File size: 26,018 Bytes
7708e74 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 |
2022-07-20 21:02:14,439 INFO [decode.py:653] Decoding started
2022-07-20 21:02:14,440 INFO [decode.py:654] {'subsampling_factor': 4, 'feature_dim': 80, 'nhead': 8, 'dim_feedforward': 2048, 'encoder_dim': 512, 'num_encoder_layers': 12, 'search_beam': 20, 'output_beam': 8, 'min_active_states': 30, 'max_active_states': 10000, 'use_double_scores': True, 'env_info': {'k2-version': '1.15.1', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': '350e449fe5d7b4231f77eb0add764782eed9f5d2', 'k2-git-date': 'Thu May 26 22:56:24 2022', 'lhotse-version': '1.3.0.dev+git.232f3eb.dirty', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'conformer_ctc2', 'icefall-git-sha1': 'd812f82-dirty', 'icefall-git-date': 'Wed Jul 20 19:47:36 2022', 'icefall-path': '/home/storage23/wangquandong/experiment/open_source/icefall', 'k2-path': '/home/storage23/wangquandong/tools/k2/k2/python/k2/__init__.py', 'lhotse-path': '/home/storage23/wangquandong/tools/lhotse/lhotse/__init__.py', 'hostname': 'tj1-asr-train-v100-30.kscn', 'IP address': '10.38.22.183'}, 'epoch': 30, 'iter': 0, 'avg': 8, 'method': 'whole-lattice-rescoring', 'use_averaged_model': True, 'num_decoder_layers': 6, 'num_paths': 100, 'nbest_scale': 0.5, 'exp_dir': PosixPath('pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp'), 'lang_dir': PosixPath('data/lang_bpe_500'), 'lm_dir': PosixPath('data/lm'), 'full_libri': True, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 20, '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'}
2022-07-20 21:02:14,840 INFO [lexicon.py:176] Loading pre-compiled data/lang_bpe_500/Linv.pt
2022-07-20 21:02:14,895 INFO [decode.py:664] device: cuda:0
2022-07-20 21:02:21,292 INFO [decode.py:730] Loading pre-compiled G_4_gram.pt
2022-07-20 21:02:26,887 INFO [decode.py:821] Calculating the averaged model over epoch range from 22 (excluded) to 30
2022-07-20 21:02:30,017 INFO [decode.py:837] Number of model parameters: 103071035
2022-07-20 21:02:30,018 INFO [asr_datamodule.py:444] About to get test-clean cuts
2022-07-20 21:02:30,022 INFO [asr_datamodule.py:451] About to get test-other cuts
2022-07-20 21:02:31,504 INFO [decode.py:588] batch 0/?, cuts processed until now is 1
2022-07-20 21:03:17,856 INFO [decode.py:588] batch 100/?, cuts processed until now is 212
2022-07-20 21:03:58,002 INFO [decode.py:588] batch 200/?, cuts processed until now is 459
2022-07-20 21:04:39,673 INFO [decode.py:588] batch 300/?, cuts processed until now is 717
2022-07-20 21:05:25,026 INFO [decode.py:588] batch 400/?, cuts processed until now is 949
2022-07-20 21:06:10,220 INFO [decode.py:588] batch 500/?, cuts processed until now is 1163
2022-07-20 21:06:51,165 INFO [decode.py:588] batch 600/?, cuts processed until now is 1411
2022-07-20 21:07:30,465 INFO [decode.py:588] batch 700/?, cuts processed until now is 1652
2022-07-20 21:08:14,019 INFO [decode.py:588] batch 800/?, cuts processed until now is 1867
2022-07-20 21:08:52,156 INFO [decode.py:588] batch 900/?, cuts processed until now is 2122
2022-07-20 21:09:33,815 INFO [decode.py:588] batch 1000/?, cuts processed until now is 2321
2022-07-20 21:10:18,006 INFO [decode.py:588] batch 1100/?, cuts processed until now is 2468
2022-07-20 21:11:00,039 INFO [decode.py:588] batch 1200/?, cuts processed until now is 2601
2022-07-20 21:11:06,549 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-clean-lm_scale_0.1.txt
2022-07-20 21:11:06,631 INFO [utils.py:416] [test-clean-lm_scale_0.1] %WER 2.82% [1482 / 52576, 214 ins, 100 del, 1168 sub ]
2022-07-20 21:11:06,864 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-clean-lm_scale_0.1.txt
2022-07-20 21:11:06,893 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-clean-lm_scale_0.2.txt
2022-07-20 21:11:06,973 INFO [utils.py:416] [test-clean-lm_scale_0.2] %WER 2.74% [1440 / 52576, 206 ins, 102 del, 1132 sub ]
2022-07-20 21:11:07,204 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-clean-lm_scale_0.2.txt
2022-07-20 21:11:07,234 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-clean-lm_scale_0.3.txt
2022-07-20 21:11:07,313 INFO [utils.py:416] [test-clean-lm_scale_0.3] %WER 2.70% [1421 / 52576, 200 ins, 105 del, 1116 sub ]
2022-07-20 21:11:07,543 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-clean-lm_scale_0.3.txt
2022-07-20 21:11:07,572 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-clean-lm_scale_0.4.txt
2022-07-20 21:11:07,649 INFO [utils.py:416] [test-clean-lm_scale_0.4] %WER 2.67% [1403 / 52576, 195 ins, 109 del, 1099 sub ]
2022-07-20 21:11:07,879 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-clean-lm_scale_0.4.txt
2022-07-20 21:11:07,907 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-clean-lm_scale_0.5.txt
2022-07-20 21:11:07,984 INFO [utils.py:416] [test-clean-lm_scale_0.5] %WER 2.66% [1400 / 52576, 188 ins, 112 del, 1100 sub ]
2022-07-20 21:11:08,213 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-clean-lm_scale_0.5.txt
2022-07-20 21:11:08,241 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-clean-lm_scale_0.6.txt
2022-07-20 21:11:08,318 INFO [utils.py:416] [test-clean-lm_scale_0.6] %WER 2.68% [1407 / 52576, 181 ins, 119 del, 1107 sub ]
2022-07-20 21:11:08,549 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-clean-lm_scale_0.6.txt
2022-07-20 21:11:08,577 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-clean-lm_scale_0.7.txt
2022-07-20 21:11:08,794 INFO [utils.py:416] [test-clean-lm_scale_0.7] %WER 2.68% [1409 / 52576, 173 ins, 130 del, 1106 sub ]
2022-07-20 21:11:09,022 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-clean-lm_scale_0.7.txt
2022-07-20 21:11:09,050 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-clean-lm_scale_0.8.txt
2022-07-20 21:11:09,127 INFO [utils.py:416] [test-clean-lm_scale_0.8] %WER 2.70% [1422 / 52576, 169 ins, 142 del, 1111 sub ]
2022-07-20 21:11:09,369 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-clean-lm_scale_0.8.txt
2022-07-20 21:11:09,400 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-clean-lm_scale_0.9.txt
2022-07-20 21:11:09,476 INFO [utils.py:416] [test-clean-lm_scale_0.9] %WER 2.74% [1443 / 52576, 156 ins, 167 del, 1120 sub ]
2022-07-20 21:11:09,705 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-clean-lm_scale_0.9.txt
2022-07-20 21:11:09,735 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-clean-lm_scale_1.0.txt
2022-07-20 21:11:09,811 INFO [utils.py:416] [test-clean-lm_scale_1.0] %WER 2.84% [1491 / 52576, 156 ins, 203 del, 1132 sub ]
2022-07-20 21:11:10,041 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-clean-lm_scale_1.0.txt
2022-07-20 21:11:10,070 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-clean-lm_scale_1.1.txt
2022-07-20 21:11:10,147 INFO [utils.py:416] [test-clean-lm_scale_1.1] %WER 2.94% [1545 / 52576, 151 ins, 246 del, 1148 sub ]
2022-07-20 21:11:10,377 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-clean-lm_scale_1.1.txt
2022-07-20 21:11:10,407 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-clean-lm_scale_1.2.txt
2022-07-20 21:11:10,485 INFO [utils.py:416] [test-clean-lm_scale_1.2] %WER 3.09% [1623 / 52576, 146 ins, 300 del, 1177 sub ]
2022-07-20 21:11:10,714 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-clean-lm_scale_1.2.txt
2022-07-20 21:11:10,743 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-clean-lm_scale_1.3.txt
2022-07-20 21:11:10,820 INFO [utils.py:416] [test-clean-lm_scale_1.3] %WER 3.30% [1734 / 52576, 149 ins, 376 del, 1209 sub ]
2022-07-20 21:11:11,203 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-clean-lm_scale_1.3.txt
2022-07-20 21:11:11,232 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-clean-lm_scale_1.4.txt
2022-07-20 21:11:11,310 INFO [utils.py:416] [test-clean-lm_scale_1.4] %WER 3.51% [1846 / 52576, 148 ins, 462 del, 1236 sub ]
2022-07-20 21:11:11,539 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-clean-lm_scale_1.4.txt
2022-07-20 21:11:11,567 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-clean-lm_scale_1.5.txt
2022-07-20 21:11:11,644 INFO [utils.py:416] [test-clean-lm_scale_1.5] %WER 3.79% [1992 / 52576, 144 ins, 567 del, 1281 sub ]
2022-07-20 21:11:11,873 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-clean-lm_scale_1.5.txt
2022-07-20 21:11:11,901 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-clean-lm_scale_1.6.txt
2022-07-20 21:11:11,978 INFO [utils.py:416] [test-clean-lm_scale_1.6] %WER 4.07% [2140 / 52576, 131 ins, 698 del, 1311 sub ]
2022-07-20 21:11:12,209 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-clean-lm_scale_1.6.txt
2022-07-20 21:11:12,237 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-clean-lm_scale_1.7.txt
2022-07-20 21:11:12,315 INFO [utils.py:416] [test-clean-lm_scale_1.7] %WER 4.38% [2303 / 52576, 131 ins, 826 del, 1346 sub ]
2022-07-20 21:11:12,544 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-clean-lm_scale_1.7.txt
2022-07-20 21:11:12,572 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-clean-lm_scale_1.8.txt
2022-07-20 21:11:12,649 INFO [utils.py:416] [test-clean-lm_scale_1.8] %WER 4.69% [2465 / 52576, 131 ins, 955 del, 1379 sub ]
2022-07-20 21:11:12,881 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-clean-lm_scale_1.8.txt
2022-07-20 21:11:12,910 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-clean-lm_scale_1.9.txt
2022-07-20 21:11:12,989 INFO [utils.py:416] [test-clean-lm_scale_1.9] %WER 4.99% [2625 / 52576, 133 ins, 1068 del, 1424 sub ]
2022-07-20 21:11:13,366 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-clean-lm_scale_1.9.txt
2022-07-20 21:11:13,405 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-clean-lm_scale_2.0.txt
2022-07-20 21:11:13,494 INFO [utils.py:416] [test-clean-lm_scale_2.0] %WER 5.31% [2794 / 52576, 130 ins, 1197 del, 1467 sub ]
2022-07-20 21:11:13,727 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-clean-lm_scale_2.0.txt
2022-07-20 21:11:13,728 INFO [decode.py:637]
For test-clean, WER of different settings are:
lm_scale_0.5 2.66 best for test-clean
lm_scale_0.4 2.67
lm_scale_0.6 2.68
lm_scale_0.7 2.68
lm_scale_0.3 2.7
lm_scale_0.8 2.7
lm_scale_0.2 2.74
lm_scale_0.9 2.74
lm_scale_0.1 2.82
lm_scale_1.0 2.84
lm_scale_1.1 2.94
lm_scale_1.2 3.09
lm_scale_1.3 3.3
lm_scale_1.4 3.51
lm_scale_1.5 3.79
lm_scale_1.6 4.07
lm_scale_1.7 4.38
lm_scale_1.8 4.69
lm_scale_1.9 4.99
lm_scale_2.0 5.31
2022-07-20 21:11:14,658 INFO [decode.py:588] batch 0/?, cuts processed until now is 2
2022-07-20 21:11:54,814 INFO [decode.py:588] batch 100/?, cuts processed until now is 249
2022-07-20 21:12:32,635 INFO [decode.py:588] batch 200/?, cuts processed until now is 540
2022-07-20 21:13:09,226 INFO [decode.py:588] batch 300/?, cuts processed until now is 834
2022-07-20 21:13:44,899 INFO [decode.py:588] batch 400/?, cuts processed until now is 1104
2022-07-20 21:14:26,260 INFO [decode.py:588] batch 500/?, cuts processed until now is 1350
2022-07-20 21:15:05,158 INFO [decode.py:588] batch 600/?, cuts processed until now is 1640
2022-07-20 21:15:40,242 INFO [decode.py:588] batch 700/?, cuts processed until now is 1933
2022-07-20 21:16:18,132 INFO [decode.py:588] batch 800/?, cuts processed until now is 2175
2022-07-20 21:16:56,866 INFO [decode.py:588] batch 900/?, cuts processed until now is 2440
2022-07-20 21:17:04,052 INFO [decode.py:783] Caught exception:
CUDA out of memory. Tried to allocate 740.00 MiB (GPU 0; 31.75 GiB total capacity; 28.28 GiB already allocated; 445.50 MiB free; 29.93 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF
2022-07-20 21:17:04,052 INFO [decode.py:789] num_arcs before pruning: 349313
2022-07-20 21:17:04,052 INFO [decode.py:792] This OOM is not an error. You can ignore it. If your model does not converge well, or --max-duration is too large, or the input sound file is difficult to decode, you will meet this exception.
2022-07-20 21:17:04,072 INFO [decode.py:803] num_arcs after pruning: 8796
2022-07-20 21:17:41,206 INFO [decode.py:588] batch 1000/?, cuts processed until now is 2656
2022-07-20 21:18:21,669 INFO [decode.py:588] batch 1100/?, cuts processed until now is 2839
2022-07-20 21:18:59,437 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-other-lm_scale_0.1.txt
2022-07-20 21:18:59,529 INFO [utils.py:416] [test-other-lm_scale_0.1] %WER 6.24% [3266 / 52343, 440 ins, 240 del, 2586 sub ]
2022-07-20 21:18:59,774 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-other-lm_scale_0.1.txt
2022-07-20 21:18:59,807 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-other-lm_scale_0.2.txt
2022-07-20 21:18:59,890 INFO [utils.py:416] [test-other-lm_scale_0.2] %WER 6.04% [3161 / 52343, 414 ins, 245 del, 2502 sub ]
2022-07-20 21:19:00,128 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-other-lm_scale_0.2.txt
2022-07-20 21:19:00,161 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-other-lm_scale_0.3.txt
2022-07-20 21:19:00,243 INFO [utils.py:416] [test-other-lm_scale_0.3] %WER 5.94% [3107 / 52343, 396 ins, 249 del, 2462 sub ]
2022-07-20 21:19:00,480 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-other-lm_scale_0.3.txt
2022-07-20 21:19:00,512 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-other-lm_scale_0.4.txt
2022-07-20 21:19:00,594 INFO [utils.py:416] [test-other-lm_scale_0.4] %WER 5.85% [3060 / 52343, 379 ins, 268 del, 2413 sub ]
2022-07-20 21:19:01,006 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-other-lm_scale_0.4.txt
2022-07-20 21:19:01,035 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-other-lm_scale_0.5.txt
2022-07-20 21:19:01,118 INFO [utils.py:416] [test-other-lm_scale_0.5] %WER 5.77% [3022 / 52343, 351 ins, 286 del, 2385 sub ]
2022-07-20 21:19:01,356 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-other-lm_scale_0.5.txt
2022-07-20 21:19:01,385 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-other-lm_scale_0.6.txt
2022-07-20 21:19:01,467 INFO [utils.py:416] [test-other-lm_scale_0.6] %WER 5.76% [3017 / 52343, 332 ins, 320 del, 2365 sub ]
2022-07-20 21:19:01,708 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-other-lm_scale_0.6.txt
2022-07-20 21:19:01,736 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-other-lm_scale_0.7.txt
2022-07-20 21:19:01,816 INFO [utils.py:416] [test-other-lm_scale_0.7] %WER 5.77% [3020 / 52343, 318 ins, 348 del, 2354 sub ]
2022-07-20 21:19:02,052 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-other-lm_scale_0.7.txt
2022-07-20 21:19:02,081 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-other-lm_scale_0.8.txt
2022-07-20 21:19:02,162 INFO [utils.py:416] [test-other-lm_scale_0.8] %WER 5.78% [3028 / 52343, 294 ins, 398 del, 2336 sub ]
2022-07-20 21:19:02,398 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-other-lm_scale_0.8.txt
2022-07-20 21:19:02,427 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-other-lm_scale_0.9.txt
2022-07-20 21:19:02,507 INFO [utils.py:416] [test-other-lm_scale_0.9] %WER 5.86% [3066 / 52343, 278 ins, 464 del, 2324 sub ]
2022-07-20 21:19:02,743 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-other-lm_scale_0.9.txt
2022-07-20 21:19:02,771 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-other-lm_scale_1.0.txt
2022-07-20 21:19:02,852 INFO [utils.py:416] [test-other-lm_scale_1.0] %WER 6.02% [3149 / 52343, 268 ins, 540 del, 2341 sub ]
2022-07-20 21:19:03,087 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-other-lm_scale_1.0.txt
2022-07-20 21:19:03,115 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-other-lm_scale_1.1.txt
2022-07-20 21:19:03,196 INFO [utils.py:416] [test-other-lm_scale_1.1] %WER 6.20% [3247 / 52343, 258 ins, 636 del, 2353 sub ]
2022-07-20 21:19:03,431 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-other-lm_scale_1.1.txt
2022-07-20 21:19:03,460 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-other-lm_scale_1.2.txt
2022-07-20 21:19:03,686 INFO [utils.py:416] [test-other-lm_scale_1.2] %WER 6.40% [3351 / 52343, 248 ins, 749 del, 2354 sub ]
2022-07-20 21:19:03,923 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-other-lm_scale_1.2.txt
2022-07-20 21:19:03,951 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-other-lm_scale_1.3.txt
2022-07-20 21:19:04,036 INFO [utils.py:416] [test-other-lm_scale_1.3] %WER 6.67% [3489 / 52343, 241 ins, 891 del, 2357 sub ]
2022-07-20 21:19:04,280 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-other-lm_scale_1.3.txt
2022-07-20 21:19:04,309 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-other-lm_scale_1.4.txt
2022-07-20 21:19:04,392 INFO [utils.py:416] [test-other-lm_scale_1.4] %WER 7.08% [3705 / 52343, 234 ins, 1071 del, 2400 sub ]
2022-07-20 21:19:04,628 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-other-lm_scale_1.4.txt
2022-07-20 21:19:04,656 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-other-lm_scale_1.5.txt
2022-07-20 21:19:04,737 INFO [utils.py:416] [test-other-lm_scale_1.5] %WER 7.42% [3886 / 52343, 224 ins, 1253 del, 2409 sub ]
2022-07-20 21:19:04,973 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-other-lm_scale_1.5.txt
2022-07-20 21:19:05,001 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-other-lm_scale_1.6.txt
2022-07-20 21:19:05,081 INFO [utils.py:416] [test-other-lm_scale_1.6] %WER 7.84% [4102 / 52343, 214 ins, 1437 del, 2451 sub ]
2022-07-20 21:19:05,317 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-other-lm_scale_1.6.txt
2022-07-20 21:19:05,346 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-other-lm_scale_1.7.txt
2022-07-20 21:19:05,426 INFO [utils.py:416] [test-other-lm_scale_1.7] %WER 8.29% [4338 / 52343, 213 ins, 1641 del, 2484 sub ]
2022-07-20 21:19:05,667 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-other-lm_scale_1.7.txt
2022-07-20 21:19:05,696 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-other-lm_scale_1.8.txt
2022-07-20 21:19:05,777 INFO [utils.py:416] [test-other-lm_scale_1.8] %WER 8.78% [4596 / 52343, 209 ins, 1859 del, 2528 sub ]
2022-07-20 21:19:06,031 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-other-lm_scale_1.8.txt
2022-07-20 21:19:06,082 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-other-lm_scale_1.9.txt
2022-07-20 21:19:06,310 INFO [utils.py:416] [test-other-lm_scale_1.9] %WER 9.27% [4854 / 52343, 205 ins, 2085 del, 2564 sub ]
2022-07-20 21:19:06,549 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-other-lm_scale_1.9.txt
2022-07-20 21:19:06,578 INFO [decode.py:609] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-other-lm_scale_2.0.txt
2022-07-20 21:19:06,659 INFO [utils.py:416] [test-other-lm_scale_2.0] %WER 9.72% [5090 / 52343, 204 ins, 2272 del, 2614 sub ]
2022-07-20 21:19:06,896 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-other-lm_scale_2.0.txt
2022-07-20 21:19:06,897 INFO [decode.py:637]
For test-other, WER of different settings are:
lm_scale_0.6 5.76 best for test-other
lm_scale_0.5 5.77
lm_scale_0.7 5.77
lm_scale_0.8 5.78
lm_scale_0.4 5.85
lm_scale_0.9 5.86
lm_scale_0.3 5.94
lm_scale_1.0 6.02
lm_scale_0.2 6.04
lm_scale_1.1 6.2
lm_scale_0.1 6.24
lm_scale_1.2 6.4
lm_scale_1.3 6.67
lm_scale_1.4 7.08
lm_scale_1.5 7.42
lm_scale_1.6 7.84
lm_scale_1.7 8.29
lm_scale_1.8 8.78
lm_scale_1.9 9.27
lm_scale_2.0 9.72
2022-07-20 21:19:06,897 INFO [decode.py:868] Done!
|