2022-07-21 23:23:55,621 INFO [decode.py:743] Decoding started 2022-07-21 23:23:55,622 INFO [decode.py:744] {'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': '1.17', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': '9e88318adfd7c80290a96f8a888d279d45dc1564', 'k2-git-date': 'Mon Jul 18 16:26:06 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': 'b9fda2c-dirty', 'icefall-git-date': 'Thu Jul 21 22:06:09 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': 'ctc-decoding', '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'), 'rnn_lm_exp_dir': 'rnn_lm/exp', 'rnn_lm_epoch': 7, 'rnn_lm_avg': 2, 'rnn_lm_embedding_dim': 2048, 'rnn_lm_hidden_dim': 2048, 'rnn_lm_num_layers': 4, 'rnn_lm_tie_weights': False, 'full_libri': True, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 200, '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'} 2022-07-21 23:23:56,049 INFO [lexicon.py:176] Loading pre-compiled data/lang_bpe_500/Linv.pt 2022-07-21 23:23:56,127 INFO [decode.py:754] device: cuda:0 2022-07-21 23:24:03,033 INFO [decode.py:916] Calculating the averaged model over epoch range from 22 (excluded) to 30 2022-07-21 23:24:05,616 INFO [decode.py:932] Number of model parameters: 103071035 2022-07-21 23:24:05,616 INFO [asr_datamodule.py:444] About to get test-clean cuts 2022-07-21 23:24:05,632 INFO [asr_datamodule.py:451] About to get test-other cuts 2022-07-21 23:24:06,872 INFO [decode.py:678] batch 0/?, cuts processed until now is 15 2022-07-21 23:24:41,077 INFO [decode.py:678] batch 100/?, cuts processed until now is 2578 2022-07-21 23:24:41,935 INFO [decode.py:699] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-clean-ctc-decoding.txt 2022-07-21 23:24:42,014 INFO [utils.py:416] [test-clean-ctc-decoding] %WER 2.98% [1568 / 52576, 162 ins, 113 del, 1293 sub ] 2022-07-21 23:24:42,241 INFO [decode.py:711] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-clean-ctc-decoding.txt 2022-07-21 23:24:42,244 INFO [decode.py:727] For test-clean, WER of different settings are: ctc-decoding 2.98 best for test-clean 2022-07-21 23:24:43,208 INFO [decode.py:678] batch 0/?, cuts processed until now is 19 2022-07-21 23:25:17,339 INFO [decode.py:678] batch 100/?, cuts processed until now is 2888 2022-07-21 23:25:17,768 INFO [decode.py:699] The transcripts are stored in pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/recogs-test-other-ctc-decoding.txt 2022-07-21 23:25:17,849 INFO [utils.py:416] [test-other-ctc-decoding] %WER 7.14% [3735 / 52343, 375 ins, 280 del, 3080 sub ] 2022-07-21 23:25:18,086 INFO [decode.py:711] Wrote detailed error stats to pruned_transducer_stateless_smooth/quandongtest_8gpu_ctc_attnetion2-exp/errs-test-other-ctc-decoding.txt 2022-07-21 23:25:18,087 INFO [decode.py:727] For test-other, WER of different settings are: ctc-decoding 7.14 best for test-other 2022-07-21 23:25:18,087 INFO [decode.py:989] Done!