decoding_results/fast_beam_search/log-decode-epoch-24-avg-5-beam-20.0-max-contexts-8-max-states-64-blank-penalty-2.2-use-averaged-model-2024-03-20-12-17-31

2024-03-20 12:17:31,401 INFO [decode_mdcc.py:651] Decoding started
2024-03-20 12:17:31,401 INFO [decode_mdcc.py:657] Device: cuda:0
2024-03-20 12:17:31,433 INFO [lexicon.py:168] Loading pre-compiled data/zh-HK/lang_char/Linv.pt
2024-03-20 12:17:31,444 INFO [decode_mdcc.py:668] {'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': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': '2989b0b1186fa6022932804f5b39fbb2781ebf42', 'k2-git-date': 'Fri Nov 24 11:34:10 2023', 'lhotse-version': '1.22.0.dev+git.d8ed1bbb.dirty', 'torch-version': '1.11.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.9', 'icefall-git-branch': 'dev/cv-zipformer', 'icefall-git-sha1': '6993183d-dirty', 'icefall-git-date': 'Fri Mar 15 19:31:35 2024', '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/k2-1.24.4.dev20231207+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-1.22.0.dev0+git.d8ed1bbb.dirty-py3.9.egg/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-1-1207150822-75498b8c5f-sb8fn', 'IP address': '10.177.74.212'}, 'epoch': 24, 'iter': 0, 'avg': 5, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_val'), 'lang_dir': PosixPath('data/zh-HK/lang_char'), 'decoding_method': 'fast_beam_search', 'beam_size': 4, 'beam': 20.0, 'ngram_lm_scale': 0.01, 'ilme_scale': 0.2, 'max_contexts': 8, 'max_states': 64, 'context_size': 1, 'max_sym_per_frame': 1, 'num_paths': 200, 'nbest_scale': 0.5, 'blank_penalty': 2.2, '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, 'language': 'en', 'cv_manifest_dir': PosixPath('data/en/fbank'), 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 200.0, '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_val/fast_beam_search'), 'suffix': 'epoch-24-avg-5-beam-20.0-max-contexts-8-max-states-64-blank-penalty-2.2-use-averaged-model', 'blank_id': 0, 'vocab_size': 3904}
2024-03-20 12:17:31,445 INFO [decode_mdcc.py:670] About to create model
2024-03-20 12:17:32,070 INFO [decode_mdcc.py:737] Calculating the averaged model over epoch range from 19 (excluded) to 24
2024-03-20 12:17:41,247 INFO [decode_mdcc.py:768] Number of model parameters: 72526519
2024-03-20 12:17:41,279 INFO [asr_datamodule.py:353] About to create dev dataset
2024-03-20 12:17:41,589 INFO [asr_datamodule.py:370] About to create dev dataloader
2024-03-20 12:17:43,400 INFO [decode_mdcc.py:547] batch 0/?, cuts processed until now is 39
2024-03-20 12:17:52,962 INFO [decode_mdcc.py:547] batch 20/?, cuts processed until now is 945
2024-03-20 12:18:01,932 INFO [decode_mdcc.py:547] batch 40/?, cuts processed until now is 1923
2024-03-20 12:18:03,417 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.4093, 3.3140, 3.3377, 3.2614, 3.5912, 2.8249, 3.3461, 3.3388],
       device='cuda:0')
2024-03-20 12:18:10,516 INFO [decode_mdcc.py:547] batch 60/?, cuts processed until now is 3088
2024-03-20 12:18:11,076 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.3791, 5.2622, 5.3177, 4.9422], device='cuda:0')
2024-03-20 12:18:16,712 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.4935, 4.1661, 3.8613, 4.2513], device='cuda:0')
2024-03-20 12:18:19,484 INFO [decode_mdcc.py:547] batch 80/?, cuts processed until now is 4031
2024-03-20 12:18:28,884 INFO [decode_mdcc.py:547] batch 100/?, cuts processed until now is 5361
2024-03-20 12:18:31,758 INFO [decode_mdcc.py:563] The transcripts are stored in zipformer/exp_val/fast_beam_search/recogs-mdcc_dev-blank_penalty_2.2_beam_20.0_max_contexts_8_max_states_64-epoch-24-avg-5-beam-20.0-max-contexts-8-max-states-64-blank-penalty-2.2-use-averaged-model.txt
2024-03-20 12:18:31,938 INFO [utils.py:656] [mdcc_dev-blank_penalty_2.2_beam_20.0_max_contexts_8_max_states_64] %WER 37.73% [25866 / 68548, 690 ins, 3266 del, 21910 sub ]
2024-03-20 12:18:32,278 INFO [decode_mdcc.py:576] Wrote detailed error stats to zipformer/exp_val/fast_beam_search/errs-mdcc_dev-blank_penalty_2.2_beam_20.0_max_contexts_8_max_states_64-epoch-24-avg-5-beam-20.0-max-contexts-8-max-states-64-blank-penalty-2.2-use-averaged-model.txt
2024-03-20 12:18:32,281 INFO [decode_mdcc.py:592] 
For mdcc_dev, WER of different settings are:
blank_penalty_2.2_beam_20.0_max_contexts_8_max_states_64	37.73	best for mdcc_dev

2024-03-20 12:18:33,878 INFO [decode_mdcc.py:547] batch 0/?, cuts processed until now is 39
2024-03-20 12:18:43,243 INFO [decode_mdcc.py:547] batch 20/?, cuts processed until now is 958
2024-03-20 12:18:51,633 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.5525, 2.9787, 3.0497, 2.8733, 2.6522, 2.9475, 3.1379, 3.0742],
       device='cuda:0')
2024-03-20 12:18:52,146 INFO [decode_mdcc.py:547] batch 40/?, cuts processed until now is 1954
2024-03-20 12:19:00,720 INFO [decode_mdcc.py:547] batch 60/?, cuts processed until now is 3138
2024-03-20 12:19:10,637 INFO [decode_mdcc.py:547] batch 80/?, cuts processed until now is 3977
2024-03-20 12:19:19,339 INFO [decode_mdcc.py:547] batch 100/?, cuts processed until now is 5175
2024-03-20 12:19:29,094 INFO [decode_mdcc.py:547] batch 120/?, cuts processed until now is 6110
2024-03-20 12:19:32,470 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([2.2357, 2.2310, 3.0345, 2.6478], device='cuda:0')
2024-03-20 12:19:37,753 INFO [decode_mdcc.py:547] batch 140/?, cuts processed until now is 7375
2024-03-20 12:19:46,063 INFO [decode_mdcc.py:547] batch 160/?, cuts processed until now is 8663
2024-03-20 12:19:54,479 INFO [decode_mdcc.py:547] batch 180/?, cuts processed until now is 9787
2024-03-20 12:19:54,521 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.1584, 3.8447, 3.6365, 3.9090], device='cuda:0')
2024-03-20 12:20:04,090 INFO [decode_mdcc.py:547] batch 200/?, cuts processed until now is 10707
2024-03-20 12:20:13,395 INFO [decode_mdcc.py:547] batch 220/?, cuts processed until now is 11738
2024-03-20 12:20:18,039 INFO [decode_mdcc.py:563] The transcripts are stored in zipformer/exp_val/fast_beam_search/recogs-mdcc_test-blank_penalty_2.2_beam_20.0_max_contexts_8_max_states_64-epoch-24-avg-5-beam-20.0-max-contexts-8-max-states-64-blank-penalty-2.2-use-averaged-model.txt
2024-03-20 12:20:18,376 INFO [utils.py:656] [mdcc_test-blank_penalty_2.2_beam_20.0_max_contexts_8_max_states_64] %WER 37.74% [56403 / 149459, 1489 ins, 7397 del, 47517 sub ]
2024-03-20 12:20:19,080 INFO [decode_mdcc.py:576] Wrote detailed error stats to zipformer/exp_val/fast_beam_search/errs-mdcc_test-blank_penalty_2.2_beam_20.0_max_contexts_8_max_states_64-epoch-24-avg-5-beam-20.0-max-contexts-8-max-states-64-blank-penalty-2.2-use-averaged-model.txt
2024-03-20 12:20:19,083 INFO [decode_mdcc.py:592] 
For mdcc_test, WER of different settings are:
blank_penalty_2.2_beam_20.0_max_contexts_8_max_states_64	37.74	best for mdcc_test

2024-03-20 12:20:19,083 INFO [decode_mdcc.py:809] Done!