pruned-transducer-stateless7-streaming-id / exp /streaming /modified_beam_search /log-decode-epoch-30-avg-9-streaming-chunk-size-32-use-averaged-model-2023-06-21-10-06-49
w11wo's picture
Added Model
9a835b2
2023-06-21 10:06:49,545 INFO [streaming_decode.py:483] Decoding started
2023-06-21 10:06:49,546 INFO [streaming_decode.py:489] Device: cuda:0
2023-06-21 10:06:49,547 INFO [lexicon.py:168] Loading pre-compiled data/lang_phone/Linv.pt
2023-06-21 10:06:49,549 INFO [streaming_decode.py:497] {'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': '1.23.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': '9426c9f730820d291f5dcb06be337662595fa7b4', 'k2-git-date': 'Sun Feb 5 17:35:01 2023', 'lhotse-version': '1.15.0.dev+git.00d3e36.clean', 'torch-version': '1.13.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'master', 'icefall-git-sha1': 'd3f5d01-dirty', 'icefall-git-date': 'Wed May 31 04:15:45 2023', 'icefall-path': '/root/icefall', 'k2-path': '/usr/local/lib/python3.10/dist-packages/k2/__init__.py', 'lhotse-path': '/root/lhotse/lhotse/__init__.py', 'hostname': 'bookbot-k2', 'IP address': '127.0.0.1'}, 'epoch': 30, 'iter': 0, 'avg': 9, 'use_averaged_model': True, 'exp_dir': PosixPath('pruned_transducer_stateless7_streaming/exp'), 'lang_dir': 'data/lang_phone', 'decoding_method': 'modified_beam_search', 'num_active_paths': 4, 'beam': 4, 'max_contexts': 4, 'max_states': 32, 'context_size': 2, 'num_decode_streams': 1500, 'num_encoder_layers': '2,4,3,2,4', 'feedforward_dims': '1024,1024,2048,2048,1024', 'nhead': '8,8,8,8,8', 'encoder_dims': '384,384,384,384,384', 'attention_dims': '192,192,192,192,192', 'encoder_unmasked_dims': '256,256,256,256,256', 'zipformer_downsampling_factors': '1,2,4,8,2', 'cnn_module_kernels': '31,31,31,31,31', 'decoder_dim': 512, 'joiner_dim': 512, 'short_chunk_size': 50, 'num_left_chunks': 4, 'decode_chunk_len': 32, '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('pruned_transducer_stateless7_streaming/exp/streaming/modified_beam_search'), 'suffix': 'epoch-30-avg-9-streaming-chunk-size-32-use-averaged-model', 'blank_id': 0, 'unk_id': 7, 'vocab_size': 33}
2023-06-21 10:06:49,550 INFO [streaming_decode.py:499] About to create model
2023-06-21 10:06:50,126 INFO [zipformer.py:405] At encoder stack 4, which has downsampling_factor=2, we will combine the outputs of layers 1 and 3, with downsampling_factors=2 and 8.
2023-06-21 10:06:50,130 INFO [streaming_decode.py:566] Calculating the averaged model over epoch range from 21 (excluded) to 30
2023-06-21 10:06:53,715 INFO [streaming_decode.py:588] Number of model parameters: 69471350
2023-06-21 10:06:53,715 INFO [multidataset.py:122] About to get LibriVox test cuts
2023-06-21 10:06:53,715 INFO [multidataset.py:124] Loading LibriVox in lazy mode
2023-06-21 10:06:53,716 INFO [multidataset.py:133] About to get FLEURS test cuts
2023-06-21 10:06:53,716 INFO [multidataset.py:135] Loading FLEURS in lazy mode
2023-06-21 10:06:53,717 INFO [multidataset.py:144] About to get Common Voice test cuts
2023-06-21 10:06:53,717 INFO [multidataset.py:146] Loading Common Voice in lazy mode
2023-06-21 10:06:53,981 INFO [streaming_decode.py:380] Cuts processed until now is 0.
2023-06-21 10:06:54,290 INFO [streaming_decode.py:380] Cuts processed until now is 50.
2023-06-21 10:06:54,603 INFO [streaming_decode.py:380] Cuts processed until now is 100.
2023-06-21 10:06:54,976 INFO [streaming_decode.py:380] Cuts processed until now is 150.
2023-06-21 10:06:55,310 INFO [streaming_decode.py:380] Cuts processed until now is 200.
2023-06-21 10:06:55,643 INFO [streaming_decode.py:380] Cuts processed until now is 250.
2023-06-21 10:06:55,975 INFO [streaming_decode.py:380] Cuts processed until now is 300.
2023-06-21 10:06:56,319 INFO [streaming_decode.py:380] Cuts processed until now is 350.
2023-06-21 10:06:56,643 INFO [streaming_decode.py:380] Cuts processed until now is 400.
2023-06-21 10:06:56,969 INFO [streaming_decode.py:380] Cuts processed until now is 450.
2023-06-21 10:06:57,308 INFO [streaming_decode.py:380] Cuts processed until now is 500.
2023-06-21 10:06:57,648 INFO [streaming_decode.py:380] Cuts processed until now is 550.
2023-06-21 10:06:57,986 INFO [streaming_decode.py:380] Cuts processed until now is 600.
2023-06-21 10:07:19,334 INFO [streaming_decode.py:425] The transcripts are stored in pruned_transducer_stateless7_streaming/exp/streaming/modified_beam_search/recogs-test-librivox-epoch-30-avg-9-streaming-chunk-size-32-use-averaged-model.txt
2023-06-21 10:07:19,369 INFO [utils.py:561] [test-librivox-num_active_paths_4] %WER 4.78% [1748 / 36594, 298 ins, 852 del, 598 sub ]
2023-06-21 10:07:19,449 INFO [streaming_decode.py:436] Wrote detailed error stats to pruned_transducer_stateless7_streaming/exp/streaming/modified_beam_search/errs-test-librivox-epoch-30-avg-9-streaming-chunk-size-32-use-averaged-model.txt
2023-06-21 10:07:19,449 INFO [streaming_decode.py:450]
For test-librivox, WER of different settings are:
num_active_paths_4 4.78 best for test-librivox
2023-06-21 10:07:19,453 INFO [streaming_decode.py:380] Cuts processed until now is 0.
2023-06-21 10:07:19,628 INFO [streaming_decode.py:380] Cuts processed until now is 50.
2023-06-21 10:07:19,788 INFO [streaming_decode.py:380] Cuts processed until now is 100.
2023-06-21 10:07:19,954 INFO [streaming_decode.py:380] Cuts processed until now is 150.
2023-06-21 10:07:20,119 INFO [streaming_decode.py:380] Cuts processed until now is 200.
2023-06-21 10:07:20,284 INFO [streaming_decode.py:380] Cuts processed until now is 250.
2023-06-21 10:07:20,442 INFO [streaming_decode.py:380] Cuts processed until now is 300.
2023-06-21 10:07:20,604 INFO [streaming_decode.py:380] Cuts processed until now is 350.
2023-06-21 10:07:20,765 INFO [streaming_decode.py:380] Cuts processed until now is 400.
2023-06-21 10:07:21,023 INFO [streaming_decode.py:380] Cuts processed until now is 450.
2023-06-21 10:07:21,186 INFO [streaming_decode.py:380] Cuts processed until now is 500.
2023-06-21 10:07:21,355 INFO [streaming_decode.py:380] Cuts processed until now is 550.
2023-06-21 10:07:21,532 INFO [streaming_decode.py:380] Cuts processed until now is 600.
2023-06-21 10:07:21,718 INFO [streaming_decode.py:380] Cuts processed until now is 650.
2023-06-21 10:08:32,133 INFO [streaming_decode.py:425] The transcripts are stored in pruned_transducer_stateless7_streaming/exp/streaming/modified_beam_search/recogs-test-fleurs-epoch-30-avg-9-streaming-chunk-size-32-use-averaged-model.txt
2023-06-21 10:08:32,229 INFO [utils.py:561] [test-fleurs-num_active_paths_4] %WER 11.83% [11074 / 93580, 1827 ins, 4283 del, 4964 sub ]
2023-06-21 10:08:32,533 INFO [streaming_decode.py:436] Wrote detailed error stats to pruned_transducer_stateless7_streaming/exp/streaming/modified_beam_search/errs-test-fleurs-epoch-30-avg-9-streaming-chunk-size-32-use-averaged-model.txt
2023-06-21 10:08:32,533 INFO [streaming_decode.py:450]
For test-fleurs, WER of different settings are:
num_active_paths_4 11.83 best for test-fleurs
2023-06-21 10:08:32,539 INFO [streaming_decode.py:380] Cuts processed until now is 0.
2023-06-21 10:08:32,785 INFO [streaming_decode.py:380] Cuts processed until now is 50.
2023-06-21 10:08:33,021 INFO [streaming_decode.py:380] Cuts processed until now is 100.
2023-06-21 10:08:33,284 INFO [streaming_decode.py:380] Cuts processed until now is 150.
2023-06-21 10:08:33,521 INFO [streaming_decode.py:380] Cuts processed until now is 200.
2023-06-21 10:08:33,766 INFO [streaming_decode.py:380] Cuts processed until now is 250.
2023-06-21 10:08:33,989 INFO [streaming_decode.py:380] Cuts processed until now is 300.
2023-06-21 10:08:34,202 INFO [streaming_decode.py:380] Cuts processed until now is 350.
2023-06-21 10:08:34,450 INFO [streaming_decode.py:380] Cuts processed until now is 400.
2023-06-21 10:08:34,674 INFO [streaming_decode.py:380] Cuts processed until now is 450.
2023-06-21 10:08:34,898 INFO [streaming_decode.py:380] Cuts processed until now is 500.
2023-06-21 10:08:35,125 INFO [streaming_decode.py:380] Cuts processed until now is 550.
2023-06-21 10:08:35,339 INFO [streaming_decode.py:380] Cuts processed until now is 600.
2023-06-21 10:08:35,569 INFO [streaming_decode.py:380] Cuts processed until now is 650.
2023-06-21 10:08:35,789 INFO [streaming_decode.py:380] Cuts processed until now is 700.
2023-06-21 10:08:36,003 INFO [streaming_decode.py:380] Cuts processed until now is 750.
2023-06-21 10:08:36,229 INFO [streaming_decode.py:380] Cuts processed until now is 800.
2023-06-21 10:08:36,467 INFO [streaming_decode.py:380] Cuts processed until now is 850.
2023-06-21 10:08:36,697 INFO [streaming_decode.py:380] Cuts processed until now is 900.
2023-06-21 10:08:36,931 INFO [streaming_decode.py:380] Cuts processed until now is 950.
2023-06-21 10:08:37,159 INFO [streaming_decode.py:380] Cuts processed until now is 1000.
2023-06-21 10:08:37,390 INFO [streaming_decode.py:380] Cuts processed until now is 1050.
2023-06-21 10:08:37,613 INFO [streaming_decode.py:380] Cuts processed until now is 1100.
2023-06-21 10:08:37,855 INFO [streaming_decode.py:380] Cuts processed until now is 1150.
2023-06-21 10:08:38,229 INFO [streaming_decode.py:380] Cuts processed until now is 1200.
2023-06-21 10:08:38,471 INFO [streaming_decode.py:380] Cuts processed until now is 1250.
2023-06-21 10:08:38,707 INFO [streaming_decode.py:380] Cuts processed until now is 1300.
2023-06-21 10:08:38,959 INFO [streaming_decode.py:380] Cuts processed until now is 1350.
2023-06-21 10:08:39,198 INFO [streaming_decode.py:380] Cuts processed until now is 1400.
2023-06-21 10:08:39,430 INFO [streaming_decode.py:380] Cuts processed until now is 1450.
2023-06-21 10:09:02,777 INFO [streaming_decode.py:380] Cuts processed until now is 1500.
2023-06-21 10:09:09,770 INFO [streaming_decode.py:380] Cuts processed until now is 1550.
2023-06-21 10:09:13,345 INFO [streaming_decode.py:380] Cuts processed until now is 1600.
2023-06-21 10:09:13,569 INFO [streaming_decode.py:380] Cuts processed until now is 1650.
2023-06-21 10:09:17,336 INFO [streaming_decode.py:380] Cuts processed until now is 1700.
2023-06-21 10:09:17,563 INFO [streaming_decode.py:380] Cuts processed until now is 1750.
2023-06-21 10:09:17,784 INFO [streaming_decode.py:380] Cuts processed until now is 1800.
2023-06-21 10:09:21,540 INFO [streaming_decode.py:380] Cuts processed until now is 1850.
2023-06-21 10:09:21,765 INFO [streaming_decode.py:380] Cuts processed until now is 1900.
2023-06-21 10:09:21,991 INFO [streaming_decode.py:380] Cuts processed until now is 1950.
2023-06-21 10:09:25,582 INFO [streaming_decode.py:380] Cuts processed until now is 2000.
2023-06-21 10:09:25,804 INFO [streaming_decode.py:380] Cuts processed until now is 2050.
2023-06-21 10:09:26,038 INFO [streaming_decode.py:380] Cuts processed until now is 2100.
2023-06-21 10:09:26,384 INFO [streaming_decode.py:380] Cuts processed until now is 2150.
2023-06-21 10:09:30,004 INFO [streaming_decode.py:380] Cuts processed until now is 2200.
2023-06-21 10:09:30,224 INFO [streaming_decode.py:380] Cuts processed until now is 2250.
2023-06-21 10:09:30,453 INFO [streaming_decode.py:380] Cuts processed until now is 2300.
2023-06-21 10:09:34,177 INFO [streaming_decode.py:380] Cuts processed until now is 2350.
2023-06-21 10:09:34,406 INFO [streaming_decode.py:380] Cuts processed until now is 2400.
2023-06-21 10:09:34,633 INFO [streaming_decode.py:380] Cuts processed until now is 2450.
2023-06-21 10:09:38,413 INFO [streaming_decode.py:380] Cuts processed until now is 2500.
2023-06-21 10:09:38,627 INFO [streaming_decode.py:380] Cuts processed until now is 2550.
2023-06-21 10:09:38,854 INFO [streaming_decode.py:380] Cuts processed until now is 2600.
2023-06-21 10:09:42,578 INFO [streaming_decode.py:380] Cuts processed until now is 2650.
2023-06-21 10:09:42,791 INFO [streaming_decode.py:380] Cuts processed until now is 2700.
2023-06-21 10:09:46,553 INFO [streaming_decode.py:380] Cuts processed until now is 2750.
2023-06-21 10:09:46,786 INFO [streaming_decode.py:380] Cuts processed until now is 2800.
2023-06-21 10:09:50,532 INFO [streaming_decode.py:380] Cuts processed until now is 2850.
2023-06-21 10:09:54,139 INFO [streaming_decode.py:380] Cuts processed until now is 2900.
2023-06-21 10:09:57,761 INFO [streaming_decode.py:380] Cuts processed until now is 2950.
2023-06-21 10:10:01,512 INFO [streaming_decode.py:380] Cuts processed until now is 3000.
2023-06-21 10:10:01,734 INFO [streaming_decode.py:380] Cuts processed until now is 3050.
2023-06-21 10:10:05,487 INFO [streaming_decode.py:380] Cuts processed until now is 3100.
2023-06-21 10:10:05,716 INFO [streaming_decode.py:380] Cuts processed until now is 3150.
2023-06-21 10:10:09,564 INFO [streaming_decode.py:380] Cuts processed until now is 3200.
2023-06-21 10:10:09,780 INFO [streaming_decode.py:380] Cuts processed until now is 3250.
2023-06-21 10:10:13,391 INFO [streaming_decode.py:380] Cuts processed until now is 3300.
2023-06-21 10:10:13,633 INFO [streaming_decode.py:380] Cuts processed until now is 3350.
2023-06-21 10:10:17,390 INFO [streaming_decode.py:380] Cuts processed until now is 3400.
2023-06-21 10:10:17,624 INFO [streaming_decode.py:380] Cuts processed until now is 3450.
2023-06-21 10:10:17,853 INFO [streaming_decode.py:380] Cuts processed until now is 3500.
2023-06-21 10:10:21,638 INFO [streaming_decode.py:380] Cuts processed until now is 3550.
2023-06-21 10:10:21,858 INFO [streaming_decode.py:380] Cuts processed until now is 3600.
2023-06-21 10:10:52,211 INFO [streaming_decode.py:425] The transcripts are stored in pruned_transducer_stateless7_streaming/exp/streaming/modified_beam_search/recogs-test-commonvoice-epoch-30-avg-9-streaming-chunk-size-32-use-averaged-model.txt
2023-06-21 10:10:52,342 INFO [utils.py:561] [test-commonvoice-num_active_paths_4] %WER 14.54% [19305 / 132787, 3354 ins, 7699 del, 8252 sub ]
2023-06-21 10:10:52,643 INFO [streaming_decode.py:436] Wrote detailed error stats to pruned_transducer_stateless7_streaming/exp/streaming/modified_beam_search/errs-test-commonvoice-epoch-30-avg-9-streaming-chunk-size-32-use-averaged-model.txt
2023-06-21 10:10:52,643 INFO [streaming_decode.py:450]
For test-commonvoice, WER of different settings are:
num_active_paths_4 14.54 best for test-commonvoice
2023-06-21 10:10:52,643 INFO [streaming_decode.py:618] Done!