2022-07-20 21:16:48,757 INFO [decode.py:653] Decoding started 2022-07-20 21:16:48,757 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/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': 'nbest-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': 30, '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 21:16:49,153 INFO [lexicon.py:176] Loading pre-compiled data/lang_bpe_500/Linv.pt 2022-07-20 21:16:49,217 INFO [decode.py:664] device: cuda:0 2022-07-20 21:16:54,771 INFO [decode.py:730] Loading pre-compiled G_4_gram.pt 2022-07-20 21:16:59,308 INFO [decode.py:821] Calculating the averaged model over epoch range from 22 (excluded) to 30 2022-07-20 21:17:01,993 INFO [decode.py:837] Number of model parameters: 103071035 2022-07-20 21:17:01,993 INFO [asr_datamodule.py:444] About to get test-clean cuts 2022-07-20 21:17:01,997 INFO [asr_datamodule.py:451] About to get test-other cuts 2022-07-20 21:17:03,382 INFO [decode.py:588] batch 0/?, cuts processed until now is 2 2022-07-20 21:17:47,578 INFO [decode.py:588] batch 100/?, cuts processed until now is 326 2022-07-20 21:18:29,336 INFO [decode.py:588] batch 200/?, cuts processed until now is 709 2022-07-20 21:19:07,578 INFO [decode.py:588] batch 300/?, cuts processed until now is 1126 2022-07-20 21:19:46,337 INFO [decode.py:588] batch 400/?, cuts processed until now is 1489 2022-07-20 21:20:27,408 INFO [decode.py:588] batch 500/?, cuts processed until now is 1823 2022-07-20 21:21:06,308 INFO [decode.py:588] batch 600/?, cuts processed until now is 2211 2022-07-20 21:21:50,370 INFO [decode.py:588] batch 700/?, cuts processed until now is 2474 2022-07-20 21:22:35,645 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 21:22:35,734 INFO [utils.py:410] [test-clean-lm_scale_0.1] %WER 2.81% [1476 / 52576, 202 ins, 102 del, 1172 sub ] 2022-07-20 21:22:36,154 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 21:22:36,184 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 21:22:36,264 INFO [utils.py:410] [test-clean-lm_scale_0.2] %WER 2.75% [1447 / 52576, 199 ins, 105 del, 1143 sub ] 2022-07-20 21:22:36,496 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 21:22:36,525 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 21:22:36,607 INFO [utils.py:410] [test-clean-lm_scale_0.3] %WER 2.72% [1428 / 52576, 195 ins, 107 del, 1126 sub ] 2022-07-20 21:22:36,840 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 21:22:36,870 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 21:22:36,949 INFO [utils.py:410] [test-clean-lm_scale_0.4] %WER 2.69% [1412 / 52576, 192 ins, 112 del, 1108 sub ] 2022-07-20 21:22:37,184 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 21:22:37,213 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 21:22:37,288 INFO [utils.py:410] [test-clean-lm_scale_0.5] %WER 2.68% [1409 / 52576, 188 ins, 115 del, 1106 sub ] 2022-07-20 21:22:37,516 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 21:22:37,544 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 21:22:37,621 INFO [utils.py:410] [test-clean-lm_scale_0.6] %WER 2.68% [1411 / 52576, 181 ins, 121 del, 1109 sub ] 2022-07-20 21:22:37,845 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 21:22:37,872 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 21:22:37,948 INFO [utils.py:410] [test-clean-lm_scale_0.7] %WER 2.68% [1409 / 52576, 173 ins, 131 del, 1105 sub ] 2022-07-20 21:22:38,179 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 21:22:38,206 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 21:22:38,282 INFO [utils.py:410] [test-clean-lm_scale_0.8] %WER 2.71% [1425 / 52576, 169 ins, 144 del, 1112 sub ] 2022-07-20 21:22:38,656 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 21:22:38,684 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 21:22:38,762 INFO [utils.py:410] [test-clean-lm_scale_0.9] %WER 2.74% [1441 / 52576, 156 ins, 164 del, 1121 sub ] 2022-07-20 21:22:38,987 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 21:22:39,016 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 21:22:39,102 INFO [utils.py:410] [test-clean-lm_scale_1.0] %WER 2.82% [1485 / 52576, 156 ins, 197 del, 1132 sub ] 2022-07-20 21:22:39,328 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 21:22:39,356 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 21:22:39,432 INFO [utils.py:410] [test-clean-lm_scale_1.1] %WER 2.91% [1530 / 52576, 153 ins, 234 del, 1143 sub ] 2022-07-20 21:22:39,667 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 21:22:39,703 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 21:22:39,779 INFO [utils.py:410] [test-clean-lm_scale_1.2] %WER 3.04% [1600 / 52576, 150 ins, 283 del, 1167 sub ] 2022-07-20 21:22:40,004 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 21:22:40,036 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 21:22:40,112 INFO [utils.py:410] [test-clean-lm_scale_1.3] %WER 3.23% [1697 / 52576, 150 ins, 347 del, 1200 sub ] 2022-07-20 21:22:40,341 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 21:22:40,372 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 21:22:40,448 INFO [utils.py:410] [test-clean-lm_scale_1.4] %WER 3.41% [1792 / 52576, 149 ins, 420 del, 1223 sub ] 2022-07-20 21:22:40,674 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 21:22:40,705 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 21:22:40,950 INFO [utils.py:410] [test-clean-lm_scale_1.5] %WER 3.62% [1902 / 52576, 146 ins, 493 del, 1263 sub ] 2022-07-20 21:22:41,181 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 21:22:41,209 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 21:22:41,286 INFO [utils.py:410] [test-clean-lm_scale_1.6] %WER 3.81% [2001 / 52576, 140 ins, 573 del, 1288 sub ] 2022-07-20 21:22:41,511 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 21:22:41,538 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 21:22:41,614 INFO [utils.py:410] [test-clean-lm_scale_1.7] %WER 3.99% [2100 / 52576, 142 ins, 651 del, 1307 sub ] 2022-07-20 21:22:41,841 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 21:22:41,868 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 21:22:41,949 INFO [utils.py:410] [test-clean-lm_scale_1.8] %WER 4.16% [2187 / 52576, 143 ins, 722 del, 1322 sub ] 2022-07-20 21:22:42,176 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 21:22:42,204 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 21:22:42,281 INFO [utils.py:410] [test-clean-lm_scale_1.9] %WER 4.32% [2272 / 52576, 141 ins, 776 del, 1355 sub ] 2022-07-20 21:22:42,508 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 21:22:42,536 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 21:22:42,612 INFO [utils.py:410] [test-clean-lm_scale_2.0] %WER 4.54% [2386 / 52576, 143 ins, 847 del, 1396 sub ] 2022-07-20 21:22:42,843 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 21:22:42,844 INFO [decode.py:637] For test-clean, WER of different settings are: lm_scale_0.5 2.68 best for test-clean lm_scale_0.6 2.68 lm_scale_0.7 2.68 lm_scale_0.4 2.69 lm_scale_0.8 2.71 lm_scale_0.3 2.72 lm_scale_0.9 2.74 lm_scale_0.2 2.75 lm_scale_0.1 2.81 lm_scale_1.0 2.82 lm_scale_1.1 2.91 lm_scale_1.2 3.04 lm_scale_1.3 3.23 lm_scale_1.4 3.41 lm_scale_1.5 3.62 lm_scale_1.6 3.81 lm_scale_1.7 3.99 lm_scale_1.8 4.16 lm_scale_1.9 4.32 lm_scale_2.0 4.54 2022-07-20 21:22:43,967 INFO [decode.py:588] batch 0/?, cuts processed until now is 2 2022-07-20 21:23:26,720 INFO [decode.py:588] batch 100/?, cuts processed until now is 377 2022-07-20 21:24:06,356 INFO [decode.py:588] batch 200/?, cuts processed until now is 811 2022-07-20 21:24:45,879 INFO [decode.py:588] batch 300/?, cuts processed until now is 1265 2022-07-20 21:25:24,826 INFO [decode.py:588] batch 400/?, cuts processed until now is 1684 2022-07-20 21:26:06,471 INFO [decode.py:588] batch 500/?, cuts processed until now is 2061 2022-07-20 21:26:50,778 INFO [decode.py:588] batch 600/?, cuts processed until now is 2495 2022-07-20 21:27:36,625 INFO [decode.py:588] batch 700/?, cuts processed until now is 2781 2022-07-20 21:28:29,094 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 21:28:29,184 INFO [utils.py:410] [test-other-lm_scale_0.1] %WER 6.16% [3224 / 52343, 397 ins, 263 del, 2564 sub ] 2022-07-20 21:28:29,431 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 21:28:29,464 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 21:28:29,553 INFO [utils.py:410] [test-other-lm_scale_0.2] %WER 6.01% [3148 / 52343, 379 ins, 262 del, 2507 sub ] 2022-07-20 21:28:29,800 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 21:28:29,830 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 21:28:29,912 INFO [utils.py:410] [test-other-lm_scale_0.3] %WER 5.93% [3105 / 52343, 365 ins, 269 del, 2471 sub ] 2022-07-20 21:28:30,172 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 21:28:30,206 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 21:28:30,286 INFO [utils.py:410] [test-other-lm_scale_0.4] %WER 5.84% [3057 / 52343, 347 ins, 291 del, 2419 sub ] 2022-07-20 21:28:30,523 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 21:28:30,556 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 21:28:30,640 INFO [utils.py:410] [test-other-lm_scale_0.5] %WER 5.77% [3021 / 52343, 333 ins, 301 del, 2387 sub ] 2022-07-20 21:28:30,885 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 21:28:30,920 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 21:28:31,004 INFO [utils.py:410] [test-other-lm_scale_0.6] %WER 5.78% [3027 / 52343, 326 ins, 333 del, 2368 sub ] 2022-07-20 21:28:31,246 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 21:28:31,285 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 21:28:31,367 INFO [utils.py:410] [test-other-lm_scale_0.7] %WER 5.82% [3046 / 52343, 314 ins, 369 del, 2363 sub ] 2022-07-20 21:28:31,607 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 21:28:31,641 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 21:28:31,732 INFO [utils.py:410] [test-other-lm_scale_0.8] %WER 5.82% [3045 / 52343, 297 ins, 412 del, 2336 sub ] 2022-07-20 21:28:31,974 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 21:28:32,006 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 21:28:32,086 INFO [utils.py:410] [test-other-lm_scale_0.9] %WER 5.89% [3084 / 52343, 285 ins, 463 del, 2336 sub ] 2022-07-20 21:28:32,543 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 21:28:32,576 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 21:28:32,660 INFO [utils.py:410] [test-other-lm_scale_1.0] %WER 6.04% [3159 / 52343, 273 ins, 530 del, 2356 sub ] 2022-07-20 21:28:32,903 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 21:28:32,936 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 21:28:33,019 INFO [utils.py:410] [test-other-lm_scale_1.1] %WER 6.20% [3244 / 52343, 268 ins, 598 del, 2378 sub ] 2022-07-20 21:28:33,282 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 21:28:33,312 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 21:28:33,397 INFO [utils.py:410] [test-other-lm_scale_1.2] %WER 6.31% [3302 / 52343, 257 ins, 667 del, 2378 sub ] 2022-07-20 21:28:33,635 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 21:28:33,667 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 21:28:33,750 INFO [utils.py:410] [test-other-lm_scale_1.3] %WER 6.58% [3444 / 52343, 250 ins, 787 del, 2407 sub ] 2022-07-20 21:28:33,989 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 21:28:34,021 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 21:28:34,105 INFO [utils.py:410] [test-other-lm_scale_1.4] %WER 6.88% [3603 / 52343, 246 ins, 921 del, 2436 sub ] 2022-07-20 21:28:34,341 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 21:28:34,373 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 21:28:34,468 INFO [utils.py:410] [test-other-lm_scale_1.5] %WER 7.09% [3713 / 52343, 235 ins, 1031 del, 2447 sub ] 2022-07-20 21:28:34,706 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 21:28:34,738 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 21:28:34,820 INFO [utils.py:410] [test-other-lm_scale_1.6] %WER 7.32% [3834 / 52343, 227 ins, 1131 del, 2476 sub ] 2022-07-20 21:28:35,258 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 21:28:35,287 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 21:28:35,380 INFO [utils.py:410] [test-other-lm_scale_1.7] %WER 7.54% [3945 / 52343, 223 ins, 1227 del, 2495 sub ] 2022-07-20 21:28:35,616 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 21:28:35,649 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 21:28:35,732 INFO [utils.py:410] [test-other-lm_scale_1.8] %WER 7.75% [4056 / 52343, 221 ins, 1320 del, 2515 sub ] 2022-07-20 21:28:35,969 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 21:28:35,998 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 21:28:36,106 INFO [utils.py:410] [test-other-lm_scale_1.9] %WER 7.97% [4170 / 52343, 222 ins, 1401 del, 2547 sub ] 2022-07-20 21:28:36,339 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 21:28:36,367 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 21:28:36,447 INFO [utils.py:410] [test-other-lm_scale_2.0] %WER 8.18% [4281 / 52343, 220 ins, 1479 del, 2582 sub ] 2022-07-20 21:28:36,681 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 21:28:36,681 INFO [decode.py:637] For test-other, WER of different settings are: lm_scale_0.5 5.77 best for test-other lm_scale_0.6 5.78 lm_scale_0.7 5.82 lm_scale_0.8 5.82 lm_scale_0.4 5.84 lm_scale_0.9 5.89 lm_scale_0.3 5.93 lm_scale_0.2 6.01 lm_scale_1.0 6.04 lm_scale_0.1 6.16 lm_scale_1.1 6.2 lm_scale_1.2 6.31 lm_scale_1.3 6.58 lm_scale_1.4 6.88 lm_scale_1.5 7.09 lm_scale_1.6 7.32 lm_scale_1.7 7.54 lm_scale_1.8 7.75 lm_scale_1.9 7.97 lm_scale_2.0 8.18 2022-07-20 21:28:36,682 INFO [decode.py:868] Done!