updated

README.md

@@ -0,0 +1 @@
+See https://github.com/k2-fsa/icefall/pull/1238

decoding_result/greedy_search/log-decode-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model-2023-08-10-10-15-51

@@ -0,0 +1,189 @@
+2023-08-10 10:15:51,972 INFO [decode.py:664] Decoding started
+2023-08-10 10:15:51,972 INFO [decode.py:670] Device: cuda:0
+2023-08-10 10:15:51,977 INFO [decode.py:680] {'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': '1.24.3', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': '821ebc378e7fb99b8adc81950227963332821e01', 'k2-git-date': 'Wed Jul 19 15:38:25 2023', 'lhotse-version': '1.16.0.dev+git.1db4d97a.dirty', 'torch-version': '1.11.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.9', 'icefall-git-branch': 'dev_multi_zh-hans', 'icefall-git-sha1': 'df6e292-clean', 'icefall-git-date': 'Wed Aug 9 09:40:02 2023', 'icefall-path': '/star-home/jinzengrui/lib/miniconda3/envs/dev39/lib/python3.9/site-packages/icefall-1.0-py3.9.egg', 'k2-path': '/star-home/jinzengrui/lib/miniconda3/envs/dev39/lib/python3.9/site-packages/k2-1.24.3.dev20230721+cuda10.2.torch1.11.0-py3.9-linux-x86_64.egg/k2/__init__.py', 'lhotse-path': '/star-home/jinzengrui/lib/miniconda3/envs/dev39/lib/python3.9/site-packages/lhotse-1.16.0.dev0+git.1db4d97a.dirty-py3.9.egg/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-0423201334-6587bbc68d-tn554', 'IP address': '10.177.74.211'}, 'epoch': 20, 'iter': 0, 'avg': 1, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp'), 'bpe_model': 'data/lang_bpe_2000/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), 'decoding_method': 'greedy_search', 'beam_size': 4, 'beam': 20.0, 'ngram_lm_scale': 0.01, 'max_contexts': 8, 'max_states': 64, 'context_size': 2, 'max_sym_per_frame': 1, 'num_paths': 200, 'nbest_scale': 0.5, 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'decoder_dim': 512, 'joiner_dim': 512, 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'use_transducer': True, 'use_ctc': False, 'full_libri': True, 'mini_libri': False, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 2000, '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', 'res_dir': PosixPath('zipformer/exp/greedy_search'), 'suffix': 'epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model', 'blank_id': 0, 'unk_id': 2, 'vocab_size': 2000}
+2023-08-10 10:15:51,977 INFO [decode.py:682] About to create model
+2023-08-10 10:15:52,532 INFO [decode.py:749] Calculating the averaged model over epoch range from 19 (excluded) to 20
+2023-08-10 10:16:02,422 INFO [decode.py:783] Number of model parameters: 68625511
+2023-08-10 10:16:02,422 INFO [multi_dataset.py:221] About to get multidataset test cuts
+2023-08-10 10:16:02,422 INFO [multi_dataset.py:224] Loading Aidatatang_200zh set in lazy mode
+2023-08-10 10:16:02,434 INFO [multi_dataset.py:233] Loading Aishell set in lazy mode
+2023-08-10 10:16:02,437 INFO [multi_dataset.py:242] Loading Aishell-2 set in lazy mode
+2023-08-10 10:16:02,439 INFO [multi_dataset.py:251] Loading Aishell-4 TEST set in lazy mode
+2023-08-10 10:16:02,440 INFO [multi_dataset.py:257] Loading Ali-Meeting set in lazy mode
+2023-08-10 10:16:02,443 INFO [multi_dataset.py:266] Loading MagicData set in lazy mode
+2023-08-10 10:16:02,445 INFO [multi_dataset.py:275] Loading KeSpeech set in lazy mode
+2023-08-10 10:16:02,449 INFO [multi_dataset.py:287] Loading WeNetSpeech set in lazy mode
+2023-08-10 10:16:08,442 WARNING [decode.py:799] Excluding cut with ID: TEST_NET_Y0000000004_0ub4ZzdHzBc_S00023 from decoding, num_frames: 8
+2023-08-10 10:16:09,393 INFO [decode.py:813] Start decoding test set: aidatatang_test
+2023-08-10 10:16:12,428 INFO [decode.py:561] batch 0/?, cuts processed until now is 535
+2023-08-10 10:17:13,609 INFO [decode.py:561] batch 50/?, cuts processed until now is 28624
+2023-08-10 10:17:25,885 INFO [zipformer.py:1728] name=None, attn_weights_entropy = tensor([3.9959, 2.1917, 4.1087, 2.0992], device='cuda:0')
+2023-08-10 10:17:50,622 INFO [decode.py:577] The transcripts are stored in zipformer/exp/greedy_search/recogs-aidatatang_test-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:17:51,570 INFO [utils.py:564] [aidatatang_test-greedy_search] %WER 3.67% [17231 / 468933, 786 ins, 4838 del, 11607 sub ]
+2023-08-10 10:17:53,346 INFO [decode.py:590] Wrote detailed error stats to zipformer/exp/greedy_search/errs-aidatatang_test-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:17:53,350 INFO [decode.py:606] 
+For aidatatang_test, WER of different settings are:
+greedy_search	3.67	best for aidatatang_test
+
+2023-08-10 10:17:53,351 INFO [decode.py:813] Start decoding test set: aidatatang_dev
+2023-08-10 10:17:56,714 INFO [decode.py:561] batch 0/?, cuts processed until now is 541
+2023-08-10 10:18:10,538 INFO [zipformer.py:1728] name=None, attn_weights_entropy = tensor([4.4447, 3.7613, 4.2179, 2.5586], device='cuda:0')
+2023-08-10 10:18:44,047 INFO [decode.py:577] The transcripts are stored in zipformer/exp/greedy_search/recogs-aidatatang_dev-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:18:44,619 INFO [utils.py:564] [aidatatang_dev-greedy_search] %WER 3.20% [7509 / 234524, 282 ins, 2330 del, 4897 sub ]
+2023-08-10 10:18:45,520 INFO [decode.py:590] Wrote detailed error stats to zipformer/exp/greedy_search/errs-aidatatang_dev-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:18:45,546 INFO [decode.py:606] 
+For aidatatang_dev, WER of different settings are:
+greedy_search	3.2	best for aidatatang_dev
+
+2023-08-10 10:18:45,547 INFO [decode.py:813] Start decoding test set: alimeeting_test
+2023-08-10 10:18:49,650 INFO [decode.py:561] batch 0/?, cuts processed until now is 295
+2023-08-10 10:18:53,473 WARNING [decode.py:799] Excluding cut with ID: R8008_M8016-8062-123 from decoding, num_frames: 6
+2023-08-10 10:19:24,921 INFO [decode.py:577] The transcripts are stored in zipformer/exp/greedy_search/recogs-alimeeting_test-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:19:25,395 INFO [utils.py:564] [alimeeting_test-greedy_search] %WER 24.78% [51991 / 209845, 3987 ins, 24986 del, 23018 sub ]
+2023-08-10 10:19:26,224 INFO [decode.py:590] Wrote detailed error stats to zipformer/exp/greedy_search/errs-alimeeting_test-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:19:26,227 INFO [decode.py:606] 
+For alimeeting_test, WER of different settings are:
+greedy_search	24.78	best for alimeeting_test
+
+2023-08-10 10:19:26,227 INFO [decode.py:813] Start decoding test set: alimeeting_eval
+2023-08-10 10:19:29,818 INFO [decode.py:561] batch 0/?, cuts processed until now is 232
+2023-08-10 10:19:42,465 INFO [decode.py:577] The transcripts are stored in zipformer/exp/greedy_search/recogs-alimeeting_eval-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:19:42,627 INFO [utils.py:564] [alimeeting_eval-greedy_search] %WER 23.15% [18781 / 81111, 1633 ins, 8231 del, 8917 sub ]
+2023-08-10 10:19:42,954 INFO [decode.py:590] Wrote detailed error stats to zipformer/exp/greedy_search/errs-alimeeting_eval-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:19:42,957 INFO [decode.py:606] 
+For alimeeting_eval, WER of different settings are:
+greedy_search	23.15	best for alimeeting_eval
+
+2023-08-10 10:19:42,957 INFO [decode.py:813] Start decoding test set: aishell_test
+2023-08-10 10:19:46,333 INFO [decode.py:561] batch 0/?, cuts processed until now is 318
+2023-08-10 10:20:09,009 INFO [decode.py:577] The transcripts are stored in zipformer/exp/greedy_search/recogs-aishell_test-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:20:09,196 INFO [utils.py:564] [aishell_test-greedy_search] %WER 3.04% [3189 / 104765, 59 ins, 1318 del, 1812 sub ]
+2023-08-10 10:20:09,552 INFO [decode.py:590] Wrote detailed error stats to zipformer/exp/greedy_search/errs-aishell_test-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:20:09,556 INFO [decode.py:606] 
+For aishell_test, WER of different settings are:
+greedy_search	3.04	best for aishell_test
+
+2023-08-10 10:20:09,556 INFO [decode.py:813] Start decoding test set: aishell_dev
+2023-08-10 10:20:14,586 INFO [decode.py:561] batch 0/?, cuts processed until now is 359
+2023-08-10 10:20:56,186 INFO [decode.py:577] The transcripts are stored in zipformer/exp/greedy_search/recogs-aishell_dev-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:20:56,546 INFO [utils.py:564] [aishell_dev-greedy_search] %WER 2.91% [5972 / 205341, 171 ins, 2361 del, 3440 sub ]
+2023-08-10 10:20:57,256 INFO [decode.py:590] Wrote detailed error stats to zipformer/exp/greedy_search/errs-aishell_dev-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:20:57,260 INFO [decode.py:606] 
+For aishell_dev, WER of different settings are:
+greedy_search	2.91	best for aishell_dev
+
+2023-08-10 10:20:57,260 INFO [decode.py:813] Start decoding test set: aishell-2_test
+2023-08-10 10:21:00,417 INFO [decode.py:561] batch 0/?, cuts processed until now is 424
+2023-08-10 10:21:08,651 INFO [decode.py:577] The transcripts are stored in zipformer/exp/greedy_search/recogs-aishell-2_test-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:21:08,744 INFO [utils.py:564] [aishell-2_test-greedy_search] %WER 4.03% [1998 / 49532, 70 ins, 384 del, 1544 sub ]
+2023-08-10 10:21:08,926 INFO [decode.py:590] Wrote detailed error stats to zipformer/exp/greedy_search/errs-aishell-2_test-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:21:08,929 INFO [decode.py:606] 
+For aishell-2_test, WER of different settings are:
+greedy_search	4.03	best for aishell-2_test
+
+2023-08-10 10:21:08,929 INFO [decode.py:813] Start decoding test set: aishell-2_dev
+2023-08-10 10:21:10,598 INFO [decode.py:561] batch 0/?, cuts processed until now is 209
+2023-08-10 10:21:15,421 INFO [decode.py:577] The transcripts are stored in zipformer/exp/greedy_search/recogs-aishell-2_dev-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:21:15,469 INFO [utils.py:564] [aishell-2_dev-greedy_search] %WER 3.59% [891 / 24802, 28 ins, 173 del, 690 sub ]
+2023-08-10 10:21:15,564 INFO [decode.py:590] Wrote detailed error stats to zipformer/exp/greedy_search/errs-aishell-2_dev-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:21:15,567 INFO [decode.py:606] 
+For aishell-2_dev, WER of different settings are:
+greedy_search	3.59	best for aishell-2_dev
+
+2023-08-10 10:21:15,568 INFO [decode.py:813] Start decoding test set: aishell-4
+2023-08-10 10:21:20,282 INFO [decode.py:561] batch 0/?, cuts processed until now is 221
+2023-08-10 10:21:33,528 INFO [zipformer.py:1728] name=None, attn_weights_entropy = tensor([2.6748, 2.9455, 3.6913, 2.8226], device='cuda:0')
+2023-08-10 10:21:54,683 INFO [decode.py:577] The transcripts are stored in zipformer/exp/greedy_search/recogs-aishell-4-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:21:55,017 INFO [utils.py:564] [aishell-4-greedy_search] %WER 15.68% [28324 / 180665, 4346 ins, 9719 del, 14259 sub ]
+2023-08-10 10:21:55,701 INFO [decode.py:590] Wrote detailed error stats to zipformer/exp/greedy_search/errs-aishell-4-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:21:55,704 INFO [decode.py:606] 
+For aishell-4, WER of different settings are:
+greedy_search	15.68	best for aishell-4
+
+2023-08-10 10:21:55,705 INFO [decode.py:813] Start decoding test set: magicdata_test
+2023-08-10 10:22:00,385 INFO [decode.py:561] batch 0/?, cuts processed until now is 384
+2023-08-10 10:22:29,383 INFO [zipformer.py:1728] name=None, attn_weights_entropy = tensor([2.3753, 3.0035, 2.7108, 2.4073, 2.9455, 2.8159, 2.9704, 2.9393],
+       device='cuda:0')
+2023-08-10 10:22:34,288 INFO [zipformer.py:1728] name=None, attn_weights_entropy = tensor([3.2779, 3.4722, 3.9441, 4.1606], device='cuda:0')
+2023-08-10 10:22:44,679 INFO [zipformer.py:1728] name=None, attn_weights_entropy = tensor([2.7041, 5.0518, 4.5722, 3.5379], device='cuda:0')
+2023-08-10 10:22:58,746 INFO [decode.py:561] batch 50/?, cuts processed until now is 21169
+2023-08-10 10:23:10,277 INFO [decode.py:577] The transcripts are stored in zipformer/exp/greedy_search/recogs-magicdata_test-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:23:10,736 INFO [utils.py:564] [magicdata_test-greedy_search] %WER 3.12% [7456 / 239091, 248 ins, 2212 del, 4996 sub ]
+2023-08-10 10:23:11,619 INFO [decode.py:590] Wrote detailed error stats to zipformer/exp/greedy_search/errs-magicdata_test-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:23:11,622 INFO [decode.py:606] 
+For magicdata_test, WER of different settings are:
+greedy_search	3.12	best for magicdata_test
+
+2023-08-10 10:23:11,623 INFO [decode.py:813] Start decoding test set: magicdata_dev
+2023-08-10 10:23:16,201 INFO [decode.py:561] batch 0/?, cuts processed until now is 353
+2023-08-10 10:23:16,722 INFO [zipformer.py:1728] name=None, attn_weights_entropy = tensor([2.4883, 3.4312, 3.6477, 3.5283], device='cuda:0')
+2023-08-10 10:23:51,750 INFO [decode.py:577] The transcripts are stored in zipformer/exp/greedy_search/recogs-magicdata_dev-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:23:51,980 INFO [utils.py:564] [magicdata_dev-greedy_search] %WER 3.68% [4302 / 116800, 231 ins, 1123 del, 2948 sub ]
+2023-08-10 10:23:52,410 INFO [decode.py:590] Wrote detailed error stats to zipformer/exp/greedy_search/errs-magicdata_dev-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:23:52,413 INFO [decode.py:606] 
+For magicdata_dev, WER of different settings are:
+greedy_search	3.68	best for magicdata_dev
+
+2023-08-10 10:23:52,414 INFO [decode.py:813] Start decoding test set: kespeech-asr_test
+2023-08-10 10:23:57,039 INFO [decode.py:561] batch 0/?, cuts processed until now is 299
+2023-08-10 10:24:39,370 INFO [zipformer.py:1728] name=None, attn_weights_entropy = tensor([1.8755, 2.2141, 2.5087, 2.4872, 2.6784, 2.3122, 2.6962, 2.7488],
+       device='cuda:0')
+2023-08-10 10:24:57,728 INFO [decode.py:561] batch 50/?, cuts processed until now is 16858
+2023-08-10 10:25:08,914 INFO [decode.py:577] The transcripts are stored in zipformer/exp/greedy_search/recogs-kespeech-asr_test-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:25:09,466 INFO [utils.py:564] [kespeech-asr_test-greedy_search] %WER 8.01% [22734 / 283772, 1235 ins, 3854 del, 17645 sub ]
+2023-08-10 10:25:10,501 INFO [decode.py:590] Wrote detailed error stats to zipformer/exp/greedy_search/errs-kespeech-asr_test-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:25:10,504 INFO [decode.py:606] 
+For kespeech-asr_test, WER of different settings are:
+greedy_search	8.01	best for kespeech-asr_test
+
+2023-08-10 10:25:10,504 INFO [decode.py:813] Start decoding test set: kespeech-asr_dev_phase1
+2023-08-10 10:25:12,832 INFO [decode.py:561] batch 0/?, cuts processed until now is 187
+2023-08-10 10:25:20,769 INFO [decode.py:577] The transcripts are stored in zipformer/exp/greedy_search/recogs-kespeech-asr_dev_phase1-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:25:20,827 INFO [utils.py:564] [kespeech-asr_dev_phase1-greedy_search] %WER 6.69% [2117 / 31634, 131 ins, 421 del, 1565 sub ]
+2023-08-10 10:25:20,953 INFO [decode.py:590] Wrote detailed error stats to zipformer/exp/greedy_search/errs-kespeech-asr_dev_phase1-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:25:20,956 INFO [decode.py:606] 
+For kespeech-asr_dev_phase1, WER of different settings are:
+greedy_search	6.69	best for kespeech-asr_dev_phase1
+
+2023-08-10 10:25:20,956 INFO [decode.py:813] Start decoding test set: kespeech-asr_dev_phase2
+2023-08-10 10:25:23,106 INFO [decode.py:561] batch 0/?, cuts processed until now is 196
+2023-08-10 10:25:23,375 INFO [zipformer.py:1728] name=None, attn_weights_entropy = tensor([4.2291, 2.7520, 3.0036, 4.2187], device='cuda:0')
+2023-08-10 10:25:30,740 INFO [decode.py:577] The transcripts are stored in zipformer/exp/greedy_search/recogs-kespeech-asr_dev_phase2-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:25:30,796 INFO [utils.py:564] [kespeech-asr_dev_phase2-greedy_search] %WER 3.19% [1020 / 31928, 56 ins, 286 del, 678 sub ]
+2023-08-10 10:25:30,909 INFO [decode.py:590] Wrote detailed error stats to zipformer/exp/greedy_search/errs-kespeech-asr_dev_phase2-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:25:30,912 INFO [decode.py:606] 
+For kespeech-asr_dev_phase2, WER of different settings are:
+greedy_search	3.19	best for kespeech-asr_dev_phase2
+
+2023-08-10 10:25:30,912 INFO [decode.py:813] Start decoding test set: wenetspeech-meeting_test
+2023-08-10 10:25:34,582 INFO [decode.py:561] batch 0/?, cuts processed until now is 187
+2023-08-10 10:26:16,279 INFO [decode.py:577] The transcripts are stored in zipformer/exp/greedy_search/recogs-wenetspeech-meeting_test-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:26:16,655 INFO [utils.py:564] [wenetspeech-meeting_test-greedy_search] %WER 7.05% [15534 / 220385, 1521 ins, 4899 del, 9114 sub ]
+2023-08-10 10:26:17,378 INFO [decode.py:590] Wrote detailed error stats to zipformer/exp/greedy_search/errs-wenetspeech-meeting_test-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:26:17,382 INFO [decode.py:606] 
+For wenetspeech-meeting_test, WER of different settings are:
+greedy_search	7.05	best for wenetspeech-meeting_test
+
+2023-08-10 10:26:17,382 INFO [decode.py:813] Start decoding test set: wenetspeech-net_test
+2023-08-10 10:26:17,588 WARNING [decode.py:799] Excluding cut with ID: TEST_NET_Y0000000004_0ub4ZzdHzBc_S00023 from decoding, num_frames: 8
+2023-08-10 10:26:21,182 INFO [decode.py:561] batch 0/?, cuts processed until now is 292
+2023-08-10 10:27:16,471 INFO [zipformer.py:1728] name=None, attn_weights_entropy = tensor([5.4299, 5.1613, 5.4692, 5.4338], device='cuda:0')
+2023-08-10 10:27:23,593 INFO [decode.py:561] batch 50/?, cuts processed until now is 24444
+2023-08-10 10:27:27,024 INFO [decode.py:577] The transcripts are stored in zipformer/exp/greedy_search/recogs-wenetspeech-net_test-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:27:27,764 INFO [utils.py:564] [wenetspeech-net_test-greedy_search] %WER 8.78% [36514 / 415746, 1493 ins, 17028 del, 17993 sub ]
+2023-08-10 10:27:29,232 INFO [decode.py:590] Wrote detailed error stats to zipformer/exp/greedy_search/errs-wenetspeech-net_test-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:27:29,236 INFO [decode.py:606] 
+For wenetspeech-net_test, WER of different settings are:
+greedy_search	8.78	best for wenetspeech-net_test
+
+2023-08-10 10:27:29,236 INFO [decode.py:813] Start decoding test set: wenetspeech_dev
+2023-08-10 10:27:31,578 INFO [zipformer.py:1728] name=None, attn_weights_entropy = tensor([2.4010, 2.7901, 3.7666, 2.6059], device='cuda:0')
+2023-08-10 10:27:32,228 INFO [decode.py:561] batch 0/?, cuts processed until now is 261
+2023-08-10 10:28:21,613 INFO [decode.py:577] The transcripts are stored in zipformer/exp/greedy_search/recogs-wenetspeech_dev-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:28:22,157 INFO [utils.py:564] [wenetspeech_dev-greedy_search] %WER 9.32% [30816 / 330498, 1337 ins, 19932 del, 9547 sub ]
+2023-08-10 10:28:23,362 INFO [decode.py:590] Wrote detailed error stats to zipformer/exp/greedy_search/errs-wenetspeech_dev-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:28:23,366 INFO [decode.py:606] 
+For wenetspeech_dev, WER of different settings are:
+greedy_search	9.32	best for wenetspeech_dev
+
+2023-08-10 10:28:23,366 INFO [decode.py:830] Done!

decoding_result/greedy_search/log-decode-epoch-20-avg-10-context-2-max-sym-per-frame-1-use-averaged-model-2023-08-10-10-11-40

@@ -0,0 +1,50 @@
+2023-08-10 10:11:40,019 INFO [decode.py:664] Decoding started
+2023-08-10 10:11:40,019 INFO [decode.py:670] Device: cuda:0
+2023-08-10 10:11:40,037 INFO [decode.py:680] {'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': '1.24.3', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': '821ebc378e7fb99b8adc81950227963332821e01', 'k2-git-date': 'Wed Jul 19 15:38:25 2023', 'lhotse-version': '1.16.0.dev+git.1db4d97a.dirty', 'torch-version': '1.11.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.9', 'icefall-git-branch': 'dev_multi_zh-hans', 'icefall-git-sha1': 'df6e292-clean', 'icefall-git-date': 'Wed Aug 9 09:40:02 2023', 'icefall-path': '/star-home/jinzengrui/lib/miniconda3/envs/dev39/lib/python3.9/site-packages/icefall-1.0-py3.9.egg', 'k2-path': '/star-home/jinzengrui/lib/miniconda3/envs/dev39/lib/python3.9/site-packages/k2-1.24.3.dev20230721+cuda10.2.torch1.11.0-py3.9-linux-x86_64.egg/k2/__init__.py', 'lhotse-path': '/star-home/jinzengrui/lib/miniconda3/envs/dev39/lib/python3.9/site-packages/lhotse-1.16.0.dev0+git.1db4d97a.dirty-py3.9.egg/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-0423201334-6587bbc68d-tn554', 'IP address': '10.177.74.211'}, 'epoch': 20, 'iter': 0, 'avg': 10, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp'), 'bpe_model': 'data/lang_bpe_2000/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), 'decoding_method': 'greedy_search', 'beam_size': 4, 'beam': 20.0, 'ngram_lm_scale': 0.01, 'max_contexts': 8, 'max_states': 64, 'context_size': 2, 'max_sym_per_frame': 1, 'num_paths': 200, 'nbest_scale': 0.5, 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'decoder_dim': 512, 'joiner_dim': 512, 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'use_transducer': True, 'use_ctc': False, 'full_libri': True, 'mini_libri': False, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 300.0, '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', 'res_dir': PosixPath('zipformer/exp/greedy_search'), 'suffix': 'epoch-20-avg-10-context-2-max-sym-per-frame-1-use-averaged-model', 'blank_id': 0, 'unk_id': 2, 'vocab_size': 2000}
+2023-08-10 10:11:40,038 INFO [decode.py:682] About to create model
+2023-08-10 10:11:40,664 INFO [decode.py:749] Calculating the averaged model over epoch range from 10 (excluded) to 20
+2023-08-10 10:11:57,318 INFO [decode.py:783] Number of model parameters: 68625511
+2023-08-10 10:11:57,318 INFO [multi_dataset.py:221] About to get multidataset test cuts
+2023-08-10 10:11:57,318 INFO [multi_dataset.py:224] Loading Aidatatang_200zh set in lazy mode
+2023-08-10 10:11:57,422 INFO [multi_dataset.py:233] Loading Aishell set in lazy mode
+2023-08-10 10:11:57,445 INFO [multi_dataset.py:242] Loading Aishell-2 set in lazy mode
+2023-08-10 10:11:57,489 INFO [multi_dataset.py:251] Loading Aishell-4 TEST set in lazy mode
+2023-08-10 10:11:57,512 INFO [multi_dataset.py:257] Loading Ali-Meeting set in lazy mode
+2023-08-10 10:11:57,567 INFO [multi_dataset.py:266] Loading MagicData set in lazy mode
+2023-08-10 10:11:57,642 INFO [multi_dataset.py:275] Loading KeSpeech set in lazy mode
+2023-08-10 10:11:57,713 INFO [multi_dataset.py:287] Loading WeNetSpeech set in lazy mode
+2023-08-10 10:12:03,719 WARNING [decode.py:799] Excluding cut with ID: TEST_NET_Y0000000004_0ub4ZzdHzBc_S00023 from decoding, num_frames: 8
+2023-08-10 10:12:04,674 INFO [decode.py:813] Start decoding test set: aidatatang_test
+2023-08-10 10:12:06,438 INFO [decode.py:561] batch 0/?, cuts processed until now is 80
+2023-08-10 10:12:21,783 INFO [decode.py:561] batch 50/?, cuts processed until now is 4543
+2023-08-10 10:12:37,452 INFO [decode.py:561] batch 100/?, cuts processed until now is 9084
+2023-08-10 10:12:43,090 INFO [zipformer.py:1728] name=None, attn_weights_entropy = tensor([2.4494, 1.5390, 2.8959, 2.0475], device='cuda:0')
+2023-08-10 10:12:43,373 INFO [zipformer.py:1728] name=None, attn_weights_entropy = tensor([2.4728, 1.5300, 2.8783, 2.0692], device='cuda:0')
+2023-08-10 10:12:52,754 INFO [decode.py:561] batch 150/?, cuts processed until now is 13880
+2023-08-10 10:13:08,182 INFO [decode.py:561] batch 200/?, cuts processed until now is 18516
+2023-08-10 10:13:11,412 INFO [zipformer.py:1728] name=None, attn_weights_entropy = tensor([4.2917, 4.4671, 2.5746, 3.9718], device='cuda:0')
+2023-08-10 10:13:23,574 INFO [decode.py:561] batch 250/?, cuts processed until now is 22994
+2023-08-10 10:13:38,251 INFO [decode.py:561] batch 300/?, cuts processed until now is 28179
+2023-08-10 10:13:41,271 INFO [zipformer.py:1728] name=None, attn_weights_entropy = tensor([4.4892, 4.6469, 2.6430, 4.0841], device='cuda:0')
+2023-08-10 10:13:49,268 INFO [zipformer.py:1728] name=None, attn_weights_entropy = tensor([4.6454, 4.3246, 4.7430, 4.7210], device='cuda:0')
+2023-08-10 10:13:53,301 INFO [decode.py:561] batch 350/?, cuts processed until now is 32886
+2023-08-10 10:13:58,613 INFO [zipformer.py:1728] name=None, attn_weights_entropy = tensor([2.9386, 2.0484, 3.7366, 2.6518], device='cuda:0')
+2023-08-10 10:14:08,460 INFO [decode.py:561] batch 400/?, cuts processed until now is 37667
+2023-08-10 10:14:10,567 INFO [zipformer.py:1728] name=None, attn_weights_entropy = tensor([4.6399, 3.9501, 4.4008, 2.7544], device='cuda:0')
+2023-08-10 10:14:20,867 INFO [zipformer.py:1728] name=None, attn_weights_entropy = tensor([2.5476, 1.7536, 2.2094, 1.9260, 2.0861, 1.8582, 2.3946, 2.2869],
+       device='cuda:0')
+2023-08-10 10:14:21,338 INFO [zipformer.py:1728] name=None, attn_weights_entropy = tensor([4.3019, 3.7027, 4.1067, 2.6600], device='cuda:0')
+2023-08-10 10:14:23,495 INFO [decode.py:561] batch 450/?, cuts processed until now is 42122
+2023-08-10 10:14:39,472 INFO [decode.py:561] batch 500/?, cuts processed until now is 46172
+2023-08-10 10:14:51,476 INFO [zipformer.py:1728] name=None, attn_weights_entropy = tensor([1.9574, 3.1089, 3.2728, 3.1859], device='cuda:0')
+2023-08-10 10:14:52,212 INFO [decode.py:577] The transcripts are stored in zipformer/exp/greedy_search/recogs-aidatatang_test-greedy_search-epoch-20-avg-10-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:14:53,135 INFO [utils.py:564] [aidatatang_test-greedy_search] %WER 6.64% [31118 / 468933, 705 ins, 17895 del, 12518 sub ]
+2023-08-10 10:14:54,879 INFO [decode.py:590] Wrote detailed error stats to zipformer/exp/greedy_search/errs-aidatatang_test-greedy_search-epoch-20-avg-10-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2023-08-10 10:14:54,883 INFO [decode.py:606] 
+For aidatatang_test, WER of different settings are:
+greedy_search	6.64	best for aidatatang_test
+
+2023-08-10 10:14:54,884 INFO [decode.py:813] Start decoding test set: aidatatang_dev
+2023-08-10 10:14:56,728 INFO [decode.py:561] batch 0/?, cuts processed until now is 81
+2023-08-10 10:15:12,011 INFO [decode.py:561] batch 50/?, cuts processed until now is 4556
+2023-08-10 10:15:27,317 INFO [decode.py:561] batch 100/?, cuts processed until now is 9077

decoding_result/greedy_search/wer-summary-aidatatang_dev-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+greedy_search	3.2

decoding_result/greedy_search/wer-summary-aidatatang_test-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+greedy_search	3.67

decoding_result/greedy_search/wer-summary-aidatatang_test-greedy_search-epoch-20-avg-10-context-2-max-sym-per-frame-1-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+greedy_search	6.64

decoding_result/greedy_search/wer-summary-aishell-2_dev-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+greedy_search	3.59

decoding_result/greedy_search/wer-summary-aishell-2_test-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+greedy_search	4.03

decoding_result/greedy_search/wer-summary-aishell-4-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+greedy_search	15.68

decoding_result/greedy_search/wer-summary-aishell_dev-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+greedy_search	2.91

decoding_result/greedy_search/wer-summary-aishell_test-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+greedy_search	3.04

decoding_result/greedy_search/wer-summary-alimeeting_eval-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+greedy_search	23.15

decoding_result/greedy_search/wer-summary-alimeeting_test-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+greedy_search	24.78

decoding_result/greedy_search/wer-summary-kespeech-asr_dev_phase1-greedy_search-epoch-20-avg-1-context-2-max-sym-per-frame-1-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+greedy_search	6.69