2024-09-27 11:13:02,392 INFO [ctc_decode_dev.py:769] Decoding started 2024-09-27 11:13:02,392 INFO [ctc_decode_dev.py:775] Device: cuda:0 2024-09-27 11:13:02,393 INFO [ctc_decode_dev.py:776] {'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, 'ignore_id': -1, 'label_smoothing': 0.1, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': '44a9d5682af9fd3ef77074777e15278ec6d390eb', 'k2-git-date': 'Wed Sep 27 11:22:55 2023', 'lhotse-version': '1.17.0.dev+git.ccfc5b2c.dirty', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'cr-ctc', 'icefall-git-sha1': 'a6eead6c-clean', 'icefall-git-date': 'Mon Sep 9 10:10:08 2024', 'icefall-path': '/star-zw/workspace/zipformer/icefall_cr_ctc', 'k2-path': '/star-zw/workspace/k2/k2/k2/python/k2/__init__.py', 'lhotse-path': '/star-zw/workspace/lhotse/lhotse/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-6-0905175136-74fb5b4b6f-p65lp', 'IP address': '10.30.18.160'}, 'frame_shift_ms': 10, 'search_beam': 20, 'output_beam': 8, 'min_active_states': 30, 'max_active_states': 10000, 'use_double_scores': True, 'epoch': 50, 'iter': 0, 'avg': 26, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp-large-cr-loss-scale-0.2-time-mask-ratio-2.5-scaled-masked-1'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), 'context_size': 2, 'decoding_method': 'ctc-greedy-search', 'num_paths': 100, 'nbest_scale': 1.0, 'hlg_scale': 0.6, 'lm_dir': PosixPath('data/lm'), 'skip_scoring': False, 'num_encoder_layers': '2,2,4,5,4,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1536,2048,1536,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,512,768,512,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,320,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'decoder_dim': 512, 'joiner_dim': 512, 'attention_decoder_dim': 512, 'attention_decoder_num_layers': 6, 'attention_decoder_attention_dim': 512, 'attention_decoder_num_heads': 8, 'attention_decoder_feedforward_dim': 2048, 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'use_transducer': False, 'use_ctc': True, 'use_attention_decoder': False, 'use_cr_ctc': True, 'full_libri': True, 'mini_libri': False, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 300, '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-large-cr-loss-scale-0.2-time-mask-ratio-2.5-scaled-masked-1/ctc-greedy-search'), 'suffix': 'epoch-50_avg-26_use-averaged-model'} 2024-09-27 11:13:02,684 INFO [lexicon.py:168] Loading pre-compiled data/lang_bpe_500/Linv.pt 2024-09-27 11:13:11,717 INFO [ctc_decode_dev.py:861] About to create model 2024-09-27 11:13:12,652 INFO [ctc_decode_dev.py:928] Calculating the averaged model over epoch range from 24 (excluded) to 50 2024-09-27 11:13:37,046 INFO [ctc_decode_dev.py:945] Number of model parameters: 147010094 2024-09-27 11:13:37,047 INFO [asr_datamodule.py:467] About to get test-clean cuts 2024-09-27 11:13:37,178 INFO [asr_datamodule.py:474] About to get test-other cuts 2024-09-27 11:13:37,179 INFO [asr_datamodule.py:453] About to get dev-clean cuts 2024-09-27 11:13:37,180 INFO [asr_datamodule.py:460] About to get dev-other cuts 2024-09-27 11:13:38,205 INFO [ctc_decode_dev.py:653] batch 0/?, cuts processed until now is 21 2024-09-27 11:13:54,280 INFO [ctc_decode_dev.py:674] The transcripts are stored in zipformer/exp-large-cr-loss-scale-0.2-time-mask-ratio-2.5-scaled-masked-1/ctc-greedy-search/recogs-test-clean-epoch-50_avg-26_use-averaged-model.txt 2024-09-27 11:13:54,376 INFO [utils.py:657] [test-clean_ctc-greedy-search] %WER 2.03% [1069 / 52576, 112 ins, 83 del, 874 sub ] 2024-09-27 11:13:54,584 INFO [ctc_decode_dev.py:701] Wrote detailed error stats to zipformer/exp-large-cr-loss-scale-0.2-time-mask-ratio-2.5-scaled-masked-1/ctc-greedy-search/errs-test-clean-epoch-50_avg-26_use-averaged-model.txt 2024-09-27 11:13:54,589 INFO [ctc_decode_dev.py:717] For test-clean, WER of different settings are: ctc-greedy-search 2.03 best for test-clean 2024-09-27 11:13:55,174 INFO [ctc_decode_dev.py:653] batch 0/?, cuts processed until now is 26 2024-09-27 11:13:58,069 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.4336, 3.6971, 2.7795, 4.6669], device='cuda:0') 2024-09-27 11:13:59,514 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([2.8737, 3.6196, 3.2033, 3.2610], device='cuda:0') 2024-09-27 11:14:02,456 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([2.8758, 3.3802, 3.0282, 3.2050, 3.3019, 3.5529, 3.6979, 3.9896], device='cuda:0') 2024-09-27 11:14:11,408 INFO [ctc_decode_dev.py:674] The transcripts are stored in zipformer/exp-large-cr-loss-scale-0.2-time-mask-ratio-2.5-scaled-masked-1/ctc-greedy-search/recogs-test-other-epoch-50_avg-26_use-averaged-model.txt 2024-09-27 11:14:11,508 INFO [utils.py:657] [test-other_ctc-greedy-search] %WER 4.37% [2286 / 52343, 218 ins, 191 del, 1877 sub ] 2024-09-27 11:14:11,730 INFO [ctc_decode_dev.py:701] Wrote detailed error stats to zipformer/exp-large-cr-loss-scale-0.2-time-mask-ratio-2.5-scaled-masked-1/ctc-greedy-search/errs-test-other-epoch-50_avg-26_use-averaged-model.txt 2024-09-27 11:14:11,743 INFO [ctc_decode_dev.py:717] For test-other, WER of different settings are: ctc-greedy-search 4.37 best for test-other 2024-09-27 11:14:12,583 INFO [ctc_decode_dev.py:653] batch 0/?, cuts processed until now is 25 2024-09-27 11:14:22,049 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.2402, 4.1981, 3.5850, 4.7534], device='cuda:0') 2024-09-27 11:14:23,708 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.2915, 3.5621, 2.6857, 4.3884], device='cuda:0') 2024-09-27 11:14:24,027 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.1753, 4.9304, 4.7392, 4.4662], device='cuda:0') 2024-09-27 11:14:27,784 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.3027, 4.2123, 3.6535, 3.6373], device='cuda:0') 2024-09-27 11:14:28,568 INFO [ctc_decode_dev.py:674] The transcripts are stored in zipformer/exp-large-cr-loss-scale-0.2-time-mask-ratio-2.5-scaled-masked-1/ctc-greedy-search/recogs-dev-clean-epoch-50_avg-26_use-averaged-model.txt 2024-09-27 11:14:28,671 INFO [utils.py:657] [dev-clean_ctc-greedy-search] %WER 1.94% [1056 / 54402, 106 ins, 99 del, 851 sub ] 2024-09-27 11:14:28,893 INFO [ctc_decode_dev.py:701] Wrote detailed error stats to zipformer/exp-large-cr-loss-scale-0.2-time-mask-ratio-2.5-scaled-masked-1/ctc-greedy-search/errs-dev-clean-epoch-50_avg-26_use-averaged-model.txt 2024-09-27 11:14:28,898 INFO [ctc_decode_dev.py:717] For dev-clean, WER of different settings are: ctc-greedy-search 1.94 best for dev-clean 2024-09-27 11:14:29,526 INFO [ctc_decode_dev.py:653] batch 0/?, cuts processed until now is 27 2024-09-27 11:14:32,217 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7798, 2.8527, 2.9879, 3.1019], device='cuda:0') 2024-09-27 11:14:43,721 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([2.6056, 3.4792, 3.1363, 3.6809], device='cuda:0') 2024-09-27 11:14:44,537 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.5547, 2.6661, 2.6173, 2.9606], device='cuda:0') 2024-09-27 11:14:45,159 INFO [ctc_decode_dev.py:674] The transcripts are stored in zipformer/exp-large-cr-loss-scale-0.2-time-mask-ratio-2.5-scaled-masked-1/ctc-greedy-search/recogs-dev-other-epoch-50_avg-26_use-averaged-model.txt 2024-09-27 11:14:45,262 INFO [utils.py:657] [dev-other_ctc-greedy-search] %WER 4.32% [2199 / 50948, 186 ins, 168 del, 1845 sub ] 2024-09-27 11:14:45,477 INFO [ctc_decode_dev.py:701] Wrote detailed error stats to zipformer/exp-large-cr-loss-scale-0.2-time-mask-ratio-2.5-scaled-masked-1/ctc-greedy-search/errs-dev-other-epoch-50_avg-26_use-averaged-model.txt 2024-09-27 11:14:45,481 INFO [ctc_decode_dev.py:717] For dev-other, WER of different settings are: ctc-greedy-search 4.32 best for dev-other 2024-09-27 11:14:45,482 INFO [ctc_decode_dev.py:989] Done!