, percent-clipped=1.0 2022-12-23 23:11:21,280 INFO [train.py:894] (3/4) Epoch 25, batch 2700, loss[loss=0.2026, simple_loss=0.28, pruned_loss=0.06257, over 18467.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2614, pruned_loss=0.04775, over 3715146.97 frames. ], batch size: 50, lr: 4.55e-03, grad_scale: 16.0 2022-12-23 23:12:05,345 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86875.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:12:09,830 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86878.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:12:37,887 INFO [train.py:894] (3/4) Epoch 25, batch 2750, loss[loss=0.2055, simple_loss=0.2903, pruned_loss=0.06037, over 18465.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2612, pruned_loss=0.0475, over 3715194.12 frames. ], batch size: 64, lr: 4.55e-03, grad_scale: 8.0 2022-12-23 23:12:41,062 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0 from training. Duration: 23.795 2022-12-23 23:12:45,484 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86902.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:12:58,824 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp1.1 from training. Duration: 21.5409375 2022-12-23 23:13:01,660 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp0.9 from training. Duration: 24.97775 2022-12-23 23:13:13,047 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0_sp0.9 from training. Duration: 23.3444375 2022-12-23 23:13:38,164 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86936.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:13:42,768 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0_sp0.9 from training. Duration: 23.2 2022-12-23 23:13:43,165 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86939.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:13:47,194 WARNING [train.py:1060] (3/4) Exclude cut with ID 5653-46179-0060-117930-0_sp0.9 from training. Duration: 21.17225 2022-12-23 23:13:52,950 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.402e+02 3.994e+02 4.847e+02 5.810e+02 1.077e+03, threshold=9.693e+02, percent-clipped=1.0 2022-12-23 23:13:54,455 INFO [train.py:894] (3/4) Epoch 25, batch 2800, loss[loss=0.1675, simple_loss=0.2452, pruned_loss=0.04491, over 18423.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2614, pruned_loss=0.04753, over 3715128.50 frames. ], batch size: 46, lr: 4.55e-03, grad_scale: 8.0 2022-12-23 23:13:58,956 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86950.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:14:05,487 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp0.9 from training. Duration: 24.6555625 2022-12-23 23:14:58,608 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0 from training. Duration: 20.44 2022-12-23 23:15:12,522 INFO [train.py:894] (3/4) Epoch 25, batch 2850, loss[loss=0.1979, simple_loss=0.2882, pruned_loss=0.05381, over 18497.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2616, pruned_loss=0.04768, over 3715308.59 frames. ], batch size: 52, lr: 4.55e-03, grad_scale: 8.0 2022-12-23 23:15:15,462 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0_sp0.9 from training. Duration: 23.45 2022-12-23 23:15:20,989 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87002.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 23:15:28,825 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.57 vs. limit=2.0 2022-12-23 23:15:46,041 WARNING [train.py:1060] (3/4) Exclude cut with ID 6945-60535-0076-12784-0_sp0.9 from training. Duration: 20.52225 2022-12-23 23:15:53,775 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0 from training. Duration: 22.19 2022-12-23 23:16:01,808 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2022-12-23 23:16:04,109 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp1.1 from training. Duration: 25.3818125 2022-12-23 23:16:21,564 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp0.9 from training. Duration: 28.0944375 2022-12-23 23:16:26,182 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.838e+02 4.269e+02 5.078e+02 6.518e+02 1.318e+03, threshold=1.016e+03, percent-clipped=2.0 2022-12-23 23:16:27,879 INFO [train.py:894] (3/4) Epoch 25, batch 2900, loss[loss=0.1611, simple_loss=0.2356, pruned_loss=0.04335, over 18598.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2613, pruned_loss=0.04769, over 3715973.87 frames. ], batch size: 45, lr: 4.55e-03, grad_scale: 8.0 2022-12-23 23:16:27,951 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0_sp0.9 from training. Duration: 22.9444375 2022-12-23 23:16:37,125 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp1.1 from training. Duration: 21.6318125 2022-12-23 23:16:53,121 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87063.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 23:16:54,057 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0 from training. Duration: 23.695 2022-12-23 23:17:20,427 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0 from training. Duration: 23.955 2022-12-23 23:17:25,417 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3235, 1.7264, 2.0260, 1.9380, 2.2910, 2.2552, 2.1589, 1.9535], device='cuda:3'), covar=tensor([0.2284, 0.3296, 0.2671, 0.2854, 0.2184, 0.1039, 0.3366, 0.1377], device='cuda:3'), in_proj_covar=tensor([0.0273, 0.0302, 0.0287, 0.0324, 0.0317, 0.0260, 0.0353, 0.0248], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 23:17:41,835 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87096.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:17:43,050 INFO [train.py:894] (3/4) Epoch 25, batch 2950, loss[loss=0.1698, simple_loss=0.2552, pruned_loss=0.04216, over 18635.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2617, pruned_loss=0.04773, over 3715132.25 frames. ], batch size: 99, lr: 4.55e-03, grad_scale: 8.0 2022-12-23 23:17:49,665 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87101.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 23:17:50,068 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.03 vs. limit=5.0 2022-12-23 23:17:53,803 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp0.9 from training. Duration: 26.438875 2022-12-23 23:18:36,021 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0021-26306-0_sp0.9 from training. Duration: 21.2444375 2022-12-23 23:18:36,050 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp0.9 from training. Duration: 31.02225 2022-12-23 23:18:46,846 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0 from training. Duration: 22.395 2022-12-23 23:18:50,755 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.43 vs. limit=5.0 2022-12-23 23:18:55,926 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=87144.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:18:59,330 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.752e+02 4.188e+02 4.950e+02 6.486e+02 1.228e+03, threshold=9.900e+02, percent-clipped=4.0 2022-12-23 23:19:00,619 INFO [train.py:894] (3/4) Epoch 25, batch 3000, loss[loss=0.187, simple_loss=0.268, pruned_loss=0.05301, over 18705.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2619, pruned_loss=0.04753, over 3715788.99 frames. ], batch size: 50, lr: 4.55e-03, grad_scale: 8.0 2022-12-23 23:19:00,619 INFO [train.py:919] (3/4) Computing validation loss 2022-12-23 23:19:04,341 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.1221, 2.7882, 2.7008, 3.1239, 2.8175, 2.7967, 3.2003, 1.0994], device='cuda:3'), covar=tensor([0.0983, 0.0803, 0.0939, 0.0881, 0.1613, 0.1322, 0.0758, 0.5201], device='cuda:3'), in_proj_covar=tensor([0.0351, 0.0234, 0.0243, 0.0278, 0.0333, 0.0274, 0.0296, 0.0289], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 23:19:11,394 INFO [train.py:928] (3/4) Epoch 25, validation: loss=0.1616, simple_loss=0.2586, pruned_loss=0.0323, over 944034.00 frames. 2022-12-23 23:19:11,395 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-23 23:19:14,469 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=87149.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:19:17,279 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0 from training. Duration: 21.075 2022-12-23 23:19:21,472 WARNING [train.py:1060] (3/4) Exclude cut with ID 6482-98857-0025-147532-0_sp0.9 from training. Duration: 20.0055625 2022-12-23 23:19:21,483 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0037-132304-0_sp0.9 from training. Duration: 22.05 2022-12-23 23:19:21,495 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0 from training. Duration: 26.8349375 2022-12-23 23:19:21,799 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5847, 1.4911, 1.5957, 1.5120, 1.1148, 3.3236, 1.3827, 1.9073], device='cuda:3'), covar=tensor([0.3226, 0.2202, 0.1998, 0.2145, 0.1612, 0.0225, 0.1687, 0.0896], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0118, 0.0125, 0.0121, 0.0105, 0.0097, 0.0090, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 23:19:26,251 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp1.1 from training. Duration: 22.1090625 2022-12-23 23:19:32,426 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp0.9 from training. Duration: 26.6166875 2022-12-23 23:19:40,601 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.09 vs. limit=5.0 2022-12-23 23:19:46,484 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87170.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:19:49,284 WARNING [train.py:1060] (3/4) Exclude cut with ID 2046-178027-0000-53705-0_sp0.9 from training. Duration: 20.3055625 2022-12-23 23:20:11,234 WARNING [train.py:1060] (3/4) Exclude cut with ID 7205-50138-0008-5373-0_sp0.9 from training. Duration: 20.7 2022-12-23 23:20:28,136 INFO [train.py:894] (3/4) Epoch 25, batch 3050, loss[loss=0.1719, simple_loss=0.2631, pruned_loss=0.04036, over 18572.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2622, pruned_loss=0.04783, over 3715991.96 frames. ], batch size: 56, lr: 4.54e-03, grad_scale: 8.0 2022-12-23 23:20:57,874 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0 from training. Duration: 22.48 2022-12-23 23:21:15,534 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp0.9 from training. Duration: 29.816625 2022-12-23 23:21:17,439 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87230.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:21:23,041 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87234.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:21:23,220 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4724, 1.8853, 0.7576, 2.1306, 2.7568, 1.7527, 2.2917, 2.2723], device='cuda:3'), covar=tensor([0.1458, 0.1960, 0.2492, 0.1362, 0.1623, 0.1704, 0.1419, 0.1640], device='cuda:3'), in_proj_covar=tensor([0.0095, 0.0099, 0.0118, 0.0097, 0.0121, 0.0093, 0.0099, 0.0095], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 23:21:25,986 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87236.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:21:35,684 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp1.1 from training. Duration: 22.7590625 2022-12-23 23:21:40,367 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0 from training. Duration: 22.555 2022-12-23 23:21:41,729 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.992e+02 4.307e+02 5.014e+02 6.364e+02 1.166e+03, threshold=1.003e+03, percent-clipped=4.0 2022-12-23 23:21:43,194 INFO [train.py:894] (3/4) Epoch 25, batch 3100, loss[loss=0.1826, simple_loss=0.2741, pruned_loss=0.04557, over 18677.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2616, pruned_loss=0.04721, over 3714946.89 frames. ], batch size: 60, lr: 4.54e-03, grad_scale: 8.0 2022-12-23 23:21:58,951 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0005-25975-0_sp0.9 from training. Duration: 21.688875 2022-12-23 23:22:35,239 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0_sp0.9 from training. Duration: 22.6 2022-12-23 23:22:40,643 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=87284.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:22:52,615 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87291.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:23:01,655 INFO [train.py:894] (3/4) Epoch 25, batch 3150, loss[loss=0.1759, simple_loss=0.2484, pruned_loss=0.05176, over 18513.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2612, pruned_loss=0.0476, over 3714826.73 frames. ], batch size: 47, lr: 4.54e-03, grad_scale: 8.0 2022-12-23 23:23:09,210 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0 from training. Duration: 24.32 2022-12-23 23:24:09,120 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-276745-0093-13116-0_sp0.9 from training. Duration: 21.061125 2022-12-23 23:24:16,160 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.705e+02 4.200e+02 4.949e+02 5.939e+02 1.054e+03, threshold=9.899e+02, percent-clipped=1.0 2022-12-23 23:24:17,839 INFO [train.py:894] (3/4) Epoch 25, batch 3200, loss[loss=0.1481, simple_loss=0.23, pruned_loss=0.03309, over 18691.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2615, pruned_loss=0.04773, over 3714266.16 frames. ], batch size: 46, lr: 4.54e-03, grad_scale: 8.0 2022-12-23 23:24:24,087 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0024-15855-0_sp0.9 from training. Duration: 20.32225 2022-12-23 23:24:34,897 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87358.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 23:24:36,392 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp1.1 from training. Duration: 0.7545625 2022-12-23 23:24:41,589 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87362.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:24:50,007 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0_sp0.9 from training. Duration: 23.9333125 2022-12-23 23:24:52,909 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.39 vs. limit=5.0 2022-12-23 23:25:21,699 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp1.1 from training. Duration: 20.17275 2022-12-23 23:25:27,096 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp1.1 from training. Duration: 20.436375 2022-12-23 23:25:33,617 INFO [train.py:894] (3/4) Epoch 25, batch 3250, loss[loss=0.175, simple_loss=0.262, pruned_loss=0.04401, over 18632.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.262, pruned_loss=0.04788, over 3714415.20 frames. ], batch size: 53, lr: 4.54e-03, grad_scale: 8.0 2022-12-23 23:26:14,314 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87423.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:26:47,660 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.020e+02 4.157e+02 5.173e+02 6.629e+02 1.386e+03, threshold=1.035e+03, percent-clipped=3.0 2022-12-23 23:26:49,124 INFO [train.py:894] (3/4) Epoch 25, batch 3300, loss[loss=0.1715, simple_loss=0.2443, pruned_loss=0.0493, over 18485.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2613, pruned_loss=0.04767, over 3714335.47 frames. ], batch size: 43, lr: 4.54e-03, grad_scale: 8.0 2022-12-23 23:26:49,234 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0_sp0.9 from training. Duration: 23.1055625 2022-12-23 23:26:50,562 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp1.1 from training. Duration: 23.5 2022-12-23 23:26:59,386 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp0.9 from training. Duration: 26.62775 2022-12-23 23:27:10,796 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87461.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:27:13,962 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0 from training. Duration: 21.105 2022-12-23 23:27:19,906 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0_sp0.9 from training. Duration: 24.411125 2022-12-23 23:27:25,132 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87470.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:27:45,450 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp1.1 from training. Duration: 21.263625 2022-12-23 23:28:05,695 INFO [train.py:894] (3/4) Epoch 25, batch 3350, loss[loss=0.1435, simple_loss=0.2237, pruned_loss=0.03162, over 18687.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2608, pruned_loss=0.04729, over 3713609.62 frames. ], batch size: 46, lr: 4.54e-03, grad_scale: 8.0 2022-12-23 23:28:18,133 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0 from training. Duration: 20.795 2022-12-23 23:28:19,270 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87505.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:28:25,329 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6279, 2.3074, 1.9875, 1.4791, 2.8897, 2.6967, 2.3580, 1.9630], device='cuda:3'), covar=tensor([0.0369, 0.0469, 0.0566, 0.0804, 0.0294, 0.0385, 0.0467, 0.0914], device='cuda:3'), in_proj_covar=tensor([0.0125, 0.0131, 0.0131, 0.0122, 0.0104, 0.0127, 0.0135, 0.0163], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 23:28:29,828 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0 from training. Duration: 24.76 2022-12-23 23:28:29,847 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0_sp0.9 from training. Duration: 22.25 2022-12-23 23:28:39,117 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=87518.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:28:45,040 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87522.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:28:54,831 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp1.1 from training. Duration: 20.5045625 2022-12-23 23:29:02,861 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87534.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:29:20,008 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.807e+02 4.240e+02 5.058e+02 6.182e+02 1.379e+03, threshold=1.012e+03, percent-clipped=2.0 2022-12-23 23:29:21,515 INFO [train.py:894] (3/4) Epoch 25, batch 3400, loss[loss=0.1713, simple_loss=0.2596, pruned_loss=0.04149, over 18515.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2604, pruned_loss=0.04722, over 3713695.79 frames. ], batch size: 58, lr: 4.54e-03, grad_scale: 8.0 2022-12-23 23:29:50,293 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87566.0, num_to_drop=1, layers_to_drop={3} 2022-12-23 23:30:13,764 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=87582.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:30:17,088 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87584.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:30:20,127 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87586.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:30:24,900 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87589.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:30:36,101 INFO [train.py:894] (3/4) Epoch 25, batch 3450, loss[loss=0.1821, simple_loss=0.271, pruned_loss=0.04653, over 18586.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2605, pruned_loss=0.047, over 3714504.39 frames. ], batch size: 56, lr: 4.53e-03, grad_scale: 8.0 2022-12-23 23:30:48,041 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87605.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:31:46,551 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87645.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 23:31:47,681 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.307e+02 4.042e+02 4.989e+02 6.137e+02 1.168e+03, threshold=9.977e+02, percent-clipped=1.0 2022-12-23 23:31:49,100 INFO [train.py:894] (3/4) Epoch 25, batch 3500, loss[loss=0.1767, simple_loss=0.2658, pruned_loss=0.04382, over 18590.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2609, pruned_loss=0.04722, over 3715372.18 frames. ], batch size: 99, lr: 4.53e-03, grad_scale: 8.0 2022-12-23 23:31:54,170 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87650.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:32:10,207 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp1.1 from training. Duration: 22.2954375 2022-12-23 23:32:20,024 INFO [train.py:894] (3/4) Epoch 26, batch 0, loss[loss=0.1733, simple_loss=0.2496, pruned_loss=0.04852, over 18533.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2496, pruned_loss=0.04852, over 18533.00 frames. ], batch size: 44, lr: 4.44e-03, grad_scale: 8.0 2022-12-23 23:32:20,024 INFO [train.py:919] (3/4) Computing validation loss 2022-12-23 23:32:28,017 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.2206, 1.2901, 1.4012, 0.8589, 1.2810, 1.3529, 1.1592, 1.5612], device='cuda:3'), covar=tensor([0.1056, 0.1991, 0.1077, 0.1399, 0.0848, 0.0914, 0.2483, 0.0595], device='cuda:3'), in_proj_covar=tensor([0.0199, 0.0218, 0.0210, 0.0195, 0.0175, 0.0219, 0.0219, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 23:32:31,399 INFO [train.py:928] (3/4) Epoch 26, validation: loss=0.163, simple_loss=0.2598, pruned_loss=0.03308, over 944034.00 frames. 2022-12-23 23:32:31,400 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-23 23:32:39,244 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87658.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 23:32:51,414 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87666.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 23:33:04,154 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5309, 2.5795, 3.0437, 1.5547, 2.9840, 3.3088, 2.1582, 3.4823], device='cuda:3'), covar=tensor([0.1411, 0.1856, 0.1434, 0.2350, 0.0856, 0.1176, 0.2351, 0.0558], device='cuda:3'), in_proj_covar=tensor([0.0200, 0.0218, 0.0211, 0.0195, 0.0175, 0.0219, 0.0219, 0.0196], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 23:33:24,341 WARNING [train.py:1060] (3/4) Exclude cut with ID 298-126791-0067-24026-0_sp0.9 from training. Duration: 21.438875 2022-12-23 23:33:28,967 WARNING [train.py:1060] (3/4) Exclude cut with ID 5652-39938-0025-23684-0_sp0.9 from training. Duration: 22.2055625 2022-12-23 23:33:46,304 INFO [train.py:894] (3/4) Epoch 26, batch 50, loss[loss=0.1809, simple_loss=0.2741, pruned_loss=0.04383, over 18683.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2583, pruned_loss=0.04071, over 837955.94 frames. ], batch size: 78, lr: 4.44e-03, grad_scale: 8.0 2022-12-23 23:33:51,525 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=87706.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 23:33:54,464 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.7659, 4.1160, 4.1007, 4.6685, 4.4030, 4.2035, 4.9257, 1.3433], device='cuda:3'), covar=tensor([0.0602, 0.0604, 0.0626, 0.0745, 0.1173, 0.1030, 0.0447, 0.5249], device='cuda:3'), in_proj_covar=tensor([0.0352, 0.0234, 0.0243, 0.0278, 0.0334, 0.0273, 0.0298, 0.0290], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 23:34:08,973 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87718.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:34:49,306 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.982e+02 3.226e+02 3.859e+02 4.611e+02 8.239e+02, threshold=7.718e+02, percent-clipped=0.0 2022-12-23 23:35:00,000 INFO [train.py:894] (3/4) Epoch 26, batch 100, loss[loss=0.1601, simple_loss=0.2398, pruned_loss=0.0402, over 18595.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2582, pruned_loss=0.04016, over 1475008.20 frames. ], batch size: 41, lr: 4.44e-03, grad_scale: 8.0 2022-12-23 23:35:54,561 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.5018, 3.8897, 3.8447, 4.4418, 4.1596, 3.9127, 4.6344, 1.5544], device='cuda:3'), covar=tensor([0.0614, 0.0651, 0.0671, 0.0692, 0.1153, 0.1010, 0.0488, 0.4790], device='cuda:3'), in_proj_covar=tensor([0.0348, 0.0232, 0.0241, 0.0275, 0.0330, 0.0270, 0.0294, 0.0287], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 23:36:16,420 INFO [train.py:894] (3/4) Epoch 26, batch 150, loss[loss=0.1676, simple_loss=0.254, pruned_loss=0.04065, over 18527.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2578, pruned_loss=0.04001, over 1970734.78 frames. ], batch size: 47, lr: 4.44e-03, grad_scale: 8.0 2022-12-23 23:36:19,774 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.8668, 3.3531, 3.3525, 3.7873, 3.5205, 3.3798, 3.9930, 1.3353], device='cuda:3'), covar=tensor([0.0862, 0.0799, 0.0781, 0.0941, 0.1529, 0.1290, 0.0741, 0.5081], device='cuda:3'), in_proj_covar=tensor([0.0348, 0.0232, 0.0241, 0.0276, 0.0330, 0.0270, 0.0294, 0.0288], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 23:36:25,614 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0 from training. Duration: 24.525 2022-12-23 23:36:38,325 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87817.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:37:00,013 WARNING [train.py:1060] (3/4) Exclude cut with ID 3699-47246-0007-3408-0_sp0.9 from training. Duration: 20.26675 2022-12-23 23:37:12,751 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp0.9 from training. Duration: 27.25 2022-12-23 23:37:19,748 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.944e+02 3.316e+02 3.922e+02 4.543e+02 1.035e+03, threshold=7.845e+02, percent-clipped=6.0 2022-12-23 23:37:31,452 INFO [train.py:894] (3/4) Epoch 26, batch 200, loss[loss=0.1488, simple_loss=0.2375, pruned_loss=0.03008, over 18456.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2581, pruned_loss=0.03974, over 2357151.51 frames. ], batch size: 50, lr: 4.44e-03, grad_scale: 8.0 2022-12-23 23:37:44,459 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87861.0, num_to_drop=1, layers_to_drop={3} 2022-12-23 23:38:20,821 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87886.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:38:24,733 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0 from training. Duration: 21.68 2022-12-23 23:38:37,699 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0 from training. Duration: 21.6300625 2022-12-23 23:38:46,983 INFO [train.py:894] (3/4) Epoch 26, batch 250, loss[loss=0.1657, simple_loss=0.2548, pruned_loss=0.03833, over 18446.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2569, pruned_loss=0.03922, over 2656538.32 frames. ], batch size: 54, lr: 4.44e-03, grad_scale: 8.0 2022-12-23 23:39:01,912 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0_sp0.9 from training. Duration: 24.033375 2022-12-23 23:39:32,893 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=87934.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:39:42,094 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87940.0, num_to_drop=1, layers_to_drop={2} 2022-12-23 23:39:49,851 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87945.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:39:50,848 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.129e+02 3.146e+02 3.956e+02 4.545e+02 7.882e+02, threshold=7.912e+02, percent-clipped=1.0 2022-12-23 23:39:59,790 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0 from training. Duration: 22.905 2022-12-23 23:40:01,209 INFO [train.py:894] (3/4) Epoch 26, batch 300, loss[loss=0.1597, simple_loss=0.2394, pruned_loss=0.04004, over 18671.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2567, pruned_loss=0.03926, over 2890692.45 frames. ], batch size: 46, lr: 4.44e-03, grad_scale: 8.0 2022-12-23 23:40:01,246 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp1.1 from training. Duration: 23.4318125 2022-12-23 23:40:13,544 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87961.0, num_to_drop=1, layers_to_drop={3} 2022-12-23 23:41:19,133 INFO [train.py:894] (3/4) Epoch 26, batch 350, loss[loss=0.2056, simple_loss=0.294, pruned_loss=0.05853, over 18515.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2574, pruned_loss=0.03953, over 3074025.26 frames. ], batch size: 58, lr: 4.43e-03, grad_scale: 8.0 2022-12-23 23:41:41,121 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88018.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:41:59,615 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp1.1 from training. Duration: 20.82275 2022-12-23 23:42:01,018 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp0.9 from training. Duration: 25.45 2022-12-23 23:42:18,659 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5710, 1.0716, 2.0267, 3.2549, 2.2770, 2.5591, 0.8126, 2.3952], device='cuda:3'), covar=tensor([0.1924, 0.1833, 0.1417, 0.0570, 0.1027, 0.1072, 0.2361, 0.0950], device='cuda:3'), in_proj_covar=tensor([0.0102, 0.0117, 0.0135, 0.0153, 0.0105, 0.0144, 0.0129, 0.0114], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-23 23:42:23,170 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.110e+02 3.234e+02 3.875e+02 5.002e+02 9.913e+02, threshold=7.750e+02, percent-clipped=2.0 2022-12-23 23:42:23,837 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.98 vs. limit=5.0 2022-12-23 23:42:33,151 INFO [train.py:894] (3/4) Epoch 26, batch 400, loss[loss=0.1761, simple_loss=0.2574, pruned_loss=0.0474, over 18705.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2571, pruned_loss=0.03903, over 3215789.89 frames. ], batch size: 46, lr: 4.43e-03, grad_scale: 8.0 2022-12-23 23:42:52,312 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=88066.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:42:54,048 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3139, 3.7434, 3.4838, 1.3547, 3.8499, 2.9279, 0.5281, 2.2362], device='cuda:3'), covar=tensor([0.2322, 0.0950, 0.1459, 0.3712, 0.0759, 0.0876, 0.5073, 0.1575], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0147, 0.0164, 0.0126, 0.0151, 0.0115, 0.0146, 0.0116], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 23:42:59,539 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0 from training. Duration: 25.775 2022-12-23 23:43:22,820 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0_sp0.9 from training. Duration: 22.25 2022-12-23 23:43:43,288 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=88100.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:43:47,791 INFO [train.py:894] (3/4) Epoch 26, batch 450, loss[loss=0.1671, simple_loss=0.2612, pruned_loss=0.03648, over 18719.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.258, pruned_loss=0.03931, over 3326847.68 frames. ], batch size: 52, lr: 4.43e-03, grad_scale: 8.0 2022-12-23 23:43:49,192 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0 from training. Duration: 26.205 2022-12-23 23:44:05,996 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp0.9 from training. Duration: 30.1555625 2022-12-23 23:44:09,381 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88117.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:44:10,549 WARNING [train.py:1060] (3/4) Exclude cut with ID 1265-135635-0050-6781-0_sp0.9 from training. Duration: 21.8333125 2022-12-23 23:44:19,540 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp1.1 from training. Duration: 20.6545625 2022-12-23 23:44:52,910 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.473e+02 3.354e+02 3.991e+02 4.865e+02 8.439e+02, threshold=7.982e+02, percent-clipped=3.0 2022-12-23 23:45:01,851 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0045-39920-0_sp0.9 from training. Duration: 20.52225 2022-12-23 23:45:03,298 INFO [train.py:894] (3/4) Epoch 26, batch 500, loss[loss=0.1673, simple_loss=0.2487, pruned_loss=0.04296, over 18407.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2592, pruned_loss=0.03961, over 3412459.13 frames. ], batch size: 46, lr: 4.43e-03, grad_scale: 8.0 2022-12-23 23:45:15,286 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88161.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 23:45:15,359 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=88161.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:45:21,114 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp0.9 from training. Duration: 29.1166875 2022-12-23 23:45:21,195 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=88165.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:46:17,802 INFO [train.py:894] (3/4) Epoch 26, batch 550, loss[loss=0.1786, simple_loss=0.2729, pruned_loss=0.04216, over 18603.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2596, pruned_loss=0.04023, over 3479350.10 frames. ], batch size: 51, lr: 4.43e-03, grad_scale: 8.0 2022-12-23 23:46:22,111 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133211-0007-59831-0_sp0.9 from training. Duration: 21.388875 2022-12-23 23:46:27,124 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=88209.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:46:56,643 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0 from training. Duration: 22.72 2022-12-23 23:46:58,031 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0_sp0.9 from training. Duration: 22.7444375 2022-12-23 23:47:13,858 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88240.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:47:18,415 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0140, 2.6725, 2.2135, 3.0878, 2.8207, 1.9097, 2.1290, 1.5862], device='cuda:3'), covar=tensor([0.1914, 0.1680, 0.1427, 0.0852, 0.1560, 0.1199, 0.2007, 0.1585], device='cuda:3'), in_proj_covar=tensor([0.0248, 0.0228, 0.0219, 0.0202, 0.0261, 0.0198, 0.0226, 0.0202], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 23:47:21,468 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88245.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:47:22,438 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.456e+02 3.440e+02 4.053e+02 4.929e+02 1.052e+03, threshold=8.106e+02, percent-clipped=3.0 2022-12-23 23:47:33,652 INFO [train.py:894] (3/4) Epoch 26, batch 600, loss[loss=0.1621, simple_loss=0.2551, pruned_loss=0.03457, over 18569.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2592, pruned_loss=0.04005, over 3531063.14 frames. ], batch size: 56, lr: 4.43e-03, grad_scale: 8.0 2022-12-23 23:47:40,061 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.28 vs. limit=5.0 2022-12-23 23:47:42,332 WARNING [train.py:1060] (3/4) Exclude cut with ID 4133-6541-0027-40495-0_sp1.1 from training. Duration: 0.9681875 2022-12-23 23:47:45,217 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0_sp0.9 from training. Duration: 22.3166875 2022-12-23 23:47:45,457 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88261.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 23:47:49,257 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133212-0015-59917-0_sp0.9 from training. Duration: 21.8166875 2022-12-23 23:48:26,450 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=88288.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:48:31,001 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=88291.0, num_to_drop=1, layers_to_drop={1} 2022-12-23 23:48:34,242 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=88293.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:48:48,665 INFO [train.py:894] (3/4) Epoch 26, batch 650, loss[loss=0.17, simple_loss=0.2593, pruned_loss=0.04038, over 18452.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2602, pruned_loss=0.04034, over 3571419.02 frames. ], batch size: 50, lr: 4.43e-03, grad_scale: 8.0 2022-12-23 23:48:57,899 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=88309.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:49:29,152 WARNING [train.py:1060] (3/4) Exclude cut with ID 4957-30119-0041-23990-0_sp0.9 from training. Duration: 20.22775 2022-12-23 23:49:53,359 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.132e+02 3.390e+02 4.009e+02 4.862e+02 9.909e+02, threshold=8.017e+02, percent-clipped=1.0 2022-12-23 23:50:02,448 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=88352.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 23:50:03,575 INFO [train.py:894] (3/4) Epoch 26, batch 700, loss[loss=0.2175, simple_loss=0.3028, pruned_loss=0.06611, over 18626.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2596, pruned_loss=0.04016, over 3603127.39 frames. ], batch size: 174, lr: 4.43e-03, grad_scale: 8.0 2022-12-23 23:50:14,904 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp1.1 from training. Duration: 24.67275 2022-12-23 23:50:16,828 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6638, 2.1339, 1.6371, 2.3739, 2.6466, 1.6763, 1.7130, 1.3365], device='cuda:3'), covar=tensor([0.1877, 0.1648, 0.1578, 0.0970, 0.1177, 0.1088, 0.2009, 0.1543], device='cuda:3'), in_proj_covar=tensor([0.0248, 0.0228, 0.0220, 0.0202, 0.0262, 0.0198, 0.0227, 0.0202], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 23:50:40,222 WARNING [train.py:1060] (3/4) Exclude cut with ID 3082-165428-0081-50734-0_sp0.9 from training. Duration: 21.8055625 2022-12-23 23:50:49,263 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7007, 1.4547, 1.6847, 1.9344, 1.6601, 3.6510, 1.3432, 1.5670], device='cuda:3'), covar=tensor([0.0829, 0.1890, 0.1079, 0.0937, 0.1537, 0.0223, 0.1515, 0.1629], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0083, 0.0072, 0.0074, 0.0092, 0.0077, 0.0085, 0.0077], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2022-12-23 23:51:18,936 INFO [train.py:894] (3/4) Epoch 26, batch 750, loss[loss=0.1513, simple_loss=0.2369, pruned_loss=0.03287, over 18689.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2591, pruned_loss=0.04046, over 3627332.42 frames. ], batch size: 46, lr: 4.42e-03, grad_scale: 8.0 2022-12-23 23:51:20,424 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0_sp0.9 from training. Duration: 22.6666875 2022-12-23 23:52:22,632 WARNING [train.py:1060] (3/4) Exclude cut with ID 2411-132532-0017-82279-0_sp1.1 from training. Duration: 0.9681875 2022-12-23 23:52:23,980 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.008e+02 3.060e+02 3.630e+02 4.640e+02 8.422e+02, threshold=7.260e+02, percent-clipped=1.0 2022-12-23 23:52:34,208 INFO [train.py:894] (3/4) Epoch 26, batch 800, loss[loss=0.1493, simple_loss=0.2334, pruned_loss=0.03265, over 18528.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2589, pruned_loss=0.04007, over 3646479.01 frames. ], batch size: 44, lr: 4.42e-03, grad_scale: 8.0 2022-12-23 23:52:38,806 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=88456.0, num_to_drop=0, layers_to_drop=set() 2022-12-23 23:52:45,578 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0 from training. Duration: 22.485 2022-12-23 23:53:11,426 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.0847, 1.0199, 0.9289, 1.2168, 1.2880, 1.1952, 1.0809, 0.9793], device='cuda:3'), covar=tensor([0.0298, 0.0247, 0.0598, 0.0230, 0.0249, 0.0374, 0.0306, 0.0336], device='cuda:3'), in_proj_covar=tensor([0.0095, 0.0127, 0.0153, 0.0122, 0.0117, 0.0121, 0.0100, 0.0127], device='cuda:3'), out_proj_covar=tensor([7.5071e-05, 1.0013e-04, 1.2512e-04, 9.6460e-05, 9.3536e-05, 9.2986e-05, 7.8119e-05, 9.9969e-05], device='cuda:3') 2022-12-23 23:53:26,546 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp1.1 from training. Duration: 23.82275 2022-12-23 23:53:41,116 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0 from training. Duration: 20.77 2022-12-23 23:53:48,544 INFO [train.py:894] (3/4) Epoch 26, batch 850, loss[loss=0.2201, simple_loss=0.302, pruned_loss=0.06908, over 18611.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2587, pruned_loss=0.04018, over 3660615.43 frames. ], batch size: 171, lr: 4.42e-03, grad_scale: 8.0 2022-12-23 23:53:48,561 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0_sp0.9 from training. Duration: 24.088875 2022-12-23 23:54:17,208 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp1.1 from training. Duration: 20.4409375 2022-12-23 23:54:53,871 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.940e+02 3.168e+02 3.813e+02 4.877e+02 1.035e+03, threshold=7.626e+02, percent-clipped=7.0 2022-12-23 23:55:03,865 INFO [train.py:894] (3/4) Epoch 26, batch 900, loss[loss=0.1811, simple_loss=0.2738, pruned_loss=0.04419, over 18532.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2597, pruned_loss=0.0403, over 3671358.20 frames. ], batch size: 97, lr: 4.42e-03, grad_scale: 8.0 2022-12-23 23:55:31,934 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0_sp0.9 from training. Duration: 22.511125 2022-12-23 23:55:33,315 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0 from training. Duration: 20.675 2022-12-23 23:56:21,084 INFO [train.py:894] (3/4) Epoch 26, batch 950, loss[loss=0.1793, simple_loss=0.2676, pruned_loss=0.04547, over 18521.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2599, pruned_loss=0.04042, over 3680378.86 frames. ], batch size: 58, lr: 4.42e-03, grad_scale: 8.0 2022-12-23 23:56:50,609 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9492, 1.8073, 1.8916, 1.8205, 1.5221, 5.0797, 2.1023, 2.3297], device='cuda:3'), covar=tensor([0.3013, 0.2077, 0.1888, 0.2049, 0.1353, 0.0083, 0.1383, 0.0857], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0119, 0.0126, 0.0122, 0.0106, 0.0097, 0.0091, 0.0091], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-23 23:57:13,745 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp0.9 from training. Duration: 24.9833125 2022-12-23 23:57:16,823 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7600, 1.1639, 0.7637, 1.2870, 2.1055, 0.9914, 1.3720, 1.5325], device='cuda:3'), covar=tensor([0.1643, 0.2210, 0.2199, 0.1572, 0.1716, 0.1851, 0.1446, 0.1769], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0098, 0.0116, 0.0096, 0.0119, 0.0092, 0.0098, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-23 23:57:22,938 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5153, 2.5533, 1.8635, 3.0238, 2.7720, 2.3657, 3.6925, 2.5358], device='cuda:3'), covar=tensor([0.0891, 0.1871, 0.2809, 0.1950, 0.1817, 0.0904, 0.0829, 0.1283], device='cuda:3'), in_proj_covar=tensor([0.0182, 0.0216, 0.0258, 0.0295, 0.0243, 0.0196, 0.0209, 0.0210], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-23 23:57:25,439 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.452e+02 3.318e+02 3.992e+02 4.999e+02 2.515e+03, threshold=7.985e+02, percent-clipped=6.0 2022-12-23 23:57:27,322 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=88647.0, num_to_drop=1, layers_to_drop={0} 2022-12-23 23:57:35,824 INFO [train.py:894] (3/4) Epoch 26, batch 1000, loss[loss=0.1527, simple_loss=0.2454, pruned_loss=0.03, over 18711.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.259, pruned_loss=0.03984, over 3688179.29 frames. ], batch size: 50, lr: 4.42e-03, grad_scale: 8.0 2022-12-23 23:57:45,892 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0 from training. Duration: 27.14 2022-12-23 23:58:00,310 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0 from training. Duration: 22.44 2022-12-23 23:58:49,453 INFO [train.py:894] (3/4) Epoch 26, batch 1050, loss[loss=0.1764, simple_loss=0.2707, pruned_loss=0.04105, over 18709.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2598, pruned_loss=0.03995, over 3694719.62 frames. ], batch size: 50, lr: 4.42e-03, grad_scale: 8.0 2022-12-23 23:59:16,372 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0060-62364-0_sp0.9 from training. Duration: 21.361125 2022-12-23 23:59:21,934 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp1.1 from training. Duration: 27.0318125 2022-12-23 23:59:26,639 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5931, 3.7413, 3.5222, 1.5932, 3.8055, 2.7007, 0.6086, 2.4665], device='cuda:3'), covar=tensor([0.2244, 0.1132, 0.1714, 0.3905, 0.0841, 0.1058, 0.5332, 0.1802], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0145, 0.0161, 0.0124, 0.0148, 0.0114, 0.0144, 0.0114], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-23 23:59:32,437 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp0.9 from training. Duration: 28.638875 2022-12-23 23:59:47,032 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0 from training. Duration: 20.4 2022-12-23 23:59:53,978 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.848e+02 3.398e+02 4.016e+02 4.877e+02 1.233e+03, threshold=8.032e+02, percent-clipped=4.0 2022-12-24 00:00:04,360 INFO [train.py:894] (3/4) Epoch 26, batch 1100, loss[loss=0.1627, simple_loss=0.2623, pruned_loss=0.03153, over 18576.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2593, pruned_loss=0.04017, over 3698404.70 frames. ], batch size: 77, lr: 4.42e-03, grad_scale: 8.0 2022-12-24 00:00:09,279 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88756.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:00:20,108 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0 from training. Duration: 20.025 2022-12-24 00:00:20,118 WARNING [train.py:1060] (3/4) Exclude cut with ID 2364-131735-0112-64612-0_sp0.9 from training. Duration: 20.488875 2022-12-24 00:00:24,856 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0 from training. Duration: 29.735 2022-12-24 00:01:20,081 INFO [train.py:894] (3/4) Epoch 26, batch 1150, loss[loss=0.1891, simple_loss=0.2802, pruned_loss=0.04896, over 18454.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2586, pruned_loss=0.03977, over 3700623.94 frames. ], batch size: 64, lr: 4.41e-03, grad_scale: 8.0 2022-12-24 00:01:21,770 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=88804.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:01:48,993 WARNING [train.py:1060] (3/4) Exclude cut with ID 7276-92427-0014-12983-0_sp0.9 from training. Duration: 21.3055625 2022-12-24 00:01:50,511 WARNING [train.py:1060] (3/4) Exclude cut with ID 1025-75365-0008-79168-0_sp0.9 from training. Duration: 22.0666875 2022-12-24 00:02:24,920 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.168e+02 3.282e+02 3.913e+02 4.660e+02 1.073e+03, threshold=7.826e+02, percent-clipped=1.0 2022-12-24 00:02:35,233 INFO [train.py:894] (3/4) Epoch 26, batch 1200, loss[loss=0.1661, simple_loss=0.2538, pruned_loss=0.0392, over 18425.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2578, pruned_loss=0.03931, over 3702917.86 frames. ], batch size: 48, lr: 4.41e-03, grad_scale: 8.0 2022-12-24 00:02:45,745 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6927, 1.5433, 1.1253, 0.2927, 1.1748, 1.5897, 1.4217, 1.4759], device='cuda:3'), covar=tensor([0.0764, 0.0580, 0.1132, 0.1733, 0.1133, 0.1820, 0.1923, 0.0760], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0188, 0.0206, 0.0191, 0.0209, 0.0202, 0.0216, 0.0202], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 00:03:09,903 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6650, 4.0873, 3.8807, 1.8891, 4.2371, 3.0844, 0.8030, 2.7463], device='cuda:3'), covar=tensor([0.2119, 0.1009, 0.1434, 0.3294, 0.0690, 0.0879, 0.4875, 0.1330], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0143, 0.0159, 0.0123, 0.0147, 0.0114, 0.0143, 0.0113], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-24 00:03:15,863 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9622, 1.6598, 1.9123, 1.8007, 1.6772, 4.9132, 2.0341, 2.4529], device='cuda:3'), covar=tensor([0.3062, 0.2261, 0.1892, 0.2118, 0.1320, 0.0098, 0.1436, 0.0829], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0119, 0.0126, 0.0123, 0.0106, 0.0097, 0.0091, 0.0091], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-24 00:03:37,342 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0 from training. Duration: 20.26 2022-12-24 00:03:50,599 INFO [train.py:894] (3/4) Epoch 26, batch 1250, loss[loss=0.1857, simple_loss=0.2838, pruned_loss=0.04381, over 18607.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2583, pruned_loss=0.03942, over 3705241.55 frames. ], batch size: 57, lr: 4.41e-03, grad_scale: 16.0 2022-12-24 00:03:50,658 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0030-9324-0_sp0.9 from training. Duration: 21.3444375 2022-12-24 00:04:00,776 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.2956, 1.7040, 1.9206, 0.9479, 1.1918, 2.0449, 1.9487, 1.6357], device='cuda:3'), covar=tensor([0.0794, 0.0321, 0.0329, 0.0421, 0.0407, 0.0429, 0.0245, 0.0677], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0174, 0.0131, 0.0142, 0.0149, 0.0145, 0.0167, 0.0177], device='cuda:3'), out_proj_covar=tensor([1.1463e-04, 1.3107e-04, 9.6821e-05, 1.0392e-04, 1.0969e-04, 1.0928e-04, 1.2657e-04, 1.3361e-04], device='cuda:3') 2022-12-24 00:04:47,169 WARNING [train.py:1060] (3/4) Exclude cut with ID 497-129325-0061-62254-0_sp1.1 from training. Duration: 0.97725 2022-12-24 00:04:54,590 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.240e+02 3.350e+02 3.895e+02 4.856e+02 7.828e+02, threshold=7.791e+02, percent-clipped=1.0 2022-12-24 00:04:55,165 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6241, 1.9208, 1.5377, 2.1877, 2.3860, 1.6221, 1.2786, 1.3286], device='cuda:3'), covar=tensor([0.1870, 0.1704, 0.1612, 0.1022, 0.1200, 0.1062, 0.2286, 0.1569], device='cuda:3'), in_proj_covar=tensor([0.0249, 0.0230, 0.0220, 0.0203, 0.0263, 0.0199, 0.0227, 0.0203], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 00:04:56,389 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88947.0, num_to_drop=1, layers_to_drop={2} 2022-12-24 00:05:04,822 INFO [train.py:894] (3/4) Epoch 26, batch 1300, loss[loss=0.18, simple_loss=0.2759, pruned_loss=0.04206, over 18568.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.259, pruned_loss=0.03971, over 3707382.86 frames. ], batch size: 56, lr: 4.41e-03, grad_scale: 16.0 2022-12-24 00:05:27,600 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0_sp0.9 from training. Duration: 22.97225 2022-12-24 00:05:59,211 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0047-39922-0_sp0.9 from training. Duration: 21.97775 2022-12-24 00:06:08,030 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=88995.0, num_to_drop=1, layers_to_drop={1} 2022-12-24 00:06:11,928 WARNING [train.py:1060] (3/4) Exclude cut with ID 1112-1043-0006-89194-0_sp0.9 from training. Duration: 21.8333125 2022-12-24 00:06:19,456 INFO [train.py:894] (3/4) Epoch 26, batch 1350, loss[loss=0.1676, simple_loss=0.2665, pruned_loss=0.03429, over 18722.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2595, pruned_loss=0.03992, over 3710062.92 frames. ], batch size: 52, lr: 4.41e-03, grad_scale: 16.0 2022-12-24 00:06:22,535 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0 from training. Duration: 20.47 2022-12-24 00:06:53,302 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2022-12-24 00:06:56,702 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89027.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:07:00,029 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0853, 1.4026, 1.7909, 1.7156, 2.0708, 2.1348, 1.8757, 1.7933], device='cuda:3'), covar=tensor([0.2342, 0.3406, 0.2809, 0.3045, 0.2209, 0.1050, 0.3430, 0.1464], device='cuda:3'), in_proj_covar=tensor([0.0273, 0.0300, 0.0287, 0.0324, 0.0317, 0.0258, 0.0353, 0.0248], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 00:07:24,338 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.964e+02 3.078e+02 3.769e+02 4.964e+02 1.403e+03, threshold=7.538e+02, percent-clipped=3.0 2022-12-24 00:07:28,662 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0037-39912-0_sp0.9 from training. Duration: 20.67225 2022-12-24 00:07:34,451 INFO [train.py:894] (3/4) Epoch 26, batch 1400, loss[loss=0.1554, simple_loss=0.252, pruned_loss=0.02937, over 18589.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2594, pruned_loss=0.03965, over 3710844.65 frames. ], batch size: 56, lr: 4.41e-03, grad_scale: 16.0 2022-12-24 00:07:48,746 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp0.9 from training. Duration: 25.2444375 2022-12-24 00:08:12,365 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0021-76797-0_sp0.9 from training. Duration: 21.1445 2022-12-24 00:08:27,506 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89088.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:08:32,606 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2022-12-24 00:08:48,468 INFO [train.py:894] (3/4) Epoch 26, batch 1450, loss[loss=0.1679, simple_loss=0.2656, pruned_loss=0.0351, over 18657.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2589, pruned_loss=0.03938, over 3711596.11 frames. ], batch size: 62, lr: 4.41e-03, grad_scale: 16.0 2022-12-24 00:09:01,918 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8802, 1.9483, 2.0604, 1.0789, 2.1494, 2.2408, 1.5642, 2.4538], device='cuda:3'), covar=tensor([0.1082, 0.1742, 0.1179, 0.1855, 0.0645, 0.0887, 0.2324, 0.0486], device='cuda:3'), in_proj_covar=tensor([0.0195, 0.0213, 0.0206, 0.0193, 0.0171, 0.0214, 0.0213, 0.0190], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 00:09:27,528 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp0.9 from training. Duration: 33.038875 2022-12-24 00:09:53,307 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.527e+02 3.198e+02 3.701e+02 4.884e+02 9.983e+02, threshold=7.403e+02, percent-clipped=2.0 2022-12-24 00:10:03,997 INFO [train.py:894] (3/4) Epoch 26, batch 1500, loss[loss=0.1799, simple_loss=0.2771, pruned_loss=0.04135, over 18451.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2584, pruned_loss=0.03918, over 3711625.89 frames. ], batch size: 54, lr: 4.41e-03, grad_scale: 16.0 2022-12-24 00:10:04,060 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64291-0000-16059-0_sp0.9 from training. Duration: 20.0944375 2022-12-24 00:10:19,302 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp1.1 from training. Duration: 20.4 2022-12-24 00:10:24,090 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0554, 1.6411, 0.9785, 1.7909, 2.2253, 1.5905, 1.7118, 1.9325], device='cuda:3'), covar=tensor([0.1372, 0.1859, 0.2137, 0.1220, 0.1660, 0.1670, 0.1304, 0.1504], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0097, 0.0116, 0.0095, 0.0119, 0.0091, 0.0098, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-24 00:10:25,178 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0 from training. Duration: 20.085 2022-12-24 00:10:28,725 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89169.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:10:37,331 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0_sp0.9 from training. Duration: 23.07775 2022-12-24 00:11:02,203 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6150, 1.3967, 1.4295, 0.8836, 1.6797, 1.6054, 1.5217, 1.3350], device='cuda:3'), covar=tensor([0.0404, 0.0616, 0.0524, 0.0794, 0.0477, 0.0457, 0.0494, 0.0974], device='cuda:3'), in_proj_covar=tensor([0.0124, 0.0131, 0.0130, 0.0119, 0.0103, 0.0126, 0.0134, 0.0161], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 00:11:03,033 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2022-12-24 00:11:14,068 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2022-12-24 00:11:18,026 INFO [train.py:894] (3/4) Epoch 26, batch 1550, loss[loss=0.1989, simple_loss=0.286, pruned_loss=0.05591, over 18631.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.259, pruned_loss=0.03977, over 3711024.88 frames. ], batch size: 60, lr: 4.40e-03, grad_scale: 16.0 2022-12-24 00:11:26,743 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp0.9 from training. Duration: 24.9333125 2022-12-24 00:11:59,397 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89230.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:12:13,865 WARNING [train.py:1060] (3/4) Exclude cut with ID 5118-111612-0016-124680-0_sp0.9 from training. Duration: 20.388875 2022-12-24 00:12:18,046 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp1.1 from training. Duration: 20.3590625 2022-12-24 00:12:22,211 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.063e+02 3.106e+02 4.096e+02 5.017e+02 1.539e+03, threshold=8.192e+02, percent-clipped=3.0 2022-12-24 00:12:32,777 INFO [train.py:894] (3/4) Epoch 26, batch 1600, loss[loss=0.1726, simple_loss=0.254, pruned_loss=0.04556, over 18679.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.259, pruned_loss=0.0398, over 3711287.78 frames. ], batch size: 48, lr: 4.40e-03, grad_scale: 16.0 2022-12-24 00:13:25,753 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0_sp1.1 from training. Duration: 0.836375 2022-12-24 00:13:33,259 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5061, 1.8897, 2.0269, 2.1272, 2.2812, 2.3834, 2.3112, 2.0457], device='cuda:3'), covar=tensor([0.2291, 0.3444, 0.2641, 0.3145, 0.2147, 0.1046, 0.3620, 0.1356], device='cuda:3'), in_proj_covar=tensor([0.0271, 0.0298, 0.0285, 0.0322, 0.0315, 0.0257, 0.0352, 0.0246], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 00:13:47,653 INFO [train.py:894] (3/4) Epoch 26, batch 1650, loss[loss=0.1558, simple_loss=0.2502, pruned_loss=0.03066, over 18712.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.259, pruned_loss=0.04012, over 3712211.78 frames. ], batch size: 50, lr: 4.40e-03, grad_scale: 16.0 2022-12-24 00:13:57,824 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2022-12-24 00:14:07,884 WARNING [train.py:1060] (3/4) Exclude cut with ID 8565-290391-0049-67394-0_sp0.9 from training. Duration: 21.3166875 2022-12-24 00:14:18,206 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2022-12-24 00:14:26,680 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.5570, 1.5472, 1.7647, 1.0846, 1.1611, 1.8629, 1.7658, 1.5090], device='cuda:3'), covar=tensor([0.0837, 0.0424, 0.0334, 0.0441, 0.0468, 0.0529, 0.0279, 0.0787], device='cuda:3'), in_proj_covar=tensor([0.0153, 0.0176, 0.0132, 0.0144, 0.0151, 0.0146, 0.0168, 0.0179], device='cuda:3'), out_proj_covar=tensor([1.1575e-04, 1.3260e-04, 9.7783e-05, 1.0553e-04, 1.1095e-04, 1.1026e-04, 1.2754e-04, 1.3494e-04], device='cuda:3') 2022-12-24 00:14:37,521 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0029-104863-0_sp0.9 from training. Duration: 22.1055625 2022-12-24 00:14:47,746 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp1.1 from training. Duration: 21.77725 2022-12-24 00:14:50,292 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.450e+02 3.582e+02 4.478e+02 5.269e+02 1.215e+03, threshold=8.956e+02, percent-clipped=2.0 2022-12-24 00:15:01,876 INFO [train.py:894] (3/4) Epoch 26, batch 1700, loss[loss=0.1847, simple_loss=0.2782, pruned_loss=0.04562, over 18537.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2597, pruned_loss=0.04135, over 3713924.12 frames. ], batch size: 58, lr: 4.40e-03, grad_scale: 16.0 2022-12-24 00:15:10,217 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp0.9 from training. Duration: 27.8166875 2022-12-24 00:15:32,948 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6442, 1.3124, 1.9014, 2.9725, 2.2803, 2.4204, 0.8071, 2.1684], device='cuda:3'), covar=tensor([0.1839, 0.1743, 0.1418, 0.0686, 0.0933, 0.1040, 0.2152, 0.1077], device='cuda:3'), in_proj_covar=tensor([0.0103, 0.0118, 0.0137, 0.0153, 0.0106, 0.0144, 0.0129, 0.0115], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2022-12-24 00:15:32,985 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89374.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:15:35,471 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp1.1 from training. Duration: 22.5090625 2022-12-24 00:15:41,280 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0 from training. Duration: 25.035 2022-12-24 00:15:45,855 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89383.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:15:59,524 WARNING [train.py:1060] (3/4) Exclude cut with ID 774-127930-0014-10412-0_sp1.1 from training. Duration: 0.95 2022-12-24 00:16:17,126 INFO [train.py:894] (3/4) Epoch 26, batch 1750, loss[loss=0.1702, simple_loss=0.2387, pruned_loss=0.05084, over 18488.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2596, pruned_loss=0.04265, over 3712545.90 frames. ], batch size: 43, lr: 4.40e-03, grad_scale: 16.0 2022-12-24 00:16:18,980 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp0.9 from training. Duration: 0.92225 2022-12-24 00:16:36,256 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.1512, 3.5907, 3.6068, 4.0724, 3.8257, 3.6610, 4.2775, 1.5628], device='cuda:3'), covar=tensor([0.0765, 0.0738, 0.0710, 0.0859, 0.1298, 0.1251, 0.0725, 0.4667], device='cuda:3'), in_proj_covar=tensor([0.0350, 0.0231, 0.0240, 0.0277, 0.0327, 0.0272, 0.0296, 0.0287], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 00:16:45,535 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0 from training. Duration: 21.97 2022-12-24 00:16:50,104 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89425.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:17:04,679 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89435.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:17:05,841 WARNING [train.py:1060] (3/4) Exclude cut with ID 7492-105653-0055-62765-0_sp0.9 from training. Duration: 21.97225 2022-12-24 00:17:07,468 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp0.9 from training. Duration: 25.3333125 2022-12-24 00:17:10,630 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89439.0, num_to_drop=1, layers_to_drop={1} 2022-12-24 00:17:16,243 WARNING [train.py:1060] (3/4) Exclude cut with ID 5172-29468-0015-19128-0_sp0.9 from training. Duration: 21.5055625 2022-12-24 00:17:21,237 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.550e+02 4.037e+02 4.930e+02 6.113e+02 1.176e+03, threshold=9.860e+02, percent-clipped=6.0 2022-12-24 00:17:27,698 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp1.1 from training. Duration: 20.72725 2022-12-24 00:17:32,060 INFO [train.py:894] (3/4) Epoch 26, batch 1800, loss[loss=0.1788, simple_loss=0.2656, pruned_loss=0.04604, over 18609.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2612, pruned_loss=0.04402, over 3713419.03 frames. ], batch size: 62, lr: 4.40e-03, grad_scale: 16.0 2022-12-24 00:17:59,419 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp0.9 from training. Duration: 26.32775 2022-12-24 00:18:21,751 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89486.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:18:33,801 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0 from training. Duration: 20.025 2022-12-24 00:18:38,733 WARNING [train.py:1060] (3/4) Exclude cut with ID 6709-74022-0004-86860-0_sp1.1 from training. Duration: 0.9409375 2022-12-24 00:18:38,742 WARNING [train.py:1060] (3/4) Exclude cut with ID 4757-1811-0023-62229-0_sp0.9 from training. Duration: 21.37775 2022-12-24 00:18:43,284 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89500.0, num_to_drop=1, layers_to_drop={2} 2022-12-24 00:18:46,970 INFO [train.py:894] (3/4) Epoch 26, batch 1850, loss[loss=0.1565, simple_loss=0.2433, pruned_loss=0.03485, over 18384.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2616, pruned_loss=0.04511, over 3712481.17 frames. ], batch size: 46, lr: 4.40e-03, grad_scale: 16.0 2022-12-24 00:18:55,989 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89509.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:19:00,027 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0004-25974-0_sp0.9 from training. Duration: 21.17225 2022-12-24 00:19:01,499 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp0.9 from training. Duration: 27.511125 2022-12-24 00:19:05,939 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4255, 1.2534, 1.8193, 2.9117, 2.1200, 2.3743, 1.0494, 2.0876], device='cuda:3'), covar=tensor([0.1993, 0.1643, 0.1388, 0.0655, 0.1029, 0.1115, 0.1889, 0.1127], device='cuda:3'), in_proj_covar=tensor([0.0102, 0.0118, 0.0136, 0.0153, 0.0106, 0.0143, 0.0129, 0.0114], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2022-12-24 00:19:19,578 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89525.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:19:32,414 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0 from training. Duration: 22.8 2022-12-24 00:19:36,723 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0 from training. Duration: 22.585 2022-12-24 00:19:51,659 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.439e+02 3.869e+02 4.361e+02 5.499e+02 1.093e+03, threshold=8.721e+02, percent-clipped=2.0 2022-12-24 00:20:01,914 INFO [train.py:894] (3/4) Epoch 26, batch 1900, loss[loss=0.1749, simple_loss=0.2537, pruned_loss=0.04803, over 18420.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2615, pruned_loss=0.04559, over 3712915.21 frames. ], batch size: 48, lr: 4.40e-03, grad_scale: 16.0 2022-12-24 00:20:03,925 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.6232, 3.2921, 2.8987, 1.3416, 2.6347, 2.4131, 2.2487, 2.7144], device='cuda:3'), covar=tensor([0.0631, 0.0557, 0.1341, 0.1921, 0.1554, 0.1411, 0.1481, 0.1078], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0188, 0.0207, 0.0191, 0.0211, 0.0204, 0.0217, 0.0203], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 00:20:04,961 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0001-146967-0_sp0.9 from training. Duration: 22.0166875 2022-12-24 00:20:21,381 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp1.1 from training. Duration: 24.395375 2022-12-24 00:20:27,924 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89570.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:20:29,022 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp0.9 from training. Duration: 27.47775 2022-12-24 00:20:32,028 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp0.9 from training. Duration: 24.8833125 2022-12-24 00:20:34,912 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0 from training. Duration: 23.39 2022-12-24 00:20:40,916 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp0.9 from training. Duration: 28.72225 2022-12-24 00:20:49,648 WARNING [train.py:1060] (3/4) Exclude cut with ID 585-294811-0110-133686-0_sp0.9 from training. Duration: 20.8944375 2022-12-24 00:21:08,100 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0_sp0.9 from training. Duration: 23.8444375 2022-12-24 00:21:17,393 INFO [train.py:894] (3/4) Epoch 26, batch 1950, loss[loss=0.1592, simple_loss=0.2359, pruned_loss=0.04126, over 18522.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2615, pruned_loss=0.04598, over 3713892.74 frames. ], batch size: 41, lr: 4.39e-03, grad_scale: 16.0 2022-12-24 00:21:29,262 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0 from training. Duration: 25.85 2022-12-24 00:21:29,274 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0 from training. Duration: 21.39 2022-12-24 00:21:40,913 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0 from training. Duration: 27.92 2022-12-24 00:22:09,260 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0039-130165-0_sp0.9 from training. Duration: 20.661125 2022-12-24 00:22:13,999 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7638, 1.8836, 1.9980, 2.4607, 2.2198, 4.7829, 1.8669, 2.1987], device='cuda:3'), covar=tensor([0.0861, 0.1721, 0.0908, 0.0828, 0.1291, 0.0212, 0.1290, 0.1346], device='cuda:3'), in_proj_covar=tensor([0.0074, 0.0083, 0.0072, 0.0075, 0.0092, 0.0077, 0.0085, 0.0077], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2022-12-24 00:22:22,081 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.800e+02 4.317e+02 5.022e+02 6.376e+02 1.135e+03, threshold=1.004e+03, percent-clipped=6.0 2022-12-24 00:22:32,554 INFO [train.py:894] (3/4) Epoch 26, batch 2000, loss[loss=0.1536, simple_loss=0.2357, pruned_loss=0.03579, over 18643.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2624, pruned_loss=0.04687, over 3713133.01 frames. ], batch size: 45, lr: 4.39e-03, grad_scale: 16.0 2022-12-24 00:22:32,630 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0043-15874-0_sp0.9 from training. Duration: 20.07225 2022-12-24 00:22:39,773 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0 from training. Duration: 21.01 2022-12-24 00:23:13,303 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89680.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:23:13,889 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2022-12-24 00:23:18,033 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=89683.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:23:46,681 INFO [train.py:894] (3/4) Epoch 26, batch 2050, loss[loss=0.1912, simple_loss=0.2756, pruned_loss=0.05335, over 18670.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2632, pruned_loss=0.04742, over 3713385.88 frames. ], batch size: 62, lr: 4.39e-03, grad_scale: 16.0 2022-12-24 00:23:48,273 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0 from training. Duration: 20.65 2022-12-24 00:23:54,798 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0 from training. Duration: 21.46 2022-12-24 00:24:28,488 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89730.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:24:29,827 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=89731.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:24:40,504 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0 from training. Duration: 0.92 2022-12-24 00:24:45,230 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89741.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:24:47,858 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0_sp0.9 from training. Duration: 23.7666875 2022-12-24 00:24:52,050 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.904e+02 4.283e+02 5.573e+02 6.773e+02 1.480e+03, threshold=1.115e+03, percent-clipped=3.0 2022-12-24 00:25:02,385 INFO [train.py:894] (3/4) Epoch 26, batch 2100, loss[loss=0.2256, simple_loss=0.292, pruned_loss=0.07954, over 18583.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2634, pruned_loss=0.04844, over 3713275.72 frames. ], batch size: 181, lr: 4.39e-03, grad_scale: 16.0 2022-12-24 00:25:22,530 WARNING [train.py:1060] (3/4) Exclude cut with ID 8544-281189-0060-101339-0_sp0.9 from training. Duration: 20.861125 2022-12-24 00:25:31,692 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0_sp0.9 from training. Duration: 22.711125 2022-12-24 00:25:40,485 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2638, 1.7330, 2.4078, 4.3642, 3.3176, 2.8246, 1.0911, 3.4662], device='cuda:3'), covar=tensor([0.1616, 0.1488, 0.1401, 0.0501, 0.0780, 0.1120, 0.2042, 0.0699], device='cuda:3'), in_proj_covar=tensor([0.0104, 0.0119, 0.0138, 0.0156, 0.0107, 0.0146, 0.0131, 0.0116], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2022-12-24 00:25:45,650 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89781.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:26:06,393 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89795.0, num_to_drop=1, layers_to_drop={3} 2022-12-24 00:26:13,149 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp1.1 from training. Duration: 22.986375 2022-12-24 00:26:18,516 INFO [train.py:894] (3/4) Epoch 26, batch 2150, loss[loss=0.1707, simple_loss=0.2616, pruned_loss=0.03985, over 18457.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2634, pruned_loss=0.0488, over 3713630.85 frames. ], batch size: 54, lr: 4.39e-03, grad_scale: 16.0 2022-12-24 00:26:30,306 WARNING [train.py:1060] (3/4) Exclude cut with ID 8040-260924-0003-80960-0_sp0.9 from training. Duration: 22.07225 2022-12-24 00:26:34,842 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0045-26330-0_sp0.9 from training. Duration: 20.3055625 2022-12-24 00:26:37,687 WARNING [train.py:1060] (3/4) Exclude cut with ID 6356-271890-0060-94317-0_sp0.9 from training. Duration: 20.72225 2022-12-24 00:26:52,349 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=89825.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:26:56,639 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp1.1 from training. Duration: 22.4818125 2022-12-24 00:27:01,694 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2022-12-24 00:27:21,012 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp0.9 from training. Duration: 25.0944375 2022-12-24 00:27:24,469 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.879e+02 4.099e+02 4.650e+02 6.605e+02 1.804e+03, threshold=9.299e+02, percent-clipped=2.0 2022-12-24 00:27:24,563 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0 from training. Duration: 21.515 2022-12-24 00:27:31,951 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp0.9 from training. Duration: 27.02225 2022-12-24 00:27:35,067 INFO [train.py:894] (3/4) Epoch 26, batch 2200, loss[loss=0.144, simple_loss=0.2194, pruned_loss=0.03431, over 18631.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2627, pruned_loss=0.04821, over 3713294.35 frames. ], batch size: 41, lr: 4.39e-03, grad_scale: 16.0 2022-12-24 00:27:36,499 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0010-62480-0_sp0.9 from training. Duration: 22.22225 2022-12-24 00:27:43,701 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0085-44554-0_sp0.9 from training. Duration: 20.85 2022-12-24 00:27:52,485 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89865.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:27:52,699 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89865.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:28:05,389 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=89873.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:28:15,506 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0 from training. Duration: 21.54 2022-12-24 00:28:19,968 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp1.1 from training. Duration: 20.5318125 2022-12-24 00:28:29,796 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0012-134311-0_sp0.9 from training. Duration: 21.9333125 2022-12-24 00:28:41,318 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.9146, 3.3566, 3.4002, 3.8393, 3.5758, 3.3836, 4.0162, 1.1695], device='cuda:3'), covar=tensor([0.0779, 0.0827, 0.0742, 0.0764, 0.1421, 0.1215, 0.0640, 0.5335], device='cuda:3'), in_proj_covar=tensor([0.0357, 0.0236, 0.0246, 0.0283, 0.0335, 0.0277, 0.0300, 0.0293], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 00:28:44,548 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5918, 1.9623, 1.5537, 2.2192, 2.5097, 1.6628, 1.3663, 1.3613], device='cuda:3'), covar=tensor([0.1877, 0.1612, 0.1568, 0.0990, 0.1070, 0.1092, 0.2164, 0.1480], device='cuda:3'), in_proj_covar=tensor([0.0250, 0.0229, 0.0221, 0.0202, 0.0262, 0.0198, 0.0227, 0.0202], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 00:28:49,712 INFO [train.py:894] (3/4) Epoch 26, batch 2250, loss[loss=0.1903, simple_loss=0.2761, pruned_loss=0.05222, over 18697.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2624, pruned_loss=0.04808, over 3712994.78 frames. ], batch size: 60, lr: 4.39e-03, grad_scale: 16.0 2022-12-24 00:29:05,341 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5387, 1.2927, 0.9444, 0.3231, 1.0272, 1.4635, 1.2341, 1.2833], device='cuda:3'), covar=tensor([0.0748, 0.0727, 0.1145, 0.1696, 0.1199, 0.1720, 0.1933, 0.0840], device='cuda:3'), in_proj_covar=tensor([0.0177, 0.0190, 0.0209, 0.0191, 0.0212, 0.0205, 0.0219, 0.0205], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 00:29:19,365 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0025-130151-0_sp0.9 from training. Duration: 21.7944375 2022-12-24 00:29:25,214 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89926.0, num_to_drop=1, layers_to_drop={1} 2022-12-24 00:29:29,446 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0_sp0.9 from training. Duration: 22.4666875 2022-12-24 00:29:35,940 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0 from training. Duration: 21.635 2022-12-24 00:29:39,888 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0_sp0.9 from training. Duration: 24.038875 2022-12-24 00:29:54,791 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.010e+02 4.413e+02 5.147e+02 7.329e+02 2.051e+03, threshold=1.029e+03, percent-clipped=6.0 2022-12-24 00:30:06,070 INFO [train.py:894] (3/4) Epoch 26, batch 2300, loss[loss=0.194, simple_loss=0.2782, pruned_loss=0.05485, over 18661.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2623, pruned_loss=0.04849, over 3713021.35 frames. ], batch size: 60, lr: 4.39e-03, grad_scale: 16.0 2022-12-24 00:30:23,794 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp1.1 from training. Duration: 21.786375 2022-12-24 00:30:35,635 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0 from training. Duration: 20.22 2022-12-24 00:31:04,946 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.7054, 1.9425, 2.1700, 1.1128, 1.6000, 2.3463, 2.0539, 1.7042], device='cuda:3'), covar=tensor([0.0891, 0.0458, 0.0355, 0.0511, 0.0401, 0.0539, 0.0302, 0.0856], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0174, 0.0132, 0.0144, 0.0151, 0.0146, 0.0168, 0.0179], device='cuda:3'), out_proj_covar=tensor([1.1526e-04, 1.3152e-04, 9.7160e-05, 1.0570e-04, 1.1082e-04, 1.1047e-04, 1.2730e-04, 1.3510e-04], device='cuda:3') 2022-12-24 00:31:25,499 INFO [train.py:894] (3/4) Epoch 26, batch 2350, loss[loss=0.1642, simple_loss=0.2496, pruned_loss=0.03939, over 18457.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2609, pruned_loss=0.04746, over 3713433.69 frames. ], batch size: 50, lr: 4.38e-03, grad_scale: 16.0 2022-12-24 00:31:35,598 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7780, 1.1751, 0.7201, 1.3401, 2.3199, 1.0718, 1.5485, 1.6478], device='cuda:3'), covar=tensor([0.1497, 0.2087, 0.2064, 0.1355, 0.1498, 0.1681, 0.1336, 0.1642], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0096, 0.0115, 0.0095, 0.0118, 0.0091, 0.0096, 0.0092], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-24 00:31:37,544 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.65 vs. limit=5.0 2022-12-24 00:31:47,021 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8080, 1.5658, 1.5016, 1.4543, 1.9214, 1.9432, 2.0299, 1.3062], device='cuda:3'), covar=tensor([0.0329, 0.0400, 0.0518, 0.0300, 0.0242, 0.0497, 0.0277, 0.0431], device='cuda:3'), in_proj_covar=tensor([0.0096, 0.0129, 0.0154, 0.0124, 0.0119, 0.0123, 0.0101, 0.0130], device='cuda:3'), out_proj_covar=tensor([7.6260e-05, 1.0206e-04, 1.2615e-04, 9.7762e-05, 9.5305e-05, 9.3730e-05, 7.8694e-05, 1.0241e-04], device='cuda:3') 2022-12-24 00:32:04,124 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4289, 1.9840, 2.1530, 2.2408, 2.3998, 2.3670, 2.3158, 1.8697], device='cuda:3'), covar=tensor([0.2297, 0.3333, 0.2528, 0.2977, 0.2153, 0.1021, 0.3591, 0.1403], device='cuda:3'), in_proj_covar=tensor([0.0272, 0.0300, 0.0286, 0.0324, 0.0315, 0.0257, 0.0352, 0.0246], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 00:32:06,633 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90030.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:32:15,655 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90036.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:32:30,329 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.765e+02 4.074e+02 4.769e+02 5.976e+02 1.069e+03, threshold=9.538e+02, percent-clipped=1.0 2022-12-24 00:32:30,794 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2462, 2.0020, 2.1135, 1.7717, 2.5592, 2.4252, 2.2811, 1.7583], device='cuda:3'), covar=tensor([0.0427, 0.0482, 0.0491, 0.0645, 0.0302, 0.0383, 0.0428, 0.0931], device='cuda:3'), in_proj_covar=tensor([0.0126, 0.0132, 0.0132, 0.0121, 0.0104, 0.0128, 0.0135, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 00:32:41,316 INFO [train.py:894] (3/4) Epoch 26, batch 2400, loss[loss=0.1688, simple_loss=0.2598, pruned_loss=0.0389, over 18518.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2608, pruned_loss=0.04733, over 3712739.65 frames. ], batch size: 55, lr: 4.38e-03, grad_scale: 16.0 2022-12-24 00:32:43,461 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0 from training. Duration: 25.285 2022-12-24 00:33:19,947 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=90078.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:33:24,633 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90081.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:33:46,053 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90095.0, num_to_drop=1, layers_to_drop={1} 2022-12-24 00:33:47,262 WARNING [train.py:1060] (3/4) Exclude cut with ID 811-130148-0001-63453-0_sp0.9 from training. Duration: 20.861125 2022-12-24 00:33:58,447 INFO [train.py:894] (3/4) Epoch 26, batch 2450, loss[loss=0.1818, simple_loss=0.2682, pruned_loss=0.04767, over 18634.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2604, pruned_loss=0.0471, over 3713358.42 frames. ], batch size: 69, lr: 4.38e-03, grad_scale: 8.0 2022-12-24 00:34:10,242 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0 from training. Duration: 20.88 2022-12-24 00:34:10,739 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7295, 1.9759, 1.7216, 2.3130, 2.6429, 1.6915, 1.6846, 1.4383], device='cuda:3'), covar=tensor([0.1821, 0.1691, 0.1499, 0.0946, 0.1181, 0.1064, 0.2007, 0.1449], device='cuda:3'), in_proj_covar=tensor([0.0251, 0.0230, 0.0221, 0.0202, 0.0264, 0.0198, 0.0228, 0.0202], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 00:34:38,146 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=90129.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:34:40,014 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90130.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:34:41,388 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0_sp0.9 from training. Duration: 23.4166875 2022-12-24 00:34:49,156 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6216, 1.6440, 0.5317, 2.0344, 2.8572, 1.6413, 2.4159, 2.3647], device='cuda:3'), covar=tensor([0.1338, 0.1972, 0.2454, 0.1286, 0.1385, 0.1754, 0.1190, 0.1569], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0097, 0.0116, 0.0096, 0.0118, 0.0091, 0.0097, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-24 00:34:59,304 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=90143.0, num_to_drop=1, layers_to_drop={0} 2022-12-24 00:35:05,269 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.816e+02 4.059e+02 5.047e+02 6.343e+02 1.627e+03, threshold=1.009e+03, percent-clipped=6.0 2022-12-24 00:35:14,330 INFO [train.py:894] (3/4) Epoch 26, batch 2500, loss[loss=0.2258, simple_loss=0.2998, pruned_loss=0.07591, over 18664.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2607, pruned_loss=0.04726, over 3713230.03 frames. ], batch size: 170, lr: 4.38e-03, grad_scale: 8.0 2022-12-24 00:35:33,035 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90165.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:35:56,754 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0 from training. Duration: 21.24 2022-12-24 00:35:56,771 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0_sp0.9 from training. Duration: 23.9055625 2022-12-24 00:36:12,519 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90191.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:36:29,985 INFO [train.py:894] (3/4) Epoch 26, batch 2550, loss[loss=0.1382, simple_loss=0.2193, pruned_loss=0.02858, over 18525.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2611, pruned_loss=0.04741, over 3713209.80 frames. ], batch size: 44, lr: 4.38e-03, grad_scale: 8.0 2022-12-24 00:36:32,610 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp0.9 from training. Duration: 25.988875 2022-12-24 00:36:41,393 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0001-134300-0_sp0.9 from training. Duration: 20.67225 2022-12-24 00:36:44,213 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=90213.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:36:44,510 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6745, 1.6149, 1.3924, 1.6003, 1.7999, 1.7486, 1.7759, 1.2376], device='cuda:3'), covar=tensor([0.0295, 0.0250, 0.0477, 0.0198, 0.0193, 0.0389, 0.0256, 0.0333], device='cuda:3'), in_proj_covar=tensor([0.0095, 0.0127, 0.0152, 0.0122, 0.0117, 0.0122, 0.0100, 0.0128], device='cuda:3'), out_proj_covar=tensor([7.5271e-05, 1.0054e-04, 1.2471e-04, 9.6383e-05, 9.4074e-05, 9.2891e-05, 7.7483e-05, 1.0060e-04], device='cuda:3') 2022-12-24 00:36:56,045 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90221.0, num_to_drop=1, layers_to_drop={3} 2022-12-24 00:37:13,120 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90232.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:37:28,611 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0 from training. Duration: 20.34 2022-12-24 00:37:35,935 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.732e+02 4.133e+02 5.034e+02 6.191e+02 1.589e+03, threshold=1.007e+03, percent-clipped=2.0 2022-12-24 00:37:44,770 INFO [train.py:894] (3/4) Epoch 26, batch 2600, loss[loss=0.1893, simple_loss=0.2681, pruned_loss=0.05527, over 18519.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2621, pruned_loss=0.04795, over 3714287.95 frames. ], batch size: 58, lr: 4.38e-03, grad_scale: 8.0 2022-12-24 00:38:40,307 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp0.9 from training. Duration: 25.061125 2022-12-24 00:38:46,791 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90293.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:38:52,336 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0 from training. Duration: 0.83 2022-12-24 00:39:01,102 INFO [train.py:894] (3/4) Epoch 26, batch 2650, loss[loss=0.1742, simple_loss=0.2572, pruned_loss=0.0456, over 18506.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2617, pruned_loss=0.04794, over 3714755.47 frames. ], batch size: 52, lr: 4.38e-03, grad_scale: 8.0 2022-12-24 00:39:15,389 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4779, 1.9015, 2.0125, 2.0911, 2.2294, 2.4207, 2.1778, 2.0521], device='cuda:3'), covar=tensor([0.2324, 0.3627, 0.2804, 0.3426, 0.2492, 0.1113, 0.4060, 0.1436], device='cuda:3'), in_proj_covar=tensor([0.0274, 0.0301, 0.0287, 0.0325, 0.0316, 0.0259, 0.0354, 0.0248], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 00:39:18,986 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0 from training. Duration: 24.73 2022-12-24 00:39:26,169 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2022-12-24 00:39:33,211 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0 from training. Duration: 23.965 2022-12-24 00:39:40,686 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0030-146996-0_sp0.9 from training. Duration: 22.088875 2022-12-24 00:39:51,482 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90336.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:39:55,532 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0_sp0.9 from training. Duration: 23.6 2022-12-24 00:40:07,446 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.288e+02 4.377e+02 5.254e+02 6.585e+02 1.615e+03, threshold=1.051e+03, percent-clipped=2.0 2022-12-24 00:40:13,367 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90351.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:40:15,949 INFO [train.py:894] (3/4) Epoch 26, batch 2700, loss[loss=0.1824, simple_loss=0.2624, pruned_loss=0.05119, over 18442.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.262, pruned_loss=0.04791, over 3714887.38 frames. ], batch size: 48, lr: 4.38e-03, grad_scale: 8.0 2022-12-24 00:40:19,856 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2022-12-24 00:41:05,167 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=90384.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:41:24,443 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0629, 1.3748, 0.9068, 1.5645, 2.4009, 1.5179, 1.7769, 1.9736], device='cuda:3'), covar=tensor([0.1575, 0.2134, 0.2239, 0.1428, 0.1671, 0.1818, 0.1417, 0.1603], device='cuda:3'), in_proj_covar=tensor([0.0095, 0.0097, 0.0117, 0.0096, 0.0119, 0.0092, 0.0098, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-24 00:41:30,008 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2022-12-24 00:41:33,748 INFO [train.py:894] (3/4) Epoch 26, batch 2750, loss[loss=0.1874, simple_loss=0.2744, pruned_loss=0.05016, over 18589.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2615, pruned_loss=0.04769, over 3713787.72 frames. ], batch size: 57, lr: 4.38e-03, grad_scale: 8.0 2022-12-24 00:41:33,751 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0 from training. Duration: 23.795 2022-12-24 00:41:48,430 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90412.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:41:49,649 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp1.1 from training. Duration: 21.5409375 2022-12-24 00:41:52,748 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp0.9 from training. Duration: 24.97775 2022-12-24 00:41:59,883 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4708, 2.6099, 1.8665, 3.1807, 3.0162, 2.4402, 3.8119, 2.5377], device='cuda:3'), covar=tensor([0.0858, 0.1816, 0.2760, 0.1894, 0.1618, 0.0870, 0.0854, 0.1218], device='cuda:3'), in_proj_covar=tensor([0.0182, 0.0216, 0.0257, 0.0294, 0.0243, 0.0196, 0.0210, 0.0210], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 00:42:04,990 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0_sp0.9 from training. Duration: 23.3444375 2022-12-24 00:42:22,501 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2022-12-24 00:42:30,688 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0_sp0.9 from training. Duration: 23.2 2022-12-24 00:42:31,584 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2022-12-24 00:42:36,781 WARNING [train.py:1060] (3/4) Exclude cut with ID 5653-46179-0060-117930-0_sp0.9 from training. Duration: 21.17225 2022-12-24 00:42:41,036 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.721e+02 4.034e+02 5.147e+02 6.608e+02 1.665e+03, threshold=1.029e+03, percent-clipped=4.0 2022-12-24 00:42:50,352 INFO [train.py:894] (3/4) Epoch 26, batch 2800, loss[loss=0.1695, simple_loss=0.2593, pruned_loss=0.03981, over 18474.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2613, pruned_loss=0.04746, over 3714344.32 frames. ], batch size: 54, lr: 4.37e-03, grad_scale: 8.0 2022-12-24 00:42:58,254 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp0.9 from training. Duration: 24.6555625 2022-12-24 00:43:00,057 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90459.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:43:40,215 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90486.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:43:49,166 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0 from training. Duration: 20.44 2022-12-24 00:43:56,131 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90496.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:44:06,644 INFO [train.py:894] (3/4) Epoch 26, batch 2850, loss[loss=0.1605, simple_loss=0.2383, pruned_loss=0.0414, over 18691.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2617, pruned_loss=0.04784, over 3714552.22 frames. ], batch size: 46, lr: 4.37e-03, grad_scale: 8.0 2022-12-24 00:44:06,700 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0_sp0.9 from training. Duration: 23.45 2022-12-24 00:44:32,433 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90520.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:44:33,770 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90521.0, num_to_drop=1, layers_to_drop={2} 2022-12-24 00:44:38,029 WARNING [train.py:1060] (3/4) Exclude cut with ID 6945-60535-0076-12784-0_sp0.9 from training. Duration: 20.52225 2022-12-24 00:44:46,657 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0 from training. Duration: 22.19 2022-12-24 00:44:58,370 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp1.1 from training. Duration: 25.3818125 2022-12-24 00:45:11,641 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.952e+02 4.260e+02 4.866e+02 6.366e+02 1.571e+03, threshold=9.732e+02, percent-clipped=4.0 2022-12-24 00:45:15,257 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp0.9 from training. Duration: 28.0944375 2022-12-24 00:45:21,394 INFO [train.py:894] (3/4) Epoch 26, batch 2900, loss[loss=0.2045, simple_loss=0.2841, pruned_loss=0.0625, over 18581.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2604, pruned_loss=0.04707, over 3713943.84 frames. ], batch size: 57, lr: 4.37e-03, grad_scale: 8.0 2022-12-24 00:45:22,934 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0_sp0.9 from training. Duration: 22.9444375 2022-12-24 00:45:27,746 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90557.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:45:30,038 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp1.1 from training. Duration: 21.6318125 2022-12-24 00:45:40,878 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.53 vs. limit=2.0 2022-12-24 00:45:46,092 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=90569.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:45:47,390 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0 from training. Duration: 23.695 2022-12-24 00:46:12,765 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0 from training. Duration: 23.955 2022-12-24 00:46:14,455 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90588.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:46:38,264 INFO [train.py:894] (3/4) Epoch 26, batch 2950, loss[loss=0.1886, simple_loss=0.2685, pruned_loss=0.05435, over 18664.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2596, pruned_loss=0.04652, over 3712584.43 frames. ], batch size: 60, lr: 4.37e-03, grad_scale: 8.0 2022-12-24 00:46:48,237 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp0.9 from training. Duration: 26.438875 2022-12-24 00:47:27,913 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0021-26306-0_sp0.9 from training. Duration: 21.2444375 2022-12-24 00:47:27,936 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp0.9 from training. Duration: 31.02225 2022-12-24 00:47:40,104 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0 from training. Duration: 22.395 2022-12-24 00:47:46,245 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.852e+02 3.883e+02 4.957e+02 5.801e+02 1.568e+03, threshold=9.915e+02, percent-clipped=5.0 2022-12-24 00:47:55,787 INFO [train.py:894] (3/4) Epoch 26, batch 3000, loss[loss=0.1763, simple_loss=0.2684, pruned_loss=0.04211, over 18539.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2598, pruned_loss=0.04671, over 3712544.57 frames. ], batch size: 55, lr: 4.37e-03, grad_scale: 8.0 2022-12-24 00:47:55,787 INFO [train.py:919] (3/4) Computing validation loss 2022-12-24 00:48:06,743 INFO [train.py:928] (3/4) Epoch 26, validation: loss=0.1627, simple_loss=0.2592, pruned_loss=0.03305, over 944034.00 frames. 2022-12-24 00:48:06,744 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-24 00:48:08,377 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0 from training. Duration: 21.075 2022-12-24 00:48:14,140 WARNING [train.py:1060] (3/4) Exclude cut with ID 6482-98857-0025-147532-0_sp0.9 from training. Duration: 20.0055625 2022-12-24 00:48:15,702 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0037-132304-0_sp0.9 from training. Duration: 22.05 2022-12-24 00:48:15,719 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0 from training. Duration: 26.8349375 2022-12-24 00:48:18,521 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp1.1 from training. Duration: 22.1090625 2022-12-24 00:48:25,637 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp0.9 from training. Duration: 26.6166875 2022-12-24 00:48:44,451 WARNING [train.py:1060] (3/4) Exclude cut with ID 2046-178027-0000-53705-0_sp0.9 from training. Duration: 20.3055625 2022-12-24 00:49:07,274 WARNING [train.py:1060] (3/4) Exclude cut with ID 7205-50138-0008-5373-0_sp0.9 from training. Duration: 20.7 2022-12-24 00:49:22,265 INFO [train.py:894] (3/4) Epoch 26, batch 3050, loss[loss=0.2058, simple_loss=0.2773, pruned_loss=0.06714, over 18528.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2596, pruned_loss=0.04664, over 3713149.08 frames. ], batch size: 98, lr: 4.37e-03, grad_scale: 8.0 2022-12-24 00:49:28,502 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90707.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:49:50,751 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0 from training. Duration: 22.48 2022-12-24 00:50:05,803 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp0.9 from training. Duration: 29.816625 2022-12-24 00:50:26,022 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp1.1 from training. Duration: 22.7590625 2022-12-24 00:50:28,928 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.601e+02 4.164e+02 4.877e+02 6.267e+02 1.480e+03, threshold=9.754e+02, percent-clipped=2.0 2022-12-24 00:50:32,189 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0 from training. Duration: 22.555 2022-12-24 00:50:37,991 INFO [train.py:894] (3/4) Epoch 26, batch 3100, loss[loss=0.1435, simple_loss=0.2216, pruned_loss=0.03276, over 18424.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.26, pruned_loss=0.04668, over 3713728.86 frames. ], batch size: 42, lr: 4.37e-03, grad_scale: 8.0 2022-12-24 00:50:38,957 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2022-12-24 00:50:52,882 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0005-25975-0_sp0.9 from training. Duration: 21.688875 2022-12-24 00:51:08,010 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0646, 2.0550, 2.2318, 2.1040, 1.9353, 3.4862, 2.0796, 2.5148], device='cuda:3'), covar=tensor([0.2611, 0.1732, 0.1511, 0.1691, 0.1121, 0.0247, 0.1608, 0.0736], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0116, 0.0125, 0.0121, 0.0105, 0.0097, 0.0090, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-24 00:51:28,627 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0_sp0.9 from training. Duration: 22.6 2022-12-24 00:51:28,947 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90786.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:51:47,291 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2022-12-24 00:51:55,243 INFO [train.py:894] (3/4) Epoch 26, batch 3150, loss[loss=0.1732, simple_loss=0.2636, pruned_loss=0.04141, over 18703.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.26, pruned_loss=0.04662, over 3712957.22 frames. ], batch size: 62, lr: 4.37e-03, grad_scale: 8.0 2022-12-24 00:52:05,863 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0 from training. Duration: 24.32 2022-12-24 00:52:13,236 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90815.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:52:42,601 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=90834.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:53:01,674 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.469e+02 4.405e+02 5.375e+02 6.469e+02 1.692e+03, threshold=1.075e+03, percent-clipped=5.0 2022-12-24 00:53:03,290 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-276745-0093-13116-0_sp0.9 from training. Duration: 21.061125 2022-12-24 00:53:05,136 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90849.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:53:09,159 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90852.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:53:10,234 INFO [train.py:894] (3/4) Epoch 26, batch 3200, loss[loss=0.1964, simple_loss=0.274, pruned_loss=0.05941, over 18464.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2596, pruned_loss=0.04616, over 3712417.84 frames. ], batch size: 54, lr: 4.36e-03, grad_scale: 8.0 2022-12-24 00:53:10,677 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7624, 1.8373, 2.0032, 1.2555, 2.0458, 2.1409, 1.5627, 2.4517], device='cuda:3'), covar=tensor([0.1155, 0.1799, 0.1188, 0.1707, 0.0681, 0.1000, 0.2206, 0.0525], device='cuda:3'), in_proj_covar=tensor([0.0199, 0.0216, 0.0209, 0.0195, 0.0173, 0.0219, 0.0217, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 00:53:19,166 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0024-15855-0_sp0.9 from training. Duration: 20.32225 2022-12-24 00:53:32,748 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp1.1 from training. Duration: 0.7545625 2022-12-24 00:53:48,815 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0_sp0.9 from training. Duration: 23.9333125 2022-12-24 00:54:01,321 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5683, 1.8638, 1.5390, 2.1314, 2.2595, 1.6046, 1.3462, 1.3136], device='cuda:3'), covar=tensor([0.1873, 0.1749, 0.1599, 0.0982, 0.1256, 0.1093, 0.2202, 0.1538], device='cuda:3'), in_proj_covar=tensor([0.0253, 0.0232, 0.0222, 0.0204, 0.0265, 0.0199, 0.0228, 0.0204], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 00:54:04,081 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90888.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:54:20,507 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp1.1 from training. Duration: 20.17275 2022-12-24 00:54:24,729 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp1.1 from training. Duration: 20.436375 2022-12-24 00:54:26,602 INFO [train.py:894] (3/4) Epoch 26, batch 3250, loss[loss=0.1712, simple_loss=0.2597, pruned_loss=0.04139, over 18387.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2592, pruned_loss=0.04583, over 3712567.75 frames. ], batch size: 53, lr: 4.36e-03, grad_scale: 8.0 2022-12-24 00:54:38,448 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90910.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:54:58,314 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0413, 2.3292, 1.9460, 2.5788, 3.1407, 1.9716, 2.1028, 1.5746], device='cuda:3'), covar=tensor([0.1712, 0.1626, 0.1411, 0.0943, 0.1220, 0.0989, 0.1792, 0.1459], device='cuda:3'), in_proj_covar=tensor([0.0253, 0.0232, 0.0222, 0.0204, 0.0265, 0.0199, 0.0228, 0.0204], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 00:55:16,729 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=90936.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:55:32,540 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.717e+02 4.387e+02 5.223e+02 6.052e+02 1.643e+03, threshold=1.045e+03, percent-clipped=3.0 2022-12-24 00:55:42,434 INFO [train.py:894] (3/4) Epoch 26, batch 3300, loss[loss=0.1975, simple_loss=0.2905, pruned_loss=0.05229, over 18545.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2586, pruned_loss=0.04545, over 3713091.92 frames. ], batch size: 78, lr: 4.36e-03, grad_scale: 8.0 2022-12-24 00:55:43,997 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0_sp0.9 from training. Duration: 23.1055625 2022-12-24 00:55:45,526 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp1.1 from training. Duration: 23.5 2022-12-24 00:55:56,368 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp0.9 from training. Duration: 26.62775 2022-12-24 00:56:09,703 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0 from training. Duration: 21.105 2022-12-24 00:56:14,071 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0_sp0.9 from training. Duration: 24.411125 2022-12-24 00:56:43,107 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp1.1 from training. Duration: 21.263625 2022-12-24 00:57:00,716 INFO [train.py:894] (3/4) Epoch 26, batch 3350, loss[loss=0.1518, simple_loss=0.2313, pruned_loss=0.03618, over 18587.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.258, pruned_loss=0.04535, over 3712770.86 frames. ], batch size: 45, lr: 4.36e-03, grad_scale: 8.0 2022-12-24 00:57:07,522 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91007.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:57:17,137 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0 from training. Duration: 20.795 2022-12-24 00:57:25,867 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0 from training. Duration: 24.76 2022-12-24 00:57:27,412 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0_sp0.9 from training. Duration: 22.25 2022-12-24 00:57:52,463 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6044, 3.6670, 3.5027, 1.6627, 3.8281, 2.8627, 0.6929, 2.4456], device='cuda:3'), covar=tensor([0.1993, 0.1208, 0.1549, 0.3220, 0.0909, 0.0881, 0.4654, 0.1411], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0148, 0.0159, 0.0125, 0.0151, 0.0116, 0.0144, 0.0115], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-24 00:57:53,692 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp1.1 from training. Duration: 20.5045625 2022-12-24 00:58:07,449 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.659e+02 3.619e+02 4.459e+02 5.984e+02 1.100e+03, threshold=8.918e+02, percent-clipped=1.0 2022-12-24 00:58:12,937 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.0758, 5.4019, 4.8779, 3.2576, 5.5390, 4.1850, 0.5230, 3.7020], device='cuda:3'), covar=tensor([0.1731, 0.1033, 0.1251, 0.2309, 0.0585, 0.0662, 0.5086, 0.1125], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0148, 0.0159, 0.0125, 0.0152, 0.0116, 0.0145, 0.0115], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-24 00:58:17,684 INFO [train.py:894] (3/4) Epoch 26, batch 3400, loss[loss=0.1299, simple_loss=0.2095, pruned_loss=0.02518, over 18515.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2579, pruned_loss=0.04507, over 3712688.58 frames. ], batch size: 44, lr: 4.36e-03, grad_scale: 8.0 2022-12-24 00:58:20,586 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=91055.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:59:29,994 INFO [train.py:894] (3/4) Epoch 26, batch 3450, loss[loss=0.1581, simple_loss=0.2336, pruned_loss=0.04131, over 18640.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2595, pruned_loss=0.04611, over 3714043.46 frames. ], batch size: 41, lr: 4.36e-03, grad_scale: 8.0 2022-12-24 00:59:47,491 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91115.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 00:59:49,135 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4384, 1.3403, 1.4256, 1.3092, 1.0500, 2.1950, 0.9492, 1.3509], device='cuda:3'), covar=tensor([0.3084, 0.2138, 0.2050, 0.2273, 0.1399, 0.0375, 0.1761, 0.0930], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0116, 0.0125, 0.0122, 0.0106, 0.0097, 0.0090, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-24 01:00:33,641 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.302e+02 4.218e+02 5.018e+02 6.316e+02 1.663e+03, threshold=1.004e+03, percent-clipped=6.0 2022-12-24 01:00:41,900 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91152.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:00:43,280 INFO [train.py:894] (3/4) Epoch 26, batch 3500, loss[loss=0.1929, simple_loss=0.2749, pruned_loss=0.05542, over 18636.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2597, pruned_loss=0.04634, over 3714051.57 frames. ], batch size: 97, lr: 4.36e-03, grad_scale: 8.0 2022-12-24 01:01:05,840 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp1.1 from training. Duration: 22.2954375 2022-12-24 01:01:14,977 INFO [train.py:894] (3/4) Epoch 27, batch 0, loss[loss=0.1692, simple_loss=0.2529, pruned_loss=0.04274, over 18665.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2529, pruned_loss=0.04274, over 18665.00 frames. ], batch size: 48, lr: 4.27e-03, grad_scale: 8.0 2022-12-24 01:01:14,977 INFO [train.py:919] (3/4) Computing validation loss 2022-12-24 01:01:19,698 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6000, 1.6413, 1.7376, 1.1300, 1.8522, 1.7950, 1.3994, 2.1489], device='cuda:3'), covar=tensor([0.1197, 0.2108, 0.1432, 0.1955, 0.0730, 0.1159, 0.2749, 0.0597], device='cuda:3'), in_proj_covar=tensor([0.0199, 0.0217, 0.0209, 0.0195, 0.0173, 0.0219, 0.0218, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 01:01:21,526 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.0452, 0.7576, 0.5064, 0.7916, 1.3750, 0.4390, 0.8226, 0.8838], device='cuda:3'), covar=tensor([0.1340, 0.1659, 0.1531, 0.1180, 0.1403, 0.1435, 0.1200, 0.1554], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0097, 0.0116, 0.0096, 0.0120, 0.0092, 0.0098, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-24 01:01:25,754 INFO [train.py:928] (3/4) Epoch 27, validation: loss=0.1651, simple_loss=0.2615, pruned_loss=0.03432, over 944034.00 frames. 2022-12-24 01:01:25,755 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-24 01:01:30,187 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=91163.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:01:32,418 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2022-12-24 01:02:14,844 WARNING [train.py:1060] (3/4) Exclude cut with ID 298-126791-0067-24026-0_sp0.9 from training. Duration: 21.438875 2022-12-24 01:02:19,562 WARNING [train.py:1060] (3/4) Exclude cut with ID 5652-39938-0025-23684-0_sp0.9 from training. Duration: 22.2055625 2022-12-24 01:02:26,711 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=91200.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:02:34,684 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91205.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:02:41,927 INFO [train.py:894] (3/4) Epoch 27, batch 50, loss[loss=0.1648, simple_loss=0.2541, pruned_loss=0.03777, over 18395.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2595, pruned_loss=0.04067, over 837767.65 frames. ], batch size: 46, lr: 4.27e-03, grad_scale: 8.0 2022-12-24 01:02:52,012 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2022-12-24 01:03:17,125 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2828, 1.7780, 0.6712, 2.0213, 2.6262, 1.8327, 2.1630, 2.3273], device='cuda:3'), covar=tensor([0.1520, 0.1930, 0.2458, 0.1374, 0.1623, 0.1676, 0.1302, 0.1687], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0097, 0.0116, 0.0096, 0.0119, 0.0092, 0.0098, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-24 01:03:38,310 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.308e+02 3.518e+02 4.263e+02 5.371e+02 1.025e+03, threshold=8.526e+02, percent-clipped=1.0 2022-12-24 01:03:57,065 INFO [train.py:894] (3/4) Epoch 27, batch 100, loss[loss=0.158, simple_loss=0.2566, pruned_loss=0.02965, over 18503.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2593, pruned_loss=0.04074, over 1474884.02 frames. ], batch size: 52, lr: 4.27e-03, grad_scale: 8.0 2022-12-24 01:04:35,490 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91285.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:05:11,680 INFO [train.py:894] (3/4) Epoch 27, batch 150, loss[loss=0.1849, simple_loss=0.2842, pruned_loss=0.04283, over 18660.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2578, pruned_loss=0.03947, over 1971561.54 frames. ], batch size: 62, lr: 4.27e-03, grad_scale: 8.0 2022-12-24 01:05:15,485 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2022-12-24 01:05:21,793 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0 from training. Duration: 24.525 2022-12-24 01:05:53,849 WARNING [train.py:1060] (3/4) Exclude cut with ID 3699-47246-0007-3408-0_sp0.9 from training. Duration: 20.26675 2022-12-24 01:06:02,763 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2022-12-24 01:06:06,755 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91346.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:06:07,770 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.293e+02 3.105e+02 3.683e+02 4.423e+02 8.436e+02, threshold=7.367e+02, percent-clipped=0.0 2022-12-24 01:06:07,787 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp0.9 from training. Duration: 27.25 2022-12-24 01:06:27,652 INFO [train.py:894] (3/4) Epoch 27, batch 200, loss[loss=0.1741, simple_loss=0.2639, pruned_loss=0.04219, over 18675.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2578, pruned_loss=0.0394, over 2358502.75 frames. ], batch size: 62, lr: 4.27e-03, grad_scale: 8.0 2022-12-24 01:07:19,972 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0 from training. Duration: 21.68 2022-12-24 01:07:31,716 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0 from training. Duration: 21.6300625 2022-12-24 01:07:35,252 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2022-12-24 01:07:41,614 INFO [train.py:894] (3/4) Epoch 27, batch 250, loss[loss=0.1506, simple_loss=0.2407, pruned_loss=0.03028, over 18571.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2563, pruned_loss=0.0387, over 2659604.12 frames. ], batch size: 49, lr: 4.27e-03, grad_scale: 8.0 2022-12-24 01:07:42,066 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8228, 2.1128, 2.2354, 1.3160, 2.2890, 2.4311, 1.7949, 2.7958], device='cuda:3'), covar=tensor([0.1411, 0.1824, 0.1492, 0.2133, 0.0773, 0.1160, 0.2351, 0.0617], device='cuda:3'), in_proj_covar=tensor([0.0198, 0.0216, 0.0208, 0.0194, 0.0172, 0.0218, 0.0217, 0.0192], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 01:07:56,721 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0_sp0.9 from training. Duration: 24.033375 2022-12-24 01:08:35,872 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.808e+02 3.295e+02 4.086e+02 4.831e+02 1.349e+03, threshold=8.172e+02, percent-clipped=3.0 2022-12-24 01:08:54,007 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0 from training. Duration: 22.905 2022-12-24 01:08:55,348 INFO [train.py:894] (3/4) Epoch 27, batch 300, loss[loss=0.1556, simple_loss=0.2425, pruned_loss=0.0344, over 18424.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2541, pruned_loss=0.03818, over 2892484.01 frames. ], batch size: 48, lr: 4.27e-03, grad_scale: 8.0 2022-12-24 01:08:55,439 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp1.1 from training. Duration: 23.4318125 2022-12-24 01:10:04,556 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91505.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:10:10,322 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2015, 1.9414, 1.4476, 0.6238, 1.5469, 1.8091, 1.3923, 1.8767], device='cuda:3'), covar=tensor([0.0660, 0.0655, 0.1162, 0.1614, 0.1106, 0.1559, 0.1901, 0.0637], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0188, 0.0209, 0.0191, 0.0210, 0.0205, 0.0219, 0.0205], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 01:10:11,308 INFO [train.py:894] (3/4) Epoch 27, batch 350, loss[loss=0.1861, simple_loss=0.2729, pruned_loss=0.04959, over 18422.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2537, pruned_loss=0.03791, over 3074754.77 frames. ], batch size: 48, lr: 4.27e-03, grad_scale: 8.0 2022-12-24 01:10:55,840 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp1.1 from training. Duration: 20.82275 2022-12-24 01:10:57,268 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp0.9 from training. Duration: 25.45 2022-12-24 01:11:00,289 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91542.0, num_to_drop=1, layers_to_drop={1} 2022-12-24 01:11:07,145 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.079e+02 3.172e+02 3.767e+02 4.627e+02 1.022e+03, threshold=7.535e+02, percent-clipped=2.0 2022-12-24 01:11:16,316 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=91553.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:11:25,878 INFO [train.py:894] (3/4) Epoch 27, batch 400, loss[loss=0.1887, simple_loss=0.2761, pruned_loss=0.05059, over 18554.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2546, pruned_loss=0.03821, over 3216359.96 frames. ], batch size: 96, lr: 4.26e-03, grad_scale: 8.0 2022-12-24 01:11:59,285 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0 from training. Duration: 25.775 2022-12-24 01:12:19,939 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0_sp0.9 from training. Duration: 22.25 2022-12-24 01:12:31,423 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91603.0, num_to_drop=1, layers_to_drop={3} 2022-12-24 01:12:31,495 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7886, 1.6953, 1.4242, 1.6488, 1.9249, 1.6741, 2.1377, 1.8825], device='cuda:3'), covar=tensor([0.0908, 0.1798, 0.2868, 0.1687, 0.1943, 0.0947, 0.1053, 0.1332], device='cuda:3'), in_proj_covar=tensor([0.0183, 0.0216, 0.0257, 0.0293, 0.0243, 0.0195, 0.0208, 0.0210], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 01:12:41,284 INFO [train.py:894] (3/4) Epoch 27, batch 450, loss[loss=0.1742, simple_loss=0.2682, pruned_loss=0.04006, over 18655.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2557, pruned_loss=0.03865, over 3326330.97 frames. ], batch size: 69, lr: 4.26e-03, grad_scale: 8.0 2022-12-24 01:12:45,261 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0 from training. Duration: 26.205 2022-12-24 01:13:02,049 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp0.9 from training. Duration: 30.1555625 2022-12-24 01:13:07,834 WARNING [train.py:1060] (3/4) Exclude cut with ID 1265-135635-0050-6781-0_sp0.9 from training. Duration: 21.8333125 2022-12-24 01:13:17,111 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp1.1 from training. Duration: 20.6545625 2022-12-24 01:13:22,147 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6893, 2.0927, 1.8201, 2.4614, 2.8848, 1.7187, 1.6089, 1.3950], device='cuda:3'), covar=tensor([0.1915, 0.1692, 0.1518, 0.0976, 0.1096, 0.1061, 0.2077, 0.1607], device='cuda:3'), in_proj_covar=tensor([0.0250, 0.0230, 0.0221, 0.0203, 0.0264, 0.0198, 0.0228, 0.0203], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 01:13:27,193 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91641.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:13:36,642 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.293e+02 3.165e+02 3.798e+02 4.275e+02 8.978e+02, threshold=7.596e+02, percent-clipped=3.0 2022-12-24 01:13:55,161 INFO [train.py:894] (3/4) Epoch 27, batch 500, loss[loss=0.1923, simple_loss=0.2788, pruned_loss=0.0529, over 18673.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2576, pruned_loss=0.0393, over 3412474.37 frames. ], batch size: 62, lr: 4.26e-03, grad_scale: 8.0 2022-12-24 01:13:59,292 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0045-39920-0_sp0.9 from training. Duration: 20.52225 2022-12-24 01:14:20,058 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp0.9 from training. Duration: 29.1166875 2022-12-24 01:14:28,616 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2022-12-24 01:15:09,274 INFO [train.py:894] (3/4) Epoch 27, batch 550, loss[loss=0.1636, simple_loss=0.2611, pruned_loss=0.03303, over 18376.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2574, pruned_loss=0.03896, over 3479251.46 frames. ], batch size: 51, lr: 4.26e-03, grad_scale: 8.0 2022-12-24 01:15:22,477 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133211-0007-59831-0_sp0.9 from training. Duration: 21.388875 2022-12-24 01:15:59,499 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0 from training. Duration: 22.72 2022-12-24 01:16:00,913 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0_sp0.9 from training. Duration: 22.7444375 2022-12-24 01:16:03,725 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.126e+02 3.350e+02 3.947e+02 4.704e+02 8.228e+02, threshold=7.893e+02, percent-clipped=1.0 2022-12-24 01:16:20,777 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2372, 2.1864, 1.6383, 2.7150, 2.5193, 2.1360, 3.0509, 2.2377], device='cuda:3'), covar=tensor([0.0833, 0.1719, 0.2713, 0.1508, 0.1588, 0.0862, 0.0831, 0.1231], device='cuda:3'), in_proj_covar=tensor([0.0182, 0.0215, 0.0254, 0.0291, 0.0241, 0.0193, 0.0207, 0.0209], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 01:16:23,305 INFO [train.py:894] (3/4) Epoch 27, batch 600, loss[loss=0.1822, simple_loss=0.2713, pruned_loss=0.04651, over 18572.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2576, pruned_loss=0.03905, over 3531961.80 frames. ], batch size: 57, lr: 4.26e-03, grad_scale: 8.0 2022-12-24 01:16:44,460 WARNING [train.py:1060] (3/4) Exclude cut with ID 4133-6541-0027-40495-0_sp1.1 from training. Duration: 0.9681875 2022-12-24 01:16:47,256 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0_sp0.9 from training. Duration: 22.3166875 2022-12-24 01:16:52,453 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133212-0015-59917-0_sp0.9 from training. Duration: 21.8166875 2022-12-24 01:17:39,239 INFO [train.py:894] (3/4) Epoch 27, batch 650, loss[loss=0.1736, simple_loss=0.2633, pruned_loss=0.04199, over 18483.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2577, pruned_loss=0.03921, over 3571603.30 frames. ], batch size: 54, lr: 4.26e-03, grad_scale: 8.0 2022-12-24 01:18:34,279 WARNING [train.py:1060] (3/4) Exclude cut with ID 4957-30119-0041-23990-0_sp0.9 from training. Duration: 20.22775 2022-12-24 01:18:35,749 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.062e+02 3.215e+02 3.943e+02 4.935e+02 9.655e+02, threshold=7.887e+02, percent-clipped=2.0 2022-12-24 01:18:55,720 INFO [train.py:894] (3/4) Epoch 27, batch 700, loss[loss=0.1535, simple_loss=0.2446, pruned_loss=0.03116, over 18464.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2587, pruned_loss=0.03961, over 3603632.43 frames. ], batch size: 50, lr: 4.26e-03, grad_scale: 8.0 2022-12-24 01:19:10,972 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2022-12-24 01:19:15,937 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp1.1 from training. Duration: 24.67275 2022-12-24 01:19:37,193 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4384, 2.1622, 1.6691, 2.2656, 1.9508, 2.0944, 1.9901, 2.4340], device='cuda:3'), covar=tensor([0.2261, 0.3281, 0.2035, 0.2660, 0.3455, 0.1097, 0.3046, 0.1009], device='cuda:3'), in_proj_covar=tensor([0.0302, 0.0299, 0.0254, 0.0349, 0.0281, 0.0236, 0.0298, 0.0222], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 01:19:41,936 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91891.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:19:42,980 WARNING [train.py:1060] (3/4) Exclude cut with ID 3082-165428-0081-50734-0_sp0.9 from training. Duration: 21.8055625 2022-12-24 01:19:51,796 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91898.0, num_to_drop=1, layers_to_drop={1} 2022-12-24 01:19:59,502 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8545, 2.2115, 2.0868, 2.7499, 2.5721, 1.8380, 1.7616, 1.5425], device='cuda:3'), covar=tensor([0.1675, 0.1538, 0.1244, 0.0811, 0.1162, 0.1013, 0.1964, 0.1443], device='cuda:3'), in_proj_covar=tensor([0.0250, 0.0230, 0.0221, 0.0203, 0.0264, 0.0198, 0.0228, 0.0204], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 01:20:09,750 INFO [train.py:894] (3/4) Epoch 27, batch 750, loss[loss=0.1836, simple_loss=0.2774, pruned_loss=0.04487, over 18505.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2588, pruned_loss=0.03966, over 3627776.30 frames. ], batch size: 52, lr: 4.26e-03, grad_scale: 8.0 2022-12-24 01:20:19,989 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0_sp0.9 from training. Duration: 22.6666875 2022-12-24 01:20:57,094 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91941.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:21:05,129 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.943e+02 3.317e+02 4.158e+02 4.971e+02 2.653e+03, threshold=8.315e+02, percent-clipped=1.0 2022-12-24 01:21:13,052 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91952.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:21:14,375 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4595, 1.1847, 0.8131, 1.2035, 1.7284, 1.0532, 1.2148, 1.3495], device='cuda:3'), covar=tensor([0.1297, 0.1783, 0.1752, 0.1261, 0.1517, 0.1676, 0.1306, 0.1433], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0096, 0.0115, 0.0095, 0.0119, 0.0092, 0.0098, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-24 01:21:17,948 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91955.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:21:20,523 WARNING [train.py:1060] (3/4) Exclude cut with ID 2411-132532-0017-82279-0_sp1.1 from training. Duration: 0.9681875 2022-12-24 01:21:24,662 INFO [train.py:894] (3/4) Epoch 27, batch 800, loss[loss=0.1538, simple_loss=0.2392, pruned_loss=0.03426, over 18398.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2599, pruned_loss=0.04023, over 3646950.75 frames. ], batch size: 46, lr: 4.26e-03, grad_scale: 8.0 2022-12-24 01:21:42,767 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0 from training. Duration: 22.485 2022-12-24 01:21:57,232 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91981.0, num_to_drop=1, layers_to_drop={1} 2022-12-24 01:22:09,394 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=91989.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:22:19,889 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp1.1 from training. Duration: 23.82275 2022-12-24 01:22:35,475 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0 from training. Duration: 20.77 2022-12-24 01:22:43,246 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4972, 1.3799, 1.4247, 1.3480, 0.9402, 2.3552, 0.9219, 1.3812], device='cuda:3'), covar=tensor([0.3328, 0.2303, 0.2182, 0.2252, 0.1587, 0.0311, 0.1723, 0.0922], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0119, 0.0126, 0.0123, 0.0107, 0.0097, 0.0091, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-24 01:22:44,322 INFO [train.py:894] (3/4) Epoch 27, batch 850, loss[loss=0.1783, simple_loss=0.2745, pruned_loss=0.04099, over 18548.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2596, pruned_loss=0.04015, over 3661272.79 frames. ], batch size: 55, lr: 4.25e-03, grad_scale: 8.0 2022-12-24 01:22:44,366 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0_sp0.9 from training. Duration: 24.088875 2022-12-24 01:22:50,502 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4382, 1.8426, 2.1248, 2.1040, 2.4608, 2.5418, 2.3653, 2.0741], device='cuda:3'), covar=tensor([0.2394, 0.3616, 0.2649, 0.3196, 0.2174, 0.0986, 0.3946, 0.1369], device='cuda:3'), in_proj_covar=tensor([0.0271, 0.0299, 0.0286, 0.0325, 0.0315, 0.0257, 0.0351, 0.0247], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 01:22:53,392 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92016.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:23:15,959 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp1.1 from training. Duration: 20.4409375 2022-12-24 01:23:32,572 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92042.0, num_to_drop=1, layers_to_drop={1} 2022-12-24 01:23:40,047 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.966e+02 3.267e+02 4.027e+02 4.878e+02 1.049e+03, threshold=8.054e+02, percent-clipped=5.0 2022-12-24 01:23:51,068 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2022-12-24 01:23:58,929 INFO [train.py:894] (3/4) Epoch 27, batch 900, loss[loss=0.1743, simple_loss=0.2666, pruned_loss=0.04101, over 18675.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2585, pruned_loss=0.03993, over 3672633.38 frames. ], batch size: 60, lr: 4.25e-03, grad_scale: 8.0 2022-12-24 01:24:23,296 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2022-12-24 01:24:32,578 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0_sp0.9 from training. Duration: 22.511125 2022-12-24 01:24:32,606 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0 from training. Duration: 20.675 2022-12-24 01:24:41,195 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1632, 2.8448, 2.7350, 1.2417, 2.9436, 2.1828, 0.5235, 1.7998], device='cuda:3'), covar=tensor([0.2542, 0.1519, 0.1872, 0.3691, 0.1183, 0.1129, 0.4967, 0.1771], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0147, 0.0160, 0.0124, 0.0151, 0.0116, 0.0144, 0.0115], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-24 01:25:09,330 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6920, 1.6658, 1.4653, 1.5732, 1.9478, 1.8681, 1.9037, 1.3193], device='cuda:3'), covar=tensor([0.0321, 0.0259, 0.0520, 0.0220, 0.0190, 0.0401, 0.0263, 0.0338], device='cuda:3'), in_proj_covar=tensor([0.0095, 0.0129, 0.0155, 0.0123, 0.0118, 0.0123, 0.0101, 0.0129], device='cuda:3'), out_proj_covar=tensor([7.5475e-05, 1.0145e-04, 1.2667e-04, 9.7036e-05, 9.4168e-05, 9.4031e-05, 7.8501e-05, 1.0151e-04], device='cuda:3') 2022-12-24 01:25:11,521 INFO [train.py:894] (3/4) Epoch 27, batch 950, loss[loss=0.1795, simple_loss=0.2712, pruned_loss=0.04394, over 18680.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2581, pruned_loss=0.03978, over 3681409.48 frames. ], batch size: 69, lr: 4.25e-03, grad_scale: 16.0 2022-12-24 01:26:09,184 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.451e+02 3.241e+02 3.825e+02 4.834e+02 9.407e+02, threshold=7.649e+02, percent-clipped=3.0 2022-12-24 01:26:09,208 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp0.9 from training. Duration: 24.9833125 2022-12-24 01:26:28,200 INFO [train.py:894] (3/4) Epoch 27, batch 1000, loss[loss=0.1524, simple_loss=0.2478, pruned_loss=0.02845, over 18401.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2574, pruned_loss=0.03925, over 3688585.73 frames. ], batch size: 53, lr: 4.25e-03, grad_scale: 16.0 2022-12-24 01:26:38,561 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0 from training. Duration: 27.14 2022-12-24 01:26:55,360 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0 from training. Duration: 22.44 2022-12-24 01:27:16,181 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3535, 2.0319, 1.6955, 2.0127, 1.8191, 2.1130, 1.8962, 2.1698], device='cuda:3'), covar=tensor([0.2419, 0.3281, 0.2123, 0.2776, 0.3953, 0.1162, 0.3282, 0.1138], device='cuda:3'), in_proj_covar=tensor([0.0301, 0.0299, 0.0254, 0.0349, 0.0281, 0.0235, 0.0298, 0.0222], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 01:27:24,994 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92198.0, num_to_drop=1, layers_to_drop={1} 2022-12-24 01:27:42,727 INFO [train.py:894] (3/4) Epoch 27, batch 1050, loss[loss=0.1904, simple_loss=0.2849, pruned_loss=0.04794, over 18662.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2578, pruned_loss=0.03927, over 3694185.65 frames. ], batch size: 60, lr: 4.25e-03, grad_scale: 16.0 2022-12-24 01:28:11,483 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0060-62364-0_sp0.9 from training. Duration: 21.361125 2022-12-24 01:28:17,146 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp1.1 from training. Duration: 27.0318125 2022-12-24 01:28:27,905 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp0.9 from training. Duration: 28.638875 2022-12-24 01:28:36,239 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=92246.0, num_to_drop=1, layers_to_drop={0} 2022-12-24 01:28:37,440 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.257e+02 3.236e+02 3.952e+02 5.116e+02 1.564e+03, threshold=7.904e+02, percent-clipped=6.0 2022-12-24 01:28:37,641 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92247.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:28:43,252 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0 from training. Duration: 20.4 2022-12-24 01:28:56,546 INFO [train.py:894] (3/4) Epoch 27, batch 1100, loss[loss=0.1826, simple_loss=0.2768, pruned_loss=0.04424, over 18586.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.258, pruned_loss=0.03928, over 3699419.71 frames. ], batch size: 56, lr: 4.25e-03, grad_scale: 16.0 2022-12-24 01:29:17,666 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0 from training. Duration: 20.025 2022-12-24 01:29:17,679 WARNING [train.py:1060] (3/4) Exclude cut with ID 2364-131735-0112-64612-0_sp0.9 from training. Duration: 20.488875 2022-12-24 01:29:23,629 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0 from training. Duration: 29.735 2022-12-24 01:29:51,387 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1714, 2.1226, 1.5674, 2.3205, 2.2648, 2.0759, 2.7801, 2.1705], device='cuda:3'), covar=tensor([0.0861, 0.1733, 0.2834, 0.1779, 0.1822, 0.0869, 0.0948, 0.1271], device='cuda:3'), in_proj_covar=tensor([0.0183, 0.0216, 0.0256, 0.0292, 0.0243, 0.0195, 0.0208, 0.0210], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 01:30:11,210 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.50 vs. limit=5.0 2022-12-24 01:30:11,848 INFO [train.py:894] (3/4) Epoch 27, batch 1150, loss[loss=0.1502, simple_loss=0.2441, pruned_loss=0.02813, over 18664.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2576, pruned_loss=0.03911, over 3702835.70 frames. ], batch size: 48, lr: 4.25e-03, grad_scale: 8.0 2022-12-24 01:30:13,479 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92311.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:30:14,802 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7693, 4.1364, 3.8156, 1.8045, 4.1974, 3.1300, 0.5586, 2.4930], device='cuda:3'), covar=tensor([0.2156, 0.1172, 0.1441, 0.3332, 0.0750, 0.0881, 0.5061, 0.1585], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0146, 0.0159, 0.0125, 0.0150, 0.0116, 0.0144, 0.0114], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-24 01:30:42,509 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6102, 2.0674, 1.6850, 2.3751, 2.6021, 1.6166, 1.5948, 1.3195], device='cuda:3'), covar=tensor([0.2026, 0.1775, 0.1614, 0.1010, 0.1258, 0.1133, 0.2194, 0.1669], device='cuda:3'), in_proj_covar=tensor([0.0248, 0.0229, 0.0219, 0.0201, 0.0262, 0.0196, 0.0226, 0.0202], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 01:30:43,454 WARNING [train.py:1060] (3/4) Exclude cut with ID 7276-92427-0014-12983-0_sp0.9 from training. Duration: 21.3055625 2022-12-24 01:30:44,942 WARNING [train.py:1060] (3/4) Exclude cut with ID 1025-75365-0008-79168-0_sp0.9 from training. Duration: 22.0666875 2022-12-24 01:30:53,162 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92337.0, num_to_drop=1, layers_to_drop={3} 2022-12-24 01:30:56,711 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2022-12-24 01:31:08,646 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.090e+02 3.040e+02 3.627e+02 4.393e+02 8.345e+02, threshold=7.254e+02, percent-clipped=1.0 2022-12-24 01:31:26,212 INFO [train.py:894] (3/4) Epoch 27, batch 1200, loss[loss=0.1461, simple_loss=0.228, pruned_loss=0.03212, over 18415.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2574, pruned_loss=0.03907, over 3705909.61 frames. ], batch size: 42, lr: 4.25e-03, grad_scale: 8.0 2022-12-24 01:32:35,709 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0 from training. Duration: 20.26 2022-12-24 01:32:42,008 INFO [train.py:894] (3/4) Epoch 27, batch 1250, loss[loss=0.1605, simple_loss=0.2511, pruned_loss=0.03492, over 18372.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2564, pruned_loss=0.03824, over 3706167.04 frames. ], batch size: 46, lr: 4.24e-03, grad_scale: 8.0 2022-12-24 01:32:51,480 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0030-9324-0_sp0.9 from training. Duration: 21.3444375 2022-12-24 01:33:39,384 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.696e+02 3.119e+02 3.679e+02 4.523e+02 1.291e+03, threshold=7.358e+02, percent-clipped=5.0 2022-12-24 01:33:45,870 WARNING [train.py:1060] (3/4) Exclude cut with ID 497-129325-0061-62254-0_sp1.1 from training. Duration: 0.97725 2022-12-24 01:33:57,910 INFO [train.py:894] (3/4) Epoch 27, batch 1300, loss[loss=0.1875, simple_loss=0.2818, pruned_loss=0.04663, over 18599.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2561, pruned_loss=0.03817, over 3707990.91 frames. ], batch size: 78, lr: 4.24e-03, grad_scale: 8.0 2022-12-24 01:34:06,078 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0660, 1.4327, 2.5468, 4.3898, 3.3194, 2.7958, 1.1944, 3.2241], device='cuda:3'), covar=tensor([0.1864, 0.1668, 0.1376, 0.0451, 0.0818, 0.1163, 0.2050, 0.0788], device='cuda:3'), in_proj_covar=tensor([0.0103, 0.0119, 0.0137, 0.0155, 0.0106, 0.0144, 0.0129, 0.0116], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2022-12-24 01:34:26,248 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2022-12-24 01:34:26,820 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0_sp0.9 from training. Duration: 22.97225 2022-12-24 01:34:45,551 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6511, 2.2324, 1.7575, 2.3197, 1.9880, 2.3244, 2.0903, 2.5039], device='cuda:3'), covar=tensor([0.1960, 0.3308, 0.2000, 0.2860, 0.3810, 0.1025, 0.3145, 0.1022], device='cuda:3'), in_proj_covar=tensor([0.0298, 0.0297, 0.0251, 0.0345, 0.0278, 0.0233, 0.0295, 0.0219], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 01:34:57,135 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2022-12-24 01:34:59,231 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0047-39922-0_sp0.9 from training. Duration: 21.97775 2022-12-24 01:35:10,048 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.69 vs. limit=5.0 2022-12-24 01:35:12,374 WARNING [train.py:1060] (3/4) Exclude cut with ID 1112-1043-0006-89194-0_sp0.9 from training. Duration: 21.8333125 2022-12-24 01:35:14,356 INFO [train.py:894] (3/4) Epoch 27, batch 1350, loss[loss=0.1539, simple_loss=0.2386, pruned_loss=0.0346, over 18393.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2569, pruned_loss=0.03852, over 3709616.13 frames. ], batch size: 46, lr: 4.24e-03, grad_scale: 8.0 2022-12-24 01:35:22,683 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0 from training. Duration: 20.47 2022-12-24 01:36:11,774 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92547.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:36:12,951 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.794e+02 3.181e+02 4.006e+02 5.114e+02 1.464e+03, threshold=8.011e+02, percent-clipped=2.0 2022-12-24 01:36:27,937 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0037-39912-0_sp0.9 from training. Duration: 20.67225 2022-12-24 01:36:30,561 INFO [train.py:894] (3/4) Epoch 27, batch 1400, loss[loss=0.1603, simple_loss=0.2413, pruned_loss=0.03965, over 18638.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2575, pruned_loss=0.03879, over 3711093.04 frames. ], batch size: 41, lr: 4.24e-03, grad_scale: 8.0 2022-12-24 01:36:46,369 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp0.9 from training. Duration: 25.2444375 2022-12-24 01:37:07,450 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0021-76797-0_sp0.9 from training. Duration: 21.1445 2022-12-24 01:37:22,427 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=92595.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:37:34,871 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2022-12-24 01:37:44,833 INFO [train.py:894] (3/4) Epoch 27, batch 1450, loss[loss=0.1811, simple_loss=0.2754, pruned_loss=0.04336, over 18536.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2579, pruned_loss=0.03892, over 3712247.59 frames. ], batch size: 55, lr: 4.24e-03, grad_scale: 8.0 2022-12-24 01:37:46,733 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92611.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:38:08,361 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92625.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:38:22,222 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp0.9 from training. Duration: 33.038875 2022-12-24 01:38:25,151 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92637.0, num_to_drop=1, layers_to_drop={2} 2022-12-24 01:38:41,491 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.186e+02 3.217e+02 4.081e+02 5.000e+02 1.032e+03, threshold=8.163e+02, percent-clipped=4.0 2022-12-24 01:38:48,146 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8657, 1.9789, 2.3685, 1.2741, 2.3442, 2.2529, 1.6569, 2.6911], device='cuda:3'), covar=tensor([0.1353, 0.1932, 0.1243, 0.2036, 0.0721, 0.1230, 0.2377, 0.0558], device='cuda:3'), in_proj_covar=tensor([0.0199, 0.0216, 0.0208, 0.0196, 0.0171, 0.0219, 0.0218, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 01:38:58,295 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=92659.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:38:59,523 INFO [train.py:894] (3/4) Epoch 27, batch 1500, loss[loss=0.1658, simple_loss=0.2649, pruned_loss=0.03335, over 18636.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2585, pruned_loss=0.03892, over 3713613.34 frames. ], batch size: 53, lr: 4.24e-03, grad_scale: 8.0 2022-12-24 01:38:59,569 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64291-0000-16059-0_sp0.9 from training. Duration: 20.0944375 2022-12-24 01:39:12,853 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp1.1 from training. Duration: 20.4 2022-12-24 01:39:20,615 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0 from training. Duration: 20.085 2022-12-24 01:39:32,537 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0_sp0.9 from training. Duration: 23.07775 2022-12-24 01:39:37,001 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=92685.0, num_to_drop=1, layers_to_drop={1} 2022-12-24 01:39:38,433 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92686.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:40:14,658 INFO [train.py:894] (3/4) Epoch 27, batch 1550, loss[loss=0.1609, simple_loss=0.2565, pruned_loss=0.03262, over 18524.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2587, pruned_loss=0.03891, over 3713274.35 frames. ], batch size: 58, lr: 4.24e-03, grad_scale: 8.0 2022-12-24 01:40:17,459 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp0.9 from training. Duration: 24.9333125 2022-12-24 01:40:29,818 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92720.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:41:00,869 WARNING [train.py:1060] (3/4) Exclude cut with ID 5118-111612-0016-124680-0_sp0.9 from training. Duration: 20.388875 2022-12-24 01:41:08,052 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp1.1 from training. Duration: 20.3590625 2022-12-24 01:41:11,048 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.280e+02 3.341e+02 4.064e+02 5.045e+02 9.351e+02, threshold=8.127e+02, percent-clipped=1.0 2022-12-24 01:41:29,805 INFO [train.py:894] (3/4) Epoch 27, batch 1600, loss[loss=0.1458, simple_loss=0.2338, pruned_loss=0.02895, over 18396.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2576, pruned_loss=0.03875, over 3712694.04 frames. ], batch size: 46, lr: 4.24e-03, grad_scale: 8.0 2022-12-24 01:42:02,895 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92781.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:42:13,517 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8516, 2.3304, 1.7511, 2.6087, 2.9610, 1.7675, 1.8448, 1.4539], device='cuda:3'), covar=tensor([0.1870, 0.1620, 0.1579, 0.1004, 0.1354, 0.1094, 0.2071, 0.1586], device='cuda:3'), in_proj_covar=tensor([0.0251, 0.0230, 0.0222, 0.0204, 0.0265, 0.0198, 0.0229, 0.0205], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 01:42:16,017 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0_sp1.1 from training. Duration: 0.836375 2022-12-24 01:42:46,229 INFO [train.py:894] (3/4) Epoch 27, batch 1650, loss[loss=0.1921, simple_loss=0.2757, pruned_loss=0.05421, over 18447.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2573, pruned_loss=0.03902, over 3713056.12 frames. ], batch size: 50, lr: 4.24e-03, grad_scale: 8.0 2022-12-24 01:42:59,485 WARNING [train.py:1060] (3/4) Exclude cut with ID 8565-290391-0049-67394-0_sp0.9 from training. Duration: 21.3166875 2022-12-24 01:43:06,840 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6835, 1.2925, 2.2260, 3.3234, 2.5485, 2.5702, 0.8816, 2.4783], device='cuda:3'), covar=tensor([0.1793, 0.1710, 0.1347, 0.0594, 0.0888, 0.1185, 0.2308, 0.0947], device='cuda:3'), in_proj_covar=tensor([0.0102, 0.0118, 0.0136, 0.0155, 0.0106, 0.0142, 0.0128, 0.0115], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-24 01:43:17,859 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92831.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:43:28,933 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0029-104863-0_sp0.9 from training. Duration: 22.1055625 2022-12-24 01:43:42,347 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp1.1 from training. Duration: 21.77725 2022-12-24 01:43:43,753 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.715e+02 3.368e+02 3.937e+02 4.829e+02 1.412e+03, threshold=7.874e+02, percent-clipped=4.0 2022-12-24 01:44:01,291 INFO [train.py:894] (3/4) Epoch 27, batch 1700, loss[loss=0.1768, simple_loss=0.2644, pruned_loss=0.04465, over 18595.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2579, pruned_loss=0.03983, over 3714812.30 frames. ], batch size: 77, lr: 4.23e-03, grad_scale: 8.0 2022-12-24 01:44:02,838 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp0.9 from training. Duration: 27.8166875 2022-12-24 01:44:28,003 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp1.1 from training. Duration: 22.5090625 2022-12-24 01:44:35,984 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0 from training. Duration: 25.035 2022-12-24 01:44:50,143 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92892.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:44:53,791 WARNING [train.py:1060] (3/4) Exclude cut with ID 774-127930-0014-10412-0_sp1.1 from training. Duration: 0.95 2022-12-24 01:45:00,589 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92899.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:45:12,212 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp0.9 from training. Duration: 0.92225 2022-12-24 01:45:16,471 INFO [train.py:894] (3/4) Epoch 27, batch 1750, loss[loss=0.1861, simple_loss=0.2776, pruned_loss=0.04726, over 18731.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2578, pruned_loss=0.04052, over 3714050.78 frames. ], batch size: 52, lr: 4.23e-03, grad_scale: 8.0 2022-12-24 01:45:38,264 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0 from training. Duration: 21.97 2022-12-24 01:45:47,666 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92931.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:45:56,532 WARNING [train.py:1060] (3/4) Exclude cut with ID 7492-105653-0055-62765-0_sp0.9 from training. Duration: 21.97225 2022-12-24 01:45:57,769 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp0.9 from training. Duration: 25.3333125 2022-12-24 01:46:09,076 WARNING [train.py:1060] (3/4) Exclude cut with ID 5172-29468-0015-19128-0_sp0.9 from training. Duration: 21.5055625 2022-12-24 01:46:13,296 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.049e+02 3.881e+02 4.504e+02 6.102e+02 9.952e+02, threshold=9.008e+02, percent-clipped=7.0 2022-12-24 01:46:17,520 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp1.1 from training. Duration: 20.72725 2022-12-24 01:46:30,515 INFO [train.py:894] (3/4) Epoch 27, batch 1800, loss[loss=0.1565, simple_loss=0.2481, pruned_loss=0.03241, over 18394.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2583, pruned_loss=0.04193, over 3713798.19 frames. ], batch size: 53, lr: 4.23e-03, grad_scale: 8.0 2022-12-24 01:46:30,948 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92960.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:46:43,786 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.76 vs. limit=5.0 2022-12-24 01:46:47,713 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp0.9 from training. Duration: 26.32775 2022-12-24 01:46:54,562 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92975.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:46:56,241 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2082, 2.0334, 1.7241, 2.0263, 2.2030, 2.0135, 2.6279, 2.2072], device='cuda:3'), covar=tensor([0.0879, 0.1690, 0.2769, 0.1785, 0.1898, 0.0918, 0.1073, 0.1268], device='cuda:3'), in_proj_covar=tensor([0.0183, 0.0216, 0.0257, 0.0293, 0.0243, 0.0195, 0.0208, 0.0209], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 01:47:03,296 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92981.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:47:20,581 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92992.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:47:21,568 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0 from training. Duration: 20.025 2022-12-24 01:47:26,937 WARNING [train.py:1060] (3/4) Exclude cut with ID 6709-74022-0004-86860-0_sp1.1 from training. Duration: 0.9409375 2022-12-24 01:47:26,944 WARNING [train.py:1060] (3/4) Exclude cut with ID 4757-1811-0023-62229-0_sp0.9 from training. Duration: 21.37775 2022-12-24 01:47:46,310 INFO [train.py:894] (3/4) Epoch 27, batch 1850, loss[loss=0.1768, simple_loss=0.2636, pruned_loss=0.04501, over 18367.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2591, pruned_loss=0.04306, over 3714577.78 frames. ], batch size: 46, lr: 4.23e-03, grad_scale: 8.0 2022-12-24 01:47:47,292 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93010.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:47:48,563 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0004-25974-0_sp0.9 from training. Duration: 21.17225 2022-12-24 01:47:48,575 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp0.9 from training. Duration: 27.511125 2022-12-24 01:47:57,504 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.2362, 1.2816, 0.8297, 1.4071, 1.5089, 2.3681, 1.2570, 1.4077], device='cuda:3'), covar=tensor([0.0860, 0.1815, 0.1122, 0.0845, 0.1439, 0.0338, 0.1417, 0.1592], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0083, 0.0072, 0.0075, 0.0091, 0.0076, 0.0084, 0.0077], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2022-12-24 01:48:21,116 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0 from training. Duration: 22.8 2022-12-24 01:48:26,264 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93036.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:48:26,324 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93036.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:48:27,225 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0 from training. Duration: 22.585 2022-12-24 01:48:28,765 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2022-12-24 01:48:44,107 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.543e+02 3.991e+02 4.524e+02 5.614e+02 9.322e+02, threshold=9.048e+02, percent-clipped=2.0 2022-12-24 01:48:55,720 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0001-146967-0_sp0.9 from training. Duration: 22.0166875 2022-12-24 01:49:02,365 INFO [train.py:894] (3/4) Epoch 27, batch 1900, loss[loss=0.1562, simple_loss=0.2363, pruned_loss=0.0381, over 18685.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2589, pruned_loss=0.0433, over 3716219.81 frames. ], batch size: 48, lr: 4.23e-03, grad_scale: 8.0 2022-12-24 01:49:03,426 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2022-12-24 01:49:13,034 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp1.1 from training. Duration: 24.395375 2022-12-24 01:49:17,567 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2022-12-24 01:49:19,683 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93071.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:49:22,222 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp0.9 from training. Duration: 27.47775 2022-12-24 01:49:26,577 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp0.9 from training. Duration: 24.8833125 2022-12-24 01:49:26,706 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93076.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:49:29,468 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0 from training. Duration: 23.39 2022-12-24 01:49:35,196 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp0.9 from training. Duration: 28.72225 2022-12-24 01:49:44,456 WARNING [train.py:1060] (3/4) Exclude cut with ID 585-294811-0110-133686-0_sp0.9 from training. Duration: 20.8944375 2022-12-24 01:49:44,873 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0236, 2.6678, 1.9213, 3.1252, 3.0373, 1.9620, 2.2398, 1.5136], device='cuda:3'), covar=tensor([0.1698, 0.1412, 0.1375, 0.0734, 0.1201, 0.0987, 0.1688, 0.1445], device='cuda:3'), in_proj_covar=tensor([0.0250, 0.0231, 0.0221, 0.0203, 0.0266, 0.0198, 0.0228, 0.0203], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 01:49:58,016 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93097.0, num_to_drop=1, layers_to_drop={0} 2022-12-24 01:49:59,152 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0_sp0.9 from training. Duration: 23.8444375 2022-12-24 01:50:12,997 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.2824, 2.8512, 2.8608, 3.2319, 2.9593, 2.8833, 3.3677, 1.0290], device='cuda:3'), covar=tensor([0.0991, 0.0917, 0.0866, 0.1115, 0.1563, 0.1360, 0.0880, 0.4964], device='cuda:3'), in_proj_covar=tensor([0.0354, 0.0234, 0.0243, 0.0282, 0.0331, 0.0273, 0.0298, 0.0290], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 01:50:16,859 INFO [train.py:894] (3/4) Epoch 27, batch 1950, loss[loss=0.1932, simple_loss=0.2852, pruned_loss=0.05058, over 18579.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2601, pruned_loss=0.04431, over 3716362.76 frames. ], batch size: 56, lr: 4.23e-03, grad_scale: 8.0 2022-12-24 01:50:22,674 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0 from training. Duration: 25.85 2022-12-24 01:50:22,680 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0 from training. Duration: 21.39 2022-12-24 01:50:24,337 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93115.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:50:33,426 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0 from training. Duration: 27.92 2022-12-24 01:51:03,001 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0039-130165-0_sp0.9 from training. Duration: 20.661125 2022-12-24 01:51:12,676 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.455e+02 4.181e+02 4.936e+02 5.974e+02 1.019e+03, threshold=9.872e+02, percent-clipped=2.0 2022-12-24 01:51:26,038 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0043-15874-0_sp0.9 from training. Duration: 20.07225 2022-12-24 01:51:31,589 INFO [train.py:894] (3/4) Epoch 27, batch 2000, loss[loss=0.1773, simple_loss=0.2662, pruned_loss=0.04422, over 18688.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2601, pruned_loss=0.04502, over 3717393.71 frames. ], batch size: 60, lr: 4.23e-03, grad_scale: 8.0 2022-12-24 01:51:31,616 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0 from training. Duration: 21.01 2022-12-24 01:51:56,092 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93176.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:52:00,704 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4963, 2.2258, 1.8964, 0.8296, 1.7239, 1.9944, 1.7335, 1.9634], device='cuda:3'), covar=tensor([0.0650, 0.0582, 0.1204, 0.1678, 0.1249, 0.1577, 0.1669, 0.0812], device='cuda:3'), in_proj_covar=tensor([0.0178, 0.0191, 0.0214, 0.0193, 0.0213, 0.0208, 0.0220, 0.0208], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 01:52:11,572 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93187.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:52:42,744 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0 from training. Duration: 20.65 2022-12-24 01:52:45,661 INFO [train.py:894] (3/4) Epoch 27, batch 2050, loss[loss=0.1804, simple_loss=0.2676, pruned_loss=0.04657, over 18514.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2608, pruned_loss=0.04581, over 3716623.00 frames. ], batch size: 58, lr: 4.23e-03, grad_scale: 8.0 2022-12-24 01:52:49,558 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0 from training. Duration: 21.46 2022-12-24 01:53:34,541 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0 from training. Duration: 0.92 2022-12-24 01:53:41,483 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.802e+02 4.249e+02 4.843e+02 5.680e+02 1.100e+03, threshold=9.685e+02, percent-clipped=2.0 2022-12-24 01:53:41,516 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0_sp0.9 from training. Duration: 23.7666875 2022-12-24 01:53:52,244 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93255.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:53:55,674 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9800, 1.8290, 1.5240, 1.5716, 1.6881, 1.8370, 1.7047, 1.7765], device='cuda:3'), covar=tensor([0.2446, 0.3180, 0.2098, 0.2539, 0.3516, 0.1150, 0.2817, 0.1122], device='cuda:3'), in_proj_covar=tensor([0.0301, 0.0299, 0.0254, 0.0348, 0.0281, 0.0234, 0.0297, 0.0221], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 01:53:59,918 INFO [train.py:894] (3/4) Epoch 27, batch 2100, loss[loss=0.1525, simple_loss=0.2288, pruned_loss=0.03814, over 18546.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2607, pruned_loss=0.04606, over 3716246.94 frames. ], batch size: 44, lr: 4.23e-03, grad_scale: 8.0 2022-12-24 01:54:18,164 WARNING [train.py:1060] (3/4) Exclude cut with ID 8544-281189-0060-101339-0_sp0.9 from training. Duration: 20.861125 2022-12-24 01:54:25,351 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2022-12-24 01:54:28,812 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0_sp0.9 from training. Duration: 22.711125 2022-12-24 01:54:31,915 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93281.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:54:40,841 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93287.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:55:10,108 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp1.1 from training. Duration: 22.986375 2022-12-24 01:55:15,632 INFO [train.py:894] (3/4) Epoch 27, batch 2150, loss[loss=0.1592, simple_loss=0.2382, pruned_loss=0.04011, over 18525.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2603, pruned_loss=0.0461, over 3716807.82 frames. ], batch size: 44, lr: 4.22e-03, grad_scale: 8.0 2022-12-24 01:55:27,367 WARNING [train.py:1060] (3/4) Exclude cut with ID 8040-260924-0003-80960-0_sp0.9 from training. Duration: 22.07225 2022-12-24 01:55:28,409 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2957, 2.0289, 1.6114, 1.9200, 1.8030, 2.0481, 1.9040, 2.1043], device='cuda:3'), covar=tensor([0.2267, 0.3085, 0.2214, 0.2637, 0.3428, 0.1156, 0.2839, 0.1097], device='cuda:3'), in_proj_covar=tensor([0.0300, 0.0299, 0.0253, 0.0348, 0.0280, 0.0233, 0.0297, 0.0220], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 01:55:32,711 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0045-26330-0_sp0.9 from training. Duration: 20.3055625 2022-12-24 01:55:35,706 WARNING [train.py:1060] (3/4) Exclude cut with ID 6356-271890-0060-94317-0_sp0.9 from training. Duration: 20.72225 2022-12-24 01:55:37,508 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4759, 1.4192, 1.3162, 1.3687, 1.7660, 1.6111, 1.6130, 1.2176], device='cuda:3'), covar=tensor([0.0307, 0.0221, 0.0507, 0.0210, 0.0187, 0.0345, 0.0230, 0.0311], device='cuda:3'), in_proj_covar=tensor([0.0096, 0.0127, 0.0154, 0.0121, 0.0117, 0.0121, 0.0100, 0.0129], device='cuda:3'), out_proj_covar=tensor([7.6112e-05, 1.0036e-04, 1.2561e-04, 9.5477e-05, 9.3542e-05, 9.2388e-05, 7.7846e-05, 1.0151e-04], device='cuda:3') 2022-12-24 01:55:44,491 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=93329.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:55:47,317 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93331.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:55:54,913 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp1.1 from training. Duration: 22.4818125 2022-12-24 01:56:11,861 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.853e+02 4.006e+02 4.754e+02 5.820e+02 9.095e+02, threshold=9.509e+02, percent-clipped=0.0 2022-12-24 01:56:20,609 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp0.9 from training. Duration: 25.0944375 2022-12-24 01:56:23,594 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0 from training. Duration: 21.515 2022-12-24 01:56:29,923 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp0.9 from training. Duration: 27.02225 2022-12-24 01:56:31,363 INFO [train.py:894] (3/4) Epoch 27, batch 2200, loss[loss=0.2098, simple_loss=0.2936, pruned_loss=0.06301, over 18688.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2602, pruned_loss=0.04618, over 3716798.96 frames. ], batch size: 69, lr: 4.22e-03, grad_scale: 8.0 2022-12-24 01:56:35,895 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0010-62480-0_sp0.9 from training. Duration: 22.22225 2022-12-24 01:56:41,803 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93366.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:56:43,013 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0085-44554-0_sp0.9 from training. Duration: 20.85 2022-12-24 01:56:56,710 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93376.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:57:04,150 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.6182, 3.9206, 3.9327, 4.4940, 4.2249, 4.0914, 4.7312, 1.4759], device='cuda:3'), covar=tensor([0.0670, 0.0764, 0.0673, 0.0863, 0.1329, 0.1163, 0.0570, 0.4931], device='cuda:3'), in_proj_covar=tensor([0.0357, 0.0236, 0.0247, 0.0285, 0.0335, 0.0276, 0.0302, 0.0293], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 01:57:07,466 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8090, 1.7528, 1.5316, 1.5827, 1.9658, 2.0672, 2.0073, 1.4043], device='cuda:3'), covar=tensor([0.0298, 0.0270, 0.0519, 0.0238, 0.0211, 0.0392, 0.0289, 0.0361], device='cuda:3'), in_proj_covar=tensor([0.0096, 0.0127, 0.0154, 0.0121, 0.0117, 0.0121, 0.0101, 0.0129], device='cuda:3'), out_proj_covar=tensor([7.6182e-05, 1.0041e-04, 1.2581e-04, 9.5550e-05, 9.3768e-05, 9.2471e-05, 7.8178e-05, 1.0163e-04], device='cuda:3') 2022-12-24 01:57:12,944 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0 from training. Duration: 21.54 2022-12-24 01:57:17,486 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp1.1 from training. Duration: 20.5318125 2022-12-24 01:57:20,669 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93392.0, num_to_drop=1, layers_to_drop={1} 2022-12-24 01:57:29,643 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0012-134311-0_sp0.9 from training. Duration: 21.9333125 2022-12-24 01:57:48,377 INFO [train.py:894] (3/4) Epoch 27, batch 2250, loss[loss=0.178, simple_loss=0.2649, pruned_loss=0.04551, over 18442.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2605, pruned_loss=0.04647, over 3715500.43 frames. ], batch size: 64, lr: 4.22e-03, grad_scale: 8.0 2022-12-24 01:58:03,064 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7810, 1.2628, 0.8744, 1.4039, 2.2226, 1.0475, 1.4667, 1.6721], device='cuda:3'), covar=tensor([0.1556, 0.2086, 0.2037, 0.1482, 0.1689, 0.1708, 0.1452, 0.1697], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0098, 0.0116, 0.0096, 0.0120, 0.0092, 0.0098, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-24 01:58:10,554 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=93424.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:58:19,113 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0025-130151-0_sp0.9 from training. Duration: 21.7944375 2022-12-24 01:58:30,617 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0_sp0.9 from training. Duration: 22.4666875 2022-12-24 01:58:37,474 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0 from training. Duration: 21.635 2022-12-24 01:58:42,382 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0_sp0.9 from training. Duration: 24.038875 2022-12-24 01:58:47,233 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.729e+02 4.386e+02 5.059e+02 6.164e+02 8.969e+02, threshold=1.012e+03, percent-clipped=0.0 2022-12-24 01:59:05,744 INFO [train.py:894] (3/4) Epoch 27, batch 2300, loss[loss=0.1595, simple_loss=0.2466, pruned_loss=0.0362, over 18452.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2593, pruned_loss=0.04607, over 3715085.46 frames. ], batch size: 50, lr: 4.22e-03, grad_scale: 8.0 2022-12-24 01:59:22,028 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93471.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 01:59:25,088 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp1.1 from training. Duration: 21.786375 2022-12-24 01:59:36,989 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0 from training. Duration: 20.22 2022-12-24 01:59:45,310 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.68 vs. limit=5.0 2022-12-24 01:59:46,489 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93487.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:00:11,898 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2022-12-24 02:00:21,176 INFO [train.py:894] (3/4) Epoch 27, batch 2350, loss[loss=0.1742, simple_loss=0.2637, pruned_loss=0.04242, over 18506.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2599, pruned_loss=0.04635, over 3714695.80 frames. ], batch size: 77, lr: 4.22e-03, grad_scale: 8.0 2022-12-24 02:00:47,118 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1682, 1.4699, 1.8114, 1.8264, 2.1248, 2.1946, 1.9578, 1.8686], device='cuda:3'), covar=tensor([0.2925, 0.4056, 0.3440, 0.3625, 0.2888, 0.1285, 0.4261, 0.1743], device='cuda:3'), in_proj_covar=tensor([0.0273, 0.0303, 0.0288, 0.0327, 0.0319, 0.0260, 0.0356, 0.0250], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 02:00:57,251 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=93535.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:01:17,317 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.947e+02 4.125e+02 5.226e+02 6.222e+02 1.366e+03, threshold=1.045e+03, percent-clipped=5.0 2022-12-24 02:01:20,596 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6174, 1.7817, 1.5725, 2.1290, 2.3977, 1.6845, 1.3555, 1.4211], device='cuda:3'), covar=tensor([0.1851, 0.1776, 0.1591, 0.1055, 0.1181, 0.1030, 0.2240, 0.1500], device='cuda:3'), in_proj_covar=tensor([0.0249, 0.0230, 0.0221, 0.0203, 0.0265, 0.0197, 0.0228, 0.0203], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 02:01:27,588 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93555.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:01:35,146 INFO [train.py:894] (3/4) Epoch 27, batch 2400, loss[loss=0.1742, simple_loss=0.2599, pruned_loss=0.04424, over 18440.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2603, pruned_loss=0.04653, over 3714299.08 frames. ], batch size: 64, lr: 4.22e-03, grad_scale: 8.0 2022-12-24 02:01:36,785 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0 from training. Duration: 25.285 2022-12-24 02:02:16,168 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93587.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:02:25,522 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8357, 1.7624, 1.7343, 1.7193, 1.5473, 3.9726, 1.7982, 2.3108], device='cuda:3'), covar=tensor([0.4276, 0.2659, 0.2422, 0.2718, 0.1419, 0.0260, 0.1529, 0.0853], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0118, 0.0126, 0.0122, 0.0107, 0.0097, 0.0091, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-24 02:02:37,713 WARNING [train.py:1060] (3/4) Exclude cut with ID 811-130148-0001-63453-0_sp0.9 from training. Duration: 20.861125 2022-12-24 02:02:41,088 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=93603.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:02:51,435 INFO [train.py:894] (3/4) Epoch 27, batch 2450, loss[loss=0.1722, simple_loss=0.2511, pruned_loss=0.04661, over 18585.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2607, pruned_loss=0.0463, over 3714033.23 frames. ], batch size: 49, lr: 4.22e-03, grad_scale: 8.0 2022-12-24 02:03:00,472 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0 from training. Duration: 20.88 2022-12-24 02:03:23,505 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93631.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:03:29,063 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=93635.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:03:30,358 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0_sp0.9 from training. Duration: 23.4166875 2022-12-24 02:03:49,295 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.298e+02 3.850e+02 4.685e+02 5.940e+02 1.003e+03, threshold=9.370e+02, percent-clipped=0.0 2022-12-24 02:04:07,630 INFO [train.py:894] (3/4) Epoch 27, batch 2500, loss[loss=0.1988, simple_loss=0.277, pruned_loss=0.0603, over 18646.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2602, pruned_loss=0.04584, over 3714276.02 frames. ], batch size: 53, lr: 4.22e-03, grad_scale: 8.0 2022-12-24 02:04:16,388 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93666.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:04:36,366 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=93679.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:04:47,224 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0 from training. Duration: 21.24 2022-12-24 02:04:47,237 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0_sp0.9 from training. Duration: 23.9055625 2022-12-24 02:04:57,074 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93692.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:05:12,046 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2022-12-24 02:05:20,488 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp0.9 from training. Duration: 25.988875 2022-12-24 02:05:23,658 INFO [train.py:894] (3/4) Epoch 27, batch 2550, loss[loss=0.1678, simple_loss=0.2589, pruned_loss=0.03836, over 18390.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2596, pruned_loss=0.04561, over 3714095.79 frames. ], batch size: 53, lr: 4.22e-03, grad_scale: 8.0 2022-12-24 02:05:27,952 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0001-134300-0_sp0.9 from training. Duration: 20.67225 2022-12-24 02:05:29,497 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=93714.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:06:09,771 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=93740.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:06:17,879 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0 from training. Duration: 20.34 2022-12-24 02:06:20,798 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.049e+02 3.985e+02 4.763e+02 6.217e+02 1.238e+03, threshold=9.527e+02, percent-clipped=6.0 2022-12-24 02:06:37,891 INFO [train.py:894] (3/4) Epoch 27, batch 2600, loss[loss=0.1893, simple_loss=0.2794, pruned_loss=0.04964, over 18533.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2589, pruned_loss=0.0455, over 3713529.30 frames. ], batch size: 58, lr: 4.21e-03, grad_scale: 8.0 2022-12-24 02:06:53,241 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.56 vs. limit=5.0 2022-12-24 02:06:54,003 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93770.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:06:55,947 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93771.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:07:06,455 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6466, 2.2201, 1.7954, 2.4062, 2.0419, 2.1798, 2.0953, 2.4352], device='cuda:3'), covar=tensor([0.2155, 0.3400, 0.2065, 0.3003, 0.3816, 0.1158, 0.3344, 0.1111], device='cuda:3'), in_proj_covar=tensor([0.0302, 0.0302, 0.0256, 0.0351, 0.0284, 0.0236, 0.0299, 0.0223], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 02:07:29,069 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp0.9 from training. Duration: 25.061125 2022-12-24 02:07:41,143 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0 from training. Duration: 0.83 2022-12-24 02:07:41,584 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3720, 1.4865, 1.2925, 1.7270, 1.6126, 1.4590, 1.0250, 1.2241], device='cuda:3'), covar=tensor([0.2036, 0.1950, 0.1804, 0.1199, 0.1346, 0.1133, 0.2334, 0.1657], device='cuda:3'), in_proj_covar=tensor([0.0250, 0.0231, 0.0223, 0.0204, 0.0266, 0.0199, 0.0230, 0.0204], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 02:07:53,329 INFO [train.py:894] (3/4) Epoch 27, batch 2650, loss[loss=0.1966, simple_loss=0.2854, pruned_loss=0.0539, over 18646.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2585, pruned_loss=0.04515, over 3713373.17 frames. ], batch size: 69, lr: 4.21e-03, grad_scale: 8.0 2022-12-24 02:08:04,729 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2022-12-24 02:08:07,512 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=93819.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:08:08,762 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0 from training. Duration: 24.73 2022-12-24 02:08:21,914 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0 from training. Duration: 23.965 2022-12-24 02:08:25,034 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93831.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:08:29,449 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2022-12-24 02:08:31,203 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0030-146996-0_sp0.9 from training. Duration: 22.088875 2022-12-24 02:08:45,811 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0_sp0.9 from training. Duration: 23.6 2022-12-24 02:08:50,008 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.431e+02 4.153e+02 4.837e+02 6.025e+02 1.293e+03, threshold=9.673e+02, percent-clipped=4.0 2022-12-24 02:09:07,695 INFO [train.py:894] (3/4) Epoch 27, batch 2700, loss[loss=0.2166, simple_loss=0.2926, pruned_loss=0.07027, over 18689.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2586, pruned_loss=0.04517, over 3713872.06 frames. ], batch size: 78, lr: 4.21e-03, grad_scale: 8.0 2022-12-24 02:10:22,945 INFO [train.py:894] (3/4) Epoch 27, batch 2750, loss[loss=0.1714, simple_loss=0.2641, pruned_loss=0.03936, over 18568.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2595, pruned_loss=0.04576, over 3714403.30 frames. ], batch size: 49, lr: 4.21e-03, grad_scale: 8.0 2022-12-24 02:10:25,840 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0 from training. Duration: 23.795 2022-12-24 02:10:41,999 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp1.1 from training. Duration: 21.5409375 2022-12-24 02:10:44,854 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp0.9 from training. Duration: 24.97775 2022-12-24 02:10:54,884 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0_sp0.9 from training. Duration: 23.3444375 2022-12-24 02:11:04,098 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.3892, 3.8034, 3.8237, 4.3003, 4.0243, 3.8659, 4.4892, 1.4942], device='cuda:3'), covar=tensor([0.0727, 0.0728, 0.0669, 0.0839, 0.1300, 0.1239, 0.0635, 0.5069], device='cuda:3'), in_proj_covar=tensor([0.0363, 0.0239, 0.0248, 0.0286, 0.0338, 0.0279, 0.0305, 0.0295], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 02:11:21,038 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.780e+02 4.032e+02 4.691e+02 5.983e+02 1.414e+03, threshold=9.382e+02, percent-clipped=4.0 2022-12-24 02:11:21,963 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.01 vs. limit=5.0 2022-12-24 02:11:22,522 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0_sp0.9 from training. Duration: 23.2 2022-12-24 02:11:28,556 WARNING [train.py:1060] (3/4) Exclude cut with ID 5653-46179-0060-117930-0_sp0.9 from training. Duration: 21.17225 2022-12-24 02:11:33,463 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2022-12-24 02:11:39,469 INFO [train.py:894] (3/4) Epoch 27, batch 2800, loss[loss=0.1704, simple_loss=0.2446, pruned_loss=0.04808, over 18394.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2588, pruned_loss=0.04551, over 3713753.36 frames. ], batch size: 46, lr: 4.21e-03, grad_scale: 8.0 2022-12-24 02:11:48,864 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp0.9 from training. Duration: 24.6555625 2022-12-24 02:12:12,011 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93981.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:12:47,363 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0 from training. Duration: 20.44 2022-12-24 02:12:59,239 INFO [train.py:894] (3/4) Epoch 27, batch 2850, loss[loss=0.1987, simple_loss=0.2734, pruned_loss=0.06193, over 18543.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2585, pruned_loss=0.04559, over 3714182.45 frames. ], batch size: 47, lr: 4.21e-03, grad_scale: 8.0 2022-12-24 02:13:02,825 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0_sp0.9 from training. Duration: 23.45 2022-12-24 02:13:04,716 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.2475, 1.5634, 1.7216, 1.0235, 1.0468, 1.7903, 1.7045, 1.5728], device='cuda:3'), covar=tensor([0.0812, 0.0338, 0.0333, 0.0388, 0.0438, 0.0482, 0.0283, 0.0709], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0171, 0.0129, 0.0141, 0.0147, 0.0142, 0.0167, 0.0177], device='cuda:3'), out_proj_covar=tensor([1.1267e-04, 1.2876e-04, 9.4777e-05, 1.0313e-04, 1.0784e-04, 1.0701e-04, 1.2587e-04, 1.3326e-04], device='cuda:3') 2022-12-24 02:13:19,932 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.86 vs. limit=5.0 2022-12-24 02:13:30,469 WARNING [train.py:1060] (3/4) Exclude cut with ID 6945-60535-0076-12784-0_sp0.9 from training. Duration: 20.52225 2022-12-24 02:13:39,373 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0 from training. Duration: 22.19 2022-12-24 02:13:47,354 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94042.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:13:51,299 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp1.1 from training. Duration: 25.3818125 2022-12-24 02:13:55,589 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.597e+02 3.768e+02 4.678e+02 5.604e+02 1.868e+03, threshold=9.356e+02, percent-clipped=5.0 2022-12-24 02:14:09,552 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp0.9 from training. Duration: 28.0944375 2022-12-24 02:14:14,011 INFO [train.py:894] (3/4) Epoch 27, batch 2900, loss[loss=0.14, simple_loss=0.2227, pruned_loss=0.02866, over 18526.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2585, pruned_loss=0.04533, over 3714408.17 frames. ], batch size: 44, lr: 4.21e-03, grad_scale: 8.0 2022-12-24 02:14:15,377 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0_sp0.9 from training. Duration: 22.9444375 2022-12-24 02:14:24,263 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp1.1 from training. Duration: 21.6318125 2022-12-24 02:14:39,691 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0 from training. Duration: 23.695 2022-12-24 02:15:05,827 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0 from training. Duration: 23.955 2022-12-24 02:15:29,433 INFO [train.py:894] (3/4) Epoch 27, batch 2950, loss[loss=0.1633, simple_loss=0.2408, pruned_loss=0.04287, over 18400.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2584, pruned_loss=0.04508, over 3714944.10 frames. ], batch size: 46, lr: 4.21e-03, grad_scale: 8.0 2022-12-24 02:15:40,131 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp0.9 from training. Duration: 26.438875 2022-12-24 02:15:53,427 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94126.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:16:22,097 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0021-26306-0_sp0.9 from training. Duration: 21.2444375 2022-12-24 02:16:23,555 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp0.9 from training. Duration: 31.02225 2022-12-24 02:16:28,309 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.796e+02 4.001e+02 4.676e+02 5.436e+02 1.154e+03, threshold=9.352e+02, percent-clipped=2.0 2022-12-24 02:16:33,419 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0 from training. Duration: 22.395 2022-12-24 02:16:45,794 INFO [train.py:894] (3/4) Epoch 27, batch 3000, loss[loss=0.1421, simple_loss=0.22, pruned_loss=0.03207, over 18485.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2582, pruned_loss=0.04514, over 3714844.39 frames. ], batch size: 43, lr: 4.21e-03, grad_scale: 8.0 2022-12-24 02:16:45,794 INFO [train.py:919] (3/4) Computing validation loss 2022-12-24 02:16:56,402 INFO [train.py:928] (3/4) Epoch 27, validation: loss=0.1648, simple_loss=0.2604, pruned_loss=0.03454, over 944034.00 frames. 2022-12-24 02:16:56,403 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-24 02:17:00,740 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0 from training. Duration: 21.075 2022-12-24 02:17:05,876 WARNING [train.py:1060] (3/4) Exclude cut with ID 6482-98857-0025-147532-0_sp0.9 from training. Duration: 20.0055625 2022-12-24 02:17:07,352 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0037-132304-0_sp0.9 from training. Duration: 22.05 2022-12-24 02:17:07,365 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0 from training. Duration: 26.8349375 2022-12-24 02:17:10,133 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp1.1 from training. Duration: 22.1090625 2022-12-24 02:17:17,647 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp0.9 from training. Duration: 26.6166875 2022-12-24 02:17:21,900 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2022-12-24 02:17:35,687 WARNING [train.py:1060] (3/4) Exclude cut with ID 2046-178027-0000-53705-0_sp0.9 from training. Duration: 20.3055625 2022-12-24 02:17:58,980 WARNING [train.py:1060] (3/4) Exclude cut with ID 7205-50138-0008-5373-0_sp0.9 from training. Duration: 20.7 2022-12-24 02:18:12,659 INFO [train.py:894] (3/4) Epoch 27, batch 3050, loss[loss=0.179, simple_loss=0.2739, pruned_loss=0.04204, over 18584.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2583, pruned_loss=0.04524, over 3714638.67 frames. ], batch size: 56, lr: 4.20e-03, grad_scale: 8.0 2022-12-24 02:18:32,208 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=94223.0, num_to_drop=1, layers_to_drop={1} 2022-12-24 02:18:43,406 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0 from training. Duration: 22.48 2022-12-24 02:18:57,300 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp0.9 from training. Duration: 29.816625 2022-12-24 02:19:11,004 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.668e+02 4.124e+02 4.922e+02 6.064e+02 1.653e+03, threshold=9.844e+02, percent-clipped=4.0 2022-12-24 02:19:17,869 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp1.1 from training. Duration: 22.7590625 2022-12-24 02:19:26,007 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0 from training. Duration: 22.555 2022-12-24 02:19:28,699 INFO [train.py:894] (3/4) Epoch 27, batch 3100, loss[loss=0.2156, simple_loss=0.3033, pruned_loss=0.06397, over 18729.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2587, pruned_loss=0.04528, over 3715242.58 frames. ], batch size: 54, lr: 4.20e-03, grad_scale: 8.0 2022-12-24 02:19:36,126 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.0307, 5.3271, 4.8312, 2.6895, 5.4795, 4.1372, 0.5484, 3.6651], device='cuda:3'), covar=tensor([0.1771, 0.0997, 0.1290, 0.2842, 0.0616, 0.0713, 0.5224, 0.1200], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0150, 0.0162, 0.0126, 0.0153, 0.0117, 0.0147, 0.0116], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-24 02:19:46,321 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0005-25975-0_sp0.9 from training. Duration: 21.688875 2022-12-24 02:20:05,566 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94284.0, num_to_drop=1, layers_to_drop={3} 2022-12-24 02:20:18,855 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0_sp0.9 from training. Duration: 22.6 2022-12-24 02:20:38,736 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.61 vs. limit=5.0 2022-12-24 02:20:43,870 INFO [train.py:894] (3/4) Epoch 27, batch 3150, loss[loss=0.1668, simple_loss=0.256, pruned_loss=0.03876, over 18512.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2587, pruned_loss=0.04512, over 3716732.47 frames. ], batch size: 52, lr: 4.20e-03, grad_scale: 16.0 2022-12-24 02:20:56,175 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0 from training. Duration: 24.32 2022-12-24 02:21:25,013 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94337.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:21:41,693 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.425e+02 3.935e+02 4.692e+02 5.963e+02 1.108e+03, threshold=9.384e+02, percent-clipped=1.0 2022-12-24 02:21:42,152 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=94348.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:21:50,682 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-276745-0093-13116-0_sp0.9 from training. Duration: 21.061125 2022-12-24 02:22:00,331 INFO [train.py:894] (3/4) Epoch 27, batch 3200, loss[loss=0.1607, simple_loss=0.2427, pruned_loss=0.03939, over 18392.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2588, pruned_loss=0.04545, over 3715084.59 frames. ], batch size: 46, lr: 4.20e-03, grad_scale: 16.0 2022-12-24 02:22:04,754 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0024-15855-0_sp0.9 from training. Duration: 20.32225 2022-12-24 02:22:18,566 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp1.1 from training. Duration: 0.7545625 2022-12-24 02:22:20,416 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.3166, 1.6456, 1.8332, 0.9854, 1.2476, 1.9631, 1.8332, 1.5562], device='cuda:3'), covar=tensor([0.0827, 0.0370, 0.0351, 0.0431, 0.0436, 0.0503, 0.0272, 0.0791], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0171, 0.0129, 0.0140, 0.0146, 0.0142, 0.0166, 0.0177], device='cuda:3'), out_proj_covar=tensor([1.1238e-04, 1.2813e-04, 9.5014e-05, 1.0262e-04, 1.0722e-04, 1.0667e-04, 1.2566e-04, 1.3297e-04], device='cuda:3') 2022-12-24 02:22:33,417 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0_sp0.9 from training. Duration: 23.9333125 2022-12-24 02:23:03,411 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp1.1 from training. Duration: 20.17275 2022-12-24 02:23:09,846 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp1.1 from training. Duration: 20.436375 2022-12-24 02:23:14,658 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94409.0, num_to_drop=1, layers_to_drop={3} 2022-12-24 02:23:15,659 INFO [train.py:894] (3/4) Epoch 27, batch 3250, loss[loss=0.18, simple_loss=0.2722, pruned_loss=0.04396, over 18459.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2578, pruned_loss=0.04487, over 3713883.08 frames. ], batch size: 54, lr: 4.20e-03, grad_scale: 16.0 2022-12-24 02:23:24,416 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.6245, 2.1423, 2.2790, 1.2444, 1.6179, 2.3677, 2.2522, 1.9156], device='cuda:3'), covar=tensor([0.0750, 0.0297, 0.0292, 0.0414, 0.0365, 0.0471, 0.0235, 0.0647], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0171, 0.0129, 0.0140, 0.0147, 0.0142, 0.0167, 0.0177], device='cuda:3'), out_proj_covar=tensor([1.1264e-04, 1.2846e-04, 9.5035e-05, 1.0286e-04, 1.0733e-04, 1.0699e-04, 1.2601e-04, 1.3308e-04], device='cuda:3') 2022-12-24 02:23:39,639 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94426.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:24:11,649 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.045e+02 4.265e+02 5.105e+02 6.397e+02 1.288e+03, threshold=1.021e+03, percent-clipped=3.0 2022-12-24 02:24:29,893 INFO [train.py:894] (3/4) Epoch 27, batch 3300, loss[loss=0.1575, simple_loss=0.2322, pruned_loss=0.0414, over 18523.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2577, pruned_loss=0.04493, over 3712891.55 frames. ], batch size: 44, lr: 4.20e-03, grad_scale: 16.0 2022-12-24 02:24:29,947 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0_sp0.9 from training. Duration: 23.1055625 2022-12-24 02:24:33,616 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp1.1 from training. Duration: 23.5 2022-12-24 02:24:43,654 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp0.9 from training. Duration: 26.62775 2022-12-24 02:24:51,271 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=94474.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:24:55,652 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0 from training. Duration: 21.105 2022-12-24 02:25:00,561 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0_sp0.9 from training. Duration: 24.411125 2022-12-24 02:25:01,368 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2022-12-24 02:25:18,758 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2965, 2.1466, 1.6331, 0.5939, 1.5181, 2.1059, 1.7559, 1.9135], device='cuda:3'), covar=tensor([0.0715, 0.0527, 0.1242, 0.1731, 0.1222, 0.1433, 0.1690, 0.0712], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0189, 0.0210, 0.0190, 0.0210, 0.0205, 0.0217, 0.0203], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 02:25:27,307 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp1.1 from training. Duration: 21.263625 2022-12-24 02:25:46,679 INFO [train.py:894] (3/4) Epoch 27, batch 3350, loss[loss=0.1756, simple_loss=0.2379, pruned_loss=0.05667, over 18480.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2585, pruned_loss=0.04521, over 3713445.86 frames. ], batch size: 43, lr: 4.20e-03, grad_scale: 16.0 2022-12-24 02:26:00,143 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0 from training. Duration: 20.795 2022-12-24 02:26:10,775 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0 from training. Duration: 24.76 2022-12-24 02:26:10,790 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0_sp0.9 from training. Duration: 22.25 2022-12-24 02:26:35,075 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2022-12-24 02:26:37,542 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp1.1 from training. Duration: 20.5045625 2022-12-24 02:26:43,831 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.518e+02 3.960e+02 4.769e+02 6.084e+02 1.782e+03, threshold=9.538e+02, percent-clipped=2.0 2022-12-24 02:27:02,428 INFO [train.py:894] (3/4) Epoch 27, batch 3400, loss[loss=0.1808, simple_loss=0.2607, pruned_loss=0.05045, over 18505.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.257, pruned_loss=0.04466, over 3713304.47 frames. ], batch size: 52, lr: 4.20e-03, grad_scale: 16.0 2022-12-24 02:27:29,484 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94579.0, num_to_drop=1, layers_to_drop={2} 2022-12-24 02:27:40,950 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7673, 1.5137, 1.7226, 2.0621, 1.8351, 3.6894, 1.4257, 1.6148], device='cuda:3'), covar=tensor([0.0839, 0.1823, 0.1009, 0.0903, 0.1398, 0.0247, 0.1449, 0.1563], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0083, 0.0072, 0.0075, 0.0091, 0.0077, 0.0085, 0.0077], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2022-12-24 02:28:13,974 INFO [train.py:894] (3/4) Epoch 27, batch 3450, loss[loss=0.2068, simple_loss=0.2803, pruned_loss=0.06666, over 18630.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.258, pruned_loss=0.04542, over 3713554.98 frames. ], batch size: 53, lr: 4.20e-03, grad_scale: 16.0 2022-12-24 02:28:18,736 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6262, 1.4350, 1.4472, 1.7907, 1.8024, 3.3014, 1.3274, 1.5724], device='cuda:3'), covar=tensor([0.0855, 0.1829, 0.1004, 0.0911, 0.1370, 0.0266, 0.1504, 0.1527], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0082, 0.0072, 0.0075, 0.0091, 0.0077, 0.0085, 0.0077], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2022-12-24 02:28:53,011 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94637.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:29:07,393 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.881e+02 4.475e+02 5.396e+02 6.310e+02 1.606e+03, threshold=1.079e+03, percent-clipped=4.0 2022-12-24 02:29:25,870 INFO [train.py:894] (3/4) Epoch 27, batch 3500, loss[loss=0.1954, simple_loss=0.2746, pruned_loss=0.05812, over 18651.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2586, pruned_loss=0.04555, over 3714595.51 frames. ], batch size: 174, lr: 4.19e-03, grad_scale: 16.0 2022-12-24 02:29:46,755 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp1.1 from training. Duration: 22.2954375 2022-12-24 02:29:55,608 INFO [train.py:894] (3/4) Epoch 28, batch 0, loss[loss=0.1931, simple_loss=0.2853, pruned_loss=0.05051, over 18567.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2853, pruned_loss=0.05051, over 18567.00 frames. ], batch size: 57, lr: 4.12e-03, grad_scale: 16.0 2022-12-24 02:29:55,608 INFO [train.py:919] (3/4) Computing validation loss 2022-12-24 02:30:06,233 INFO [train.py:928] (3/4) Epoch 28, validation: loss=0.1635, simple_loss=0.2596, pruned_loss=0.03369, over 944034.00 frames. 2022-12-24 02:30:06,234 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-24 02:30:34,550 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=94685.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:30:57,081 WARNING [train.py:1060] (3/4) Exclude cut with ID 298-126791-0067-24026-0_sp0.9 from training. Duration: 21.438875 2022-12-24 02:31:01,676 WARNING [train.py:1060] (3/4) Exclude cut with ID 5652-39938-0025-23684-0_sp0.9 from training. Duration: 22.2055625 2022-12-24 02:31:03,308 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94704.0, num_to_drop=1, layers_to_drop={1} 2022-12-24 02:31:20,489 INFO [train.py:894] (3/4) Epoch 28, batch 50, loss[loss=0.1563, simple_loss=0.2394, pruned_loss=0.0366, over 18462.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2599, pruned_loss=0.04036, over 836502.87 frames. ], batch size: 43, lr: 4.12e-03, grad_scale: 16.0 2022-12-24 02:32:09,504 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.789e+02 3.263e+02 4.050e+02 5.069e+02 1.061e+03, threshold=8.099e+02, percent-clipped=0.0 2022-12-24 02:32:34,822 INFO [train.py:894] (3/4) Epoch 28, batch 100, loss[loss=0.1682, simple_loss=0.2592, pruned_loss=0.03857, over 18460.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2584, pruned_loss=0.03931, over 1475153.93 frames. ], batch size: 50, lr: 4.11e-03, grad_scale: 16.0 2022-12-24 02:33:48,277 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0556, 1.5959, 1.0292, 1.5356, 2.3030, 1.4146, 1.6618, 1.9350], device='cuda:3'), covar=tensor([0.1434, 0.1858, 0.2079, 0.1454, 0.1581, 0.1742, 0.1405, 0.1521], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0097, 0.0116, 0.0096, 0.0120, 0.0093, 0.0099, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-24 02:33:50,520 INFO [train.py:894] (3/4) Epoch 28, batch 150, loss[loss=0.1794, simple_loss=0.2679, pruned_loss=0.04546, over 18517.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2582, pruned_loss=0.0398, over 1970654.44 frames. ], batch size: 64, lr: 4.11e-03, grad_scale: 16.0 2022-12-24 02:33:54,715 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0375, 1.8521, 1.5484, 1.5039, 1.6879, 1.8758, 1.6745, 1.8116], device='cuda:3'), covar=tensor([0.2412, 0.3354, 0.2139, 0.2885, 0.3675, 0.1178, 0.3054, 0.1203], device='cuda:3'), in_proj_covar=tensor([0.0298, 0.0298, 0.0253, 0.0348, 0.0280, 0.0235, 0.0296, 0.0221], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 02:33:58,521 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0 from training. Duration: 24.525 2022-12-24 02:34:35,165 WARNING [train.py:1060] (3/4) Exclude cut with ID 3699-47246-0007-3408-0_sp0.9 from training. Duration: 20.26675 2022-12-24 02:34:39,510 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.175e+02 2.935e+02 3.702e+02 4.603e+02 1.159e+03, threshold=7.404e+02, percent-clipped=1.0 2022-12-24 02:34:47,201 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.8465, 2.5628, 2.1378, 1.0825, 2.0745, 2.1980, 1.9499, 2.2736], device='cuda:3'), covar=tensor([0.0598, 0.0603, 0.1355, 0.1743, 0.1432, 0.1467, 0.1531, 0.0889], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0190, 0.0211, 0.0191, 0.0211, 0.0205, 0.0217, 0.0204], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 02:34:48,315 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp0.9 from training. Duration: 27.25 2022-12-24 02:35:05,999 INFO [train.py:894] (3/4) Epoch 28, batch 200, loss[loss=0.1656, simple_loss=0.2534, pruned_loss=0.03891, over 18549.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2573, pruned_loss=0.03911, over 2357140.68 frames. ], batch size: 47, lr: 4.11e-03, grad_scale: 16.0 2022-12-24 02:35:25,635 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94879.0, num_to_drop=1, layers_to_drop={2} 2022-12-24 02:35:59,896 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0 from training. Duration: 21.68 2022-12-24 02:36:09,942 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0 from training. Duration: 21.6300625 2022-12-24 02:36:20,718 INFO [train.py:894] (3/4) Epoch 28, batch 250, loss[loss=0.1527, simple_loss=0.2505, pruned_loss=0.02743, over 18583.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2561, pruned_loss=0.03819, over 2657065.86 frames. ], batch size: 56, lr: 4.11e-03, grad_scale: 16.0 2022-12-24 02:36:32,597 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0_sp0.9 from training. Duration: 24.033375 2022-12-24 02:36:37,038 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=94927.0, num_to_drop=1, layers_to_drop={1} 2022-12-24 02:36:45,333 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8803, 1.4582, 0.9753, 1.4523, 2.2213, 1.3218, 1.7311, 1.7932], device='cuda:3'), covar=tensor([0.1566, 0.1914, 0.2037, 0.1455, 0.1675, 0.1786, 0.1296, 0.1676], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0097, 0.0116, 0.0096, 0.0120, 0.0093, 0.0099, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-24 02:36:51,145 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2386, 2.0075, 1.4486, 0.5481, 1.4486, 1.9027, 1.6863, 1.8036], device='cuda:3'), covar=tensor([0.0710, 0.0617, 0.1257, 0.1814, 0.1235, 0.1703, 0.1736, 0.0779], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0189, 0.0209, 0.0189, 0.0210, 0.0204, 0.0215, 0.0203], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 02:37:07,817 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.987e+02 2.888e+02 3.443e+02 4.147e+02 1.178e+03, threshold=6.886e+02, percent-clipped=2.0 2022-12-24 02:37:08,962 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.48 vs. limit=5.0 2022-12-24 02:37:31,384 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0 from training. Duration: 22.905 2022-12-24 02:37:31,417 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp1.1 from training. Duration: 23.4318125 2022-12-24 02:37:34,419 INFO [train.py:894] (3/4) Epoch 28, batch 300, loss[loss=0.1506, simple_loss=0.244, pruned_loss=0.02862, over 18467.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2557, pruned_loss=0.03783, over 2892073.11 frames. ], batch size: 50, lr: 4.11e-03, grad_scale: 16.0 2022-12-24 02:38:31,538 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95004.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:38:33,060 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7456, 1.9267, 2.1511, 1.2457, 1.9918, 2.0490, 1.5579, 2.4840], device='cuda:3'), covar=tensor([0.1351, 0.1957, 0.1393, 0.2101, 0.0861, 0.1286, 0.2593, 0.0643], device='cuda:3'), in_proj_covar=tensor([0.0200, 0.0217, 0.0209, 0.0197, 0.0173, 0.0218, 0.0218, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 02:38:48,898 INFO [train.py:894] (3/4) Epoch 28, batch 350, loss[loss=0.1847, simple_loss=0.28, pruned_loss=0.04465, over 18626.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2546, pruned_loss=0.0378, over 3074614.46 frames. ], batch size: 53, lr: 4.11e-03, grad_scale: 16.0 2022-12-24 02:38:55,525 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5537, 1.9030, 1.5771, 2.2790, 2.4507, 1.5695, 1.5594, 1.2884], device='cuda:3'), covar=tensor([0.1915, 0.1846, 0.1606, 0.1021, 0.1390, 0.1161, 0.2161, 0.1620], device='cuda:3'), in_proj_covar=tensor([0.0250, 0.0231, 0.0222, 0.0203, 0.0264, 0.0199, 0.0230, 0.0204], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 02:39:26,949 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp1.1 from training. Duration: 20.82275 2022-12-24 02:39:28,521 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp0.9 from training. Duration: 25.45 2022-12-24 02:39:35,531 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.241e+02 3.092e+02 3.846e+02 4.799e+02 1.160e+03, threshold=7.692e+02, percent-clipped=9.0 2022-12-24 02:39:41,470 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=95052.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:39:44,960 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2022-12-24 02:40:01,455 INFO [train.py:894] (3/4) Epoch 28, batch 400, loss[loss=0.1607, simple_loss=0.2567, pruned_loss=0.03238, over 18518.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2563, pruned_loss=0.03876, over 3215939.45 frames. ], batch size: 52, lr: 4.11e-03, grad_scale: 16.0 2022-12-24 02:40:17,723 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2022-12-24 02:40:28,892 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0 from training. Duration: 25.775 2022-12-24 02:40:40,802 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6129, 1.5883, 1.7043, 1.5564, 1.1508, 3.0717, 1.1576, 1.6380], device='cuda:3'), covar=tensor([0.3143, 0.2108, 0.1908, 0.2153, 0.1541, 0.0210, 0.1822, 0.0938], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0118, 0.0125, 0.0123, 0.0107, 0.0097, 0.0090, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-24 02:40:51,363 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.41 vs. limit=5.0 2022-12-24 02:40:51,870 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0_sp0.9 from training. Duration: 22.25 2022-12-24 02:40:59,120 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95105.0, num_to_drop=1, layers_to_drop={1} 2022-12-24 02:41:15,542 INFO [train.py:894] (3/4) Epoch 28, batch 450, loss[loss=0.18, simple_loss=0.2713, pruned_loss=0.04434, over 18568.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2575, pruned_loss=0.03901, over 3325765.91 frames. ], batch size: 49, lr: 4.11e-03, grad_scale: 16.0 2022-12-24 02:41:15,609 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0 from training. Duration: 26.205 2022-12-24 02:41:33,403 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp0.9 from training. Duration: 30.1555625 2022-12-24 02:41:39,056 WARNING [train.py:1060] (3/4) Exclude cut with ID 1265-135635-0050-6781-0_sp0.9 from training. Duration: 21.8333125 2022-12-24 02:41:47,817 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp1.1 from training. Duration: 20.6545625 2022-12-24 02:42:03,579 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.356e+02 3.308e+02 4.017e+02 5.104e+02 8.552e+02, threshold=8.035e+02, percent-clipped=1.0 2022-12-24 02:42:27,067 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0045-39920-0_sp0.9 from training. Duration: 20.52225 2022-12-24 02:42:30,401 INFO [train.py:894] (3/4) Epoch 28, batch 500, loss[loss=0.179, simple_loss=0.2707, pruned_loss=0.0437, over 18387.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2572, pruned_loss=0.03877, over 3412128.35 frames. ], batch size: 53, lr: 4.11e-03, grad_scale: 16.0 2022-12-24 02:42:30,816 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95166.0, num_to_drop=1, layers_to_drop={1} 2022-12-24 02:42:47,823 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp0.9 from training. Duration: 29.1166875 2022-12-24 02:42:55,297 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3085, 1.7736, 1.9866, 1.9772, 2.2256, 2.2997, 2.1462, 1.9826], device='cuda:3'), covar=tensor([0.2214, 0.3353, 0.2644, 0.2971, 0.2238, 0.1022, 0.3390, 0.1374], device='cuda:3'), in_proj_covar=tensor([0.0270, 0.0300, 0.0286, 0.0325, 0.0318, 0.0259, 0.0352, 0.0248], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 02:43:40,607 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.0355, 3.1665, 2.1367, 1.7838, 3.4830, 3.5382, 3.0139, 2.5067], device='cuda:3'), covar=tensor([0.0387, 0.0369, 0.0605, 0.0751, 0.0247, 0.0368, 0.0432, 0.0763], device='cuda:3'), in_proj_covar=tensor([0.0125, 0.0131, 0.0130, 0.0118, 0.0104, 0.0127, 0.0133, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 02:43:45,713 INFO [train.py:894] (3/4) Epoch 28, batch 550, loss[loss=0.1788, simple_loss=0.2763, pruned_loss=0.04067, over 18702.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2576, pruned_loss=0.03868, over 3479062.95 frames. ], batch size: 52, lr: 4.11e-03, grad_scale: 16.0 2022-12-24 02:43:46,082 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0286, 1.7498, 1.9846, 2.4169, 2.3132, 4.6912, 1.9293, 2.1776], device='cuda:3'), covar=tensor([0.0716, 0.1728, 0.0991, 0.0869, 0.1257, 0.0149, 0.1259, 0.1383], device='cuda:3'), in_proj_covar=tensor([0.0072, 0.0082, 0.0072, 0.0075, 0.0091, 0.0076, 0.0084, 0.0077], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2022-12-24 02:43:47,094 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133211-0007-59831-0_sp0.9 from training. Duration: 21.388875 2022-12-24 02:44:21,314 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0 from training. Duration: 22.72 2022-12-24 02:44:22,831 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0_sp0.9 from training. Duration: 22.7444375 2022-12-24 02:44:33,105 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.381e+02 3.520e+02 3.982e+02 4.755e+02 1.366e+03, threshold=7.964e+02, percent-clipped=2.0 2022-12-24 02:44:51,890 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95260.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:45:00,340 INFO [train.py:894] (3/4) Epoch 28, batch 600, loss[loss=0.1702, simple_loss=0.2556, pruned_loss=0.04245, over 18531.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2569, pruned_loss=0.03853, over 3530351.82 frames. ], batch size: 47, lr: 4.10e-03, grad_scale: 16.0 2022-12-24 02:45:03,660 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95268.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:45:06,626 WARNING [train.py:1060] (3/4) Exclude cut with ID 4133-6541-0027-40495-0_sp1.1 from training. Duration: 0.9681875 2022-12-24 02:45:10,277 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0_sp0.9 from training. Duration: 22.3166875 2022-12-24 02:45:15,903 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133212-0015-59917-0_sp0.9 from training. Duration: 21.8166875 2022-12-24 02:45:38,772 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2022-12-24 02:46:15,547 INFO [train.py:894] (3/4) Epoch 28, batch 650, loss[loss=0.1847, simple_loss=0.2809, pruned_loss=0.04422, over 18703.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.257, pruned_loss=0.03833, over 3571467.37 frames. ], batch size: 98, lr: 4.10e-03, grad_scale: 16.0 2022-12-24 02:46:23,572 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95321.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:46:29,956 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4459, 1.8849, 1.5041, 2.1110, 2.6994, 1.5530, 1.5840, 1.2233], device='cuda:3'), covar=tensor([0.2218, 0.1989, 0.1890, 0.1195, 0.1237, 0.1256, 0.2307, 0.1812], device='cuda:3'), in_proj_covar=tensor([0.0251, 0.0232, 0.0223, 0.0204, 0.0265, 0.0200, 0.0231, 0.0205], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 02:46:35,502 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95329.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:46:49,754 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95339.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:46:58,431 WARNING [train.py:1060] (3/4) Exclude cut with ID 4957-30119-0041-23990-0_sp0.9 from training. Duration: 20.22775 2022-12-24 02:47:03,200 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.842e+02 3.223e+02 3.713e+02 4.578e+02 8.119e+02, threshold=7.427e+02, percent-clipped=2.0 2022-12-24 02:47:29,690 INFO [train.py:894] (3/4) Epoch 28, batch 700, loss[loss=0.1619, simple_loss=0.2493, pruned_loss=0.03727, over 18433.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2575, pruned_loss=0.03871, over 3603519.15 frames. ], batch size: 48, lr: 4.10e-03, grad_scale: 16.0 2022-12-24 02:47:33,488 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3622, 2.2434, 1.9161, 1.2932, 2.6532, 2.4811, 2.2392, 1.7719], device='cuda:3'), covar=tensor([0.0397, 0.0450, 0.0531, 0.0777, 0.0316, 0.0405, 0.0450, 0.0880], device='cuda:3'), in_proj_covar=tensor([0.0125, 0.0132, 0.0130, 0.0118, 0.0104, 0.0128, 0.0134, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 02:47:42,368 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp1.1 from training. Duration: 24.67275 2022-12-24 02:48:10,435 WARNING [train.py:1060] (3/4) Exclude cut with ID 3082-165428-0081-50734-0_sp0.9 from training. Duration: 21.8055625 2022-12-24 02:48:21,867 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95400.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:48:44,506 INFO [train.py:894] (3/4) Epoch 28, batch 750, loss[loss=0.1658, simple_loss=0.2573, pruned_loss=0.03716, over 18597.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2569, pruned_loss=0.03848, over 3628134.43 frames. ], batch size: 57, lr: 4.10e-03, grad_scale: 16.0 2022-12-24 02:48:47,737 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0_sp0.9 from training. Duration: 22.6666875 2022-12-24 02:49:32,040 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.240e+02 3.241e+02 3.902e+02 5.298e+02 1.359e+03, threshold=7.804e+02, percent-clipped=7.0 2022-12-24 02:49:32,388 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5896, 1.4796, 1.5353, 1.3721, 0.9037, 2.3406, 0.9289, 1.4661], device='cuda:3'), covar=tensor([0.3262, 0.2258, 0.2100, 0.2298, 0.1688, 0.0365, 0.1740, 0.0915], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0118, 0.0125, 0.0123, 0.0107, 0.0097, 0.0090, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-24 02:49:43,141 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6342, 1.4859, 1.4917, 0.8422, 1.6984, 1.6276, 1.5735, 1.2417], device='cuda:3'), covar=tensor([0.0435, 0.0575, 0.0494, 0.0806, 0.0454, 0.0437, 0.0476, 0.1079], device='cuda:3'), in_proj_covar=tensor([0.0126, 0.0132, 0.0131, 0.0118, 0.0105, 0.0128, 0.0134, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 02:49:50,290 WARNING [train.py:1060] (3/4) Exclude cut with ID 2411-132532-0017-82279-0_sp1.1 from training. Duration: 0.9681875 2022-12-24 02:49:51,889 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95461.0, num_to_drop=1, layers_to_drop={2} 2022-12-24 02:50:00,125 INFO [train.py:894] (3/4) Epoch 28, batch 800, loss[loss=0.1935, simple_loss=0.284, pruned_loss=0.05149, over 18472.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2574, pruned_loss=0.03878, over 3647050.92 frames. ], batch size: 54, lr: 4.10e-03, grad_scale: 16.0 2022-12-24 02:50:17,021 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0 from training. Duration: 22.485 2022-12-24 02:50:53,596 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp1.1 from training. Duration: 23.82275 2022-12-24 02:51:05,171 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0 from training. Duration: 20.77 2022-12-24 02:51:13,290 INFO [train.py:894] (3/4) Epoch 28, batch 850, loss[loss=0.1692, simple_loss=0.2608, pruned_loss=0.03879, over 18580.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2576, pruned_loss=0.03892, over 3661604.79 frames. ], batch size: 51, lr: 4.10e-03, grad_scale: 16.0 2022-12-24 02:51:14,691 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0_sp0.9 from training. Duration: 24.088875 2022-12-24 02:51:42,227 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp1.1 from training. Duration: 20.4409375 2022-12-24 02:51:59,546 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.124e+02 3.261e+02 3.961e+02 5.080e+02 1.056e+03, threshold=7.922e+02, percent-clipped=6.0 2022-12-24 02:52:09,381 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0052, 1.8649, 1.5250, 1.5590, 1.6925, 1.8481, 1.6759, 1.7006], device='cuda:3'), covar=tensor([0.2379, 0.3127, 0.2248, 0.2702, 0.3754, 0.1197, 0.3028, 0.1273], device='cuda:3'), in_proj_covar=tensor([0.0300, 0.0301, 0.0255, 0.0350, 0.0282, 0.0236, 0.0299, 0.0222], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 02:52:14,925 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.2767, 2.9095, 2.8440, 3.2388, 2.9854, 2.9307, 3.3779, 1.0584], device='cuda:3'), covar=tensor([0.1033, 0.0863, 0.0915, 0.1030, 0.1667, 0.1280, 0.0876, 0.4779], device='cuda:3'), in_proj_covar=tensor([0.0358, 0.0235, 0.0244, 0.0283, 0.0337, 0.0274, 0.0301, 0.0292], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 02:52:27,516 INFO [train.py:894] (3/4) Epoch 28, batch 900, loss[loss=0.1556, simple_loss=0.2473, pruned_loss=0.03198, over 18725.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2573, pruned_loss=0.03882, over 3672614.07 frames. ], batch size: 52, lr: 4.10e-03, grad_scale: 16.0 2022-12-24 02:52:43,430 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8048, 1.8716, 1.6310, 1.7544, 2.0927, 2.0566, 2.0398, 1.4241], device='cuda:3'), covar=tensor([0.0338, 0.0265, 0.0486, 0.0212, 0.0181, 0.0380, 0.0279, 0.0347], device='cuda:3'), in_proj_covar=tensor([0.0098, 0.0130, 0.0157, 0.0121, 0.0118, 0.0123, 0.0102, 0.0131], device='cuda:3'), out_proj_covar=tensor([7.7244e-05, 1.0245e-04, 1.2787e-04, 9.6001e-05, 9.4439e-05, 9.4462e-05, 7.9182e-05, 1.0281e-04], device='cuda:3') 2022-12-24 02:53:00,658 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0_sp0.9 from training. Duration: 22.511125 2022-12-24 02:53:00,680 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0 from training. Duration: 20.675 2022-12-24 02:53:43,113 INFO [train.py:894] (3/4) Epoch 28, batch 950, loss[loss=0.1653, simple_loss=0.261, pruned_loss=0.03476, over 18712.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2587, pruned_loss=0.0389, over 3682249.00 frames. ], batch size: 60, lr: 4.10e-03, grad_scale: 16.0 2022-12-24 02:53:43,295 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95616.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:53:43,615 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6516, 1.9466, 1.4803, 2.2019, 2.6044, 1.6650, 1.6310, 1.3011], device='cuda:3'), covar=tensor([0.1909, 0.1834, 0.1756, 0.1140, 0.1234, 0.1122, 0.2158, 0.1580], device='cuda:3'), in_proj_covar=tensor([0.0251, 0.0232, 0.0223, 0.0204, 0.0265, 0.0200, 0.0230, 0.0204], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 02:53:45,007 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95617.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:53:52,564 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5650, 1.4911, 2.0040, 3.3742, 2.5586, 2.6588, 1.1778, 2.4793], device='cuda:3'), covar=tensor([0.1960, 0.1540, 0.1444, 0.0554, 0.0941, 0.1118, 0.2020, 0.0976], device='cuda:3'), in_proj_covar=tensor([0.0103, 0.0119, 0.0137, 0.0156, 0.0106, 0.0145, 0.0130, 0.0116], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2022-12-24 02:53:55,226 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95624.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:54:30,165 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.028e+02 3.050e+02 3.792e+02 4.585e+02 1.055e+03, threshold=7.584e+02, percent-clipped=2.0 2022-12-24 02:54:31,529 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2022-12-24 02:54:36,448 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp0.9 from training. Duration: 24.9833125 2022-12-24 02:54:57,701 INFO [train.py:894] (3/4) Epoch 28, batch 1000, loss[loss=0.1684, simple_loss=0.2635, pruned_loss=0.0366, over 18510.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2589, pruned_loss=0.03883, over 3689957.56 frames. ], batch size: 52, lr: 4.10e-03, grad_scale: 16.0 2022-12-24 02:55:08,403 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0 from training. Duration: 27.14 2022-12-24 02:55:15,650 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95678.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:55:24,935 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0 from training. Duration: 22.44 2022-12-24 02:55:40,530 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95695.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:56:11,207 INFO [train.py:894] (3/4) Epoch 28, batch 1050, loss[loss=0.1535, simple_loss=0.2361, pruned_loss=0.03541, over 18532.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2592, pruned_loss=0.03905, over 3695227.91 frames. ], batch size: 47, lr: 4.09e-03, grad_scale: 16.0 2022-12-24 02:56:27,215 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5766, 2.1614, 1.6956, 2.3627, 1.9685, 2.1125, 2.0327, 2.4505], device='cuda:3'), covar=tensor([0.2205, 0.3436, 0.2212, 0.2828, 0.3767, 0.1201, 0.3458, 0.1096], device='cuda:3'), in_proj_covar=tensor([0.0300, 0.0301, 0.0255, 0.0350, 0.0282, 0.0235, 0.0299, 0.0222], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 02:56:38,452 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0060-62364-0_sp0.9 from training. Duration: 21.361125 2022-12-24 02:56:45,134 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp1.1 from training. Duration: 27.0318125 2022-12-24 02:56:55,921 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp0.9 from training. Duration: 28.638875 2022-12-24 02:56:58,766 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.978e+02 3.368e+02 4.220e+02 5.256e+02 1.036e+03, threshold=8.440e+02, percent-clipped=6.0 2022-12-24 02:57:10,906 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0 from training. Duration: 20.4 2022-12-24 02:57:18,850 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95761.0, num_to_drop=1, layers_to_drop={1} 2022-12-24 02:57:25,879 INFO [train.py:894] (3/4) Epoch 28, batch 1100, loss[loss=0.1646, simple_loss=0.2603, pruned_loss=0.03451, over 18677.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2586, pruned_loss=0.03902, over 3698657.08 frames. ], batch size: 99, lr: 4.09e-03, grad_scale: 16.0 2022-12-24 02:57:42,343 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3882, 2.0474, 1.6580, 2.0829, 1.8268, 2.1498, 1.9114, 2.2057], device='cuda:3'), covar=tensor([0.2199, 0.3275, 0.2110, 0.2586, 0.3868, 0.1046, 0.3111, 0.1057], device='cuda:3'), in_proj_covar=tensor([0.0298, 0.0299, 0.0254, 0.0348, 0.0281, 0.0234, 0.0297, 0.0221], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 02:57:43,333 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0 from training. Duration: 20.025 2022-12-24 02:57:43,346 WARNING [train.py:1060] (3/4) Exclude cut with ID 2364-131735-0112-64612-0_sp0.9 from training. Duration: 20.488875 2022-12-24 02:57:49,144 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0 from training. Duration: 29.735 2022-12-24 02:58:30,416 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=95809.0, num_to_drop=1, layers_to_drop={1} 2022-12-24 02:58:34,610 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95812.0, num_to_drop=1, layers_to_drop={1} 2022-12-24 02:58:40,337 INFO [train.py:894] (3/4) Epoch 28, batch 1150, loss[loss=0.1662, simple_loss=0.2466, pruned_loss=0.04292, over 18581.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2578, pruned_loss=0.0388, over 3702728.96 frames. ], batch size: 41, lr: 4.09e-03, grad_scale: 16.0 2022-12-24 02:59:07,924 WARNING [train.py:1060] (3/4) Exclude cut with ID 7276-92427-0014-12983-0_sp0.9 from training. Duration: 21.3055625 2022-12-24 02:59:09,737 WARNING [train.py:1060] (3/4) Exclude cut with ID 1025-75365-0008-79168-0_sp0.9 from training. Duration: 22.0666875 2022-12-24 02:59:25,858 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95847.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 02:59:26,970 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.213e+02 3.423e+02 3.966e+02 5.306e+02 8.843e+02, threshold=7.932e+02, percent-clipped=1.0 2022-12-24 02:59:54,524 INFO [train.py:894] (3/4) Epoch 28, batch 1200, loss[loss=0.17, simple_loss=0.2634, pruned_loss=0.03829, over 18731.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2571, pruned_loss=0.03871, over 3704572.93 frames. ], batch size: 54, lr: 4.09e-03, grad_scale: 16.0 2022-12-24 03:00:04,925 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95873.0, num_to_drop=1, layers_to_drop={3} 2022-12-24 03:00:55,631 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0 from training. Duration: 20.26 2022-12-24 03:00:57,512 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95908.0, num_to_drop=1, layers_to_drop={1} 2022-12-24 03:01:08,841 INFO [train.py:894] (3/4) Epoch 28, batch 1250, loss[loss=0.1526, simple_loss=0.2425, pruned_loss=0.03134, over 18462.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2564, pruned_loss=0.03849, over 3706003.45 frames. ], batch size: 43, lr: 4.09e-03, grad_scale: 16.0 2022-12-24 03:01:09,168 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95916.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:01:09,569 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2022-12-24 03:01:10,317 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0030-9324-0_sp0.9 from training. Duration: 21.3444375 2022-12-24 03:01:19,640 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.20 vs. limit=5.0 2022-12-24 03:01:20,523 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95924.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:01:38,497 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4387, 2.0782, 1.8800, 2.4122, 2.0045, 2.0949, 1.9426, 2.4456], device='cuda:3'), covar=tensor([0.1969, 0.3048, 0.1728, 0.2261, 0.3178, 0.1067, 0.3216, 0.0939], device='cuda:3'), in_proj_covar=tensor([0.0299, 0.0300, 0.0255, 0.0349, 0.0283, 0.0235, 0.0298, 0.0222], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 03:01:55,407 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.867e+02 3.056e+02 3.610e+02 4.347e+02 1.155e+03, threshold=7.221e+02, percent-clipped=3.0 2022-12-24 03:02:09,557 WARNING [train.py:1060] (3/4) Exclude cut with ID 497-129325-0061-62254-0_sp1.1 from training. Duration: 0.97725 2022-12-24 03:02:19,706 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=95964.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:02:22,335 INFO [train.py:894] (3/4) Epoch 28, batch 1300, loss[loss=0.1501, simple_loss=0.2251, pruned_loss=0.03752, over 18523.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2559, pruned_loss=0.03813, over 3707472.80 frames. ], batch size: 44, lr: 4.09e-03, grad_scale: 16.0 2022-12-24 03:02:30,980 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=95972.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:02:32,641 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95973.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:02:34,602 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2022-12-24 03:02:41,404 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.1713, 1.4470, 1.6657, 0.8480, 1.0015, 1.7766, 1.7042, 1.5110], device='cuda:3'), covar=tensor([0.0850, 0.0343, 0.0356, 0.0409, 0.0447, 0.0478, 0.0264, 0.0711], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0173, 0.0131, 0.0141, 0.0148, 0.0142, 0.0166, 0.0178], device='cuda:3'), out_proj_covar=tensor([1.1293e-04, 1.2994e-04, 9.6114e-05, 1.0341e-04, 1.0792e-04, 1.0684e-04, 1.2526e-04, 1.3413e-04], device='cuda:3') 2022-12-24 03:02:42,566 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95980.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:02:53,559 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0_sp0.9 from training. Duration: 22.97225 2022-12-24 03:03:05,003 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95995.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:03:07,951 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5241, 2.4586, 2.0075, 1.4195, 2.7177, 2.6425, 2.2956, 1.8185], device='cuda:3'), covar=tensor([0.0384, 0.0411, 0.0504, 0.0768, 0.0311, 0.0364, 0.0451, 0.0899], device='cuda:3'), in_proj_covar=tensor([0.0125, 0.0132, 0.0130, 0.0119, 0.0104, 0.0128, 0.0134, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 03:03:17,288 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2022-12-24 03:03:29,272 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0047-39922-0_sp0.9 from training. Duration: 21.97775 2022-12-24 03:03:39,139 INFO [train.py:894] (3/4) Epoch 28, batch 1350, loss[loss=0.1767, simple_loss=0.2689, pruned_loss=0.04227, over 18565.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2569, pruned_loss=0.03823, over 3708704.39 frames. ], batch size: 57, lr: 4.09e-03, grad_scale: 16.0 2022-12-24 03:03:42,531 WARNING [train.py:1060] (3/4) Exclude cut with ID 1112-1043-0006-89194-0_sp0.9 from training. Duration: 21.8333125 2022-12-24 03:03:52,525 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0 from training. Duration: 20.47 2022-12-24 03:04:17,357 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96041.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:04:20,239 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=96043.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:04:27,235 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.260e+02 3.035e+02 3.982e+02 4.869e+02 1.123e+03, threshold=7.963e+02, percent-clipped=2.0 2022-12-24 03:04:54,708 INFO [train.py:894] (3/4) Epoch 28, batch 1400, loss[loss=0.1398, simple_loss=0.2244, pruned_loss=0.02763, over 18593.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2573, pruned_loss=0.03852, over 3710037.61 frames. ], batch size: 45, lr: 4.09e-03, grad_scale: 16.0 2022-12-24 03:04:57,598 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0037-39912-0_sp0.9 from training. Duration: 20.67225 2022-12-24 03:05:10,567 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.17 vs. limit=5.0 2022-12-24 03:05:15,989 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp0.9 from training. Duration: 25.2444375 2022-12-24 03:05:16,379 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96080.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:05:24,531 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4948, 2.9069, 2.9291, 1.5609, 3.0354, 3.1004, 2.2139, 3.2114], device='cuda:3'), covar=tensor([0.1257, 0.1508, 0.1317, 0.2302, 0.0653, 0.1063, 0.2051, 0.0565], device='cuda:3'), in_proj_covar=tensor([0.0197, 0.0213, 0.0206, 0.0194, 0.0171, 0.0215, 0.0215, 0.0192], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 03:05:37,382 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0021-76797-0_sp0.9 from training. Duration: 21.1445 2022-12-24 03:06:11,370 INFO [train.py:894] (3/4) Epoch 28, batch 1450, loss[loss=0.1533, simple_loss=0.2373, pruned_loss=0.03468, over 18392.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2567, pruned_loss=0.03806, over 3710274.40 frames. ], batch size: 46, lr: 4.09e-03, grad_scale: 16.0 2022-12-24 03:06:47,640 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp0.9 from training. Duration: 33.038875 2022-12-24 03:06:48,037 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96141.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:06:58,572 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.188e+02 3.324e+02 3.803e+02 4.702e+02 9.088e+02, threshold=7.606e+02, percent-clipped=2.0 2022-12-24 03:07:22,826 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96165.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:07:23,683 INFO [train.py:894] (3/4) Epoch 28, batch 1500, loss[loss=0.1798, simple_loss=0.2743, pruned_loss=0.04265, over 18473.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2569, pruned_loss=0.03807, over 3710608.68 frames. ], batch size: 64, lr: 4.09e-03, grad_scale: 16.0 2022-12-24 03:07:26,530 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64291-0000-16059-0_sp0.9 from training. Duration: 20.0944375 2022-12-24 03:07:26,635 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96168.0, num_to_drop=1, layers_to_drop={0} 2022-12-24 03:07:42,012 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp1.1 from training. Duration: 20.4 2022-12-24 03:07:49,124 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0 from training. Duration: 20.085 2022-12-24 03:07:59,170 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0_sp0.9 from training. Duration: 23.07775 2022-12-24 03:08:19,079 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96203.0, num_to_drop=1, layers_to_drop={2} 2022-12-24 03:08:37,045 INFO [train.py:894] (3/4) Epoch 28, batch 1550, loss[loss=0.1802, simple_loss=0.2727, pruned_loss=0.04383, over 18606.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2571, pruned_loss=0.03813, over 3710697.46 frames. ], batch size: 56, lr: 4.08e-03, grad_scale: 16.0 2022-12-24 03:08:42,529 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp0.9 from training. Duration: 24.9333125 2022-12-24 03:08:49,659 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5978, 1.4086, 1.9792, 3.1458, 2.2844, 2.4392, 1.2481, 2.2857], device='cuda:3'), covar=tensor([0.1826, 0.1527, 0.1356, 0.0560, 0.0937, 0.1282, 0.1859, 0.1029], device='cuda:3'), in_proj_covar=tensor([0.0103, 0.0119, 0.0137, 0.0156, 0.0105, 0.0145, 0.0128, 0.0115], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:3') 2022-12-24 03:08:52,712 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96226.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:08:57,018 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5031, 3.0078, 3.3083, 1.3419, 2.8750, 3.7096, 2.9480, 2.7870], device='cuda:3'), covar=tensor([0.0735, 0.0339, 0.0329, 0.0525, 0.0380, 0.0352, 0.0314, 0.0658], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0174, 0.0130, 0.0141, 0.0148, 0.0143, 0.0166, 0.0178], device='cuda:3'), out_proj_covar=tensor([1.1237e-04, 1.3067e-04, 9.5802e-05, 1.0298e-04, 1.0782e-04, 1.0701e-04, 1.2509e-04, 1.3415e-04], device='cuda:3') 2022-12-24 03:09:24,987 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.260e+02 3.285e+02 3.978e+02 4.636e+02 1.099e+03, threshold=7.955e+02, percent-clipped=3.0 2022-12-24 03:09:26,621 WARNING [train.py:1060] (3/4) Exclude cut with ID 5118-111612-0016-124680-0_sp0.9 from training. Duration: 20.388875 2022-12-24 03:09:33,779 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp1.1 from training. Duration: 20.3590625 2022-12-24 03:09:50,896 INFO [train.py:894] (3/4) Epoch 28, batch 1600, loss[loss=0.1821, simple_loss=0.2717, pruned_loss=0.04623, over 18665.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2562, pruned_loss=0.03778, over 3711387.72 frames. ], batch size: 62, lr: 4.08e-03, grad_scale: 16.0 2022-12-24 03:09:51,695 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2022-12-24 03:09:58,263 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5174, 1.4582, 1.4922, 1.4015, 0.9055, 2.2783, 0.8621, 1.3011], device='cuda:3'), covar=tensor([0.3268, 0.2202, 0.2088, 0.2250, 0.1582, 0.0349, 0.1864, 0.0966], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0118, 0.0126, 0.0124, 0.0108, 0.0096, 0.0090, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-24 03:10:02,722 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96273.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:10:37,213 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.2112, 3.6626, 3.6056, 4.1814, 3.8562, 3.7218, 4.3780, 1.3490], device='cuda:3'), covar=tensor([0.0705, 0.0773, 0.0658, 0.0778, 0.1316, 0.1131, 0.0553, 0.5282], device='cuda:3'), in_proj_covar=tensor([0.0360, 0.0237, 0.0246, 0.0285, 0.0339, 0.0276, 0.0305, 0.0296], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 03:10:40,944 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0_sp1.1 from training. Duration: 0.836375 2022-12-24 03:10:57,169 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7107, 2.3627, 1.9111, 0.7787, 1.8558, 2.2393, 1.9078, 2.1120], device='cuda:3'), covar=tensor([0.0625, 0.0621, 0.1295, 0.1716, 0.1238, 0.1323, 0.1538, 0.0853], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0190, 0.0209, 0.0189, 0.0210, 0.0204, 0.0217, 0.0204], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 03:11:06,338 INFO [train.py:894] (3/4) Epoch 28, batch 1650, loss[loss=0.1347, simple_loss=0.2155, pruned_loss=0.02696, over 18595.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2571, pruned_loss=0.03841, over 3713087.19 frames. ], batch size: 45, lr: 4.08e-03, grad_scale: 32.0 2022-12-24 03:11:14,264 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=96321.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:11:18,732 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96324.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:11:24,085 WARNING [train.py:1060] (3/4) Exclude cut with ID 8565-290391-0049-67394-0_sp0.9 from training. Duration: 21.3166875 2022-12-24 03:11:36,048 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96336.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:11:54,047 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.974e+02 3.462e+02 4.232e+02 5.237e+02 1.043e+03, threshold=8.464e+02, percent-clipped=3.0 2022-12-24 03:11:55,433 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0029-104863-0_sp0.9 from training. Duration: 22.1055625 2022-12-24 03:12:05,636 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2022-12-24 03:12:06,222 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp1.1 from training. Duration: 21.77725 2022-12-24 03:12:19,864 INFO [train.py:894] (3/4) Epoch 28, batch 1700, loss[loss=0.1744, simple_loss=0.2664, pruned_loss=0.04123, over 18502.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.257, pruned_loss=0.03943, over 3713117.27 frames. ], batch size: 52, lr: 4.08e-03, grad_scale: 32.0 2022-12-24 03:12:25,906 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp0.9 from training. Duration: 27.8166875 2022-12-24 03:12:27,592 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5145, 1.4316, 1.3950, 1.4108, 1.7330, 1.6249, 1.5911, 1.2605], device='cuda:3'), covar=tensor([0.0312, 0.0244, 0.0561, 0.0226, 0.0219, 0.0377, 0.0273, 0.0349], device='cuda:3'), in_proj_covar=tensor([0.0098, 0.0130, 0.0157, 0.0123, 0.0119, 0.0124, 0.0103, 0.0132], device='cuda:3'), out_proj_covar=tensor([7.7546e-05, 1.0285e-04, 1.2795e-04, 9.6989e-05, 9.5266e-05, 9.4799e-05, 7.9849e-05, 1.0347e-04], device='cuda:3') 2022-12-24 03:12:46,973 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96385.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:12:48,267 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp1.1 from training. Duration: 22.5090625 2022-12-24 03:12:55,569 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96390.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:12:56,698 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0 from training. Duration: 25.035 2022-12-24 03:13:01,992 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2022-12-24 03:13:14,389 WARNING [train.py:1060] (3/4) Exclude cut with ID 774-127930-0014-10412-0_sp1.1 from training. Duration: 0.95 2022-12-24 03:13:32,284 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp0.9 from training. Duration: 0.92225 2022-12-24 03:13:34,216 INFO [train.py:894] (3/4) Epoch 28, batch 1750, loss[loss=0.184, simple_loss=0.2629, pruned_loss=0.05259, over 18706.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2582, pruned_loss=0.04086, over 3713644.69 frames. ], batch size: 46, lr: 4.08e-03, grad_scale: 32.0 2022-12-24 03:13:50,957 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.48 vs. limit=2.0 2022-12-24 03:13:57,571 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0 from training. Duration: 21.97 2022-12-24 03:14:04,642 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96436.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:14:16,157 WARNING [train.py:1060] (3/4) Exclude cut with ID 7492-105653-0055-62765-0_sp0.9 from training. Duration: 21.97225 2022-12-24 03:14:17,449 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp0.9 from training. Duration: 25.3333125 2022-12-24 03:14:21,833 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.827e+02 4.069e+02 4.851e+02 6.169e+02 1.123e+03, threshold=9.703e+02, percent-clipped=7.0 2022-12-24 03:14:26,401 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96451.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:14:27,436 WARNING [train.py:1060] (3/4) Exclude cut with ID 5172-29468-0015-19128-0_sp0.9 from training. Duration: 21.5055625 2022-12-24 03:14:35,980 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp1.1 from training. Duration: 20.72725 2022-12-24 03:14:48,930 INFO [train.py:894] (3/4) Epoch 28, batch 1800, loss[loss=0.2023, simple_loss=0.2863, pruned_loss=0.05916, over 18596.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2587, pruned_loss=0.04195, over 3713314.02 frames. ], batch size: 78, lr: 4.08e-03, grad_scale: 32.0 2022-12-24 03:14:52,270 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96468.0, num_to_drop=1, layers_to_drop={1} 2022-12-24 03:15:08,561 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp0.9 from training. Duration: 26.32775 2022-12-24 03:15:40,618 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0 from training. Duration: 20.025 2022-12-24 03:15:45,109 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96503.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:15:46,596 WARNING [train.py:1060] (3/4) Exclude cut with ID 6709-74022-0004-86860-0_sp1.1 from training. Duration: 0.9409375 2022-12-24 03:15:46,606 WARNING [train.py:1060] (3/4) Exclude cut with ID 4757-1811-0023-62229-0_sp0.9 from training. Duration: 21.37775 2022-12-24 03:16:05,110 INFO [train.py:894] (3/4) Epoch 28, batch 1850, loss[loss=0.1646, simple_loss=0.2475, pruned_loss=0.04079, over 18591.00 frames. ], tot_loss[loss=0.172, simple_loss=0.258, pruned_loss=0.04294, over 3712292.09 frames. ], batch size: 56, lr: 4.08e-03, grad_scale: 32.0 2022-12-24 03:16:05,275 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=96516.0, num_to_drop=1, layers_to_drop={0} 2022-12-24 03:16:09,501 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0004-25974-0_sp0.9 from training. Duration: 21.17225 2022-12-24 03:16:09,513 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp0.9 from training. Duration: 27.511125 2022-12-24 03:16:12,260 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96521.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:16:42,509 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0 from training. Duration: 22.8 2022-12-24 03:16:46,794 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0 from training. Duration: 22.585 2022-12-24 03:16:52,436 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.179e+02 4.172e+02 5.277e+02 6.658e+02 9.431e+02, threshold=1.055e+03, percent-clipped=0.0 2022-12-24 03:16:56,802 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=96551.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:17:17,024 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0001-146967-0_sp0.9 from training. Duration: 22.0166875 2022-12-24 03:17:20,004 INFO [train.py:894] (3/4) Epoch 28, batch 1900, loss[loss=0.1839, simple_loss=0.2674, pruned_loss=0.05026, over 18649.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2587, pruned_loss=0.04405, over 3713290.27 frames. ], batch size: 69, lr: 4.08e-03, grad_scale: 32.0 2022-12-24 03:17:32,213 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp1.1 from training. Duration: 24.395375 2022-12-24 03:17:39,810 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp0.9 from training. Duration: 27.47775 2022-12-24 03:17:44,629 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp0.9 from training. Duration: 24.8833125 2022-12-24 03:17:47,639 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0 from training. Duration: 23.39 2022-12-24 03:17:53,359 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp0.9 from training. Duration: 28.72225 2022-12-24 03:18:03,455 WARNING [train.py:1060] (3/4) Exclude cut with ID 585-294811-0110-133686-0_sp0.9 from training. Duration: 20.8944375 2022-12-24 03:18:18,248 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0_sp0.9 from training. Duration: 23.8444375 2022-12-24 03:18:36,187 INFO [train.py:894] (3/4) Epoch 28, batch 1950, loss[loss=0.1714, simple_loss=0.2613, pruned_loss=0.04075, over 18666.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2593, pruned_loss=0.04485, over 3713450.55 frames. ], batch size: 60, lr: 4.08e-03, grad_scale: 32.0 2022-12-24 03:18:41,747 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0 from training. Duration: 25.85 2022-12-24 03:18:41,757 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0 from training. Duration: 21.39 2022-12-24 03:18:53,562 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0 from training. Duration: 27.92 2022-12-24 03:19:06,168 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96636.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:19:19,733 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0039-130165-0_sp0.9 from training. Duration: 20.661125 2022-12-24 03:19:23,237 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.734e+02 3.791e+02 4.616e+02 5.491e+02 1.283e+03, threshold=9.233e+02, percent-clipped=2.0 2022-12-24 03:19:44,160 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0043-15874-0_sp0.9 from training. Duration: 20.07225 2022-12-24 03:19:50,084 INFO [train.py:894] (3/4) Epoch 28, batch 2000, loss[loss=0.183, simple_loss=0.2709, pruned_loss=0.04748, over 18730.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2596, pruned_loss=0.04474, over 3713940.78 frames. ], batch size: 54, lr: 4.07e-03, grad_scale: 32.0 2022-12-24 03:19:51,559 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0 from training. Duration: 21.01 2022-12-24 03:20:11,425 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96680.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:20:17,243 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=96684.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:20:51,594 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2022-12-24 03:20:59,230 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0 from training. Duration: 20.65 2022-12-24 03:21:05,004 INFO [train.py:894] (3/4) Epoch 28, batch 2050, loss[loss=0.2095, simple_loss=0.2869, pruned_loss=0.06611, over 18638.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2593, pruned_loss=0.04504, over 3714101.04 frames. ], batch size: 53, lr: 4.07e-03, grad_scale: 32.0 2022-12-24 03:21:07,305 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0 from training. Duration: 21.46 2022-12-24 03:21:17,980 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5665, 1.9265, 1.6396, 2.2849, 2.3120, 1.7159, 1.5122, 1.3920], device='cuda:3'), covar=tensor([0.2054, 0.1825, 0.1632, 0.1049, 0.1306, 0.1129, 0.2296, 0.1594], device='cuda:3'), in_proj_covar=tensor([0.0251, 0.0232, 0.0223, 0.0204, 0.0265, 0.0199, 0.0230, 0.0205], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 03:21:35,335 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96736.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:21:50,886 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96746.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:21:52,002 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0 from training. Duration: 0.92 2022-12-24 03:21:53,231 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.630e+02 4.069e+02 4.932e+02 5.906e+02 1.093e+03, threshold=9.865e+02, percent-clipped=4.0 2022-12-24 03:21:59,000 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0_sp0.9 from training. Duration: 23.7666875 2022-12-24 03:22:19,500 INFO [train.py:894] (3/4) Epoch 28, batch 2100, loss[loss=0.212, simple_loss=0.286, pruned_loss=0.06902, over 18652.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2594, pruned_loss=0.04559, over 3714028.10 frames. ], batch size: 184, lr: 4.07e-03, grad_scale: 32.0 2022-12-24 03:22:26,183 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.0883, 0.9923, 0.9530, 1.1707, 1.2858, 1.1558, 1.0962, 0.9796], device='cuda:3'), covar=tensor([0.0322, 0.0271, 0.0616, 0.0234, 0.0258, 0.0447, 0.0319, 0.0363], device='cuda:3'), in_proj_covar=tensor([0.0098, 0.0130, 0.0156, 0.0123, 0.0119, 0.0124, 0.0103, 0.0132], device='cuda:3'), out_proj_covar=tensor([7.7322e-05, 1.0261e-04, 1.2707e-04, 9.6788e-05, 9.5187e-05, 9.4826e-05, 7.9617e-05, 1.0360e-04], device='cuda:3') 2022-12-24 03:22:35,010 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96776.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:22:36,126 WARNING [train.py:1060] (3/4) Exclude cut with ID 8544-281189-0060-101339-0_sp0.9 from training. Duration: 20.861125 2022-12-24 03:22:42,393 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2022-12-24 03:22:44,747 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0_sp0.9 from training. Duration: 22.711125 2022-12-24 03:22:46,441 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=96784.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:23:26,835 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp1.1 from training. Duration: 22.986375 2022-12-24 03:23:34,599 INFO [train.py:894] (3/4) Epoch 28, batch 2150, loss[loss=0.2019, simple_loss=0.2776, pruned_loss=0.06313, over 18651.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2593, pruned_loss=0.04592, over 3713617.30 frames. ], batch size: 184, lr: 4.07e-03, grad_scale: 16.0 2022-12-24 03:23:42,516 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96821.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:23:43,142 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2022-12-24 03:23:43,490 WARNING [train.py:1060] (3/4) Exclude cut with ID 8040-260924-0003-80960-0_sp0.9 from training. Duration: 22.07225 2022-12-24 03:23:47,735 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0045-26330-0_sp0.9 from training. Duration: 20.3055625 2022-12-24 03:23:49,246 WARNING [train.py:1060] (3/4) Exclude cut with ID 6356-271890-0060-94317-0_sp0.9 from training. Duration: 20.72225 2022-12-24 03:23:56,898 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([5.7045, 4.7721, 4.9808, 5.6654, 5.2463, 5.0822, 5.8537, 1.6470], device='cuda:3'), covar=tensor([0.0638, 0.0743, 0.0541, 0.0724, 0.1297, 0.1014, 0.0387, 0.4947], device='cuda:3'), in_proj_covar=tensor([0.0361, 0.0239, 0.0248, 0.0286, 0.0342, 0.0278, 0.0305, 0.0297], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 03:24:05,649 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96837.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:24:08,654 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp1.1 from training. Duration: 22.4818125 2022-12-24 03:24:24,111 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.321e+02 3.798e+02 4.750e+02 6.031e+02 1.281e+03, threshold=9.499e+02, percent-clipped=3.0 2022-12-24 03:24:31,136 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5956, 3.7159, 3.4928, 1.3212, 3.8187, 2.8535, 0.7212, 2.4500], device='cuda:3'), covar=tensor([0.2083, 0.1384, 0.1558, 0.3866, 0.0905, 0.0926, 0.4885, 0.1508], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0151, 0.0164, 0.0126, 0.0154, 0.0118, 0.0147, 0.0118], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-24 03:24:32,412 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp0.9 from training. Duration: 25.0944375 2022-12-24 03:24:36,077 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0 from training. Duration: 21.515 2022-12-24 03:24:43,883 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp0.9 from training. Duration: 27.02225 2022-12-24 03:24:49,650 INFO [train.py:894] (3/4) Epoch 28, batch 2200, loss[loss=0.135, simple_loss=0.2225, pruned_loss=0.02376, over 18390.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.259, pruned_loss=0.04566, over 3713783.35 frames. ], batch size: 46, lr: 4.07e-03, grad_scale: 16.0 2022-12-24 03:24:49,675 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0010-62480-0_sp0.9 from training. Duration: 22.22225 2022-12-24 03:24:53,015 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3010, 2.0989, 2.0172, 1.2676, 2.6033, 2.3814, 2.1145, 1.8447], device='cuda:3'), covar=tensor([0.0402, 0.0499, 0.0476, 0.0752, 0.0308, 0.0420, 0.0438, 0.0861], device='cuda:3'), in_proj_covar=tensor([0.0126, 0.0132, 0.0130, 0.0118, 0.0104, 0.0128, 0.0134, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 03:24:54,245 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=96869.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:24:55,870 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0085-44554-0_sp0.9 from training. Duration: 20.85 2022-12-24 03:25:30,298 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0 from training. Duration: 21.54 2022-12-24 03:25:34,638 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp1.1 from training. Duration: 20.5318125 2022-12-24 03:25:44,758 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0012-134311-0_sp0.9 from training. Duration: 21.9333125 2022-12-24 03:26:04,709 INFO [train.py:894] (3/4) Epoch 28, batch 2250, loss[loss=0.1665, simple_loss=0.257, pruned_loss=0.03802, over 18510.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2592, pruned_loss=0.04573, over 3714337.22 frames. ], batch size: 52, lr: 4.07e-03, grad_scale: 16.0 2022-12-24 03:26:18,638 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.14 vs. limit=5.0 2022-12-24 03:26:30,508 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0025-130151-0_sp0.9 from training. Duration: 21.7944375 2022-12-24 03:26:43,757 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0_sp0.9 from training. Duration: 22.4666875 2022-12-24 03:26:49,590 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0 from training. Duration: 21.635 2022-12-24 03:26:54,245 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.577e+02 4.074e+02 4.693e+02 5.920e+02 1.379e+03, threshold=9.385e+02, percent-clipped=3.0 2022-12-24 03:26:55,988 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0_sp0.9 from training. Duration: 24.038875 2022-12-24 03:27:19,895 INFO [train.py:894] (3/4) Epoch 28, batch 2300, loss[loss=0.1605, simple_loss=0.2378, pruned_loss=0.04161, over 18500.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2585, pruned_loss=0.04552, over 3714610.17 frames. ], batch size: 43, lr: 4.07e-03, grad_scale: 16.0 2022-12-24 03:27:39,473 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp1.1 from training. Duration: 21.786375 2022-12-24 03:27:41,167 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96980.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:27:52,278 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0 from training. Duration: 20.22 2022-12-24 03:28:10,515 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97000.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:28:33,343 INFO [train.py:894] (3/4) Epoch 28, batch 2350, loss[loss=0.1624, simple_loss=0.2548, pruned_loss=0.035, over 18484.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2592, pruned_loss=0.04588, over 3713246.83 frames. ], batch size: 64, lr: 4.07e-03, grad_scale: 16.0 2022-12-24 03:28:52,061 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=97028.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:29:19,448 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97046.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:29:23,492 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.576e+02 4.083e+02 5.105e+02 6.686e+02 1.127e+03, threshold=1.021e+03, percent-clipped=9.0 2022-12-24 03:29:42,426 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97061.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:29:48,000 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0 from training. Duration: 25.285 2022-12-24 03:29:49,351 INFO [train.py:894] (3/4) Epoch 28, batch 2400, loss[loss=0.236, simple_loss=0.3009, pruned_loss=0.08552, over 18629.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2582, pruned_loss=0.04557, over 3712492.20 frames. ], batch size: 178, lr: 4.07e-03, grad_scale: 16.0 2022-12-24 03:30:21,564 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.2593, 1.3636, 1.5415, 0.9289, 1.3277, 1.4434, 1.2163, 1.6900], device='cuda:3'), covar=tensor([0.1050, 0.1956, 0.1091, 0.1490, 0.0796, 0.0950, 0.2459, 0.0598], device='cuda:3'), in_proj_covar=tensor([0.0198, 0.0215, 0.0208, 0.0195, 0.0171, 0.0218, 0.0217, 0.0194], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 03:30:30,448 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=97094.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:30:35,759 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.51 vs. limit=2.0 2022-12-24 03:30:52,964 WARNING [train.py:1060] (3/4) Exclude cut with ID 811-130148-0001-63453-0_sp0.9 from training. Duration: 20.861125 2022-12-24 03:31:03,779 INFO [train.py:894] (3/4) Epoch 28, batch 2450, loss[loss=0.166, simple_loss=0.2582, pruned_loss=0.03691, over 18533.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2589, pruned_loss=0.04555, over 3712740.41 frames. ], batch size: 55, lr: 4.07e-03, grad_scale: 16.0 2022-12-24 03:31:15,318 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0 from training. Duration: 20.88 2022-12-24 03:31:27,040 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97132.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:31:46,204 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0_sp0.9 from training. Duration: 23.4166875 2022-12-24 03:31:53,012 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.282e+02 3.799e+02 4.558e+02 5.930e+02 1.895e+03, threshold=9.116e+02, percent-clipped=2.0 2022-12-24 03:32:18,244 INFO [train.py:894] (3/4) Epoch 28, batch 2500, loss[loss=0.1522, simple_loss=0.226, pruned_loss=0.03922, over 18477.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2588, pruned_loss=0.0459, over 3712682.13 frames. ], batch size: 43, lr: 4.06e-03, grad_scale: 16.0 2022-12-24 03:33:01,473 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0 from training. Duration: 21.24 2022-12-24 03:33:02,974 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0_sp0.9 from training. Duration: 23.9055625 2022-12-24 03:33:31,820 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6417, 1.5998, 1.6239, 1.6176, 1.1695, 2.9742, 1.1733, 1.8647], device='cuda:3'), covar=tensor([0.3268, 0.2088, 0.2003, 0.2146, 0.1542, 0.0270, 0.1894, 0.0882], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0118, 0.0125, 0.0123, 0.0108, 0.0097, 0.0091, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-24 03:33:32,878 INFO [train.py:894] (3/4) Epoch 28, batch 2550, loss[loss=0.1847, simple_loss=0.2637, pruned_loss=0.05287, over 18544.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2589, pruned_loss=0.04566, over 3713094.93 frames. ], batch size: 57, lr: 4.06e-03, grad_scale: 16.0 2022-12-24 03:33:34,334 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp0.9 from training. Duration: 25.988875 2022-12-24 03:33:43,083 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0001-134300-0_sp0.9 from training. Duration: 20.67225 2022-12-24 03:34:21,826 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.183e+02 3.973e+02 4.624e+02 5.734e+02 1.222e+03, threshold=9.247e+02, percent-clipped=1.0 2022-12-24 03:34:27,746 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0 from training. Duration: 20.34 2022-12-24 03:34:39,360 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97260.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:34:48,187 INFO [train.py:894] (3/4) Epoch 28, batch 2600, loss[loss=0.1669, simple_loss=0.2509, pruned_loss=0.04142, over 18561.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2586, pruned_loss=0.04589, over 3713453.07 frames. ], batch size: 49, lr: 4.06e-03, grad_scale: 16.0 2022-12-24 03:34:54,488 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6357, 3.0279, 3.3913, 1.3326, 3.0071, 3.7241, 3.0133, 2.8531], device='cuda:3'), covar=tensor([0.0750, 0.0413, 0.0304, 0.0549, 0.0347, 0.0374, 0.0332, 0.0691], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0175, 0.0132, 0.0142, 0.0149, 0.0145, 0.0168, 0.0181], device='cuda:3'), out_proj_covar=tensor([1.1373e-04, 1.3127e-04, 9.6700e-05, 1.0390e-04, 1.0906e-04, 1.0907e-04, 1.2711e-04, 1.3569e-04], device='cuda:3') 2022-12-24 03:35:43,776 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2022-12-24 03:35:44,204 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp0.9 from training. Duration: 25.061125 2022-12-24 03:35:55,040 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0 from training. Duration: 0.83 2022-12-24 03:36:04,431 INFO [train.py:894] (3/4) Epoch 28, batch 2650, loss[loss=0.1984, simple_loss=0.2799, pruned_loss=0.05845, over 18603.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2582, pruned_loss=0.0456, over 3713205.91 frames. ], batch size: 53, lr: 4.06e-03, grad_scale: 16.0 2022-12-24 03:36:12,305 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97321.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:36:21,025 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0 from training. Duration: 24.73 2022-12-24 03:36:31,829 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0 from training. Duration: 23.965 2022-12-24 03:36:40,933 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0030-146996-0_sp0.9 from training. Duration: 22.088875 2022-12-24 03:36:44,046 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97342.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:36:51,778 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97347.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:36:54,148 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.790e+02 3.885e+02 4.802e+02 6.141e+02 1.138e+03, threshold=9.603e+02, percent-clipped=2.0 2022-12-24 03:36:57,269 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0_sp0.9 from training. Duration: 23.6 2022-12-24 03:37:05,652 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97356.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:37:20,346 INFO [train.py:894] (3/4) Epoch 28, batch 2700, loss[loss=0.1827, simple_loss=0.2688, pruned_loss=0.04833, over 18647.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2587, pruned_loss=0.04561, over 3712769.15 frames. ], batch size: 78, lr: 4.06e-03, grad_scale: 16.0 2022-12-24 03:37:32,576 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97374.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:38:17,599 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97403.0, num_to_drop=1, layers_to_drop={3} 2022-12-24 03:38:24,743 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97408.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:38:36,577 INFO [train.py:894] (3/4) Epoch 28, batch 2750, loss[loss=0.1825, simple_loss=0.2659, pruned_loss=0.0495, over 18666.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2586, pruned_loss=0.04543, over 3713335.61 frames. ], batch size: 62, lr: 4.06e-03, grad_scale: 16.0 2022-12-24 03:38:39,300 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0 from training. Duration: 23.795 2022-12-24 03:38:57,748 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp1.1 from training. Duration: 21.5409375 2022-12-24 03:38:59,059 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp0.9 from training. Duration: 24.97775 2022-12-24 03:39:00,761 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97432.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:39:05,192 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97435.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:39:10,607 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0_sp0.9 from training. Duration: 23.3444375 2022-12-24 03:39:26,355 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.529e+02 3.785e+02 4.498e+02 5.381e+02 9.968e+02, threshold=8.997e+02, percent-clipped=1.0 2022-12-24 03:39:36,799 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0_sp0.9 from training. Duration: 23.2 2022-12-24 03:39:42,994 WARNING [train.py:1060] (3/4) Exclude cut with ID 5653-46179-0060-117930-0_sp0.9 from training. Duration: 21.17225 2022-12-24 03:39:52,088 INFO [train.py:894] (3/4) Epoch 28, batch 2800, loss[loss=0.1784, simple_loss=0.2707, pruned_loss=0.04304, over 18675.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2587, pruned_loss=0.0454, over 3713261.30 frames. ], batch size: 69, lr: 4.06e-03, grad_scale: 16.0 2022-12-24 03:40:03,079 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp0.9 from training. Duration: 24.6555625 2022-12-24 03:40:13,276 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=97480.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:40:30,830 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.49 vs. limit=2.0 2022-12-24 03:40:58,315 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0 from training. Duration: 20.44 2022-12-24 03:41:06,757 INFO [train.py:894] (3/4) Epoch 28, batch 2850, loss[loss=0.2019, simple_loss=0.2895, pruned_loss=0.05717, over 18585.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2585, pruned_loss=0.04529, over 3713167.37 frames. ], batch size: 57, lr: 4.06e-03, grad_scale: 16.0 2022-12-24 03:41:11,459 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0_sp0.9 from training. Duration: 23.45 2022-12-24 03:41:17,226 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.74 vs. limit=5.0 2022-12-24 03:41:42,351 WARNING [train.py:1060] (3/4) Exclude cut with ID 6945-60535-0076-12784-0_sp0.9 from training. Duration: 20.52225 2022-12-24 03:41:48,207 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6127, 1.0775, 0.7230, 1.2300, 1.9775, 0.6402, 1.2485, 1.4159], device='cuda:3'), covar=tensor([0.1641, 0.2176, 0.1833, 0.1475, 0.1824, 0.1789, 0.1481, 0.1747], device='cuda:3'), in_proj_covar=tensor([0.0095, 0.0097, 0.0117, 0.0096, 0.0121, 0.0093, 0.0099, 0.0095], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-24 03:41:49,321 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0 from training. Duration: 22.19 2022-12-24 03:41:56,692 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.835e+02 4.191e+02 5.018e+02 6.319e+02 1.560e+03, threshold=1.004e+03, percent-clipped=3.0 2022-12-24 03:41:59,639 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp1.1 from training. Duration: 25.3818125 2022-12-24 03:42:06,137 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97555.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:42:13,006 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp0.9 from training. Duration: 28.0944375 2022-12-24 03:42:16,502 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.1827, 3.1641, 2.3112, 1.7016, 3.6299, 3.4779, 3.0643, 2.6510], device='cuda:3'), covar=tensor([0.0380, 0.0398, 0.0563, 0.0790, 0.0220, 0.0367, 0.0445, 0.0727], device='cuda:3'), in_proj_covar=tensor([0.0124, 0.0131, 0.0128, 0.0116, 0.0103, 0.0126, 0.0133, 0.0161], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 03:42:17,883 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0_sp0.9 from training. Duration: 22.9444375 2022-12-24 03:42:22,874 INFO [train.py:894] (3/4) Epoch 28, batch 2900, loss[loss=0.1961, simple_loss=0.2714, pruned_loss=0.06042, over 18618.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2593, pruned_loss=0.04581, over 3713953.49 frames. ], batch size: 180, lr: 4.06e-03, grad_scale: 16.0 2022-12-24 03:42:27,104 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp1.1 from training. Duration: 21.6318125 2022-12-24 03:42:46,845 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0 from training. Duration: 23.695 2022-12-24 03:43:13,168 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0 from training. Duration: 23.955 2022-12-24 03:43:20,860 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4401, 2.5103, 1.8034, 2.9461, 2.7823, 2.3324, 3.4892, 2.5075], device='cuda:3'), covar=tensor([0.0929, 0.1791, 0.2828, 0.1885, 0.1776, 0.0868, 0.0902, 0.1266], device='cuda:3'), in_proj_covar=tensor([0.0185, 0.0220, 0.0260, 0.0294, 0.0245, 0.0198, 0.0209, 0.0211], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 03:43:37,272 INFO [train.py:894] (3/4) Epoch 28, batch 2950, loss[loss=0.1591, simple_loss=0.2346, pruned_loss=0.04181, over 18467.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2584, pruned_loss=0.04505, over 3713761.48 frames. ], batch size: 43, lr: 4.05e-03, grad_scale: 16.0 2022-12-24 03:43:37,455 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97616.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:43:37,669 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97616.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:43:46,381 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp0.9 from training. Duration: 26.438875 2022-12-24 03:44:26,829 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.787e+02 3.883e+02 4.695e+02 5.764e+02 1.290e+03, threshold=9.391e+02, percent-clipped=3.0 2022-12-24 03:44:31,196 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0021-26306-0_sp0.9 from training. Duration: 21.2444375 2022-12-24 03:44:31,222 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp0.9 from training. Duration: 31.02225 2022-12-24 03:44:37,407 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97656.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:44:42,149 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0 from training. Duration: 22.395 2022-12-24 03:44:52,228 INFO [train.py:894] (3/4) Epoch 28, batch 3000, loss[loss=0.1763, simple_loss=0.2645, pruned_loss=0.04405, over 18556.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2581, pruned_loss=0.04488, over 3713281.37 frames. ], batch size: 57, lr: 4.05e-03, grad_scale: 16.0 2022-12-24 03:44:52,229 INFO [train.py:919] (3/4) Computing validation loss 2022-12-24 03:45:02,930 INFO [train.py:928] (3/4) Epoch 28, validation: loss=0.1624, simple_loss=0.2586, pruned_loss=0.03309, over 944034.00 frames. 2022-12-24 03:45:02,931 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-24 03:45:10,307 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0 from training. Duration: 21.075 2022-12-24 03:45:14,617 WARNING [train.py:1060] (3/4) Exclude cut with ID 6482-98857-0025-147532-0_sp0.9 from training. Duration: 20.0055625 2022-12-24 03:45:14,625 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0037-132304-0_sp0.9 from training. Duration: 22.05 2022-12-24 03:45:14,636 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0 from training. Duration: 26.8349375 2022-12-24 03:45:15,016 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97674.0, num_to_drop=1, layers_to_drop={0} 2022-12-24 03:45:18,859 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp1.1 from training. Duration: 22.1090625 2022-12-24 03:45:26,410 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp0.9 from training. Duration: 26.6166875 2022-12-24 03:45:44,602 WARNING [train.py:1060] (3/4) Exclude cut with ID 2046-178027-0000-53705-0_sp0.9 from training. Duration: 20.3055625 2022-12-24 03:45:50,673 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97698.0, num_to_drop=1, layers_to_drop={2} 2022-12-24 03:45:58,313 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97703.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:45:59,742 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=97704.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:46:07,136 WARNING [train.py:1060] (3/4) Exclude cut with ID 7205-50138-0008-5373-0_sp0.9 from training. Duration: 20.7 2022-12-24 03:46:12,055 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7711, 2.0254, 1.7730, 2.3893, 2.6355, 1.7850, 1.7151, 1.4590], device='cuda:3'), covar=tensor([0.1826, 0.1701, 0.1574, 0.1028, 0.1210, 0.1052, 0.2086, 0.1490], device='cuda:3'), in_proj_covar=tensor([0.0252, 0.0231, 0.0223, 0.0205, 0.0265, 0.0199, 0.0231, 0.0205], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 03:46:17,324 INFO [train.py:894] (3/4) Epoch 28, batch 3050, loss[loss=0.1777, simple_loss=0.2636, pruned_loss=0.04597, over 18391.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2573, pruned_loss=0.04455, over 3712716.41 frames. ], batch size: 53, lr: 4.05e-03, grad_scale: 16.0 2022-12-24 03:46:31,005 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.6716, 4.0460, 4.0250, 4.5766, 4.2760, 4.0908, 4.8384, 1.3966], device='cuda:3'), covar=tensor([0.0813, 0.0742, 0.0705, 0.1000, 0.1518, 0.1334, 0.0596, 0.5661], device='cuda:3'), in_proj_covar=tensor([0.0367, 0.0240, 0.0251, 0.0291, 0.0345, 0.0282, 0.0309, 0.0300], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 03:46:38,877 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97730.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:46:46,695 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97735.0, num_to_drop=1, layers_to_drop={3} 2022-12-24 03:46:47,655 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0 from training. Duration: 22.48 2022-12-24 03:47:04,979 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp0.9 from training. Duration: 29.816625 2022-12-24 03:47:08,411 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.276e+02 3.752e+02 4.841e+02 5.897e+02 1.196e+03, threshold=9.682e+02, percent-clipped=5.0 2022-12-24 03:47:11,855 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6221, 2.4752, 2.0775, 1.2191, 1.9881, 2.1865, 1.9280, 2.1843], device='cuda:3'), covar=tensor([0.0590, 0.0536, 0.1183, 0.1520, 0.1171, 0.1234, 0.1477, 0.0761], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0189, 0.0209, 0.0189, 0.0210, 0.0205, 0.0217, 0.0204], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 03:47:24,751 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp1.1 from training. Duration: 22.7590625 2022-12-24 03:47:30,650 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0 from training. Duration: 22.555 2022-12-24 03:47:33,711 INFO [train.py:894] (3/4) Epoch 28, batch 3100, loss[loss=0.1759, simple_loss=0.2707, pruned_loss=0.0406, over 18503.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2577, pruned_loss=0.0448, over 3712169.24 frames. ], batch size: 52, lr: 4.05e-03, grad_scale: 16.0 2022-12-24 03:47:51,735 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0005-25975-0_sp0.9 from training. Duration: 21.688875 2022-12-24 03:47:58,296 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0207, 1.1976, 1.8432, 1.6825, 2.0633, 2.0680, 1.7694, 1.8201], device='cuda:3'), covar=tensor([0.2408, 0.3404, 0.2691, 0.3043, 0.2209, 0.1080, 0.3340, 0.1429], device='cuda:3'), in_proj_covar=tensor([0.0274, 0.0304, 0.0290, 0.0330, 0.0323, 0.0261, 0.0358, 0.0252], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 03:48:09,241 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2595, 2.0458, 1.6305, 2.0234, 1.8194, 2.0096, 1.9233, 2.2466], device='cuda:3'), covar=tensor([0.2313, 0.3036, 0.2129, 0.2553, 0.3609, 0.1162, 0.2918, 0.1096], device='cuda:3'), in_proj_covar=tensor([0.0302, 0.0304, 0.0258, 0.0352, 0.0284, 0.0236, 0.0299, 0.0224], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 03:48:24,231 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0_sp0.9 from training. Duration: 22.6 2022-12-24 03:48:50,865 INFO [train.py:894] (3/4) Epoch 28, batch 3150, loss[loss=0.1895, simple_loss=0.2674, pruned_loss=0.05584, over 18691.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2573, pruned_loss=0.04451, over 3713567.64 frames. ], batch size: 183, lr: 4.05e-03, grad_scale: 16.0 2022-12-24 03:49:01,532 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0 from training. Duration: 24.32 2022-12-24 03:49:41,081 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.574e+02 3.893e+02 4.688e+02 5.671e+02 9.844e+02, threshold=9.376e+02, percent-clipped=1.0 2022-12-24 03:49:58,968 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-276745-0093-13116-0_sp0.9 from training. Duration: 21.061125 2022-12-24 03:50:06,572 INFO [train.py:894] (3/4) Epoch 28, batch 3200, loss[loss=0.1529, simple_loss=0.2394, pruned_loss=0.03319, over 18723.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2582, pruned_loss=0.04499, over 3714414.08 frames. ], batch size: 52, lr: 4.05e-03, grad_scale: 16.0 2022-12-24 03:50:14,044 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0024-15855-0_sp0.9 from training. Duration: 20.32225 2022-12-24 03:50:26,309 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp1.1 from training. Duration: 0.7545625 2022-12-24 03:50:34,275 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7916, 2.4540, 2.1606, 1.0334, 2.1205, 2.0932, 1.6158, 2.2495], device='cuda:3'), covar=tensor([0.0602, 0.0699, 0.1196, 0.1662, 0.1212, 0.1414, 0.1775, 0.0827], device='cuda:3'), in_proj_covar=tensor([0.0177, 0.0189, 0.0209, 0.0189, 0.0210, 0.0205, 0.0218, 0.0204], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 03:50:39,826 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0_sp0.9 from training. Duration: 23.9333125 2022-12-24 03:50:40,209 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97888.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:50:58,439 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1099, 1.3224, 0.5552, 1.7741, 2.1925, 1.7020, 1.6077, 1.8901], device='cuda:3'), covar=tensor([0.2158, 0.2931, 0.2939, 0.1861, 0.2360, 0.2250, 0.2141, 0.2467], device='cuda:3'), in_proj_covar=tensor([0.0095, 0.0098, 0.0118, 0.0097, 0.0122, 0.0093, 0.0100, 0.0095], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-24 03:51:09,486 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2022-12-24 03:51:11,994 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp1.1 from training. Duration: 20.17275 2022-12-24 03:51:14,914 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97911.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:51:17,555 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp1.1 from training. Duration: 20.436375 2022-12-24 03:51:22,062 INFO [train.py:894] (3/4) Epoch 28, batch 3250, loss[loss=0.2201, simple_loss=0.296, pruned_loss=0.07211, over 18555.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2583, pruned_loss=0.04509, over 3713819.73 frames. ], batch size: 177, lr: 4.05e-03, grad_scale: 16.0 2022-12-24 03:51:22,349 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97916.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:52:13,183 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 3.219e+02 4.264e+02 4.990e+02 6.052e+02 1.194e+03, threshold=9.979e+02, percent-clipped=2.0 2022-12-24 03:52:14,235 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97949.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:52:33,134 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0_sp0.9 from training. Duration: 23.1055625 2022-12-24 03:52:35,137 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp1.1 from training. Duration: 23.5 2022-12-24 03:52:36,936 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=97964.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:52:39,766 INFO [train.py:894] (3/4) Epoch 28, batch 3300, loss[loss=0.2077, simple_loss=0.2826, pruned_loss=0.06636, over 18584.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2592, pruned_loss=0.04584, over 3715295.86 frames. ], batch size: 174, lr: 4.05e-03, grad_scale: 16.0 2022-12-24 03:52:44,372 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp0.9 from training. Duration: 26.62775 2022-12-24 03:52:59,154 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0 from training. Duration: 21.105 2022-12-24 03:53:02,171 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0_sp0.9 from training. Duration: 24.411125 2022-12-24 03:53:28,589 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97998.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:53:29,879 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp1.1 from training. Duration: 21.263625 2022-12-24 03:53:39,585 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98003.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:54:00,164 INFO [train.py:894] (3/4) Epoch 28, batch 3350, loss[loss=0.1712, simple_loss=0.2531, pruned_loss=0.04472, over 18517.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2578, pruned_loss=0.04485, over 3714655.91 frames. ], batch size: 47, lr: 4.05e-03, grad_scale: 16.0 2022-12-24 03:54:07,606 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0 from training. Duration: 20.795 2022-12-24 03:54:09,578 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6540, 2.2830, 1.7798, 2.4085, 2.0465, 2.2205, 2.2017, 2.6723], device='cuda:3'), covar=tensor([0.2066, 0.3404, 0.2081, 0.2900, 0.4022, 0.1099, 0.3045, 0.0990], device='cuda:3'), in_proj_covar=tensor([0.0300, 0.0302, 0.0256, 0.0350, 0.0283, 0.0235, 0.0298, 0.0224], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 03:54:16,520 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0 from training. Duration: 24.76 2022-12-24 03:54:16,535 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0_sp0.9 from training. Duration: 22.25 2022-12-24 03:54:21,845 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98030.0, num_to_drop=1, layers_to_drop={3} 2022-12-24 03:54:22,018 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98030.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:54:42,825 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp1.1 from training. Duration: 20.5045625 2022-12-24 03:54:46,008 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=98046.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:54:49,990 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.669e+02 3.930e+02 4.880e+02 5.841e+02 1.288e+03, threshold=9.759e+02, percent-clipped=4.0 2022-12-24 03:54:53,410 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=98051.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:55:15,427 INFO [train.py:894] (3/4) Epoch 28, batch 3400, loss[loss=0.1722, simple_loss=0.2475, pruned_loss=0.04846, over 18609.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2573, pruned_loss=0.04494, over 3714408.36 frames. ], batch size: 45, lr: 4.05e-03, grad_scale: 16.0 2022-12-24 03:55:33,213 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=98078.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:55:33,616 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4200, 2.1398, 1.7591, 1.9389, 1.8863, 2.1147, 2.0022, 2.2292], device='cuda:3'), covar=tensor([0.2081, 0.3375, 0.2115, 0.2729, 0.3762, 0.1084, 0.2976, 0.1129], device='cuda:3'), in_proj_covar=tensor([0.0299, 0.0302, 0.0256, 0.0350, 0.0282, 0.0235, 0.0298, 0.0223], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 03:56:14,232 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98106.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:56:28,565 INFO [train.py:894] (3/4) Epoch 28, batch 3450, loss[loss=0.1908, simple_loss=0.2685, pruned_loss=0.05651, over 18618.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.257, pruned_loss=0.04466, over 3714003.88 frames. ], batch size: 184, lr: 4.04e-03, grad_scale: 16.0 2022-12-24 03:57:15,592 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.690e+02 3.760e+02 4.523e+02 5.539e+02 7.784e+02, threshold=9.046e+02, percent-clipped=1.0 2022-12-24 03:57:40,699 INFO [train.py:894] (3/4) Epoch 28, batch 3500, loss[loss=0.1933, simple_loss=0.2814, pruned_loss=0.05255, over 18608.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2577, pruned_loss=0.04482, over 3713858.15 frames. ], batch size: 98, lr: 4.04e-03, grad_scale: 16.0 2022-12-24 03:57:42,649 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98167.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:58:00,178 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp1.1 from training. Duration: 22.2954375 2022-12-24 03:58:08,569 INFO [train.py:894] (3/4) Epoch 29, batch 0, loss[loss=0.1649, simple_loss=0.2558, pruned_loss=0.03703, over 18464.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2558, pruned_loss=0.03703, over 18464.00 frames. ], batch size: 50, lr: 3.97e-03, grad_scale: 16.0 2022-12-24 03:58:08,569 INFO [train.py:919] (3/4) Computing validation loss 2022-12-24 03:58:19,243 INFO [train.py:928] (3/4) Epoch 29, validation: loss=0.163, simple_loss=0.259, pruned_loss=0.03348, over 944034.00 frames. 2022-12-24 03:58:19,245 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-24 03:59:12,518 WARNING [train.py:1060] (3/4) Exclude cut with ID 298-126791-0067-24026-0_sp0.9 from training. Duration: 21.438875 2022-12-24 03:59:17,020 WARNING [train.py:1060] (3/4) Exclude cut with ID 5652-39938-0025-23684-0_sp0.9 from training. Duration: 22.2055625 2022-12-24 03:59:18,610 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98211.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 03:59:34,736 INFO [train.py:894] (3/4) Epoch 29, batch 50, loss[loss=0.1555, simple_loss=0.239, pruned_loss=0.03603, over 18591.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.259, pruned_loss=0.03975, over 836999.58 frames. ], batch size: 45, lr: 3.97e-03, grad_scale: 16.0 2022-12-24 03:59:36,658 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4822, 2.2130, 2.0551, 2.1623, 2.6231, 3.1793, 2.8735, 2.1349], device='cuda:3'), covar=tensor([0.0381, 0.0301, 0.0429, 0.0239, 0.0228, 0.0321, 0.0291, 0.0323], device='cuda:3'), in_proj_covar=tensor([0.0098, 0.0129, 0.0154, 0.0122, 0.0118, 0.0123, 0.0101, 0.0129], device='cuda:3'), out_proj_covar=tensor([7.7267e-05, 1.0209e-04, 1.2621e-04, 9.6400e-05, 9.4068e-05, 9.4295e-05, 7.8484e-05, 1.0157e-04], device='cuda:3') 2022-12-24 03:59:52,763 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2022-12-24 04:00:08,511 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98244.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:00:15,073 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.323e+02 3.149e+02 4.026e+02 5.359e+02 1.218e+03, threshold=8.052e+02, percent-clipped=4.0 2022-12-24 04:00:29,802 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=98259.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:00:38,980 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3169, 2.5855, 2.8871, 0.8699, 2.6223, 3.3957, 2.4728, 2.5317], device='cuda:3'), covar=tensor([0.0903, 0.0458, 0.0429, 0.0679, 0.0394, 0.0480, 0.0428, 0.0784], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0174, 0.0131, 0.0143, 0.0150, 0.0144, 0.0168, 0.0180], device='cuda:3'), out_proj_covar=tensor([1.1340e-04, 1.3037e-04, 9.6099e-05, 1.0424e-04, 1.0951e-04, 1.0815e-04, 1.2659e-04, 1.3515e-04], device='cuda:3') 2022-12-24 04:00:49,564 INFO [train.py:894] (3/4) Epoch 29, batch 100, loss[loss=0.1765, simple_loss=0.2766, pruned_loss=0.0382, over 18572.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2588, pruned_loss=0.03959, over 1474611.45 frames. ], batch size: 57, lr: 3.97e-03, grad_scale: 16.0 2022-12-24 04:01:37,758 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2022-12-24 04:02:06,027 INFO [train.py:894] (3/4) Epoch 29, batch 150, loss[loss=0.1715, simple_loss=0.2647, pruned_loss=0.03918, over 18684.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2551, pruned_loss=0.03784, over 1968877.97 frames. ], batch size: 78, lr: 3.97e-03, grad_scale: 16.0 2022-12-24 04:02:18,101 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0 from training. Duration: 24.525 2022-12-24 04:02:18,334 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98330.0, num_to_drop=1, layers_to_drop={2} 2022-12-24 04:02:47,186 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.956e+02 3.151e+02 3.628e+02 4.247e+02 8.996e+02, threshold=7.256e+02, percent-clipped=1.0 2022-12-24 04:02:49,006 WARNING [train.py:1060] (3/4) Exclude cut with ID 3699-47246-0007-3408-0_sp0.9 from training. Duration: 20.26675 2022-12-24 04:03:01,286 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp0.9 from training. Duration: 27.25 2022-12-24 04:03:23,392 INFO [train.py:894] (3/4) Epoch 29, batch 200, loss[loss=0.1695, simple_loss=0.2698, pruned_loss=0.03459, over 18606.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2548, pruned_loss=0.0376, over 2356192.17 frames. ], batch size: 56, lr: 3.97e-03, grad_scale: 16.0 2022-12-24 04:03:32,499 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=98378.0, num_to_drop=1, layers_to_drop={1} 2022-12-24 04:04:18,030 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0 from training. Duration: 21.68 2022-12-24 04:04:19,586 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98410.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:04:29,331 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0 from training. Duration: 21.6300625 2022-12-24 04:04:37,731 INFO [train.py:894] (3/4) Epoch 29, batch 250, loss[loss=0.1499, simple_loss=0.2354, pruned_loss=0.03222, over 18687.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2526, pruned_loss=0.03668, over 2657030.31 frames. ], batch size: 48, lr: 3.97e-03, grad_scale: 16.0 2022-12-24 04:04:45,509 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2022-12-24 04:04:54,941 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0_sp0.9 from training. Duration: 24.033375 2022-12-24 04:04:58,710 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2022-12-24 04:05:17,892 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.102e+02 2.945e+02 3.476e+02 4.234e+02 1.493e+03, threshold=6.951e+02, percent-clipped=2.0 2022-12-24 04:05:37,606 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98462.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:05:50,820 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0 from training. Duration: 22.905 2022-12-24 04:05:52,352 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp1.1 from training. Duration: 23.4318125 2022-12-24 04:05:52,827 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98471.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:05:53,823 INFO [train.py:894] (3/4) Epoch 29, batch 300, loss[loss=0.1391, simple_loss=0.2296, pruned_loss=0.02434, over 18388.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2526, pruned_loss=0.03693, over 2891677.20 frames. ], batch size: 46, lr: 3.97e-03, grad_scale: 16.0 2022-12-24 04:06:11,960 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2022-12-24 04:07:07,861 INFO [train.py:894] (3/4) Epoch 29, batch 350, loss[loss=0.1833, simple_loss=0.2793, pruned_loss=0.04367, over 18512.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2523, pruned_loss=0.03672, over 3072779.01 frames. ], batch size: 64, lr: 3.96e-03, grad_scale: 16.0 2022-12-24 04:07:16,875 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2372, 2.2808, 2.3817, 2.2950, 2.1143, 4.0411, 2.3827, 2.6504], device='cuda:3'), covar=tensor([0.2565, 0.1702, 0.1456, 0.1674, 0.1054, 0.0156, 0.1506, 0.0759], device='cuda:3'), in_proj_covar=tensor([0.0131, 0.0117, 0.0125, 0.0122, 0.0106, 0.0096, 0.0090, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-24 04:07:39,680 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98543.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:07:40,886 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98544.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:07:42,439 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9318, 1.8313, 1.7408, 1.7476, 2.1589, 2.0914, 2.0797, 1.5227], device='cuda:3'), covar=tensor([0.0341, 0.0253, 0.0453, 0.0218, 0.0178, 0.0386, 0.0280, 0.0332], device='cuda:3'), in_proj_covar=tensor([0.0099, 0.0130, 0.0156, 0.0123, 0.0119, 0.0125, 0.0103, 0.0130], device='cuda:3'), out_proj_covar=tensor([7.7909e-05, 1.0295e-04, 1.2786e-04, 9.7303e-05, 9.5186e-05, 9.5487e-05, 7.9538e-05, 1.0244e-04], device='cuda:3') 2022-12-24 04:07:46,956 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7935, 1.7615, 1.3973, 1.7083, 1.9194, 1.6932, 2.0883, 1.9294], device='cuda:3'), covar=tensor([0.0946, 0.1739, 0.2857, 0.1709, 0.1925, 0.0939, 0.1022, 0.1247], device='cuda:3'), in_proj_covar=tensor([0.0183, 0.0217, 0.0257, 0.0291, 0.0242, 0.0195, 0.0205, 0.0209], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 04:07:47,848 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.942e+02 3.244e+02 3.661e+02 4.280e+02 7.325e+02, threshold=7.323e+02, percent-clipped=2.0 2022-12-24 04:07:52,145 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp1.1 from training. Duration: 20.82275 2022-12-24 04:07:52,194 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp0.9 from training. Duration: 25.45 2022-12-24 04:08:22,543 INFO [train.py:894] (3/4) Epoch 29, batch 400, loss[loss=0.1637, simple_loss=0.2624, pruned_loss=0.03252, over 18704.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2536, pruned_loss=0.03719, over 3215813.00 frames. ], batch size: 98, lr: 3.96e-03, grad_scale: 16.0 2022-12-24 04:08:32,747 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8839, 0.7136, 1.7198, 1.5074, 1.9272, 1.9550, 1.6076, 1.7634], device='cuda:3'), covar=tensor([0.2246, 0.3446, 0.2631, 0.2726, 0.2123, 0.1048, 0.3060, 0.1368], device='cuda:3'), in_proj_covar=tensor([0.0273, 0.0302, 0.0289, 0.0328, 0.0321, 0.0261, 0.0357, 0.0251], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 04:08:52,266 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=98592.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:08:53,650 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0 from training. Duration: 25.775 2022-12-24 04:09:10,220 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98604.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:09:16,405 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0_sp0.9 from training. Duration: 22.25 2022-12-24 04:09:37,589 INFO [train.py:894] (3/4) Epoch 29, batch 450, loss[loss=0.1707, simple_loss=0.2616, pruned_loss=0.03991, over 18421.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2541, pruned_loss=0.03749, over 3326294.93 frames. ], batch size: 48, lr: 3.96e-03, grad_scale: 16.0 2022-12-24 04:09:43,371 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0 from training. Duration: 26.205 2022-12-24 04:10:02,910 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp0.9 from training. Duration: 30.1555625 2022-12-24 04:10:08,648 WARNING [train.py:1060] (3/4) Exclude cut with ID 1265-135635-0050-6781-0_sp0.9 from training. Duration: 21.8333125 2022-12-24 04:10:16,398 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp1.1 from training. Duration: 20.6545625 2022-12-24 04:10:17,782 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.919e+02 3.115e+02 3.804e+02 4.610e+02 9.757e+02, threshold=7.608e+02, percent-clipped=2.0 2022-12-24 04:10:53,238 INFO [train.py:894] (3/4) Epoch 29, batch 500, loss[loss=0.1404, simple_loss=0.2223, pruned_loss=0.02927, over 18523.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2541, pruned_loss=0.03765, over 3411491.88 frames. ], batch size: 41, lr: 3.96e-03, grad_scale: 16.0 2022-12-24 04:10:57,548 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0045-39920-0_sp0.9 from training. Duration: 20.52225 2022-12-24 04:11:18,705 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp0.9 from training. Duration: 29.1166875 2022-12-24 04:12:08,558 INFO [train.py:894] (3/4) Epoch 29, batch 550, loss[loss=0.1479, simple_loss=0.2307, pruned_loss=0.03261, over 18540.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.255, pruned_loss=0.03782, over 3478894.75 frames. ], batch size: 41, lr: 3.96e-03, grad_scale: 16.0 2022-12-24 04:12:17,581 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133211-0007-59831-0_sp0.9 from training. Duration: 21.388875 2022-12-24 04:12:49,076 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.526e+02 3.198e+02 3.726e+02 4.559e+02 9.672e+02, threshold=7.451e+02, percent-clipped=1.0 2022-12-24 04:12:52,787 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0 from training. Duration: 22.72 2022-12-24 04:12:53,943 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0_sp0.9 from training. Duration: 22.7444375 2022-12-24 04:13:10,034 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98762.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:13:15,400 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98766.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:13:21,804 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.8704, 3.3550, 3.3532, 3.8038, 3.5449, 3.3792, 4.0068, 1.2664], device='cuda:3'), covar=tensor([0.0811, 0.0827, 0.0788, 0.0799, 0.1504, 0.1294, 0.0706, 0.5165], device='cuda:3'), in_proj_covar=tensor([0.0365, 0.0238, 0.0249, 0.0286, 0.0339, 0.0278, 0.0306, 0.0296], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 04:13:24,497 INFO [train.py:894] (3/4) Epoch 29, batch 600, loss[loss=0.1616, simple_loss=0.2566, pruned_loss=0.03326, over 18722.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.257, pruned_loss=0.03871, over 3531000.72 frames. ], batch size: 52, lr: 3.96e-03, grad_scale: 16.0 2022-12-24 04:13:33,238 WARNING [train.py:1060] (3/4) Exclude cut with ID 4133-6541-0027-40495-0_sp1.1 from training. Duration: 0.9681875 2022-12-24 04:13:37,709 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0_sp0.9 from training. Duration: 22.3166875 2022-12-24 04:13:43,972 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133212-0015-59917-0_sp0.9 from training. Duration: 21.8166875 2022-12-24 04:13:48,955 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8193, 1.7079, 1.3913, 1.6434, 1.8936, 1.6958, 2.1249, 1.8819], device='cuda:3'), covar=tensor([0.0907, 0.1801, 0.2849, 0.1773, 0.1918, 0.0955, 0.1032, 0.1331], device='cuda:3'), in_proj_covar=tensor([0.0184, 0.0218, 0.0258, 0.0292, 0.0243, 0.0196, 0.0206, 0.0211], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 04:13:48,988 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7572, 0.7201, 1.5470, 1.4479, 1.7781, 1.8382, 1.4730, 1.6245], device='cuda:3'), covar=tensor([0.2907, 0.4154, 0.3493, 0.3513, 0.2875, 0.1378, 0.4172, 0.1836], device='cuda:3'), in_proj_covar=tensor([0.0271, 0.0299, 0.0286, 0.0326, 0.0320, 0.0258, 0.0354, 0.0249], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 04:14:23,236 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=98810.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:14:40,756 INFO [train.py:894] (3/4) Epoch 29, batch 650, loss[loss=0.1684, simple_loss=0.263, pruned_loss=0.03696, over 18586.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2577, pruned_loss=0.039, over 3572277.36 frames. ], batch size: 56, lr: 3.96e-03, grad_scale: 32.0 2022-12-24 04:15:06,110 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.2110, 1.4760, 1.7041, 0.9638, 0.9857, 1.7899, 1.7536, 1.5796], device='cuda:3'), covar=tensor([0.0852, 0.0349, 0.0365, 0.0421, 0.0492, 0.0570, 0.0260, 0.0733], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0175, 0.0132, 0.0143, 0.0150, 0.0144, 0.0168, 0.0181], device='cuda:3'), out_proj_covar=tensor([1.1366e-04, 1.3091e-04, 9.6700e-05, 1.0478e-04, 1.0937e-04, 1.0824e-04, 1.2653e-04, 1.3598e-04], device='cuda:3') 2022-12-24 04:15:21,393 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.842e+02 3.342e+02 3.913e+02 4.819e+02 8.981e+02, threshold=7.826e+02, percent-clipped=4.0 2022-12-24 04:15:25,665 WARNING [train.py:1060] (3/4) Exclude cut with ID 4957-30119-0041-23990-0_sp0.9 from training. Duration: 20.22775 2022-12-24 04:15:55,855 INFO [train.py:894] (3/4) Epoch 29, batch 700, loss[loss=0.1508, simple_loss=0.2366, pruned_loss=0.03255, over 18673.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2583, pruned_loss=0.03909, over 3603605.32 frames. ], batch size: 41, lr: 3.96e-03, grad_scale: 16.0 2022-12-24 04:16:09,376 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp1.1 from training. Duration: 24.67275 2022-12-24 04:16:22,659 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98890.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:16:30,565 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98895.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:16:36,258 WARNING [train.py:1060] (3/4) Exclude cut with ID 3082-165428-0081-50734-0_sp0.9 from training. Duration: 21.8055625 2022-12-24 04:16:36,366 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98899.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:16:52,308 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9419, 1.6177, 2.0612, 2.1864, 1.9806, 4.4265, 1.7128, 1.8574], device='cuda:3'), covar=tensor([0.0791, 0.1774, 0.1014, 0.0919, 0.1337, 0.0174, 0.1393, 0.1519], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0083, 0.0073, 0.0075, 0.0091, 0.0077, 0.0085, 0.0078], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2022-12-24 04:17:10,930 INFO [train.py:894] (3/4) Epoch 29, batch 750, loss[loss=0.1763, simple_loss=0.2713, pruned_loss=0.04069, over 18635.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.259, pruned_loss=0.03923, over 3629178.89 frames. ], batch size: 62, lr: 3.96e-03, grad_scale: 16.0 2022-12-24 04:17:12,298 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0_sp0.9 from training. Duration: 22.6666875 2022-12-24 04:17:53,677 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.092e+02 3.312e+02 3.981e+02 4.974e+02 8.508e+02, threshold=7.962e+02, percent-clipped=2.0 2022-12-24 04:17:56,128 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98951.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:18:03,414 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98956.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:18:14,983 WARNING [train.py:1060] (3/4) Exclude cut with ID 2411-132532-0017-82279-0_sp1.1 from training. Duration: 0.9681875 2022-12-24 04:18:25,317 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98971.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:18:26,282 INFO [train.py:894] (3/4) Epoch 29, batch 800, loss[loss=0.1633, simple_loss=0.2498, pruned_loss=0.03833, over 18704.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2584, pruned_loss=0.03894, over 3648254.64 frames. ], batch size: 50, lr: 3.96e-03, grad_scale: 16.0 2022-12-24 04:18:32,718 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2022-12-24 04:18:41,345 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0 from training. Duration: 22.485 2022-12-24 04:19:19,625 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp1.1 from training. Duration: 23.82275 2022-12-24 04:19:34,177 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0 from training. Duration: 20.77 2022-12-24 04:19:40,110 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0_sp0.9 from training. Duration: 24.088875 2022-12-24 04:19:41,614 INFO [train.py:894] (3/4) Epoch 29, batch 850, loss[loss=0.1349, simple_loss=0.2198, pruned_loss=0.02495, over 18496.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2585, pruned_loss=0.03897, over 3661705.61 frames. ], batch size: 43, lr: 3.95e-03, grad_scale: 16.0 2022-12-24 04:19:57,600 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99032.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:20:12,093 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp1.1 from training. Duration: 20.4409375 2022-12-24 04:20:13,948 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5985, 2.0695, 1.6423, 2.3118, 2.4611, 1.5927, 1.7170, 1.2933], device='cuda:3'), covar=tensor([0.2149, 0.1851, 0.1785, 0.1133, 0.1578, 0.1338, 0.2296, 0.1822], device='cuda:3'), in_proj_covar=tensor([0.0248, 0.0231, 0.0223, 0.0203, 0.0263, 0.0198, 0.0229, 0.0205], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 04:20:24,862 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.082e+02 3.230e+02 3.807e+02 4.729e+02 8.839e+02, threshold=7.614e+02, percent-clipped=1.0 2022-12-24 04:20:26,602 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99051.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:20:47,150 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99066.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:20:55,347 INFO [train.py:894] (3/4) Epoch 29, batch 900, loss[loss=0.177, simple_loss=0.2663, pruned_loss=0.04383, over 18562.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2586, pruned_loss=0.03906, over 3671920.27 frames. ], batch size: 77, lr: 3.95e-03, grad_scale: 16.0 2022-12-24 04:21:28,305 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0_sp0.9 from training. Duration: 22.511125 2022-12-24 04:21:28,330 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0 from training. Duration: 20.675 2022-12-24 04:21:55,092 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99112.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:21:57,799 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99114.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:22:09,356 INFO [train.py:894] (3/4) Epoch 29, batch 950, loss[loss=0.1606, simple_loss=0.2565, pruned_loss=0.03237, over 18671.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2579, pruned_loss=0.03866, over 3680562.60 frames. ], batch size: 97, lr: 3.95e-03, grad_scale: 16.0 2022-12-24 04:22:51,783 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.071e+02 3.012e+02 3.640e+02 4.456e+02 1.368e+03, threshold=7.280e+02, percent-clipped=4.0 2022-12-24 04:23:02,332 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3663, 2.0832, 1.9770, 1.2070, 2.5193, 2.4114, 2.1175, 1.7490], device='cuda:3'), covar=tensor([0.0357, 0.0468, 0.0468, 0.0815, 0.0337, 0.0368, 0.0438, 0.0843], device='cuda:3'), in_proj_covar=tensor([0.0125, 0.0131, 0.0128, 0.0116, 0.0104, 0.0128, 0.0134, 0.0161], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 04:23:04,794 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp0.9 from training. Duration: 24.9833125 2022-12-24 04:23:05,238 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8656, 0.7686, 1.7068, 1.4665, 1.9229, 1.9592, 1.5845, 1.7630], device='cuda:3'), covar=tensor([0.2431, 0.3561, 0.2793, 0.2985, 0.2256, 0.1081, 0.3358, 0.1432], device='cuda:3'), in_proj_covar=tensor([0.0272, 0.0301, 0.0288, 0.0328, 0.0321, 0.0259, 0.0356, 0.0249], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 04:23:12,022 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5545, 2.5429, 2.0635, 2.3246, 2.8166, 3.3681, 3.0280, 2.2765], device='cuda:3'), covar=tensor([0.0394, 0.0318, 0.0493, 0.0256, 0.0241, 0.0289, 0.0302, 0.0336], device='cuda:3'), in_proj_covar=tensor([0.0099, 0.0131, 0.0158, 0.0125, 0.0120, 0.0126, 0.0104, 0.0132], device='cuda:3'), out_proj_covar=tensor([7.8119e-05, 1.0337e-04, 1.2877e-04, 9.8752e-05, 9.5795e-05, 9.6317e-05, 8.0173e-05, 1.0344e-04], device='cuda:3') 2022-12-24 04:23:23,376 INFO [train.py:894] (3/4) Epoch 29, batch 1000, loss[loss=0.1601, simple_loss=0.2605, pruned_loss=0.0299, over 18651.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2572, pruned_loss=0.03807, over 3686971.36 frames. ], batch size: 99, lr: 3.95e-03, grad_scale: 16.0 2022-12-24 04:23:36,942 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0 from training. Duration: 27.14 2022-12-24 04:23:49,143 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9548, 2.3786, 1.9769, 2.6352, 3.2699, 1.9345, 2.1346, 1.5901], device='cuda:3'), covar=tensor([0.1812, 0.1722, 0.1516, 0.1036, 0.1167, 0.1049, 0.1843, 0.1515], device='cuda:3'), in_proj_covar=tensor([0.0247, 0.0230, 0.0222, 0.0202, 0.0261, 0.0197, 0.0227, 0.0203], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 04:23:53,443 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0 from training. Duration: 22.44 2022-12-24 04:24:04,830 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99199.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:24:37,474 INFO [train.py:894] (3/4) Epoch 29, batch 1050, loss[loss=0.1671, simple_loss=0.2619, pruned_loss=0.03611, over 18727.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2574, pruned_loss=0.03815, over 3694096.68 frames. ], batch size: 52, lr: 3.95e-03, grad_scale: 16.0 2022-12-24 04:24:48,913 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99229.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:24:52,387 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8188, 2.1757, 1.8438, 2.4473, 3.1871, 1.8044, 1.9959, 1.4848], device='cuda:3'), covar=tensor([0.1905, 0.1735, 0.1537, 0.1051, 0.1227, 0.1087, 0.1918, 0.1550], device='cuda:3'), in_proj_covar=tensor([0.0248, 0.0231, 0.0222, 0.0202, 0.0262, 0.0198, 0.0227, 0.0203], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 04:25:09,371 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6028, 1.4217, 1.4497, 0.7739, 1.6723, 1.5144, 1.4970, 1.2643], device='cuda:3'), covar=tensor([0.0429, 0.0626, 0.0495, 0.0864, 0.0515, 0.0467, 0.0477, 0.1079], device='cuda:3'), in_proj_covar=tensor([0.0125, 0.0131, 0.0129, 0.0116, 0.0104, 0.0128, 0.0135, 0.0161], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 04:25:12,492 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0060-62364-0_sp0.9 from training. Duration: 21.361125 2022-12-24 04:25:15,433 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99246.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:25:16,894 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99247.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:25:18,227 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp1.1 from training. Duration: 27.0318125 2022-12-24 04:25:20,995 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.030e+02 3.056e+02 3.707e+02 4.620e+02 9.577e+02, threshold=7.414e+02, percent-clipped=2.0 2022-12-24 04:25:22,708 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99251.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:25:27,974 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp0.9 from training. Duration: 28.638875 2022-12-24 04:25:41,318 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6786, 1.6159, 1.5074, 1.5870, 1.8821, 1.8117, 1.8446, 1.3037], device='cuda:3'), covar=tensor([0.0316, 0.0237, 0.0482, 0.0209, 0.0197, 0.0424, 0.0248, 0.0354], device='cuda:3'), in_proj_covar=tensor([0.0099, 0.0130, 0.0157, 0.0124, 0.0120, 0.0125, 0.0103, 0.0131], device='cuda:3'), out_proj_covar=tensor([7.7792e-05, 1.0278e-04, 1.2821e-04, 9.8336e-05, 9.5539e-05, 9.5866e-05, 7.9759e-05, 1.0316e-04], device='cuda:3') 2022-12-24 04:25:42,293 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0 from training. Duration: 20.4 2022-12-24 04:25:46,950 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99268.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:25:52,968 INFO [train.py:894] (3/4) Epoch 29, batch 1100, loss[loss=0.1603, simple_loss=0.2415, pruned_loss=0.03958, over 18676.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2575, pruned_loss=0.03817, over 3699038.76 frames. ], batch size: 46, lr: 3.95e-03, grad_scale: 16.0 2022-12-24 04:25:57,786 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1574, 1.8594, 1.8450, 1.1130, 2.2454, 2.1138, 1.9242, 1.5581], device='cuda:3'), covar=tensor([0.0376, 0.0550, 0.0464, 0.0829, 0.0392, 0.0402, 0.0442, 0.0960], device='cuda:3'), in_proj_covar=tensor([0.0125, 0.0131, 0.0129, 0.0116, 0.0105, 0.0128, 0.0135, 0.0161], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 04:26:17,132 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0 from training. Duration: 20.025 2022-12-24 04:26:17,146 WARNING [train.py:1060] (3/4) Exclude cut with ID 2364-131735-0112-64612-0_sp0.9 from training. Duration: 20.488875 2022-12-24 04:26:20,245 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0 from training. Duration: 29.735 2022-12-24 04:26:21,087 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99290.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:26:32,478 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7070, 1.6187, 1.5990, 1.6224, 1.8937, 1.8237, 1.8612, 1.3362], device='cuda:3'), covar=tensor([0.0351, 0.0267, 0.0460, 0.0228, 0.0206, 0.0430, 0.0311, 0.0349], device='cuda:3'), in_proj_covar=tensor([0.0099, 0.0131, 0.0158, 0.0125, 0.0120, 0.0126, 0.0103, 0.0132], device='cuda:3'), out_proj_covar=tensor([7.8161e-05, 1.0318e-04, 1.2865e-04, 9.8704e-05, 9.5861e-05, 9.6438e-05, 8.0077e-05, 1.0371e-04], device='cuda:3') 2022-12-24 04:27:08,761 INFO [train.py:894] (3/4) Epoch 29, batch 1150, loss[loss=0.1568, simple_loss=0.257, pruned_loss=0.02828, over 18466.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2572, pruned_loss=0.03797, over 3701447.49 frames. ], batch size: 54, lr: 3.95e-03, grad_scale: 16.0 2022-12-24 04:27:12,638 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99324.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:27:16,533 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99327.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:27:16,950 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0803, 1.3155, 1.8392, 1.7616, 2.1357, 2.1754, 1.8759, 1.8416], device='cuda:3'), covar=tensor([0.2431, 0.3605, 0.2814, 0.3038, 0.2336, 0.1091, 0.3517, 0.1505], device='cuda:3'), in_proj_covar=tensor([0.0273, 0.0301, 0.0288, 0.0328, 0.0321, 0.0260, 0.0356, 0.0250], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 04:27:19,603 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99329.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:27:39,228 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6282, 4.1768, 3.9380, 1.7684, 4.3332, 3.0621, 0.4047, 2.6495], device='cuda:3'), covar=tensor([0.1969, 0.0962, 0.1292, 0.3361, 0.0642, 0.0891, 0.5222, 0.1530], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0149, 0.0160, 0.0124, 0.0151, 0.0116, 0.0145, 0.0118], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-24 04:27:48,289 WARNING [train.py:1060] (3/4) Exclude cut with ID 7276-92427-0014-12983-0_sp0.9 from training. Duration: 21.3055625 2022-12-24 04:27:49,569 WARNING [train.py:1060] (3/4) Exclude cut with ID 1025-75365-0008-79168-0_sp0.9 from training. Duration: 22.0666875 2022-12-24 04:27:51,035 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.723e+02 3.197e+02 3.795e+02 4.539e+02 9.903e+02, threshold=7.590e+02, percent-clipped=2.0 2022-12-24 04:28:23,582 INFO [train.py:894] (3/4) Epoch 29, batch 1200, loss[loss=0.1483, simple_loss=0.2336, pruned_loss=0.03154, over 18669.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2572, pruned_loss=0.03817, over 3705143.85 frames. ], batch size: 48, lr: 3.95e-03, grad_scale: 16.0 2022-12-24 04:28:43,359 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99385.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:28:47,631 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.2936, 1.4045, 1.4622, 1.0658, 1.3545, 1.4618, 1.2589, 1.6989], device='cuda:3'), covar=tensor([0.1014, 0.1895, 0.1143, 0.1308, 0.0800, 0.0995, 0.2318, 0.0575], device='cuda:3'), in_proj_covar=tensor([0.0197, 0.0214, 0.0208, 0.0194, 0.0170, 0.0217, 0.0215, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 04:29:15,578 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99407.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:29:32,784 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0 from training. Duration: 20.26 2022-12-24 04:29:35,387 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99420.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:29:37,889 INFO [train.py:894] (3/4) Epoch 29, batch 1250, loss[loss=0.1781, simple_loss=0.2657, pruned_loss=0.0453, over 18700.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2581, pruned_loss=0.03842, over 3707471.67 frames. ], batch size: 60, lr: 3.95e-03, grad_scale: 16.0 2022-12-24 04:29:48,582 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0030-9324-0_sp0.9 from training. Duration: 21.3444375 2022-12-24 04:29:51,874 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3289, 1.7091, 1.4452, 2.0154, 1.9306, 1.5222, 1.2571, 1.2672], device='cuda:3'), covar=tensor([0.2280, 0.2034, 0.1856, 0.1174, 0.1630, 0.1290, 0.2511, 0.1760], device='cuda:3'), in_proj_covar=tensor([0.0250, 0.0232, 0.0223, 0.0203, 0.0264, 0.0199, 0.0228, 0.0205], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 04:30:04,519 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2022-12-24 04:30:19,587 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.792e+02 3.036e+02 3.747e+02 4.299e+02 8.271e+02, threshold=7.494e+02, percent-clipped=1.0 2022-12-24 04:30:44,491 WARNING [train.py:1060] (3/4) Exclude cut with ID 497-129325-0061-62254-0_sp1.1 from training. Duration: 0.97725 2022-12-24 04:30:47,216 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99468.0, num_to_drop=1, layers_to_drop={1} 2022-12-24 04:30:52,825 INFO [train.py:894] (3/4) Epoch 29, batch 1300, loss[loss=0.1497, simple_loss=0.2314, pruned_loss=0.03404, over 18582.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2581, pruned_loss=0.03807, over 3709648.33 frames. ], batch size: 45, lr: 3.95e-03, grad_scale: 16.0 2022-12-24 04:31:06,829 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99481.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:31:26,366 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0_sp0.9 from training. Duration: 22.97225 2022-12-24 04:31:55,822 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0047-39922-0_sp0.9 from training. Duration: 21.97775 2022-12-24 04:32:08,086 INFO [train.py:894] (3/4) Epoch 29, batch 1350, loss[loss=0.168, simple_loss=0.2598, pruned_loss=0.03804, over 18715.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2572, pruned_loss=0.03788, over 3709808.40 frames. ], batch size: 54, lr: 3.94e-03, grad_scale: 16.0 2022-12-24 04:32:09,685 WARNING [train.py:1060] (3/4) Exclude cut with ID 1112-1043-0006-89194-0_sp0.9 from training. Duration: 21.8333125 2022-12-24 04:32:19,367 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99529.0, num_to_drop=1, layers_to_drop={3} 2022-12-24 04:32:20,268 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0 from training. Duration: 20.47 2022-12-24 04:32:44,365 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99546.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:32:50,067 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.781e+02 3.126e+02 3.882e+02 4.637e+02 1.252e+03, threshold=7.765e+02, percent-clipped=4.0 2022-12-24 04:32:51,784 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99551.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:33:16,925 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99568.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:33:18,507 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5643, 2.1550, 1.8216, 2.1156, 1.9724, 2.2059, 2.0430, 2.3500], device='cuda:3'), covar=tensor([0.2065, 0.3243, 0.2101, 0.2929, 0.4044, 0.1130, 0.3114, 0.1072], device='cuda:3'), in_proj_covar=tensor([0.0299, 0.0301, 0.0257, 0.0349, 0.0281, 0.0235, 0.0299, 0.0224], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 04:33:22,791 INFO [train.py:894] (3/4) Epoch 29, batch 1400, loss[loss=0.145, simple_loss=0.2359, pruned_loss=0.02703, over 18383.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.257, pruned_loss=0.0381, over 3710416.55 frames. ], batch size: 46, lr: 3.94e-03, grad_scale: 16.0 2022-12-24 04:33:24,482 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0037-39912-0_sp0.9 from training. Duration: 20.67225 2022-12-24 04:33:42,940 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99585.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:33:44,570 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp0.9 from training. Duration: 25.2444375 2022-12-24 04:33:49,048 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99589.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:33:55,858 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99594.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:34:02,718 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99599.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:34:05,951 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0021-76797-0_sp0.9 from training. Duration: 21.1445 2022-12-24 04:34:09,475 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0472, 1.8965, 1.5951, 1.4574, 1.7531, 1.9010, 1.7233, 1.7115], device='cuda:3'), covar=tensor([0.2440, 0.3275, 0.2203, 0.2954, 0.3822, 0.1193, 0.3139, 0.1262], device='cuda:3'), in_proj_covar=tensor([0.0299, 0.0301, 0.0256, 0.0349, 0.0281, 0.0235, 0.0299, 0.0223], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 04:34:30,902 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2022-12-24 04:34:38,032 INFO [train.py:894] (3/4) Epoch 29, batch 1450, loss[loss=0.1385, simple_loss=0.2362, pruned_loss=0.0204, over 18428.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2567, pruned_loss=0.03775, over 3712049.95 frames. ], batch size: 48, lr: 3.94e-03, grad_scale: 16.0 2022-12-24 04:34:41,297 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99624.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:34:46,116 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99627.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:34:49,965 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99629.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:35:19,633 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.877e+02 3.031e+02 3.905e+02 4.844e+02 1.106e+03, threshold=7.811e+02, percent-clipped=1.0 2022-12-24 04:35:20,105 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99650.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:35:21,081 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp0.9 from training. Duration: 33.038875 2022-12-24 04:35:53,198 INFO [train.py:894] (3/4) Epoch 29, batch 1500, loss[loss=0.1637, simple_loss=0.2594, pruned_loss=0.03399, over 18536.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2569, pruned_loss=0.03813, over 3712532.32 frames. ], batch size: 97, lr: 3.94e-03, grad_scale: 16.0 2022-12-24 04:35:55,584 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64291-0000-16059-0_sp0.9 from training. Duration: 20.0944375 2022-12-24 04:35:58,688 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99675.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:36:06,129 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99680.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:36:06,289 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99680.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:36:10,104 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp1.1 from training. Duration: 20.4 2022-12-24 04:36:18,566 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0 from training. Duration: 20.085 2022-12-24 04:36:26,986 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0_sp0.9 from training. Duration: 23.07775 2022-12-24 04:36:44,794 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99707.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:36:44,976 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7578, 1.7865, 1.5565, 1.6000, 1.9868, 1.9434, 1.9947, 1.3986], device='cuda:3'), covar=tensor([0.0362, 0.0276, 0.0507, 0.0244, 0.0207, 0.0458, 0.0278, 0.0367], device='cuda:3'), in_proj_covar=tensor([0.0099, 0.0131, 0.0158, 0.0125, 0.0121, 0.0126, 0.0103, 0.0131], device='cuda:3'), out_proj_covar=tensor([7.8329e-05, 1.0355e-04, 1.2883e-04, 9.8721e-05, 9.6361e-05, 9.6551e-05, 8.0006e-05, 1.0323e-04], device='cuda:3') 2022-12-24 04:37:07,845 INFO [train.py:894] (3/4) Epoch 29, batch 1550, loss[loss=0.1639, simple_loss=0.2482, pruned_loss=0.03975, over 18515.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2572, pruned_loss=0.03799, over 3713736.12 frames. ], batch size: 47, lr: 3.94e-03, grad_scale: 16.0 2022-12-24 04:37:12,940 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp0.9 from training. Duration: 24.9333125 2022-12-24 04:37:36,327 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99741.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:37:49,144 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.330e+02 3.376e+02 3.911e+02 4.789e+02 1.084e+03, threshold=7.821e+02, percent-clipped=4.0 2022-12-24 04:37:53,650 WARNING [train.py:1060] (3/4) Exclude cut with ID 5118-111612-0016-124680-0_sp0.9 from training. Duration: 20.388875 2022-12-24 04:37:56,713 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99755.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:37:59,893 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp1.1 from training. Duration: 20.3590625 2022-12-24 04:38:23,630 INFO [train.py:894] (3/4) Epoch 29, batch 1600, loss[loss=0.1484, simple_loss=0.237, pruned_loss=0.02988, over 18676.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2573, pruned_loss=0.03803, over 3712940.17 frames. ], batch size: 48, lr: 3.94e-03, grad_scale: 16.0 2022-12-24 04:38:29,407 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99776.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:39:10,533 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0_sp1.1 from training. Duration: 0.836375 2022-12-24 04:39:37,378 INFO [train.py:894] (3/4) Epoch 29, batch 1650, loss[loss=0.1979, simple_loss=0.2761, pruned_loss=0.05981, over 18553.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2573, pruned_loss=0.03846, over 3712630.23 frames. ], batch size: 49, lr: 3.94e-03, grad_scale: 16.0 2022-12-24 04:39:40,327 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99824.0, num_to_drop=1, layers_to_drop={2} 2022-12-24 04:39:52,120 WARNING [train.py:1060] (3/4) Exclude cut with ID 8565-290391-0049-67394-0_sp0.9 from training. Duration: 21.3166875 2022-12-24 04:40:17,779 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.398e+02 3.479e+02 4.265e+02 5.308e+02 1.026e+03, threshold=8.530e+02, percent-clipped=3.0 2022-12-24 04:40:22,135 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0029-104863-0_sp0.9 from training. Duration: 22.1055625 2022-12-24 04:40:33,442 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp1.1 from training. Duration: 21.77725 2022-12-24 04:40:52,338 INFO [train.py:894] (3/4) Epoch 29, batch 1700, loss[loss=0.1518, simple_loss=0.2332, pruned_loss=0.03523, over 18583.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2583, pruned_loss=0.03954, over 3712608.29 frames. ], batch size: 45, lr: 3.94e-03, grad_scale: 16.0 2022-12-24 04:40:54,057 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp0.9 from training. Duration: 27.8166875 2022-12-24 04:41:11,450 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99885.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:41:12,038 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2022-12-24 04:41:18,376 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp1.1 from training. Duration: 22.5090625 2022-12-24 04:41:21,387 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.5974, 1.9435, 2.2925, 1.2438, 1.5067, 2.4309, 2.1948, 1.8835], device='cuda:3'), covar=tensor([0.0798, 0.0368, 0.0266, 0.0426, 0.0363, 0.0435, 0.0238, 0.0653], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0175, 0.0132, 0.0142, 0.0149, 0.0144, 0.0169, 0.0181], device='cuda:3'), out_proj_covar=tensor([1.1426e-04, 1.3090e-04, 9.7189e-05, 1.0395e-04, 1.0894e-04, 1.0789e-04, 1.2701e-04, 1.3608e-04], device='cuda:3') 2022-12-24 04:41:23,699 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0 from training. Duration: 25.035 2022-12-24 04:41:41,839 WARNING [train.py:1060] (3/4) Exclude cut with ID 774-127930-0014-10412-0_sp1.1 from training. Duration: 0.95 2022-12-24 04:42:00,814 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp0.9 from training. Duration: 0.92225 2022-12-24 04:42:06,474 INFO [train.py:894] (3/4) Epoch 29, batch 1750, loss[loss=0.1669, simple_loss=0.2541, pruned_loss=0.03986, over 18466.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2581, pruned_loss=0.04067, over 3713565.49 frames. ], batch size: 50, lr: 3.94e-03, grad_scale: 16.0 2022-12-24 04:42:09,566 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99924.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:42:09,670 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99924.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:42:22,311 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99933.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:42:23,693 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0 from training. Duration: 21.97 2022-12-24 04:42:40,154 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99945.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:42:44,111 WARNING [train.py:1060] (3/4) Exclude cut with ID 7492-105653-0055-62765-0_sp0.9 from training. Duration: 21.97225 2022-12-24 04:42:44,141 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp0.9 from training. Duration: 25.3333125 2022-12-24 04:42:47,626 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.605e+02 3.896e+02 4.661e+02 6.128e+02 1.064e+03, threshold=9.321e+02, percent-clipped=8.0 2022-12-24 04:42:56,447 WARNING [train.py:1060] (3/4) Exclude cut with ID 5172-29468-0015-19128-0_sp0.9 from training. Duration: 21.5055625 2022-12-24 04:42:56,775 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7357, 1.6157, 1.8017, 1.6373, 1.1471, 3.7040, 1.6239, 2.0876], device='cuda:3'), covar=tensor([0.3107, 0.2080, 0.1901, 0.2133, 0.1579, 0.0189, 0.1686, 0.0893], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0118, 0.0125, 0.0123, 0.0107, 0.0096, 0.0091, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-24 04:43:07,230 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp1.1 from training. Duration: 20.72725 2022-12-24 04:43:08,107 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2022-12-24 04:43:21,511 INFO [train.py:894] (3/4) Epoch 29, batch 1800, loss[loss=0.18, simple_loss=0.269, pruned_loss=0.04554, over 18716.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2587, pruned_loss=0.04185, over 3713740.78 frames. ], batch size: 65, lr: 3.94e-03, grad_scale: 16.0 2022-12-24 04:43:21,707 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99972.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:43:31,817 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8197, 1.8177, 2.3220, 1.3638, 2.2526, 2.4490, 1.4798, 2.7617], device='cuda:3'), covar=tensor([0.1355, 0.2112, 0.1430, 0.2041, 0.0776, 0.1095, 0.2589, 0.0538], device='cuda:3'), in_proj_covar=tensor([0.0198, 0.0215, 0.0209, 0.0196, 0.0171, 0.0218, 0.0217, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 04:43:33,003 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99980.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:43:40,169 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp0.9 from training. Duration: 26.32775 2022-12-24 04:44:16,714 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0 from training. Duration: 20.025 2022-12-24 04:44:17,041 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100007.0, num_to_drop=1, layers_to_drop={1} 2022-12-24 04:44:21,059 WARNING [train.py:1060] (3/4) Exclude cut with ID 6709-74022-0004-86860-0_sp1.1 from training. Duration: 0.9409375 2022-12-24 04:44:21,065 WARNING [train.py:1060] (3/4) Exclude cut with ID 4757-1811-0023-62229-0_sp0.9 from training. Duration: 21.37775 2022-12-24 04:44:39,150 INFO [train.py:894] (3/4) Epoch 29, batch 1850, loss[loss=0.1782, simple_loss=0.2575, pruned_loss=0.04944, over 18424.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2587, pruned_loss=0.04266, over 3713455.62 frames. ], batch size: 48, lr: 3.93e-03, grad_scale: 16.0 2022-12-24 04:44:42,158 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0004-25974-0_sp0.9 from training. Duration: 21.17225 2022-12-24 04:44:42,169 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp0.9 from training. Duration: 27.511125 2022-12-24 04:44:48,242 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=100028.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:44:59,265 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100036.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:45:12,916 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0 from training. Duration: 22.8 2022-12-24 04:45:17,170 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0 from training. Duration: 22.585 2022-12-24 04:45:20,321 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.870e+02 3.993e+02 4.913e+02 5.637e+02 1.150e+03, threshold=9.826e+02, percent-clipped=4.0 2022-12-24 04:45:31,986 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5997, 1.1731, 0.8487, 1.2601, 2.0608, 0.7211, 1.3258, 1.4643], device='cuda:3'), covar=tensor([0.1596, 0.2074, 0.1732, 0.1479, 0.1727, 0.1843, 0.1380, 0.1655], device='cuda:3'), in_proj_covar=tensor([0.0095, 0.0098, 0.0117, 0.0097, 0.0122, 0.0093, 0.0098, 0.0095], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-24 04:45:48,576 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100068.0, num_to_drop=1, layers_to_drop={1} 2022-12-24 04:45:49,512 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0001-146967-0_sp0.9 from training. Duration: 22.0166875 2022-12-24 04:45:53,703 INFO [train.py:894] (3/4) Epoch 29, batch 1900, loss[loss=0.1955, simple_loss=0.2778, pruned_loss=0.05658, over 18483.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2591, pruned_loss=0.04347, over 3712325.69 frames. ], batch size: 64, lr: 3.93e-03, grad_scale: 16.0 2022-12-24 04:45:59,879 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100076.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:46:07,047 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp1.1 from training. Duration: 24.395375 2022-12-24 04:46:14,254 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp0.9 from training. Duration: 27.47775 2022-12-24 04:46:19,640 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp0.9 from training. Duration: 24.8833125 2022-12-24 04:46:22,585 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0 from training. Duration: 23.39 2022-12-24 04:46:27,460 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp0.9 from training. Duration: 28.72225 2022-12-24 04:46:38,373 WARNING [train.py:1060] (3/4) Exclude cut with ID 585-294811-0110-133686-0_sp0.9 from training. Duration: 20.8944375 2022-12-24 04:46:38,883 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6202, 2.0055, 1.7519, 2.3104, 2.5805, 1.7523, 1.5924, 1.4318], device='cuda:3'), covar=tensor([0.1916, 0.1707, 0.1502, 0.0968, 0.1179, 0.1061, 0.2104, 0.1503], device='cuda:3'), in_proj_covar=tensor([0.0252, 0.0235, 0.0225, 0.0206, 0.0267, 0.0201, 0.0230, 0.0206], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 04:46:49,161 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9938, 1.8961, 2.0424, 1.8711, 1.4229, 5.0129, 1.9951, 2.6711], device='cuda:3'), covar=tensor([0.3027, 0.1930, 0.1755, 0.2037, 0.1408, 0.0104, 0.1485, 0.0762], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0118, 0.0125, 0.0123, 0.0107, 0.0096, 0.0091, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-24 04:46:53,125 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0_sp0.9 from training. Duration: 23.8444375 2022-12-24 04:47:07,844 INFO [train.py:894] (3/4) Epoch 29, batch 1950, loss[loss=0.1728, simple_loss=0.2555, pruned_loss=0.04504, over 18706.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2589, pruned_loss=0.0436, over 3713110.24 frames. ], batch size: 50, lr: 3.93e-03, grad_scale: 8.0 2022-12-24 04:47:11,087 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=100124.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:47:11,241 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100124.0, num_to_drop=1, layers_to_drop={2} 2022-12-24 04:47:18,148 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0 from training. Duration: 25.85 2022-12-24 04:47:18,159 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0 from training. Duration: 21.39 2022-12-24 04:47:29,869 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0 from training. Duration: 27.92 2022-12-24 04:47:50,708 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.524e+02 3.680e+02 4.686e+02 5.699e+02 9.097e+02, threshold=9.372e+02, percent-clipped=0.0 2022-12-24 04:47:57,159 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0039-130165-0_sp0.9 from training. Duration: 20.661125 2022-12-24 04:48:22,914 INFO [train.py:894] (3/4) Epoch 29, batch 2000, loss[loss=0.1801, simple_loss=0.2684, pruned_loss=0.04586, over 18700.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2586, pruned_loss=0.04367, over 3714368.20 frames. ], batch size: 60, lr: 3.93e-03, grad_scale: 8.0 2022-12-24 04:48:22,964 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0043-15874-0_sp0.9 from training. Duration: 20.07225 2022-12-24 04:48:23,059 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=100172.0, num_to_drop=1, layers_to_drop={1} 2022-12-24 04:48:29,712 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0 from training. Duration: 21.01 2022-12-24 04:49:34,540 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0 from training. Duration: 20.65 2022-12-24 04:49:37,252 INFO [train.py:894] (3/4) Epoch 29, batch 2050, loss[loss=0.1664, simple_loss=0.2558, pruned_loss=0.03847, over 18715.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2599, pruned_loss=0.04493, over 3715470.04 frames. ], batch size: 54, lr: 3.93e-03, grad_scale: 8.0 2022-12-24 04:49:40,810 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100224.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:49:43,237 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0 from training. Duration: 21.46 2022-12-24 04:50:00,984 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100238.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:50:11,241 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100245.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:50:20,512 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.560e+02 3.985e+02 4.677e+02 5.805e+02 1.123e+03, threshold=9.354e+02, percent-clipped=1.0 2022-12-24 04:50:26,245 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0 from training. Duration: 0.92 2022-12-24 04:50:33,910 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0_sp0.9 from training. Duration: 23.7666875 2022-12-24 04:50:42,994 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2022-12-24 04:50:52,856 INFO [train.py:894] (3/4) Epoch 29, batch 2100, loss[loss=0.1912, simple_loss=0.2747, pruned_loss=0.05382, over 18662.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2601, pruned_loss=0.04513, over 3714326.74 frames. ], batch size: 183, lr: 3.93e-03, grad_scale: 8.0 2022-12-24 04:50:53,016 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=100272.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:51:09,518 WARNING [train.py:1060] (3/4) Exclude cut with ID 8544-281189-0060-101339-0_sp0.9 from training. Duration: 20.861125 2022-12-24 04:51:19,030 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0_sp0.9 from training. Duration: 22.711125 2022-12-24 04:51:26,293 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=100293.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:51:35,018 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100299.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:52:00,252 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp1.1 from training. Duration: 22.986375 2022-12-24 04:52:09,441 INFO [train.py:894] (3/4) Epoch 29, batch 2150, loss[loss=0.1888, simple_loss=0.2665, pruned_loss=0.05549, over 18649.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2594, pruned_loss=0.04519, over 3713998.06 frames. ], batch size: 179, lr: 3.93e-03, grad_scale: 8.0 2022-12-24 04:52:15,323 WARNING [train.py:1060] (3/4) Exclude cut with ID 8040-260924-0003-80960-0_sp0.9 from training. Duration: 22.07225 2022-12-24 04:52:21,778 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0045-26330-0_sp0.9 from training. Duration: 20.3055625 2022-12-24 04:52:23,445 WARNING [train.py:1060] (3/4) Exclude cut with ID 6356-271890-0060-94317-0_sp0.9 from training. Duration: 20.72225 2022-12-24 04:52:32,872 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100336.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:52:43,335 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp1.1 from training. Duration: 22.4818125 2022-12-24 04:52:54,789 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.550e+02 4.110e+02 5.035e+02 6.020e+02 1.639e+03, threshold=1.007e+03, percent-clipped=1.0 2022-12-24 04:53:08,332 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp0.9 from training. Duration: 25.0944375 2022-12-24 04:53:11,287 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0 from training. Duration: 21.515 2022-12-24 04:53:12,852 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100363.0, num_to_drop=1, layers_to_drop={3} 2022-12-24 04:53:18,602 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp0.9 from training. Duration: 27.02225 2022-12-24 04:53:22,964 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0010-62480-0_sp0.9 from training. Duration: 22.22225 2022-12-24 04:53:25,697 INFO [train.py:894] (3/4) Epoch 29, batch 2200, loss[loss=0.1781, simple_loss=0.2716, pruned_loss=0.04231, over 18513.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2592, pruned_loss=0.04516, over 3712348.34 frames. ], batch size: 52, lr: 3.93e-03, grad_scale: 8.0 2022-12-24 04:53:29,288 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0085-44554-0_sp0.9 from training. Duration: 20.85 2022-12-24 04:53:31,513 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.12 vs. limit=5.0 2022-12-24 04:53:45,597 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=100384.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:54:01,963 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0 from training. Duration: 21.54 2022-12-24 04:54:05,144 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp1.1 from training. Duration: 20.5318125 2022-12-24 04:54:13,159 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100402.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:54:15,899 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0012-134311-0_sp0.9 from training. Duration: 21.9333125 2022-12-24 04:54:43,502 INFO [train.py:894] (3/4) Epoch 29, batch 2250, loss[loss=0.1923, simple_loss=0.2734, pruned_loss=0.0556, over 18588.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2588, pruned_loss=0.04493, over 3712762.05 frames. ], batch size: 57, lr: 3.93e-03, grad_scale: 8.0 2022-12-24 04:54:57,095 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9859, 1.5648, 1.7389, 2.2994, 1.8975, 3.7250, 1.4459, 1.6425], device='cuda:3'), covar=tensor([0.0770, 0.1868, 0.1050, 0.0814, 0.1403, 0.0246, 0.1539, 0.1645], device='cuda:3'), in_proj_covar=tensor([0.0074, 0.0083, 0.0073, 0.0075, 0.0092, 0.0077, 0.0086, 0.0079], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0004], device='cuda:3') 2022-12-24 04:55:05,002 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0025-130151-0_sp0.9 from training. Duration: 21.7944375 2022-12-24 04:55:16,404 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0_sp0.9 from training. Duration: 22.4666875 2022-12-24 04:55:23,978 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0 from training. Duration: 21.635 2022-12-24 04:55:26,873 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.445e+02 4.059e+02 4.926e+02 5.682e+02 1.179e+03, threshold=9.852e+02, percent-clipped=4.0 2022-12-24 04:55:28,606 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0_sp0.9 from training. Duration: 24.038875 2022-12-24 04:55:45,371 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100463.0, num_to_drop=1, layers_to_drop={1} 2022-12-24 04:56:00,252 INFO [train.py:894] (3/4) Epoch 29, batch 2300, loss[loss=0.1531, simple_loss=0.2442, pruned_loss=0.03105, over 18592.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.259, pruned_loss=0.04479, over 3714171.73 frames. ], batch size: 51, lr: 3.93e-03, grad_scale: 8.0 2022-12-24 04:56:11,107 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp1.1 from training. Duration: 21.786375 2022-12-24 04:56:22,771 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0 from training. Duration: 20.22 2022-12-24 04:57:15,047 INFO [train.py:894] (3/4) Epoch 29, batch 2350, loss[loss=0.1783, simple_loss=0.2668, pruned_loss=0.04491, over 18466.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2592, pruned_loss=0.04469, over 3712955.31 frames. ], batch size: 50, lr: 3.92e-03, grad_scale: 8.0 2022-12-24 04:57:56,843 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.255e+02 3.961e+02 4.890e+02 6.164e+02 1.267e+03, threshold=9.780e+02, percent-clipped=4.0 2022-12-24 04:58:23,567 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5338, 2.8616, 3.2678, 1.7576, 2.7171, 3.1627, 2.1904, 3.4512], device='cuda:3'), covar=tensor([0.1359, 0.1554, 0.1304, 0.2028, 0.0800, 0.1209, 0.2165, 0.0593], device='cuda:3'), in_proj_covar=tensor([0.0199, 0.0215, 0.0210, 0.0195, 0.0171, 0.0218, 0.0217, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 04:58:25,807 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0 from training. Duration: 25.285 2022-12-24 04:58:29,159 INFO [train.py:894] (3/4) Epoch 29, batch 2400, loss[loss=0.1926, simple_loss=0.2733, pruned_loss=0.05596, over 18660.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2587, pruned_loss=0.0444, over 3714252.65 frames. ], batch size: 60, lr: 3.92e-03, grad_scale: 8.0 2022-12-24 04:58:41,121 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6831, 4.0346, 3.9053, 1.5951, 4.0519, 3.1677, 0.5580, 2.7349], device='cuda:3'), covar=tensor([0.2041, 0.1235, 0.1307, 0.3551, 0.0970, 0.0826, 0.5152, 0.1422], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0150, 0.0161, 0.0125, 0.0152, 0.0117, 0.0146, 0.0118], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-24 04:59:01,584 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100594.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 04:59:29,520 WARNING [train.py:1060] (3/4) Exclude cut with ID 811-130148-0001-63453-0_sp0.9 from training. Duration: 20.861125 2022-12-24 04:59:44,337 INFO [train.py:894] (3/4) Epoch 29, batch 2450, loss[loss=0.1601, simple_loss=0.2407, pruned_loss=0.03973, over 18531.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.258, pruned_loss=0.04419, over 3714740.97 frames. ], batch size: 47, lr: 3.92e-03, grad_scale: 8.0 2022-12-24 04:59:48,948 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0 from training. Duration: 20.88 2022-12-24 04:59:52,664 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1486, 1.3512, 1.9298, 1.7898, 2.1634, 2.1794, 1.9699, 1.8696], device='cuda:3'), covar=tensor([0.2302, 0.3438, 0.2569, 0.2757, 0.2048, 0.0977, 0.3112, 0.1374], device='cuda:3'), in_proj_covar=tensor([0.0274, 0.0302, 0.0290, 0.0329, 0.0321, 0.0261, 0.0358, 0.0251], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 05:00:21,488 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0_sp0.9 from training. Duration: 23.4166875 2022-12-24 05:00:27,740 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.767e+02 3.997e+02 4.879e+02 5.983e+02 1.573e+03, threshold=9.759e+02, percent-clipped=4.0 2022-12-24 05:00:46,134 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100663.0, num_to_drop=1, layers_to_drop={0} 2022-12-24 05:00:59,034 INFO [train.py:894] (3/4) Epoch 29, batch 2500, loss[loss=0.2127, simple_loss=0.2812, pruned_loss=0.07208, over 18626.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2578, pruned_loss=0.04421, over 3713958.95 frames. ], batch size: 184, lr: 3.92e-03, grad_scale: 8.0 2022-12-24 05:01:40,066 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0 from training. Duration: 21.24 2022-12-24 05:01:40,080 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0_sp0.9 from training. Duration: 23.9055625 2022-12-24 05:01:53,619 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7783, 2.7249, 2.0216, 3.3574, 2.9283, 2.4639, 3.7851, 2.7779], device='cuda:3'), covar=tensor([0.0799, 0.1811, 0.2756, 0.1760, 0.1803, 0.0883, 0.0903, 0.1138], device='cuda:3'), in_proj_covar=tensor([0.0186, 0.0221, 0.0262, 0.0295, 0.0245, 0.0198, 0.0210, 0.0213], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 05:01:57,799 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=100711.0, num_to_drop=1, layers_to_drop={0} 2022-12-24 05:02:13,672 INFO [train.py:894] (3/4) Epoch 29, batch 2550, loss[loss=0.1506, simple_loss=0.2317, pruned_loss=0.03475, over 18470.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2574, pruned_loss=0.04389, over 3713751.76 frames. ], batch size: 50, lr: 3.92e-03, grad_scale: 8.0 2022-12-24 05:02:13,725 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp0.9 from training. Duration: 25.988875 2022-12-24 05:02:23,998 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0001-134300-0_sp0.9 from training. Duration: 20.67225 2022-12-24 05:02:25,866 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100730.0, num_to_drop=1, layers_to_drop={0} 2022-12-24 05:02:34,309 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9453, 1.3551, 0.7955, 1.4275, 2.3619, 1.2499, 1.6169, 1.8049], device='cuda:3'), covar=tensor([0.1517, 0.2047, 0.2180, 0.1442, 0.1621, 0.1848, 0.1429, 0.1567], device='cuda:3'), in_proj_covar=tensor([0.0095, 0.0097, 0.0116, 0.0097, 0.0121, 0.0093, 0.0098, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-24 05:02:55,944 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.707e+02 4.004e+02 4.626e+02 5.678e+02 1.065e+03, threshold=9.253e+02, percent-clipped=1.0 2022-12-24 05:03:06,071 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7299, 1.5948, 1.6720, 1.5761, 1.2175, 3.7637, 1.5336, 2.0189], device='cuda:3'), covar=tensor([0.3221, 0.2155, 0.2035, 0.2201, 0.1551, 0.0215, 0.1670, 0.0926], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0118, 0.0126, 0.0124, 0.0108, 0.0097, 0.0091, 0.0091], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-24 05:03:07,221 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100758.0, num_to_drop=1, layers_to_drop={1} 2022-12-24 05:03:12,181 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0 from training. Duration: 20.34 2022-12-24 05:03:28,306 INFO [train.py:894] (3/4) Epoch 29, batch 2600, loss[loss=0.1296, simple_loss=0.2109, pruned_loss=0.02417, over 18472.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2573, pruned_loss=0.04387, over 3713009.90 frames. ], batch size: 43, lr: 3.92e-03, grad_scale: 8.0 2022-12-24 05:03:56,092 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100791.0, num_to_drop=1, layers_to_drop={2} 2022-12-24 05:04:24,470 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp0.9 from training. Duration: 25.061125 2022-12-24 05:04:35,160 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0 from training. Duration: 0.83 2022-12-24 05:04:43,735 INFO [train.py:894] (3/4) Epoch 29, batch 2650, loss[loss=0.1602, simple_loss=0.241, pruned_loss=0.03967, over 18427.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.257, pruned_loss=0.04368, over 3712610.89 frames. ], batch size: 48, lr: 3.92e-03, grad_scale: 8.0 2022-12-24 05:04:57,929 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0 from training. Duration: 24.73 2022-12-24 05:05:11,149 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0 from training. Duration: 23.965 2022-12-24 05:05:19,148 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0030-146996-0_sp0.9 from training. Duration: 22.088875 2022-12-24 05:05:26,552 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2022-12-24 05:05:27,247 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.255e+02 4.318e+02 5.283e+02 6.216e+02 2.785e+03, threshold=1.057e+03, percent-clipped=2.0 2022-12-24 05:05:38,026 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0_sp0.9 from training. Duration: 23.6 2022-12-24 05:05:57,956 INFO [train.py:894] (3/4) Epoch 29, batch 2700, loss[loss=0.1962, simple_loss=0.2775, pruned_loss=0.05741, over 18609.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2569, pruned_loss=0.04346, over 3713033.78 frames. ], batch size: 180, lr: 3.92e-03, grad_scale: 8.0 2022-12-24 05:06:31,393 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100894.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 05:07:12,432 INFO [train.py:894] (3/4) Epoch 29, batch 2750, loss[loss=0.1781, simple_loss=0.2561, pruned_loss=0.05007, over 18508.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2575, pruned_loss=0.04396, over 3713165.57 frames. ], batch size: 47, lr: 3.92e-03, grad_scale: 8.0 2022-12-24 05:07:13,974 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0 from training. Duration: 23.795 2022-12-24 05:07:30,425 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp1.1 from training. Duration: 21.5409375 2022-12-24 05:07:33,526 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp0.9 from training. Duration: 24.97775 2022-12-24 05:07:43,261 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=100942.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 05:07:44,774 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0_sp0.9 from training. Duration: 23.3444375 2022-12-24 05:07:57,585 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.603e+02 4.059e+02 4.995e+02 5.667e+02 1.400e+03, threshold=9.990e+02, percent-clipped=2.0 2022-12-24 05:08:14,003 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0_sp0.9 from training. Duration: 23.2 2022-12-24 05:08:20,140 WARNING [train.py:1060] (3/4) Exclude cut with ID 5653-46179-0060-117930-0_sp0.9 from training. Duration: 21.17225 2022-12-24 05:08:21,814 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5036, 1.1163, 0.8193, 1.1878, 2.0061, 0.7333, 1.2493, 1.3952], device='cuda:3'), covar=tensor([0.1658, 0.2144, 0.1828, 0.1517, 0.1814, 0.1832, 0.1501, 0.1754], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0097, 0.0116, 0.0097, 0.0120, 0.0092, 0.0098, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-24 05:08:28,842 INFO [train.py:894] (3/4) Epoch 29, batch 2800, loss[loss=0.1825, simple_loss=0.2699, pruned_loss=0.04761, over 18560.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2558, pruned_loss=0.04362, over 3712571.26 frames. ], batch size: 57, lr: 3.92e-03, grad_scale: 8.0 2022-12-24 05:08:37,020 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.23 vs. limit=5.0 2022-12-24 05:08:38,896 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp0.9 from training. Duration: 24.6555625 2022-12-24 05:08:53,368 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6010, 3.3474, 3.2616, 1.3542, 3.5754, 2.6938, 0.5984, 2.2573], device='cuda:3'), covar=tensor([0.2078, 0.1449, 0.1522, 0.3786, 0.0904, 0.0983, 0.5104, 0.1617], device='cuda:3'), in_proj_covar=tensor([0.0155, 0.0153, 0.0164, 0.0127, 0.0154, 0.0119, 0.0148, 0.0120], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-24 05:09:34,867 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0 from training. Duration: 20.44 2022-12-24 05:09:45,293 INFO [train.py:894] (3/4) Epoch 29, batch 2850, loss[loss=0.1856, simple_loss=0.2657, pruned_loss=0.05277, over 18659.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2557, pruned_loss=0.04333, over 3712848.35 frames. ], batch size: 69, lr: 3.92e-03, grad_scale: 8.0 2022-12-24 05:09:46,068 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2022-12-24 05:09:49,953 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0_sp0.9 from training. Duration: 23.45 2022-12-24 05:10:19,339 WARNING [train.py:1060] (3/4) Exclude cut with ID 6945-60535-0076-12784-0_sp0.9 from training. Duration: 20.52225 2022-12-24 05:10:25,192 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0 from training. Duration: 22.19 2022-12-24 05:10:31,130 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.944e+02 4.277e+02 5.213e+02 6.121e+02 9.835e+02, threshold=1.043e+03, percent-clipped=0.0 2022-12-24 05:10:35,779 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp1.1 from training. Duration: 25.3818125 2022-12-24 05:10:41,866 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101058.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 05:10:52,117 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp0.9 from training. Duration: 28.0944375 2022-12-24 05:10:59,630 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0_sp0.9 from training. Duration: 22.9444375 2022-12-24 05:11:01,374 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8208, 1.2933, 0.7844, 1.4894, 2.1759, 1.3754, 1.7232, 1.8417], device='cuda:3'), covar=tensor([0.1710, 0.2177, 0.2284, 0.1518, 0.1810, 0.1810, 0.1403, 0.1681], device='cuda:3'), in_proj_covar=tensor([0.0095, 0.0097, 0.0116, 0.0097, 0.0121, 0.0092, 0.0098, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-24 05:11:02,463 INFO [train.py:894] (3/4) Epoch 29, batch 2900, loss[loss=0.1633, simple_loss=0.2559, pruned_loss=0.03537, over 18482.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2561, pruned_loss=0.04364, over 3713908.08 frames. ], batch size: 54, lr: 3.91e-03, grad_scale: 8.0 2022-12-24 05:11:09,344 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp1.1 from training. Duration: 21.6318125 2022-12-24 05:11:24,515 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=101086.0, num_to_drop=1, layers_to_drop={2} 2022-12-24 05:11:27,364 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0 from training. Duration: 23.695 2022-12-24 05:11:32,119 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([3.2008, 2.4748, 1.8963, 2.9243, 2.2591, 2.4013, 2.4487, 3.2843], device='cuda:3'), covar=tensor([0.2057, 0.3409, 0.2208, 0.3141, 0.4241, 0.1147, 0.3427, 0.0907], device='cuda:3'), in_proj_covar=tensor([0.0305, 0.0305, 0.0259, 0.0355, 0.0286, 0.0239, 0.0303, 0.0226], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 05:11:50,598 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0 from training. Duration: 23.955 2022-12-24 05:11:53,593 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=101106.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 05:11:55,272 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101107.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 05:12:17,127 INFO [train.py:894] (3/4) Epoch 29, batch 2950, loss[loss=0.1823, simple_loss=0.2771, pruned_loss=0.04375, over 18600.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2568, pruned_loss=0.04382, over 3714239.69 frames. ], batch size: 56, lr: 3.91e-03, grad_scale: 8.0 2022-12-24 05:12:24,844 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp0.9 from training. Duration: 26.438875 2022-12-24 05:13:00,821 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.294e+02 3.879e+02 5.080e+02 6.075e+02 1.162e+03, threshold=1.016e+03, percent-clipped=1.0 2022-12-24 05:13:03,847 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0021-26306-0_sp0.9 from training. Duration: 21.2444375 2022-12-24 05:13:05,416 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp0.9 from training. Duration: 31.02225 2022-12-24 05:13:14,396 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0 from training. Duration: 22.395 2022-12-24 05:13:28,292 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101168.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 05:13:30,947 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0 from training. Duration: 21.075 2022-12-24 05:13:34,506 INFO [train.py:894] (3/4) Epoch 29, batch 3000, loss[loss=0.1852, simple_loss=0.2798, pruned_loss=0.04531, over 18481.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2577, pruned_loss=0.04441, over 3714062.54 frames. ], batch size: 77, lr: 3.91e-03, grad_scale: 8.0 2022-12-24 05:13:34,506 INFO [train.py:919] (3/4) Computing validation loss 2022-12-24 05:13:45,185 INFO [train.py:928] (3/4) Epoch 29, validation: loss=0.1637, simple_loss=0.2594, pruned_loss=0.034, over 944034.00 frames. 2022-12-24 05:13:45,186 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-24 05:13:49,726 WARNING [train.py:1060] (3/4) Exclude cut with ID 6482-98857-0025-147532-0_sp0.9 from training. Duration: 20.0055625 2022-12-24 05:13:49,737 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0037-132304-0_sp0.9 from training. Duration: 22.05 2022-12-24 05:13:49,747 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0 from training. Duration: 26.8349375 2022-12-24 05:13:52,744 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp1.1 from training. Duration: 22.1090625 2022-12-24 05:13:59,681 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp0.9 from training. Duration: 26.6166875 2022-12-24 05:14:16,736 WARNING [train.py:1060] (3/4) Exclude cut with ID 2046-178027-0000-53705-0_sp0.9 from training. Duration: 20.3055625 2022-12-24 05:14:40,824 WARNING [train.py:1060] (3/4) Exclude cut with ID 7205-50138-0008-5373-0_sp0.9 from training. Duration: 20.7 2022-12-24 05:14:59,866 INFO [train.py:894] (3/4) Epoch 29, batch 3050, loss[loss=0.1608, simple_loss=0.2535, pruned_loss=0.03404, over 18523.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2575, pruned_loss=0.04447, over 3713022.31 frames. ], batch size: 55, lr: 3.91e-03, grad_scale: 8.0 2022-12-24 05:15:05,877 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0096, 1.9029, 1.6126, 1.5977, 1.7727, 1.8966, 1.7717, 1.7983], device='cuda:3'), covar=tensor([0.2275, 0.2994, 0.2058, 0.2463, 0.3403, 0.1156, 0.2691, 0.1122], device='cuda:3'), in_proj_covar=tensor([0.0303, 0.0302, 0.0257, 0.0351, 0.0285, 0.0237, 0.0300, 0.0225], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 05:15:22,298 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0 from training. Duration: 22.48 2022-12-24 05:15:40,329 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp0.9 from training. Duration: 29.816625 2022-12-24 05:15:43,407 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.815e+02 4.122e+02 4.831e+02 6.036e+02 1.084e+03, threshold=9.661e+02, percent-clipped=3.0 2022-12-24 05:16:01,379 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp1.1 from training. Duration: 22.7590625 2022-12-24 05:16:07,340 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0 from training. Duration: 22.555 2022-12-24 05:16:14,400 INFO [train.py:894] (3/4) Epoch 29, batch 3100, loss[loss=0.1744, simple_loss=0.2466, pruned_loss=0.0511, over 18512.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2578, pruned_loss=0.0451, over 3713737.92 frames. ], batch size: 43, lr: 3.91e-03, grad_scale: 8.0 2022-12-24 05:16:27,313 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0005-25975-0_sp0.9 from training. Duration: 21.688875 2022-12-24 05:17:00,264 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0_sp0.9 from training. Duration: 22.6 2022-12-24 05:17:29,624 INFO [train.py:894] (3/4) Epoch 29, batch 3150, loss[loss=0.1812, simple_loss=0.2729, pruned_loss=0.04471, over 18713.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2572, pruned_loss=0.0445, over 3714084.12 frames. ], batch size: 62, lr: 3.91e-03, grad_scale: 8.0 2022-12-24 05:17:37,258 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0 from training. Duration: 24.32 2022-12-24 05:18:13,841 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.460e+02 3.834e+02 4.753e+02 5.855e+02 1.523e+03, threshold=9.507e+02, percent-clipped=2.0 2022-12-24 05:18:38,976 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-276745-0093-13116-0_sp0.9 from training. Duration: 21.061125 2022-12-24 05:18:39,840 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 2022-12-24 05:18:44,382 INFO [train.py:894] (3/4) Epoch 29, batch 3200, loss[loss=0.1741, simple_loss=0.249, pruned_loss=0.04963, over 18542.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2568, pruned_loss=0.04418, over 3713736.27 frames. ], batch size: 47, lr: 3.91e-03, grad_scale: 8.0 2022-12-24 05:18:51,431 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0024-15855-0_sp0.9 from training. Duration: 20.32225 2022-12-24 05:18:53,736 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2022-12-24 05:19:01,683 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp1.1 from training. Duration: 0.7545625 2022-12-24 05:19:04,865 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101386.0, num_to_drop=1, layers_to_drop={2} 2022-12-24 05:19:12,157 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2022-12-24 05:19:18,680 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0_sp0.9 from training. Duration: 23.9333125 2022-12-24 05:19:31,584 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9339, 1.7312, 1.9995, 2.1195, 1.9391, 2.9272, 1.7930, 1.7859], device='cuda:3'), covar=tensor([0.0780, 0.1433, 0.1060, 0.0779, 0.1201, 0.0361, 0.1143, 0.1294], device='cuda:3'), in_proj_covar=tensor([0.0074, 0.0083, 0.0073, 0.0075, 0.0092, 0.0077, 0.0086, 0.0078], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0004], device='cuda:3') 2022-12-24 05:19:49,091 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp1.1 from training. Duration: 20.17275 2022-12-24 05:19:55,365 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp1.1 from training. Duration: 20.436375 2022-12-24 05:19:59,785 INFO [train.py:894] (3/4) Epoch 29, batch 3250, loss[loss=0.1612, simple_loss=0.2417, pruned_loss=0.04037, over 18434.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2572, pruned_loss=0.04403, over 3713935.70 frames. ], batch size: 48, lr: 3.91e-03, grad_scale: 8.0 2022-12-24 05:20:19,096 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=101434.0, num_to_drop=1, layers_to_drop={0} 2022-12-24 05:20:31,570 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9279, 2.4343, 1.8821, 2.8039, 3.1135, 1.7934, 2.0260, 1.5817], device='cuda:3'), covar=tensor([0.1901, 0.1642, 0.1514, 0.0930, 0.1293, 0.1098, 0.1931, 0.1522], device='cuda:3'), in_proj_covar=tensor([0.0252, 0.0235, 0.0225, 0.0206, 0.0267, 0.0199, 0.0231, 0.0206], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 05:20:43,365 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.471e+02 3.720e+02 4.479e+02 6.030e+02 1.410e+03, threshold=8.958e+02, percent-clipped=2.0 2022-12-24 05:21:01,320 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=101463.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 05:21:12,868 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0_sp0.9 from training. Duration: 23.1055625 2022-12-24 05:21:14,202 INFO [train.py:894] (3/4) Epoch 29, batch 3300, loss[loss=0.1765, simple_loss=0.2634, pruned_loss=0.04478, over 18717.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2575, pruned_loss=0.04404, over 3715290.70 frames. ], batch size: 52, lr: 3.91e-03, grad_scale: 8.0 2022-12-24 05:21:16,155 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp1.1 from training. Duration: 23.5 2022-12-24 05:21:20,260 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2022-12-24 05:21:28,344 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp0.9 from training. Duration: 26.62775 2022-12-24 05:21:40,952 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0 from training. Duration: 21.105 2022-12-24 05:21:45,021 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0_sp0.9 from training. Duration: 24.411125 2022-12-24 05:22:07,910 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2022-12-24 05:22:12,494 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp1.1 from training. Duration: 21.263625 2022-12-24 05:22:29,738 INFO [train.py:894] (3/4) Epoch 29, batch 3350, loss[loss=0.1603, simple_loss=0.2362, pruned_loss=0.04226, over 18548.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2572, pruned_loss=0.0443, over 3714351.81 frames. ], batch size: 44, lr: 3.91e-03, grad_scale: 8.0 2022-12-24 05:22:43,766 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0 from training. Duration: 20.795 2022-12-24 05:22:55,862 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0 from training. Duration: 24.76 2022-12-24 05:22:55,878 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0_sp0.9 from training. Duration: 22.25 2022-12-24 05:23:12,764 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.686e+02 3.942e+02 4.707e+02 5.878e+02 1.188e+03, threshold=9.415e+02, percent-clipped=4.0 2022-12-24 05:23:20,038 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp1.1 from training. Duration: 20.5045625 2022-12-24 05:23:43,805 INFO [train.py:894] (3/4) Epoch 29, batch 3400, loss[loss=0.1925, simple_loss=0.2737, pruned_loss=0.05564, over 18639.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2563, pruned_loss=0.04358, over 3713607.94 frames. ], batch size: 60, lr: 3.90e-03, grad_scale: 8.0 2022-12-24 05:24:29,154 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4873, 1.9590, 1.4861, 2.2956, 2.3578, 1.5950, 1.4417, 1.2284], device='cuda:3'), covar=tensor([0.2199, 0.1889, 0.1860, 0.1069, 0.1445, 0.1229, 0.2453, 0.1803], device='cuda:3'), in_proj_covar=tensor([0.0251, 0.0234, 0.0224, 0.0205, 0.0267, 0.0199, 0.0231, 0.0205], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 05:24:50,631 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6317, 1.6615, 1.8834, 1.0578, 1.7820, 1.8337, 1.4563, 2.2512], device='cuda:3'), covar=tensor([0.1088, 0.1971, 0.1177, 0.1747, 0.0704, 0.1138, 0.2489, 0.0557], device='cuda:3'), in_proj_covar=tensor([0.0199, 0.0217, 0.0211, 0.0197, 0.0172, 0.0221, 0.0219, 0.0197], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 05:24:57,345 INFO [train.py:894] (3/4) Epoch 29, batch 3450, loss[loss=0.1639, simple_loss=0.2439, pruned_loss=0.04193, over 18679.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2565, pruned_loss=0.04372, over 3714091.03 frames. ], batch size: 46, lr: 3.90e-03, grad_scale: 8.0 2022-12-24 05:25:13,984 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2022-12-24 05:25:38,890 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.398e+02 3.927e+02 4.688e+02 5.596e+02 9.547e+02, threshold=9.377e+02, percent-clipped=1.0 2022-12-24 05:26:07,229 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0856, 2.1127, 2.2093, 2.0181, 1.8661, 4.5680, 2.1167, 2.6098], device='cuda:3'), covar=tensor([0.2938, 0.1862, 0.1759, 0.2007, 0.1322, 0.0163, 0.1527, 0.0840], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0118, 0.0126, 0.0124, 0.0108, 0.0097, 0.0091, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-24 05:26:10,010 INFO [train.py:894] (3/4) Epoch 29, batch 3500, loss[loss=0.1842, simple_loss=0.268, pruned_loss=0.05018, over 18623.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2566, pruned_loss=0.04358, over 3714043.36 frames. ], batch size: 169, lr: 3.90e-03, grad_scale: 8.0 2022-12-24 05:26:30,674 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp1.1 from training. Duration: 22.2954375 2022-12-24 05:26:36,019 INFO [train.py:894] (3/4) Epoch 30, batch 0, loss[loss=0.1611, simple_loss=0.2559, pruned_loss=0.03317, over 18553.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2559, pruned_loss=0.03317, over 18553.00 frames. ], batch size: 49, lr: 3.84e-03, grad_scale: 8.0 2022-12-24 05:26:36,020 INFO [train.py:919] (3/4) Computing validation loss 2022-12-24 05:26:44,252 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8802, 1.5227, 1.8676, 2.2508, 1.9379, 3.4791, 1.7039, 1.6211], device='cuda:3'), covar=tensor([0.0884, 0.1963, 0.1057, 0.0921, 0.1520, 0.0236, 0.1431, 0.1756], device='cuda:3'), in_proj_covar=tensor([0.0074, 0.0083, 0.0073, 0.0075, 0.0092, 0.0077, 0.0086, 0.0078], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0004], device='cuda:3') 2022-12-24 05:26:46,587 INFO [train.py:928] (3/4) Epoch 30, validation: loss=0.1625, simple_loss=0.2589, pruned_loss=0.03306, over 944034.00 frames. 2022-12-24 05:26:46,588 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-24 05:26:55,650 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101684.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 05:27:35,104 WARNING [train.py:1060] (3/4) Exclude cut with ID 298-126791-0067-24026-0_sp0.9 from training. Duration: 21.438875 2022-12-24 05:27:39,826 WARNING [train.py:1060] (3/4) Exclude cut with ID 5652-39938-0025-23684-0_sp0.9 from training. Duration: 22.2055625 2022-12-24 05:27:54,976 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2022-12-24 05:28:01,817 INFO [train.py:894] (3/4) Epoch 30, batch 50, loss[loss=0.16, simple_loss=0.2547, pruned_loss=0.03268, over 18580.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2631, pruned_loss=0.04085, over 838319.43 frames. ], batch size: 51, lr: 3.83e-03, grad_scale: 8.0 2022-12-24 05:28:23,948 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.60 vs. limit=5.0 2022-12-24 05:28:26,580 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101745.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 05:28:29,529 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7961, 1.8653, 2.1003, 1.2321, 2.1690, 2.2075, 1.4856, 2.5675], device='cuda:3'), covar=tensor([0.1399, 0.2124, 0.1531, 0.2159, 0.0796, 0.1312, 0.2654, 0.0613], device='cuda:3'), in_proj_covar=tensor([0.0199, 0.0217, 0.0211, 0.0197, 0.0172, 0.0220, 0.0218, 0.0197], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 05:28:35,183 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.974e+02 3.070e+02 3.838e+02 4.860e+02 9.164e+02, threshold=7.675e+02, percent-clipped=0.0 2022-12-24 05:28:52,105 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101763.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 05:29:14,270 INFO [train.py:894] (3/4) Epoch 30, batch 100, loss[loss=0.1861, simple_loss=0.2733, pruned_loss=0.0495, over 18700.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2586, pruned_loss=0.0393, over 1475379.86 frames. ], batch size: 97, lr: 3.83e-03, grad_scale: 8.0 2022-12-24 05:29:26,631 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6140, 1.5404, 1.6658, 1.5731, 1.2509, 3.7648, 1.4969, 2.0125], device='cuda:3'), covar=tensor([0.3411, 0.2226, 0.2066, 0.2204, 0.1547, 0.0169, 0.1745, 0.0977], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0118, 0.0125, 0.0123, 0.0107, 0.0097, 0.0091, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-24 05:30:03,209 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=101811.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 05:30:30,026 INFO [train.py:894] (3/4) Epoch 30, batch 150, loss[loss=0.1662, simple_loss=0.2575, pruned_loss=0.03742, over 18464.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2569, pruned_loss=0.03815, over 1970807.22 frames. ], batch size: 50, lr: 3.83e-03, grad_scale: 8.0 2022-12-24 05:30:38,824 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0 from training. Duration: 24.525 2022-12-24 05:31:03,396 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.089e+02 3.245e+02 3.691e+02 4.719e+02 9.898e+02, threshold=7.382e+02, percent-clipped=3.0 2022-12-24 05:31:09,294 WARNING [train.py:1060] (3/4) Exclude cut with ID 3699-47246-0007-3408-0_sp0.9 from training. Duration: 20.26675 2022-12-24 05:31:23,826 WARNING [train.py:1060] (3/4) Exclude cut with ID 7859-102521-0017-7548-0_sp0.9 from training. Duration: 27.25 2022-12-24 05:31:45,375 INFO [train.py:894] (3/4) Epoch 30, batch 200, loss[loss=0.1729, simple_loss=0.2669, pruned_loss=0.03944, over 18387.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2558, pruned_loss=0.03748, over 2356750.76 frames. ], batch size: 53, lr: 3.83e-03, grad_scale: 8.0 2022-12-24 05:32:22,034 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2022-12-24 05:32:35,371 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4582, 1.6873, 1.5046, 2.0295, 2.2224, 1.5206, 1.2907, 1.2966], device='cuda:3'), covar=tensor([0.1935, 0.1845, 0.1614, 0.1069, 0.1200, 0.1131, 0.2307, 0.1549], device='cuda:3'), in_proj_covar=tensor([0.0250, 0.0234, 0.0224, 0.0205, 0.0265, 0.0199, 0.0230, 0.0205], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 05:32:37,801 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0 from training. Duration: 21.68 2022-12-24 05:32:47,007 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101920.0, num_to_drop=1, layers_to_drop={1} 2022-12-24 05:32:48,258 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0 from training. Duration: 21.6300625 2022-12-24 05:32:58,775 INFO [train.py:894] (3/4) Epoch 30, batch 250, loss[loss=0.1686, simple_loss=0.2661, pruned_loss=0.03556, over 18636.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.255, pruned_loss=0.0372, over 2657151.07 frames. ], batch size: 69, lr: 3.83e-03, grad_scale: 8.0 2022-12-24 05:33:12,792 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0007-59342-0_sp0.9 from training. Duration: 24.033375 2022-12-24 05:33:31,317 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.846e+02 3.129e+02 3.860e+02 4.670e+02 7.401e+02, threshold=7.720e+02, percent-clipped=1.0 2022-12-24 05:33:34,992 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=3.05 vs. limit=5.0 2022-12-24 05:33:41,928 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6112, 1.5014, 1.5611, 1.4049, 1.0290, 2.3953, 0.8889, 1.5319], device='cuda:3'), covar=tensor([0.3110, 0.2159, 0.1955, 0.2241, 0.1526, 0.0312, 0.1798, 0.0877], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0118, 0.0125, 0.0123, 0.0107, 0.0097, 0.0091, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-24 05:34:07,699 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0 from training. Duration: 22.905 2022-12-24 05:34:09,907 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp1.1 from training. Duration: 23.4318125 2022-12-24 05:34:13,087 INFO [train.py:894] (3/4) Epoch 30, batch 300, loss[loss=0.1809, simple_loss=0.2692, pruned_loss=0.04635, over 18702.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2549, pruned_loss=0.03713, over 2890919.32 frames. ], batch size: 60, lr: 3.83e-03, grad_scale: 8.0 2022-12-24 05:34:17,771 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101981.0, num_to_drop=1, layers_to_drop={0} 2022-12-24 05:34:26,745 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101987.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 05:34:50,749 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2022-12-24 05:35:32,112 INFO [train.py:894] (3/4) Epoch 30, batch 350, loss[loss=0.1675, simple_loss=0.2511, pruned_loss=0.04191, over 18437.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2545, pruned_loss=0.03722, over 3072570.84 frames. ], batch size: 48, lr: 3.83e-03, grad_scale: 8.0 2022-12-24 05:35:49,662 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102040.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 05:36:01,139 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102048.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 05:36:05,012 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.772e+02 3.223e+02 3.909e+02 4.619e+02 7.719e+02, threshold=7.817e+02, percent-clipped=0.0 2022-12-24 05:36:09,514 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp1.1 from training. Duration: 20.82275 2022-12-24 05:36:11,124 WARNING [train.py:1060] (3/4) Exclude cut with ID 4278-13270-0009-59344-0_sp0.9 from training. Duration: 25.45 2022-12-24 05:36:32,987 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2022-12-24 05:36:45,861 INFO [train.py:894] (3/4) Epoch 30, batch 400, loss[loss=0.1503, simple_loss=0.2357, pruned_loss=0.03241, over 18685.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2552, pruned_loss=0.03744, over 3214969.14 frames. ], batch size: 46, lr: 3.83e-03, grad_scale: 8.0 2022-12-24 05:37:06,316 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.2315, 3.5993, 3.6595, 4.1703, 3.8860, 3.6620, 4.3682, 1.3382], device='cuda:3'), covar=tensor([0.0683, 0.0722, 0.0712, 0.0783, 0.1255, 0.1112, 0.0570, 0.5007], device='cuda:3'), in_proj_covar=tensor([0.0364, 0.0238, 0.0251, 0.0286, 0.0341, 0.0278, 0.0305, 0.0296], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 05:37:13,415 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0 from training. Duration: 25.775 2022-12-24 05:37:26,611 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2022-12-24 05:37:34,914 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0_sp0.9 from training. Duration: 22.25 2022-12-24 05:38:01,028 INFO [train.py:894] (3/4) Epoch 30, batch 450, loss[loss=0.158, simple_loss=0.2592, pruned_loss=0.02837, over 18721.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2564, pruned_loss=0.03803, over 3326466.62 frames. ], batch size: 54, lr: 3.83e-03, grad_scale: 16.0 2022-12-24 05:38:01,068 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0 from training. Duration: 26.205 2022-12-24 05:38:18,723 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp0.9 from training. Duration: 30.1555625 2022-12-24 05:38:20,578 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9803, 2.0088, 2.3447, 1.2924, 2.3305, 2.3376, 1.6715, 2.7430], device='cuda:3'), covar=tensor([0.1282, 0.2058, 0.1375, 0.2178, 0.0743, 0.1275, 0.2583, 0.0567], device='cuda:3'), in_proj_covar=tensor([0.0197, 0.0216, 0.0209, 0.0195, 0.0171, 0.0219, 0.0217, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 05:38:24,344 WARNING [train.py:1060] (3/4) Exclude cut with ID 1265-135635-0050-6781-0_sp0.9 from training. Duration: 21.8333125 2022-12-24 05:38:34,420 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.124e+02 3.155e+02 3.798e+02 4.778e+02 1.868e+03, threshold=7.596e+02, percent-clipped=6.0 2022-12-24 05:38:34,472 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp1.1 from training. Duration: 20.6545625 2022-12-24 05:39:15,305 INFO [train.py:894] (3/4) Epoch 30, batch 500, loss[loss=0.1285, simple_loss=0.2128, pruned_loss=0.02204, over 18400.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2553, pruned_loss=0.03752, over 3411904.02 frames. ], batch size: 42, lr: 3.83e-03, grad_scale: 16.0 2022-12-24 05:39:16,979 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0045-39920-0_sp0.9 from training. Duration: 20.52225 2022-12-24 05:39:35,483 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp0.9 from training. Duration: 29.1166875 2022-12-24 05:40:31,525 INFO [train.py:894] (3/4) Epoch 30, batch 550, loss[loss=0.1708, simple_loss=0.2585, pruned_loss=0.04152, over 18575.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2561, pruned_loss=0.03779, over 3478482.40 frames. ], batch size: 49, lr: 3.83e-03, grad_scale: 16.0 2022-12-24 05:40:34,973 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133211-0007-59831-0_sp0.9 from training. Duration: 21.388875 2022-12-24 05:41:05,079 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.220e+02 3.027e+02 3.739e+02 4.714e+02 8.555e+02, threshold=7.479e+02, percent-clipped=1.0 2022-12-24 05:41:12,816 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0 from training. Duration: 22.72 2022-12-24 05:41:14,702 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0_sp0.9 from training. Duration: 22.7444375 2022-12-24 05:41:39,501 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6073, 1.4891, 1.5817, 1.4895, 0.9833, 3.0140, 1.1445, 1.7241], device='cuda:3'), covar=tensor([0.3154, 0.2243, 0.2017, 0.2199, 0.1690, 0.0226, 0.1850, 0.0920], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0119, 0.0125, 0.0123, 0.0108, 0.0097, 0.0091, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-24 05:41:43,492 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102276.0, num_to_drop=1, layers_to_drop={3} 2022-12-24 05:41:46,135 INFO [train.py:894] (3/4) Epoch 30, batch 600, loss[loss=0.1616, simple_loss=0.2574, pruned_loss=0.03295, over 18602.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2563, pruned_loss=0.03802, over 3530833.24 frames. ], batch size: 57, lr: 3.82e-03, grad_scale: 16.0 2022-12-24 05:41:56,078 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2022-12-24 05:41:58,163 WARNING [train.py:1060] (3/4) Exclude cut with ID 4133-6541-0027-40495-0_sp1.1 from training. Duration: 0.9681875 2022-12-24 05:42:00,993 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0_sp0.9 from training. Duration: 22.3166875 2022-12-24 05:42:06,755 WARNING [train.py:1060] (3/4) Exclude cut with ID 543-133212-0015-59917-0_sp0.9 from training. Duration: 21.8166875 2022-12-24 05:42:47,482 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7803, 1.6949, 1.7795, 1.6478, 1.5521, 3.6856, 1.7138, 2.1898], device='cuda:3'), covar=tensor([0.3090, 0.2059, 0.1918, 0.2156, 0.1373, 0.0183, 0.1610, 0.0842], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0118, 0.0125, 0.0123, 0.0108, 0.0096, 0.0091, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-24 05:43:02,703 INFO [train.py:894] (3/4) Epoch 30, batch 650, loss[loss=0.1925, simple_loss=0.2816, pruned_loss=0.05174, over 18536.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2564, pruned_loss=0.03782, over 3571251.89 frames. ], batch size: 55, lr: 3.82e-03, grad_scale: 8.0 2022-12-24 05:43:21,011 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102340.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 05:43:25,347 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102343.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 05:43:26,725 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2325, 2.8585, 2.7442, 1.1286, 3.0557, 2.1505, 0.4736, 1.8021], device='cuda:3'), covar=tensor([0.2203, 0.1527, 0.1793, 0.3713, 0.1097, 0.1094, 0.4677, 0.1706], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0151, 0.0163, 0.0126, 0.0153, 0.0117, 0.0146, 0.0119], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-24 05:43:38,572 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.067e+02 2.967e+02 3.701e+02 4.563e+02 8.252e+02, threshold=7.403e+02, percent-clipped=2.0 2022-12-24 05:43:48,944 WARNING [train.py:1060] (3/4) Exclude cut with ID 4957-30119-0041-23990-0_sp0.9 from training. Duration: 20.22775 2022-12-24 05:44:16,265 INFO [train.py:894] (3/4) Epoch 30, batch 700, loss[loss=0.1962, simple_loss=0.2852, pruned_loss=0.05361, over 18668.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2561, pruned_loss=0.03774, over 3602975.90 frames. ], batch size: 78, lr: 3.82e-03, grad_scale: 8.0 2022-12-24 05:44:31,288 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=102388.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 05:44:32,515 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0_sp1.1 from training. Duration: 24.67275 2022-12-24 05:44:48,456 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.87 vs. limit=5.0 2022-12-24 05:45:01,277 WARNING [train.py:1060] (3/4) Exclude cut with ID 3082-165428-0081-50734-0_sp0.9 from training. Duration: 21.8055625 2022-12-24 05:45:32,301 INFO [train.py:894] (3/4) Epoch 30, batch 750, loss[loss=0.1614, simple_loss=0.2565, pruned_loss=0.0332, over 18649.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2562, pruned_loss=0.03782, over 3626543.24 frames. ], batch size: 69, lr: 3.82e-03, grad_scale: 8.0 2022-12-24 05:45:36,639 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0_sp0.9 from training. Duration: 22.6666875 2022-12-24 05:46:05,139 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5155, 2.3216, 1.7810, 0.7370, 1.6523, 2.0570, 1.8201, 1.9739], device='cuda:3'), covar=tensor([0.0705, 0.0566, 0.1350, 0.1801, 0.1341, 0.1505, 0.1698, 0.0828], device='cuda:3'), in_proj_covar=tensor([0.0177, 0.0190, 0.0211, 0.0190, 0.0212, 0.0206, 0.0219, 0.0205], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 05:46:07,507 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.493e+02 3.194e+02 3.640e+02 4.304e+02 1.220e+03, threshold=7.280e+02, percent-clipped=2.0 2022-12-24 05:46:39,086 WARNING [train.py:1060] (3/4) Exclude cut with ID 2411-132532-0017-82279-0_sp1.1 from training. Duration: 0.9681875 2022-12-24 05:46:47,414 INFO [train.py:894] (3/4) Epoch 30, batch 800, loss[loss=0.1928, simple_loss=0.2825, pruned_loss=0.05153, over 18638.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2564, pruned_loss=0.03805, over 3646568.62 frames. ], batch size: 62, lr: 3.82e-03, grad_scale: 8.0 2022-12-24 05:47:04,501 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0 from training. Duration: 22.485 2022-12-24 05:47:38,290 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2022-12-24 05:47:44,920 WARNING [train.py:1060] (3/4) Exclude cut with ID 3972-170212-0014-23379-0_sp1.1 from training. Duration: 23.82275 2022-12-24 05:47:58,221 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0 from training. Duration: 20.77 2022-12-24 05:48:00,506 INFO [train.py:894] (3/4) Epoch 30, batch 850, loss[loss=0.1893, simple_loss=0.2778, pruned_loss=0.0504, over 18499.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2561, pruned_loss=0.03802, over 3660836.94 frames. ], batch size: 58, lr: 3.82e-03, grad_scale: 8.0 2022-12-24 05:48:05,875 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64292-0017-15984-0_sp0.9 from training. Duration: 24.088875 2022-12-24 05:48:34,379 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp1.1 from training. Duration: 20.4409375 2022-12-24 05:48:35,719 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.988e+02 3.397e+02 3.948e+02 4.845e+02 7.696e+02, threshold=7.896e+02, percent-clipped=1.0 2022-12-24 05:49:11,959 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102576.0, num_to_drop=1, layers_to_drop={0} 2022-12-24 05:49:14,489 INFO [train.py:894] (3/4) Epoch 30, batch 900, loss[loss=0.1655, simple_loss=0.2604, pruned_loss=0.03532, over 18496.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2565, pruned_loss=0.0379, over 3672831.68 frames. ], batch size: 52, lr: 3.82e-03, grad_scale: 8.0 2022-12-24 05:49:16,636 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2022-12-24 05:49:48,898 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0_sp0.9 from training. Duration: 22.511125 2022-12-24 05:49:48,916 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0 from training. Duration: 20.675 2022-12-24 05:50:23,371 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=102624.0, num_to_drop=1, layers_to_drop={1} 2022-12-24 05:50:28,617 INFO [train.py:894] (3/4) Epoch 30, batch 950, loss[loss=0.1774, simple_loss=0.2652, pruned_loss=0.04483, over 18516.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2567, pruned_loss=0.03782, over 3681555.74 frames. ], batch size: 58, lr: 3.82e-03, grad_scale: 8.0 2022-12-24 05:50:51,225 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102643.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 05:51:03,634 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.806e+02 3.148e+02 4.122e+02 4.963e+02 1.598e+03, threshold=8.245e+02, percent-clipped=4.0 2022-12-24 05:51:27,731 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62850-0007-91323-0_sp0.9 from training. Duration: 24.9833125 2022-12-24 05:51:42,752 INFO [train.py:894] (3/4) Epoch 30, batch 1000, loss[loss=0.14, simple_loss=0.2256, pruned_loss=0.02716, over 18574.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.256, pruned_loss=0.03741, over 3690460.74 frames. ], batch size: 41, lr: 3.82e-03, grad_scale: 8.0 2022-12-24 05:51:59,625 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0047-9341-0 from training. Duration: 27.14 2022-12-24 05:52:01,634 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=102691.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 05:52:06,813 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6940, 1.8367, 2.0774, 1.2414, 2.0476, 2.2274, 1.5824, 2.5322], device='cuda:3'), covar=tensor([0.1314, 0.2059, 0.1273, 0.1838, 0.0784, 0.1007, 0.2467, 0.0494], device='cuda:3'), in_proj_covar=tensor([0.0195, 0.0214, 0.0206, 0.0193, 0.0170, 0.0216, 0.0214, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 05:52:13,638 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0 from training. Duration: 22.44 2022-12-24 05:52:57,216 INFO [train.py:894] (3/4) Epoch 30, batch 1050, loss[loss=0.1436, simple_loss=0.2407, pruned_loss=0.02323, over 18679.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2557, pruned_loss=0.03725, over 3695367.89 frames. ], batch size: 48, lr: 3.82e-03, grad_scale: 8.0 2022-12-24 05:53:32,764 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.729e+02 3.103e+02 3.715e+02 4.329e+02 7.586e+02, threshold=7.430e+02, percent-clipped=0.0 2022-12-24 05:53:35,455 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0060-62364-0_sp0.9 from training. Duration: 21.361125 2022-12-24 05:53:41,427 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp1.1 from training. Duration: 27.0318125 2022-12-24 05:53:50,884 WARNING [train.py:1060] (3/4) Exclude cut with ID 5622-44585-0006-90525-0_sp0.9 from training. Duration: 28.638875 2022-12-24 05:54:05,516 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0054-76830-0 from training. Duration: 20.4 2022-12-24 05:54:11,562 INFO [train.py:894] (3/4) Epoch 30, batch 1100, loss[loss=0.1746, simple_loss=0.2643, pruned_loss=0.0424, over 18673.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2562, pruned_loss=0.03751, over 3698516.34 frames. ], batch size: 62, lr: 3.82e-03, grad_scale: 8.0 2022-12-24 05:54:38,057 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0071-62375-0 from training. Duration: 20.025 2022-12-24 05:54:38,069 WARNING [train.py:1060] (3/4) Exclude cut with ID 2364-131735-0112-64612-0_sp0.9 from training. Duration: 20.488875 2022-12-24 05:54:43,817 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0 from training. Duration: 29.735 2022-12-24 05:54:46,412 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2022-12-24 05:54:50,504 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3104, 2.1497, 2.5733, 2.5922, 2.5088, 3.8075, 2.1716, 2.2319], device='cuda:3'), covar=tensor([0.0723, 0.1357, 0.0919, 0.0745, 0.1049, 0.0272, 0.1079, 0.1191], device='cuda:3'), in_proj_covar=tensor([0.0074, 0.0083, 0.0073, 0.0075, 0.0092, 0.0078, 0.0086, 0.0078], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0004], device='cuda:3') 2022-12-24 05:55:26,094 INFO [train.py:894] (3/4) Epoch 30, batch 1150, loss[loss=0.1711, simple_loss=0.2683, pruned_loss=0.03696, over 18603.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2567, pruned_loss=0.03764, over 3702314.37 frames. ], batch size: 56, lr: 3.81e-03, grad_scale: 8.0 2022-12-24 05:55:59,368 WARNING [train.py:1060] (3/4) Exclude cut with ID 7276-92427-0014-12983-0_sp0.9 from training. Duration: 21.3055625 2022-12-24 05:56:01,178 WARNING [train.py:1060] (3/4) Exclude cut with ID 1025-75365-0008-79168-0_sp0.9 from training. Duration: 22.0666875 2022-12-24 05:56:01,493 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=102851.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 05:56:02,554 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.993e+02 3.311e+02 3.980e+02 4.684e+02 1.228e+03, threshold=7.959e+02, percent-clipped=4.0 2022-12-24 05:56:41,191 INFO [train.py:894] (3/4) Epoch 30, batch 1200, loss[loss=0.1436, simple_loss=0.2312, pruned_loss=0.02797, over 18427.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2567, pruned_loss=0.0376, over 3705745.86 frames. ], batch size: 48, lr: 3.81e-03, grad_scale: 8.0 2022-12-24 05:56:55,395 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=102887.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 05:57:33,192 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102912.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 05:57:47,529 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0062-62366-0 from training. Duration: 20.26 2022-12-24 05:57:51,095 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2022-12-24 05:57:54,756 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.6322, 1.6412, 1.7819, 1.1675, 1.8873, 1.9463, 1.3881, 2.2273], device='cuda:3'), covar=tensor([0.1102, 0.2043, 0.1279, 0.1704, 0.0743, 0.1005, 0.2535, 0.0558], device='cuda:3'), in_proj_covar=tensor([0.0195, 0.0214, 0.0207, 0.0193, 0.0170, 0.0216, 0.0215, 0.0194], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 05:57:55,701 INFO [train.py:894] (3/4) Epoch 30, batch 1250, loss[loss=0.1882, simple_loss=0.2823, pruned_loss=0.04703, over 18710.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.257, pruned_loss=0.0376, over 3708297.11 frames. ], batch size: 65, lr: 3.81e-03, grad_scale: 8.0 2022-12-24 05:58:00,452 WARNING [train.py:1060] (3/4) Exclude cut with ID 5239-32139-0030-9324-0_sp0.9 from training. Duration: 21.3444375 2022-12-24 05:58:11,061 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.1165, 2.0040, 2.0544, 1.9962, 1.8397, 5.1944, 1.8913, 2.4327], device='cuda:3'), covar=tensor([0.2990, 0.2009, 0.1856, 0.2067, 0.1253, 0.0083, 0.1583, 0.0845], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0119, 0.0125, 0.0124, 0.0108, 0.0097, 0.0092, 0.0091], device='cuda:3'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2022-12-24 05:58:26,654 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102948.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 05:58:32,825 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.010e+02 3.139e+02 3.911e+02 4.609e+02 1.084e+03, threshold=7.821e+02, percent-clipped=2.0 2022-12-24 05:58:55,634 WARNING [train.py:1060] (3/4) Exclude cut with ID 497-129325-0061-62254-0_sp1.1 from training. Duration: 0.97725 2022-12-24 05:59:12,452 INFO [train.py:894] (3/4) Epoch 30, batch 1300, loss[loss=0.1597, simple_loss=0.2595, pruned_loss=0.02991, over 18379.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2559, pruned_loss=0.03713, over 3708643.92 frames. ], batch size: 51, lr: 3.81e-03, grad_scale: 8.0 2022-12-24 05:59:37,129 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0031-39906-0_sp0.9 from training. Duration: 22.97225 2022-12-24 06:00:08,786 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0047-39922-0_sp0.9 from training. Duration: 21.97775 2022-12-24 06:00:16,369 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.50 vs. limit=2.0 2022-12-24 06:00:24,071 WARNING [train.py:1060] (3/4) Exclude cut with ID 1112-1043-0006-89194-0_sp0.9 from training. Duration: 21.8333125 2022-12-24 06:00:27,019 INFO [train.py:894] (3/4) Epoch 30, batch 1350, loss[loss=0.1465, simple_loss=0.2326, pruned_loss=0.03016, over 18526.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2563, pruned_loss=0.03703, over 3710045.51 frames. ], batch size: 47, lr: 3.81e-03, grad_scale: 8.0 2022-12-24 06:00:34,223 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0031-94921-0 from training. Duration: 20.47 2022-12-24 06:01:01,611 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.043e+02 3.138e+02 3.681e+02 4.641e+02 8.476e+02, threshold=7.361e+02, percent-clipped=1.0 2022-12-24 06:01:37,462 WARNING [train.py:1060] (3/4) Exclude cut with ID 7395-89880-0037-39912-0_sp0.9 from training. Duration: 20.67225 2022-12-24 06:01:41,954 INFO [train.py:894] (3/4) Epoch 30, batch 1400, loss[loss=0.16, simple_loss=0.261, pruned_loss=0.02954, over 18593.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2554, pruned_loss=0.0368, over 3710925.73 frames. ], batch size: 56, lr: 3.81e-03, grad_scale: 8.0 2022-12-24 06:01:55,104 WARNING [train.py:1060] (3/4) Exclude cut with ID 1914-133440-0024-94914-0_sp0.9 from training. Duration: 25.2444375 2022-12-24 06:02:11,529 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103098.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:02:20,438 WARNING [train.py:1060] (3/4) Exclude cut with ID 3340-169293-0021-76797-0_sp0.9 from training. Duration: 21.1445 2022-12-24 06:02:23,542 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9404, 1.3656, 0.9355, 1.3596, 2.1275, 1.1639, 1.6274, 1.6577], device='cuda:3'), covar=tensor([0.1421, 0.1996, 0.2078, 0.1500, 0.1722, 0.1876, 0.1321, 0.1736], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0097, 0.0115, 0.0097, 0.0119, 0.0091, 0.0098, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-24 06:02:35,443 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6194, 2.2292, 1.6844, 2.4208, 2.0211, 2.2528, 2.1916, 2.6767], device='cuda:3'), covar=tensor([0.2298, 0.3568, 0.2262, 0.3090, 0.4248, 0.1148, 0.3346, 0.1079], device='cuda:3'), in_proj_covar=tensor([0.0303, 0.0304, 0.0258, 0.0352, 0.0284, 0.0238, 0.0301, 0.0227], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 06:02:55,826 INFO [train.py:894] (3/4) Epoch 30, batch 1450, loss[loss=0.1482, simple_loss=0.2289, pruned_loss=0.03373, over 18525.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2562, pruned_loss=0.03693, over 3712540.03 frames. ], batch size: 44, lr: 3.81e-03, grad_scale: 8.0 2022-12-24 06:03:01,890 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103132.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:03:19,522 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2022-12-24 06:03:31,248 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.686e+02 3.036e+02 3.905e+02 4.908e+02 8.550e+02, threshold=7.811e+02, percent-clipped=3.0 2022-12-24 06:03:34,143 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0079-62383-0_sp0.9 from training. Duration: 33.038875 2022-12-24 06:03:41,431 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103159.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:04:09,806 INFO [train.py:894] (3/4) Epoch 30, batch 1500, loss[loss=0.1748, simple_loss=0.2636, pruned_loss=0.04295, over 18608.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2557, pruned_loss=0.03658, over 3712500.22 frames. ], batch size: 78, lr: 3.81e-03, grad_scale: 8.0 2022-12-24 06:04:14,001 WARNING [train.py:1060] (3/4) Exclude cut with ID 6426-64291-0000-16059-0_sp0.9 from training. Duration: 20.0944375 2022-12-24 06:04:27,004 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp1.1 from training. Duration: 20.4 2022-12-24 06:04:31,967 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103193.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:04:34,188 WARNING [train.py:1060] (3/4) Exclude cut with ID 6330-62851-0022-91297-0 from training. Duration: 20.085 2022-12-24 06:04:47,534 WARNING [train.py:1060] (3/4) Exclude cut with ID 4860-13185-0032-76709-0_sp0.9 from training. Duration: 23.07775 2022-12-24 06:04:53,223 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103207.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:05:10,711 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2022-12-24 06:05:24,575 INFO [train.py:894] (3/4) Epoch 30, batch 1550, loss[loss=0.1739, simple_loss=0.2693, pruned_loss=0.03924, over 18583.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2553, pruned_loss=0.03635, over 3712794.99 frames. ], batch size: 57, lr: 3.81e-03, grad_scale: 8.0 2022-12-24 06:05:31,817 WARNING [train.py:1060] (3/4) Exclude cut with ID 2929-85685-0044-62348-0_sp0.9 from training. Duration: 24.9333125 2022-12-24 06:05:46,996 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103243.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:06:00,766 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 1.963e+02 3.194e+02 3.778e+02 4.720e+02 8.597e+02, threshold=7.556e+02, percent-clipped=1.0 2022-12-24 06:06:02,553 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.5236, 1.8322, 0.8150, 1.7955, 2.6147, 1.8027, 2.2382, 2.2540], device='cuda:3'), covar=tensor([0.1409, 0.1902, 0.2325, 0.1425, 0.1533, 0.1765, 0.1224, 0.1648], device='cuda:3'), in_proj_covar=tensor([0.0095, 0.0098, 0.0117, 0.0098, 0.0120, 0.0092, 0.0099, 0.0095], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-24 06:06:13,966 WARNING [train.py:1060] (3/4) Exclude cut with ID 5118-111612-0016-124680-0_sp0.9 from training. Duration: 20.388875 2022-12-24 06:06:20,273 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp1.1 from training. Duration: 20.3590625 2022-12-24 06:06:25,536 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3855, 1.4976, 1.2852, 1.6188, 1.6722, 2.9869, 1.5345, 1.5215], device='cuda:3'), covar=tensor([0.0883, 0.1758, 0.1075, 0.0914, 0.1369, 0.0235, 0.1228, 0.1542], device='cuda:3'), in_proj_covar=tensor([0.0074, 0.0083, 0.0073, 0.0075, 0.0091, 0.0077, 0.0085, 0.0078], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2022-12-24 06:06:39,351 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2022-12-24 06:06:39,748 INFO [train.py:894] (3/4) Epoch 30, batch 1600, loss[loss=0.1573, simple_loss=0.2379, pruned_loss=0.03836, over 18379.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.256, pruned_loss=0.03717, over 3713396.26 frames. ], batch size: 46, lr: 3.81e-03, grad_scale: 8.0 2022-12-24 06:07:26,612 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0_sp1.1 from training. Duration: 0.836375 2022-12-24 06:07:53,613 INFO [train.py:894] (3/4) Epoch 30, batch 1650, loss[loss=0.1952, simple_loss=0.2858, pruned_loss=0.05229, over 18590.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2568, pruned_loss=0.0381, over 3713703.40 frames. ], batch size: 56, lr: 3.81e-03, grad_scale: 8.0 2022-12-24 06:08:08,736 WARNING [train.py:1060] (3/4) Exclude cut with ID 8565-290391-0049-67394-0_sp0.9 from training. Duration: 21.3166875 2022-12-24 06:08:29,720 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.022e+02 3.417e+02 4.120e+02 5.329e+02 9.621e+02, threshold=8.239e+02, percent-clipped=4.0 2022-12-24 06:08:39,709 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0029-104863-0_sp0.9 from training. Duration: 22.1055625 2022-12-24 06:08:51,035 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp1.1 from training. Duration: 21.77725 2022-12-24 06:08:51,337 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103366.0, num_to_drop=1, layers_to_drop={0} 2022-12-24 06:09:08,247 INFO [train.py:894] (3/4) Epoch 30, batch 1700, loss[loss=0.1562, simple_loss=0.2383, pruned_loss=0.03706, over 18382.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2568, pruned_loss=0.03887, over 3713272.67 frames. ], batch size: 46, lr: 3.80e-03, grad_scale: 8.0 2022-12-24 06:09:11,471 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp0.9 from training. Duration: 27.8166875 2022-12-24 06:09:35,254 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp1.1 from training. Duration: 22.5090625 2022-12-24 06:09:41,466 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0 from training. Duration: 25.035 2022-12-24 06:09:59,289 WARNING [train.py:1060] (3/4) Exclude cut with ID 774-127930-0014-10412-0_sp1.1 from training. Duration: 0.95 2022-12-24 06:10:16,933 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp0.9 from training. Duration: 0.92225 2022-12-24 06:10:23,385 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103427.0, num_to_drop=1, layers_to_drop={1} 2022-12-24 06:10:24,381 INFO [train.py:894] (3/4) Epoch 30, batch 1750, loss[loss=0.1454, simple_loss=0.2312, pruned_loss=0.02978, over 18378.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2568, pruned_loss=0.03965, over 3713053.32 frames. ], batch size: 46, lr: 3.80e-03, grad_scale: 8.0 2022-12-24 06:10:26,157 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4969, 1.0749, 2.0743, 3.1843, 2.2870, 2.6415, 1.0651, 2.2837], device='cuda:3'), covar=tensor([0.1960, 0.1805, 0.1377, 0.0699, 0.1030, 0.1039, 0.2020, 0.1067], device='cuda:3'), in_proj_covar=tensor([0.0104, 0.0120, 0.0139, 0.0159, 0.0108, 0.0145, 0.0131, 0.0117], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2022-12-24 06:10:44,807 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0 from training. Duration: 21.97 2022-12-24 06:11:00,831 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.344e+02 3.614e+02 4.053e+02 5.108e+02 1.085e+03, threshold=8.106e+02, percent-clipped=5.0 2022-12-24 06:11:02,262 WARNING [train.py:1060] (3/4) Exclude cut with ID 7492-105653-0055-62765-0_sp0.9 from training. Duration: 21.97225 2022-12-24 06:11:04,093 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp0.9 from training. Duration: 25.3333125 2022-12-24 06:11:04,187 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103454.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:11:15,026 WARNING [train.py:1060] (3/4) Exclude cut with ID 5172-29468-0015-19128-0_sp0.9 from training. Duration: 21.5055625 2022-12-24 06:11:25,484 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0_sp1.1 from training. Duration: 20.72725 2022-12-24 06:11:40,432 INFO [train.py:894] (3/4) Epoch 30, batch 1800, loss[loss=0.1677, simple_loss=0.2575, pruned_loss=0.03896, over 18593.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2577, pruned_loss=0.04111, over 3713225.83 frames. ], batch size: 51, lr: 3.80e-03, grad_scale: 8.0 2022-12-24 06:11:55,379 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103488.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:11:59,406 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp0.9 from training. Duration: 26.32775 2022-12-24 06:12:19,787 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.9785, 2.5553, 2.3435, 3.0632, 2.5579, 2.5723, 2.3523, 3.1979], device='cuda:3'), covar=tensor([0.1828, 0.2864, 0.1564, 0.2387, 0.3072, 0.1025, 0.3321, 0.0778], device='cuda:3'), in_proj_covar=tensor([0.0302, 0.0303, 0.0257, 0.0351, 0.0283, 0.0237, 0.0299, 0.0225], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 06:12:24,073 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103507.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:12:30,012 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0 from training. Duration: 20.025 2022-12-24 06:12:34,283 WARNING [train.py:1060] (3/4) Exclude cut with ID 6709-74022-0004-86860-0_sp1.1 from training. Duration: 0.9409375 2022-12-24 06:12:35,859 WARNING [train.py:1060] (3/4) Exclude cut with ID 4757-1811-0023-62229-0_sp0.9 from training. Duration: 21.37775 2022-12-24 06:12:55,188 INFO [train.py:894] (3/4) Epoch 30, batch 1850, loss[loss=0.1796, simple_loss=0.2684, pruned_loss=0.04545, over 18583.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2576, pruned_loss=0.04184, over 3713248.90 frames. ], batch size: 56, lr: 3.80e-03, grad_scale: 8.0 2022-12-24 06:12:55,231 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0004-25974-0_sp0.9 from training. Duration: 21.17225 2022-12-24 06:12:55,242 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0_sp0.9 from training. Duration: 27.511125 2022-12-24 06:13:14,492 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.51 vs. limit=5.0 2022-12-24 06:13:18,049 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103543.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:13:31,001 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.198e+02 4.176e+02 5.067e+02 6.015e+02 1.138e+03, threshold=1.013e+03, percent-clipped=6.0 2022-12-24 06:13:31,060 WARNING [train.py:1060] (3/4) Exclude cut with ID 453-131332-0000-47844-0 from training. Duration: 22.8 2022-12-24 06:13:35,318 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0 from training. Duration: 22.585 2022-12-24 06:13:36,002 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=103555.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:14:06,214 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0001-146967-0_sp0.9 from training. Duration: 22.0166875 2022-12-24 06:14:11,251 INFO [train.py:894] (3/4) Epoch 30, batch 1900, loss[loss=0.213, simple_loss=0.2908, pruned_loss=0.0676, over 18708.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2583, pruned_loss=0.04291, over 3715070.39 frames. ], batch size: 79, lr: 3.80e-03, grad_scale: 8.0 2022-12-24 06:14:22,746 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp1.1 from training. Duration: 24.395375 2022-12-24 06:14:28,650 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp0.9 from training. Duration: 27.47775 2022-12-24 06:14:30,243 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=103591.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:14:33,097 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0_sp0.9 from training. Duration: 24.8833125 2022-12-24 06:14:36,062 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0 from training. Duration: 23.39 2022-12-24 06:14:42,295 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp0.9 from training. Duration: 28.72225 2022-12-24 06:14:51,623 WARNING [train.py:1060] (3/4) Exclude cut with ID 585-294811-0110-133686-0_sp0.9 from training. Duration: 20.8944375 2022-12-24 06:14:57,897 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.3984, 1.1989, 1.6691, 2.3481, 1.6566, 2.4532, 0.8007, 1.8237], device='cuda:3'), covar=tensor([0.1733, 0.1564, 0.1143, 0.0779, 0.1119, 0.0777, 0.1783, 0.1041], device='cuda:3'), in_proj_covar=tensor([0.0104, 0.0120, 0.0139, 0.0159, 0.0108, 0.0146, 0.0131, 0.0117], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2022-12-24 06:15:05,749 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0_sp0.9 from training. Duration: 23.8444375 2022-12-24 06:15:26,314 INFO [train.py:894] (3/4) Epoch 30, batch 1950, loss[loss=0.2104, simple_loss=0.2929, pruned_loss=0.06398, over 18653.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2587, pruned_loss=0.04369, over 3715639.10 frames. ], batch size: 60, lr: 3.80e-03, grad_scale: 8.0 2022-12-24 06:15:30,771 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0 from training. Duration: 25.85 2022-12-24 06:15:30,784 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0 from training. Duration: 21.39 2022-12-24 06:15:41,178 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0 from training. Duration: 27.92 2022-12-24 06:16:04,199 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.424e+02 3.831e+02 4.811e+02 6.471e+02 1.162e+03, threshold=9.622e+02, percent-clipped=1.0 2022-12-24 06:16:09,881 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0039-130165-0_sp0.9 from training. Duration: 20.661125 2022-12-24 06:16:37,205 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0043-15874-0_sp0.9 from training. Duration: 20.07225 2022-12-24 06:16:41,392 INFO [train.py:894] (3/4) Epoch 30, batch 2000, loss[loss=0.1799, simple_loss=0.2726, pruned_loss=0.04359, over 18551.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.258, pruned_loss=0.04364, over 3715621.87 frames. ], batch size: 57, lr: 3.80e-03, grad_scale: 8.0 2022-12-24 06:16:44,418 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0 from training. Duration: 21.01 2022-12-24 06:17:48,406 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0 from training. Duration: 20.65 2022-12-24 06:17:48,511 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103722.0, num_to_drop=1, layers_to_drop={3} 2022-12-24 06:17:57,499 INFO [train.py:894] (3/4) Epoch 30, batch 2050, loss[loss=0.1857, simple_loss=0.2708, pruned_loss=0.05025, over 18684.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2584, pruned_loss=0.04404, over 3715453.21 frames. ], batch size: 60, lr: 3.80e-03, grad_scale: 8.0 2022-12-24 06:17:57,535 WARNING [train.py:1060] (3/4) Exclude cut with ID 5796-66357-0007-116447-0 from training. Duration: 21.46 2022-12-24 06:18:15,884 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103740.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:18:34,846 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.766e+02 4.041e+02 4.718e+02 5.719e+02 1.210e+03, threshold=9.436e+02, percent-clipped=5.0 2022-12-24 06:18:36,572 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103754.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:18:41,999 WARNING [train.py:1060] (3/4) Exclude cut with ID 3557-8342-0013-54691-0 from training. Duration: 0.92 2022-12-24 06:18:42,372 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.6929, 2.2190, 1.7993, 0.8257, 1.9155, 2.0928, 1.5803, 1.9795], device='cuda:3'), covar=tensor([0.0723, 0.0872, 0.1739, 0.2213, 0.1559, 0.1764, 0.2370, 0.1083], device='cuda:3'), in_proj_covar=tensor([0.0178, 0.0191, 0.0213, 0.0192, 0.0214, 0.0208, 0.0220, 0.0207], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 06:18:48,445 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0023-13010-0_sp0.9 from training. Duration: 23.7666875 2022-12-24 06:18:53,304 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103765.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:18:59,538 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7674, 2.3520, 1.8880, 0.8852, 2.0336, 2.2522, 1.7268, 2.0800], device='cuda:3'), covar=tensor([0.0621, 0.0690, 0.1316, 0.1788, 0.1182, 0.1453, 0.1917, 0.0837], device='cuda:3'), in_proj_covar=tensor([0.0177, 0.0191, 0.0213, 0.0192, 0.0213, 0.0207, 0.0220, 0.0207], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 06:19:11,914 INFO [train.py:894] (3/4) Epoch 30, batch 2100, loss[loss=0.1685, simple_loss=0.2478, pruned_loss=0.0446, over 18670.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2581, pruned_loss=0.04453, over 3714831.69 frames. ], batch size: 48, lr: 3.80e-03, grad_scale: 8.0 2022-12-24 06:19:24,122 WARNING [train.py:1060] (3/4) Exclude cut with ID 8544-281189-0060-101339-0_sp0.9 from training. Duration: 20.861125 2022-12-24 06:19:28,384 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103788.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:19:35,216 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0_sp0.9 from training. Duration: 22.711125 2022-12-24 06:19:43,026 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4911, 1.4286, 1.3601, 1.4508, 1.7368, 1.5548, 1.5769, 1.2462], device='cuda:3'), covar=tensor([0.0344, 0.0268, 0.0571, 0.0240, 0.0221, 0.0492, 0.0317, 0.0373], device='cuda:3'), in_proj_covar=tensor([0.0100, 0.0130, 0.0160, 0.0125, 0.0121, 0.0126, 0.0104, 0.0132], device='cuda:3'), out_proj_covar=tensor([7.8668e-05, 1.0290e-04, 1.3054e-04, 9.9300e-05, 9.6250e-05, 9.6230e-05, 8.0519e-05, 1.0396e-04], device='cuda:3') 2022-12-24 06:19:47,500 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103801.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:19:48,678 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=103802.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:20:16,727 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp1.1 from training. Duration: 22.986375 2022-12-24 06:20:24,147 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103826.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:20:26,361 INFO [train.py:894] (3/4) Epoch 30, batch 2150, loss[loss=0.1733, simple_loss=0.2503, pruned_loss=0.0482, over 18473.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2585, pruned_loss=0.04451, over 3715429.63 frames. ], batch size: 43, lr: 3.80e-03, grad_scale: 8.0 2022-12-24 06:20:30,650 WARNING [train.py:1060] (3/4) Exclude cut with ID 8040-260924-0003-80960-0_sp0.9 from training. Duration: 22.07225 2022-12-24 06:20:38,265 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0045-26330-0_sp0.9 from training. Duration: 20.3055625 2022-12-24 06:20:39,877 WARNING [train.py:1060] (3/4) Exclude cut with ID 6356-271890-0060-94317-0_sp0.9 from training. Duration: 20.72225 2022-12-24 06:20:39,981 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=103836.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:20:58,662 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0_sp1.1 from training. Duration: 22.4818125 2022-12-24 06:21:04,439 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.179e+02 3.826e+02 4.710e+02 5.981e+02 1.265e+03, threshold=9.420e+02, percent-clipped=2.0 2022-12-24 06:21:24,661 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp0.9 from training. Duration: 25.0944375 2022-12-24 06:21:28,830 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0 from training. Duration: 21.515 2022-12-24 06:21:34,322 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2022-12-24 06:21:36,254 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp0.9 from training. Duration: 27.02225 2022-12-24 06:21:39,503 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103876.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:21:40,722 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0010-62480-0_sp0.9 from training. Duration: 22.22225 2022-12-24 06:21:42,074 INFO [train.py:894] (3/4) Epoch 30, batch 2200, loss[loss=0.1732, simple_loss=0.2661, pruned_loss=0.04013, over 18582.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2582, pruned_loss=0.04443, over 3715243.88 frames. ], batch size: 51, lr: 3.79e-03, grad_scale: 8.0 2022-12-24 06:21:47,798 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0085-44554-0_sp0.9 from training. Duration: 20.85 2022-12-24 06:22:00,999 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.83 vs. limit=5.0 2022-12-24 06:22:04,749 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([4.8403, 4.1884, 4.3590, 4.7342, 4.5644, 4.3205, 4.9181, 2.7328], device='cuda:3'), covar=tensor([0.0550, 0.0623, 0.0546, 0.0769, 0.0962, 0.0941, 0.0531, 0.3474], device='cuda:3'), in_proj_covar=tensor([0.0367, 0.0240, 0.0253, 0.0290, 0.0343, 0.0280, 0.0305, 0.0299], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 06:22:22,460 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0 from training. Duration: 21.54 2022-12-24 06:22:28,077 WARNING [train.py:1060] (3/4) Exclude cut with ID 4964-30587-0040-44509-0_sp1.1 from training. Duration: 20.5318125 2022-12-24 06:22:36,708 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0012-134311-0_sp0.9 from training. Duration: 21.9333125 2022-12-24 06:22:50,915 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8818, 2.2224, 1.7840, 2.4661, 2.8873, 1.8330, 1.9666, 1.4741], device='cuda:3'), covar=tensor([0.1913, 0.1792, 0.1594, 0.1046, 0.1334, 0.1094, 0.1936, 0.1560], device='cuda:3'), in_proj_covar=tensor([0.0255, 0.0237, 0.0227, 0.0207, 0.0269, 0.0203, 0.0234, 0.0208], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 06:22:56,282 INFO [train.py:894] (3/4) Epoch 30, batch 2250, loss[loss=0.1938, simple_loss=0.2751, pruned_loss=0.05625, over 18659.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.257, pruned_loss=0.04388, over 3714575.02 frames. ], batch size: 60, lr: 3.79e-03, grad_scale: 8.0 2022-12-24 06:23:11,443 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103937.0, num_to_drop=1, layers_to_drop={0} 2022-12-24 06:23:21,608 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0025-130151-0_sp0.9 from training. Duration: 21.7944375 2022-12-24 06:23:33,967 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0_sp0.9 from training. Duration: 22.4666875 2022-12-24 06:23:35,550 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.692e+02 4.009e+02 4.873e+02 5.788e+02 1.098e+03, threshold=9.745e+02, percent-clipped=7.0 2022-12-24 06:23:41,340 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0 from training. Duration: 21.635 2022-12-24 06:23:47,210 WARNING [train.py:1060] (3/4) Exclude cut with ID 6121-9014-0076-24124-0_sp0.9 from training. Duration: 24.038875 2022-12-24 06:24:13,203 INFO [train.py:894] (3/4) Epoch 30, batch 2300, loss[loss=0.1723, simple_loss=0.2611, pruned_loss=0.04172, over 18502.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2573, pruned_loss=0.04416, over 3714360.12 frames. ], batch size: 52, lr: 3.79e-03, grad_scale: 8.0 2022-12-24 06:24:30,906 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103990.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:24:31,983 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp1.1 from training. Duration: 21.786375 2022-12-24 06:24:41,689 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103997.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:24:44,119 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0002-12989-0 from training. Duration: 20.22 2022-12-24 06:25:17,094 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([0.2481, 1.4502, 1.6907, 1.0385, 1.0169, 1.7607, 1.6833, 1.5545], device='cuda:3'), covar=tensor([0.0832, 0.0349, 0.0355, 0.0376, 0.0441, 0.0543, 0.0258, 0.0677], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0176, 0.0134, 0.0144, 0.0149, 0.0146, 0.0169, 0.0184], device='cuda:3'), out_proj_covar=tensor([1.1366e-04, 1.3165e-04, 9.8217e-05, 1.0508e-04, 1.0846e-04, 1.0951e-04, 1.2731e-04, 1.3770e-04], device='cuda:3') 2022-12-24 06:25:21,709 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104022.0, num_to_drop=1, layers_to_drop={2} 2022-12-24 06:25:31,240 INFO [train.py:894] (3/4) Epoch 30, batch 2350, loss[loss=0.1724, simple_loss=0.2525, pruned_loss=0.04614, over 18556.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2569, pruned_loss=0.0438, over 3714594.34 frames. ], batch size: 49, lr: 3.79e-03, grad_scale: 8.0 2022-12-24 06:26:02,087 INFO [zipformer.py:660] (3/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=104049.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:26:05,090 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104051.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:26:07,809 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.777e+02 3.890e+02 4.874e+02 6.166e+02 1.500e+03, threshold=9.748e+02, percent-clipped=4.0 2022-12-24 06:26:16,545 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104058.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:26:32,721 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2022-12-24 06:26:34,470 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=104070.0, num_to_drop=1, layers_to_drop={1} 2022-12-24 06:26:45,789 INFO [train.py:894] (3/4) Epoch 30, batch 2400, loss[loss=0.1778, simple_loss=0.2521, pruned_loss=0.05178, over 18548.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2565, pruned_loss=0.04386, over 3714564.72 frames. ], batch size: 47, lr: 3.79e-03, grad_scale: 8.0 2022-12-24 06:26:47,262 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0 from training. Duration: 25.285 2022-12-24 06:27:12,775 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104096.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:27:29,091 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.4775, 2.6266, 1.9828, 2.9829, 2.7832, 2.4430, 3.6261, 2.5768], device='cuda:3'), covar=tensor([0.0838, 0.1903, 0.2692, 0.1889, 0.1826, 0.0857, 0.0915, 0.1177], device='cuda:3'), in_proj_covar=tensor([0.0185, 0.0220, 0.0260, 0.0293, 0.0245, 0.0198, 0.0208, 0.0212], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 06:27:33,393 INFO [zipformer.py:660] (3/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104110.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:27:47,816 WARNING [train.py:1060] (3/4) Exclude cut with ID 811-130148-0001-63453-0_sp0.9 from training. Duration: 20.861125 2022-12-24 06:27:50,967 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104121.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:27:59,247 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2022-12-24 06:28:01,169 INFO [train.py:894] (3/4) Epoch 30, batch 2450, loss[loss=0.1581, simple_loss=0.2294, pruned_loss=0.04339, over 18669.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2574, pruned_loss=0.04447, over 3714257.44 frames. ], batch size: 46, lr: 3.79e-03, grad_scale: 8.0 2022-12-24 06:28:08,958 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0 from training. Duration: 20.88 2022-12-24 06:28:24,241 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.7537, 1.5810, 1.5651, 1.6102, 1.9105, 1.8143, 1.7786, 1.3654], device='cuda:3'), covar=tensor([0.0295, 0.0249, 0.0468, 0.0215, 0.0192, 0.0400, 0.0261, 0.0326], device='cuda:3'), in_proj_covar=tensor([0.0099, 0.0129, 0.0159, 0.0125, 0.0119, 0.0126, 0.0103, 0.0131], device='cuda:3'), out_proj_covar=tensor([7.8070e-05, 1.0199e-04, 1.2948e-04, 9.8768e-05, 9.5137e-05, 9.5678e-05, 7.9834e-05, 1.0269e-04], device='cuda:3') 2022-12-24 06:28:39,222 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.777e+02 4.495e+02 5.196e+02 6.656e+02 1.409e+03, threshold=1.039e+03, percent-clipped=6.0 2022-12-24 06:28:40,811 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0_sp0.9 from training. Duration: 23.4166875 2022-12-24 06:28:49,550 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.3919, 1.5986, 0.8726, 1.8143, 2.5276, 1.8852, 1.9873, 2.2565], device='cuda:3'), covar=tensor([0.1440, 0.2042, 0.2416, 0.1392, 0.1671, 0.1730, 0.1380, 0.1553], device='cuda:3'), in_proj_covar=tensor([0.0095, 0.0098, 0.0117, 0.0098, 0.0121, 0.0093, 0.0099, 0.0095], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-24 06:29:16,539 INFO [train.py:894] (3/4) Epoch 30, batch 2500, loss[loss=0.1549, simple_loss=0.2425, pruned_loss=0.03371, over 18658.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2576, pruned_loss=0.04465, over 3715014.42 frames. ], batch size: 48, lr: 3.79e-03, grad_scale: 8.0 2022-12-24 06:29:41,225 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2022-12-24 06:30:01,296 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0 from training. Duration: 21.24 2022-12-24 06:30:01,315 WARNING [train.py:1060] (3/4) Exclude cut with ID 6533-399-0047-104881-0_sp0.9 from training. Duration: 23.9055625 2022-12-24 06:30:29,470 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2022-12-24 06:30:31,028 INFO [train.py:894] (3/4) Epoch 30, batch 2550, loss[loss=0.196, simple_loss=0.274, pruned_loss=0.059, over 18623.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2583, pruned_loss=0.04488, over 3715108.65 frames. ], batch size: 53, lr: 3.79e-03, grad_scale: 8.0 2022-12-24 06:30:32,995 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp0.9 from training. Duration: 25.988875 2022-12-24 06:30:37,577 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104232.0, num_to_drop=1, layers_to_drop={2} 2022-12-24 06:30:37,899 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.2429, 2.2466, 1.8177, 2.5077, 2.3722, 2.1826, 2.9829, 2.3345], device='cuda:3'), covar=tensor([0.0971, 0.1732, 0.2730, 0.1717, 0.1878, 0.0929, 0.0923, 0.1312], device='cuda:3'), in_proj_covar=tensor([0.0185, 0.0220, 0.0260, 0.0292, 0.0244, 0.0198, 0.0207, 0.0212], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 06:30:42,347 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0001-134300-0_sp0.9 from training. Duration: 20.67225 2022-12-24 06:31:10,006 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.524e+02 4.032e+02 4.876e+02 5.931e+02 1.170e+03, threshold=9.752e+02, percent-clipped=3.0 2022-12-24 06:31:29,730 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0 from training. Duration: 20.34 2022-12-24 06:31:46,936 INFO [train.py:894] (3/4) Epoch 30, batch 2600, loss[loss=0.1394, simple_loss=0.2219, pruned_loss=0.02847, over 18468.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2579, pruned_loss=0.04484, over 3715154.34 frames. ], batch size: 43, lr: 3.79e-03, grad_scale: 8.0 2022-12-24 06:32:41,936 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp0.9 from training. Duration: 25.061125 2022-12-24 06:32:54,330 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0 from training. Duration: 0.83 2022-12-24 06:33:01,507 INFO [train.py:894] (3/4) Epoch 30, batch 2650, loss[loss=0.2002, simple_loss=0.2777, pruned_loss=0.06138, over 18650.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2583, pruned_loss=0.04471, over 3714781.39 frames. ], batch size: 69, lr: 3.79e-03, grad_scale: 8.0 2022-12-24 06:33:19,782 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0017-41203-0 from training. Duration: 24.73 2022-12-24 06:33:29,486 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104346.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:33:30,839 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0 from training. Duration: 23.965 2022-12-24 06:33:39,922 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.438e+02 4.320e+02 5.325e+02 6.686e+02 2.435e+03, threshold=1.065e+03, percent-clipped=5.0 2022-12-24 06:33:39,992 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0030-146996-0_sp0.9 from training. Duration: 22.088875 2022-12-24 06:33:40,122 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104353.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:33:55,276 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0006-134305-0_sp0.9 from training. Duration: 23.6 2022-12-24 06:34:14,679 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.9234, 1.7178, 1.8546, 2.1269, 1.9826, 3.3347, 1.8155, 1.7949], device='cuda:3'), covar=tensor([0.0811, 0.1668, 0.1020, 0.0821, 0.1297, 0.0323, 0.1264, 0.1411], device='cuda:3'), in_proj_covar=tensor([0.0074, 0.0083, 0.0072, 0.0074, 0.0091, 0.0077, 0.0085, 0.0078], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2022-12-24 06:34:17,276 INFO [train.py:894] (3/4) Epoch 30, batch 2700, loss[loss=0.1433, simple_loss=0.2243, pruned_loss=0.0312, over 18538.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2577, pruned_loss=0.04463, over 3713825.36 frames. ], batch size: 44, lr: 3.79e-03, grad_scale: 8.0 2022-12-24 06:34:46,187 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104396.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:34:58,882 INFO [zipformer.py:660] (3/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104405.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:35:22,320 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104421.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:35:32,648 INFO [train.py:894] (3/4) Epoch 30, batch 2750, loss[loss=0.1596, simple_loss=0.2447, pruned_loss=0.03728, over 18637.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.258, pruned_loss=0.04448, over 3713401.60 frames. ], batch size: 53, lr: 3.78e-03, grad_scale: 8.0 2022-12-24 06:35:36,934 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0 from training. Duration: 23.795 2022-12-24 06:35:51,761 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2022-12-24 06:35:54,460 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0_sp1.1 from training. Duration: 21.5409375 2022-12-24 06:35:57,357 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp0.9 from training. Duration: 24.97775 2022-12-24 06:35:57,447 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=104444.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:36:06,956 WARNING [train.py:1060] (3/4) Exclude cut with ID 1085-156170-0017-128270-0_sp0.9 from training. Duration: 23.3444375 2022-12-24 06:36:09,549 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.603e+02 3.880e+02 4.528e+02 5.642e+02 1.210e+03, threshold=9.056e+02, percent-clipped=1.0 2022-12-24 06:36:32,939 WARNING [train.py:1060] (3/4) Exclude cut with ID 6010-56788-0055-90261-0_sp0.9 from training. Duration: 23.2 2022-12-24 06:36:34,532 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=104469.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:36:40,419 WARNING [train.py:1060] (3/4) Exclude cut with ID 5653-46179-0060-117930-0_sp0.9 from training. Duration: 21.17225 2022-12-24 06:36:47,359 INFO [train.py:894] (3/4) Epoch 30, batch 2800, loss[loss=0.1536, simple_loss=0.2308, pruned_loss=0.03821, over 18447.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2575, pruned_loss=0.04425, over 3713725.61 frames. ], batch size: 42, lr: 3.78e-03, grad_scale: 8.0 2022-12-24 06:37:00,005 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp0.9 from training. Duration: 24.6555625 2022-12-24 06:37:43,518 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.7457, 4.1697, 4.0732, 1.8671, 4.3821, 3.4945, 0.8975, 3.0899], device='cuda:3'), covar=tensor([0.2136, 0.1585, 0.1576, 0.3445, 0.0846, 0.0734, 0.4945, 0.1313], device='cuda:3'), in_proj_covar=tensor([0.0156, 0.0155, 0.0165, 0.0127, 0.0156, 0.0120, 0.0148, 0.0122], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2022-12-24 06:37:54,746 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-65654-0031-41259-0 from training. Duration: 20.44 2022-12-24 06:38:02,533 INFO [train.py:894] (3/4) Epoch 30, batch 2850, loss[loss=0.1856, simple_loss=0.2603, pruned_loss=0.05547, over 18577.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2577, pruned_loss=0.0442, over 3713961.92 frames. ], batch size: 49, lr: 3.78e-03, grad_scale: 8.0 2022-12-24 06:38:08,360 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104532.0, num_to_drop=1, layers_to_drop={2} 2022-12-24 06:38:09,761 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0_sp0.9 from training. Duration: 23.45 2022-12-24 06:38:29,959 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.4655, 1.3861, 1.1517, 1.2021, 1.2239, 1.3137, 1.2262, 1.3519], device='cuda:3'), covar=tensor([0.1785, 0.2147, 0.1570, 0.1946, 0.2458, 0.0888, 0.2135, 0.0907], device='cuda:3'), in_proj_covar=tensor([0.0304, 0.0303, 0.0259, 0.0353, 0.0285, 0.0238, 0.0301, 0.0227], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 06:38:39,966 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.496e+02 3.637e+02 4.715e+02 6.506e+02 1.297e+03, threshold=9.431e+02, percent-clipped=5.0 2022-12-24 06:38:40,009 WARNING [train.py:1060] (3/4) Exclude cut with ID 6945-60535-0076-12784-0_sp0.9 from training. Duration: 20.52225 2022-12-24 06:38:47,344 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0 from training. Duration: 22.19 2022-12-24 06:38:57,723 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp1.1 from training. Duration: 25.3818125 2022-12-24 06:39:05,676 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=2.42 vs. limit=5.0 2022-12-24 06:39:12,920 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0043-132310-0_sp0.9 from training. Duration: 28.0944375 2022-12-24 06:39:16,974 INFO [train.py:894] (3/4) Epoch 30, batch 2900, loss[loss=0.155, simple_loss=0.2338, pruned_loss=0.03813, over 18683.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2575, pruned_loss=0.04448, over 3713912.21 frames. ], batch size: 48, lr: 3.78e-03, grad_scale: 8.0 2022-12-24 06:39:20,643 WARNING [train.py:1060] (3/4) Exclude cut with ID 2195-150901-0045-59933-0_sp0.9 from training. Duration: 22.9444375 2022-12-24 06:39:20,772 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=104580.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:39:27,904 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp1.1 from training. Duration: 21.6318125 2022-12-24 06:39:29,666 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([2.0628, 2.0007, 2.3667, 1.3316, 2.2556, 2.4139, 1.6374, 2.7140], device='cuda:3'), covar=tensor([0.1217, 0.1906, 0.1242, 0.2043, 0.0801, 0.1177, 0.2536, 0.0569], device='cuda:3'), in_proj_covar=tensor([0.0197, 0.0216, 0.0209, 0.0194, 0.0173, 0.0219, 0.0218, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 06:39:44,957 WARNING [train.py:1060] (3/4) Exclude cut with ID 8631-249866-0030-130156-0 from training. Duration: 23.695 2022-12-24 06:40:10,368 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0 from training. Duration: 23.955 2022-12-24 06:40:14,435 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2022-12-24 06:40:32,906 INFO [train.py:894] (3/4) Epoch 30, batch 2950, loss[loss=0.175, simple_loss=0.2575, pruned_loss=0.04623, over 18705.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2581, pruned_loss=0.04449, over 3715235.97 frames. ], batch size: 41, lr: 3.78e-03, grad_scale: 8.0 2022-12-24 06:40:43,205 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0024-13011-0_sp0.9 from training. Duration: 26.438875 2022-12-24 06:40:59,831 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104646.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:41:09,934 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.578e+02 3.847e+02 4.605e+02 6.019e+02 2.047e+03, threshold=9.211e+02, percent-clipped=4.0 2022-12-24 06:41:10,363 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104653.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:41:28,278 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0021-26306-0_sp0.9 from training. Duration: 21.2444375 2022-12-24 06:41:28,299 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0014-15845-0_sp0.9 from training. Duration: 31.02225 2022-12-24 06:41:38,760 WARNING [train.py:1060] (3/4) Exclude cut with ID 432-122774-0017-62487-0 from training. Duration: 22.395 2022-12-24 06:41:47,953 INFO [train.py:894] (3/4) Epoch 30, batch 3000, loss[loss=0.1416, simple_loss=0.2302, pruned_loss=0.02652, over 18673.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2571, pruned_loss=0.04392, over 3715721.58 frames. ], batch size: 48, lr: 3.78e-03, grad_scale: 8.0 2022-12-24 06:41:47,953 INFO [train.py:919] (3/4) Computing validation loss 2022-12-24 06:41:58,660 INFO [train.py:928] (3/4) Epoch 30, validation: loss=0.163, simple_loss=0.2585, pruned_loss=0.0338, over 944034.00 frames. 2022-12-24 06:41:58,661 INFO [train.py:929] (3/4) Maximum memory allocated so far is 24809MB 2022-12-24 06:42:05,967 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0045-15876-0 from training. Duration: 21.075 2022-12-24 06:42:11,933 WARNING [train.py:1060] (3/4) Exclude cut with ID 6482-98857-0025-147532-0_sp0.9 from training. Duration: 20.0055625 2022-12-24 06:42:11,946 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0037-132304-0_sp0.9 from training. Duration: 22.05 2022-12-24 06:42:11,959 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0 from training. Duration: 26.8349375 2022-12-24 06:42:15,015 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0_sp1.1 from training. Duration: 22.1090625 2022-12-24 06:42:15,345 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8888, 1.2472, 0.8226, 1.4369, 2.2135, 1.1584, 1.5361, 1.7503], device='cuda:3'), covar=tensor([0.1541, 0.2202, 0.2254, 0.1502, 0.1787, 0.1872, 0.1486, 0.1672], device='cuda:3'), in_proj_covar=tensor([0.0096, 0.0098, 0.0117, 0.0098, 0.0122, 0.0093, 0.0100, 0.0095], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:3') 2022-12-24 06:42:22,340 WARNING [train.py:1060] (3/4) Exclude cut with ID 7699-105389-0094-26379-0_sp0.9 from training. Duration: 26.6166875 2022-12-24 06:42:22,501 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=104694.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:42:32,971 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=104701.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:42:39,408 INFO [zipformer.py:660] (3/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104705.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:42:40,733 WARNING [train.py:1060] (3/4) Exclude cut with ID 2046-178027-0000-53705-0_sp0.9 from training. Duration: 20.3055625 2022-12-24 06:43:01,698 WARNING [train.py:1060] (3/4) Exclude cut with ID 7205-50138-0008-5373-0_sp0.9 from training. Duration: 20.7 2022-12-24 06:43:13,899 INFO [train.py:894] (3/4) Epoch 30, batch 3050, loss[loss=0.1752, simple_loss=0.2644, pruned_loss=0.04302, over 18474.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2569, pruned_loss=0.04362, over 3715920.07 frames. ], batch size: 54, lr: 3.78e-03, grad_scale: 8.0 2022-12-24 06:43:44,642 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0 from training. Duration: 22.48 2022-12-24 06:43:51,975 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.496e+02 3.921e+02 4.500e+02 5.468e+02 9.473e+02, threshold=9.001e+02, percent-clipped=1.0 2022-12-24 06:43:52,164 INFO [zipformer.py:660] (3/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=104753.0, num_to_drop=0, layers_to_drop=set() 2022-12-24 06:43:59,252 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0003-134302-0_sp0.9 from training. Duration: 29.816625 2022-12-24 06:44:19,593 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0005-134304-0_sp1.1 from training. Duration: 22.7590625 2022-12-24 06:44:25,507 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0 from training. Duration: 22.555 2022-12-24 06:44:29,629 INFO [train.py:894] (3/4) Epoch 30, batch 3100, loss[loss=0.1664, simple_loss=0.2487, pruned_loss=0.04206, over 18701.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2567, pruned_loss=0.04389, over 3715294.95 frames. ], batch size: 50, lr: 3.78e-03, grad_scale: 8.0 2022-12-24 06:44:44,954 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2022-12-24 06:44:45,388 WARNING [train.py:1060] (3/4) Exclude cut with ID 1250-135782-0005-25975-0_sp0.9 from training. Duration: 21.688875 2022-12-24 06:45:18,368 WARNING [train.py:1060] (3/4) Exclude cut with ID 3488-85273-0038-41224-0_sp0.9 from training. Duration: 22.6 2022-12-24 06:45:44,264 INFO [train.py:894] (3/4) Epoch 30, batch 3150, loss[loss=0.1813, simple_loss=0.2676, pruned_loss=0.04753, over 18533.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2566, pruned_loss=0.04368, over 3715475.34 frames. ], batch size: 55, lr: 3.78e-03, grad_scale: 8.0 2022-12-24 06:45:55,376 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0009-15840-0 from training. Duration: 24.32 2022-12-24 06:46:21,956 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.960e+02 3.977e+02 4.589e+02 5.752e+02 1.230e+03, threshold=9.178e+02, percent-clipped=2.0 2022-12-24 06:46:53,114 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-276745-0093-13116-0_sp0.9 from training. Duration: 21.061125 2022-12-24 06:46:59,064 INFO [train.py:894] (3/4) Epoch 30, batch 3200, loss[loss=0.1624, simple_loss=0.2483, pruned_loss=0.03823, over 18699.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.257, pruned_loss=0.04378, over 3714614.12 frames. ], batch size: 60, lr: 3.78e-03, grad_scale: 8.0 2022-12-24 06:47:05,950 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.8159, 1.8884, 1.5817, 1.7048, 2.0483, 2.1526, 2.0822, 1.4629], device='cuda:3'), covar=tensor([0.0385, 0.0262, 0.0499, 0.0235, 0.0196, 0.0340, 0.0261, 0.0340], device='cuda:3'), in_proj_covar=tensor([0.0099, 0.0129, 0.0158, 0.0125, 0.0120, 0.0125, 0.0103, 0.0130], device='cuda:3'), out_proj_covar=tensor([7.8277e-05, 1.0149e-04, 1.2898e-04, 9.8498e-05, 9.5275e-05, 9.5328e-05, 7.9721e-05, 1.0226e-04], device='cuda:3') 2022-12-24 06:47:06,979 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0024-15855-0_sp0.9 from training. Duration: 20.32225 2022-12-24 06:47:20,289 WARNING [train.py:1060] (3/4) Exclude cut with ID 3033-130750-0096-55598-0_sp1.1 from training. Duration: 0.7545625 2022-12-24 06:47:35,014 WARNING [train.py:1060] (3/4) Exclude cut with ID 4295-39940-0007-92567-0_sp0.9 from training. Duration: 23.9333125 2022-12-24 06:48:07,400 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0008-134307-0_sp1.1 from training. Duration: 20.17275 2022-12-24 06:48:12,081 WARNING [train.py:1060] (3/4) Exclude cut with ID 6978-92210-0019-146985-0_sp1.1 from training. Duration: 20.436375 2022-12-24 06:48:13,575 INFO [train.py:894] (3/4) Epoch 30, batch 3250, loss[loss=0.1528, simple_loss=0.2459, pruned_loss=0.02991, over 18462.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2574, pruned_loss=0.04417, over 3713918.69 frames. ], batch size: 50, lr: 3.78e-03, grad_scale: 8.0 2022-12-24 06:48:51,164 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.506e+02 3.857e+02 4.754e+02 5.995e+02 1.031e+03, threshold=9.507e+02, percent-clipped=5.0 2022-12-24 06:49:28,775 INFO [train.py:894] (3/4) Epoch 30, batch 3300, loss[loss=0.1432, simple_loss=0.2281, pruned_loss=0.02919, over 18585.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.256, pruned_loss=0.04343, over 3713591.63 frames. ], batch size: 45, lr: 3.78e-03, grad_scale: 8.0 2022-12-24 06:49:33,996 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0_sp0.9 from training. Duration: 23.1055625 2022-12-24 06:49:35,317 WARNING [train.py:1060] (3/4) Exclude cut with ID 8291-282929-0007-12994-0_sp1.1 from training. Duration: 23.5 2022-12-24 06:49:47,635 WARNING [train.py:1060] (3/4) Exclude cut with ID 7255-291500-0009-134308-0_sp0.9 from training. Duration: 26.62775 2022-12-24 06:49:59,128 WARNING [train.py:1060] (3/4) Exclude cut with ID 6951-79737-0018-132285-0 from training. Duration: 21.105 2022-12-24 06:50:04,339 WARNING [train.py:1060] (3/4) Exclude cut with ID 4511-76322-0006-80011-0_sp0.9 from training. Duration: 24.411125 2022-12-24 06:50:18,660 INFO [scaling.py:679] (3/4) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2022-12-24 06:50:29,008 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2022-12-24 06:50:30,924 WARNING [train.py:1060] (3/4) Exclude cut with ID 6758-72288-0033-108368-0_sp1.1 from training. Duration: 21.263625 2022-12-24 06:50:45,221 INFO [train.py:894] (3/4) Epoch 30, batch 3350, loss[loss=0.1727, simple_loss=0.2606, pruned_loss=0.04242, over 18544.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2553, pruned_loss=0.04328, over 3714099.52 frames. ], batch size: 47, lr: 3.77e-03, grad_scale: 8.0 2022-12-24 06:51:06,249 WARNING [train.py:1060] (3/4) Exclude cut with ID 4234-40345-0022-142709-0 from training. Duration: 20.795 2022-12-24 06:51:15,718 INFO [zipformer.py:1480] (3/4) attn_weights_entropy = tensor([1.5896, 1.4805, 1.3550, 0.7934, 1.6701, 1.5275, 1.4894, 1.3664], device='cuda:3'), covar=tensor([0.0433, 0.0568, 0.0552, 0.0806, 0.0482, 0.0481, 0.0496, 0.1009], device='cuda:3'), in_proj_covar=tensor([0.0125, 0.0131, 0.0128, 0.0118, 0.0105, 0.0127, 0.0134, 0.0163], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2022-12-24 06:51:17,010 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0021-15852-0 from training. Duration: 24.76 2022-12-24 06:51:17,609 WARNING [train.py:1060] (3/4) Exclude cut with ID 3867-173237-0077-144769-0_sp0.9 from training. Duration: 22.25 2022-12-24 06:51:23,332 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.673e+02 4.112e+02 5.188e+02 6.129e+02 1.207e+03, threshold=1.038e+03, percent-clipped=1.0 2022-12-24 06:51:39,605 WARNING [train.py:1060] (3/4) Exclude cut with ID 7357-94126-0026-15857-0_sp1.1 from training. Duration: 20.5045625 2022-12-24 06:52:01,705 INFO [train.py:894] (3/4) Epoch 30, batch 3400, loss[loss=0.1996, simple_loss=0.287, pruned_loss=0.05605, over 18490.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2553, pruned_loss=0.04306, over 3714430.10 frames. ], batch size: 52, lr: 3.77e-03, grad_scale: 8.0 2022-12-24 06:52:42,526 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2022-12-24 06:53:12,277 INFO [train.py:894] (3/4) Epoch 30, batch 3450, loss[loss=0.1752, simple_loss=0.2673, pruned_loss=0.04159, over 18669.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2561, pruned_loss=0.04353, over 3715236.04 frames. ], batch size: 62, lr: 3.77e-03, grad_scale: 8.0 2022-12-24 06:53:38,477 INFO [scaling.py:679] (3/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2022-12-24 06:53:49,269 INFO [optim.py:369] (3/4) Clipping_scale=2.0, grad-norm quartiles 2.749e+02 3.937e+02 4.807e+02 6.235e+02 1.177e+03, threshold=9.613e+02, percent-clipped=2.0 2022-12-24 06:54:25,725 INFO [train.py:894] (3/4) Epoch 30, batch 3500, loss[loss=0.1818, simple_loss=0.2669, pruned_loss=0.04832, over 18669.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2569, pruned_loss=0.04405, over 3715112.65 frames. ], batch size: 184, lr: 3.77e-03, grad_scale: 8.0 2022-12-24 06:54:36,653 INFO [train.py:1158] (3/4) Done!