add models

.gitattributes

@@ -33,3 +33,6 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+exp_large_v2/epoch-10-avg-6.pt filter=lfs diff=lfs merge=lfs -text
+exp_large_v3/epoch-5-avg-3.pt filter=lfs diff=lfs merge=lfs -text
+exp_medium/epoch-10-avg-4.pt filter=lfs diff=lfs merge=lfs -text

exp_large_v2/epoch-10-avg-6.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a5820bc5b288df5b0f48fcde2d0fedf954d84fcf9c275209268c3634b910aac
+size 6173671693

exp_large_v2/log/cer-summary-test-epoch-999-avg-1.txt

@@ -0,0 +1,2 @@
+settings	CER
+beam-search	2.47

exp_large_v2/log/cer-summary-valid-epoch-999-avg-1.txt

@@ -0,0 +1,2 @@
+settings	CER
+beam-search	2.3

exp_large_v2/log/log-decode-epoch-999-avg-1-2024-01-15-06-50-33

@@ -0,0 +1,23 @@
+2024-01-15 06:50:33,774 INFO [decode.py:423] Decoding started
+2024-01-15 06:50:33,776 INFO [decode.py:424] {'subsampling_factor': 4, 'feature_dim': 80, 'nhead': 4, 'attention_dim': 512, 'num_encoder_layers': 12, 'num_decoder_layers': 6, 'vgg_frontend': False, 'use_feat_batchnorm': True, 'search_beam': 20, 'output_beam': 7, 'min_active_states': 30, 'max_active_states': 10000, 'use_double_scores': True, 'env_info': {'k2-version': '1.24.3', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': '279b0c87015a615b81b147251814d737a548f397', 'k2-git-date': 'Wed May 24 22:24:09 2023', 'lhotse-version': '1.20.0.dev+git.0089643.clean', 'torch-version': '2.0.1+cu118', 'torch-cuda-available': True, 'torch-cuda-version': '11.8', 'python-version': '3.1', 'icefall-git-branch': None, 'icefall-git-sha1': None, 'icefall-git-date': None, 'icefall-path': '/workspace/icefall', 'k2-path': '/usr/local/lib/python3.10/dist-packages/k2/__init__.py', 'lhotse-path': '/usr/local/lib/python3.10/dist-packages/lhotse/__init__.py', 'hostname': '5767730', 'IP address': '0.88.2.50'}, 'epoch': 999, 'avg': 1, 'method': 'beam-search', 'beam_size': 10, 'exp_dir': PosixPath('whisper/exp_large_v2'), 'model_name': 'large-v2', 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 50, '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': False, 'suffix': 'epoch-999-avg-1', 'decoding_options': DecodingOptions(task='transcribe', language='zh', temperature=0.0, sample_len=None, best_of=None, beam_size=10, patience=None, length_penalty=None, prompt=None, prefix=None, suppress_tokens='-1', suppress_blank=True, without_timestamps=True, max_initial_timestamp=1.0, fp16=True), 'cleaner': <whisper.normalizers.basic.BasicTextNormalizer object at 0x7fd26e637c70>, 'normalizer': <tn.chinese.normalizer.Normalizer object at 0x7fd26e654310>}
+2024-01-15 06:50:33,776 INFO [decode.py:430] device: cuda
+2024-01-15 06:50:54,834 INFO [decode.py:470] Number of model parameters: 1541384960
+2024-01-15 06:50:54,834 INFO [asr_datamodule.py:380] About to get test cuts
+2024-01-15 06:50:54,843 INFO [asr_datamodule.py:345] About to create test dataset
+2024-01-15 06:50:59,471 INFO [decode.py:356] batch 0/?, cuts processed until now is 8
+2024-01-15 06:55:17,355 INFO [decode.py:356] batch 100/?, cuts processed until now is 873
+2024-01-15 06:59:30,069 INFO [decode.py:356] batch 200/?, cuts processed until now is 1792
+2024-01-15 07:03:47,097 INFO [decode.py:356] batch 300/?, cuts processed until now is 2729
+2024-01-15 07:08:04,988 INFO [decode.py:356] batch 400/?, cuts processed until now is 3639
+2024-01-15 07:12:19,440 INFO [decode.py:356] batch 500/?, cuts processed until now is 4492
+2024-01-15 07:16:36,995 INFO [decode.py:356] batch 600/?, cuts processed until now is 5407
+2024-01-15 07:20:56,794 INFO [decode.py:356] batch 700/?, cuts processed until now is 6300
+2024-01-15 07:25:16,990 INFO [decode.py:356] batch 800/?, cuts processed until now is 7105
+2024-01-15 07:25:44,885 INFO [decode.py:373] The transcripts are stored in whisper/exp_large_v2/recogs-test-beam-search-epoch-999-avg-1.txt
+2024-01-15 07:25:45,123 INFO [utils.py:564] [test-beam-search] %WER 2.47% [2585 / 104765, 68 ins, 93 del, 2424 sub ]
+2024-01-15 07:25:45,383 INFO [decode.py:389] Wrote detailed error stats to whisper/exp_large_v2/errs-test-beam-search-epoch-999-avg-1.txt
+2024-01-15 07:25:45,387 INFO [decode.py:403] 
+For test, CER of different settings are:
+beam-search	2.47	best for test
+
+2024-01-15 07:25:45,392 INFO [decode.py:491] Done!

exp_large_v2/log/log-decode-epoch-999-avg-1-2024-01-15-07-38-29

@@ -0,0 +1,32 @@
+2024-01-15 07:38:29,472 INFO [decode.py:423] Decoding started
+2024-01-15 07:38:29,473 INFO [decode.py:424] {'subsampling_factor': 4, 'feature_dim': 80, 'nhead': 4, 'attention_dim': 512, 'num_encoder_layers': 12, 'num_decoder_layers': 6, 'vgg_frontend': False, 'use_feat_batchnorm': True, 'search_beam': 20, 'output_beam': 7, 'min_active_states': 30, 'max_active_states': 10000, 'use_double_scores': True, 'env_info': {'k2-version': '1.24.3', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': '279b0c87015a615b81b147251814d737a548f397', 'k2-git-date': 'Wed May 24 22:24:09 2023', 'lhotse-version': '1.20.0.dev+git.0089643.clean', 'torch-version': '2.0.1+cu118', 'torch-cuda-available': True, 'torch-cuda-version': '11.8', 'python-version': '3.1', 'icefall-git-branch': None, 'icefall-git-sha1': None, 'icefall-git-date': None, 'icefall-path': '/workspace/icefall', 'k2-path': '/usr/local/lib/python3.10/dist-packages/k2/__init__.py', 'lhotse-path': '/usr/local/lib/python3.10/dist-packages/lhotse/__init__.py', 'hostname': '5767730', 'IP address': '0.88.2.50'}, 'epoch': 999, 'avg': 1, 'method': 'beam-search', 'beam_size': 10, 'exp_dir': PosixPath('whisper/exp_large_v2'), 'model_name': 'large-v2', 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 50, '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': False, 'suffix': 'epoch-999-avg-1', 'decoding_options': DecodingOptions(task='transcribe', language='zh', temperature=0.0, sample_len=None, best_of=None, beam_size=10, patience=None, length_penalty=None, prompt=None, prefix=None, suppress_tokens='-1', suppress_blank=True, without_timestamps=True, max_initial_timestamp=1.0, fp16=True), 'cleaner': <whisper.normalizers.basic.BasicTextNormalizer object at 0x7f1413f64610>, 'normalizer': <tn.chinese.normalizer.Normalizer object at 0x7f1413f644c0>}
+2024-01-15 07:38:29,474 INFO [decode.py:430] device: cuda
+2024-01-15 07:38:48,476 INFO [decode.py:468] Number of model parameters: 1541384960
+2024-01-15 07:38:48,477 INFO [asr_datamodule.py:380] About to get test cuts
+2024-01-15 07:38:48,479 INFO [asr_datamodule.py:345] About to create test dataset
+2024-01-15 07:38:48,743 INFO [asr_datamodule.py:375] About to get dev cuts
+2024-01-15 07:38:48,751 INFO [asr_datamodule.py:314] About to create dev dataset
+2024-01-15 07:38:49,112 INFO [asr_datamodule.py:333] About to create dev dataloader
+2024-01-15 07:38:53,617 INFO [decode.py:356] batch 0/?, cuts processed until now is 9
+2024-01-15 07:43:25,975 INFO [decode.py:356] batch 100/?, cuts processed until now is 967
+2024-01-15 07:47:55,818 INFO [decode.py:356] batch 200/?, cuts processed until now is 1977
+2024-01-15 07:52:23,520 INFO [decode.py:356] batch 300/?, cuts processed until now is 3009
+2024-01-15 07:56:54,941 INFO [decode.py:356] batch 400/?, cuts processed until now is 4013
+2024-01-15 08:01:28,296 INFO [decode.py:356] batch 500/?, cuts processed until now is 4971
+2024-01-15 08:05:58,501 INFO [decode.py:356] batch 600/?, cuts processed until now is 5984
+2024-01-15 08:10:26,461 INFO [decode.py:356] batch 700/?, cuts processed until now is 7015
+2024-01-15 08:15:02,342 INFO [decode.py:356] batch 800/?, cuts processed until now is 7989
+2024-01-15 08:19:37,343 INFO [decode.py:356] batch 900/?, cuts processed until now is 9002
+2024-01-15 08:24:08,775 INFO [decode.py:356] batch 1000/?, cuts processed until now is 9990
+2024-01-15 08:28:37,200 INFO [decode.py:356] batch 1100/?, cuts processed until now is 10963
+2024-01-15 08:33:06,093 INFO [decode.py:356] batch 1200/?, cuts processed until now is 11989
+2024-01-15 08:37:33,633 INFO [decode.py:356] batch 1300/?, cuts processed until now is 12955
+2024-01-15 08:42:08,035 INFO [decode.py:356] batch 1400/?, cuts processed until now is 13900
+2024-01-15 08:44:45,978 INFO [decode.py:373] The transcripts are stored in whisper/exp_large_v2/recogs-valid-beam-search-epoch-999-avg-1.txt
+2024-01-15 08:44:46,394 INFO [utils.py:564] [valid-beam-search] %WER 2.30% [4714 / 205341, 186 ins, 152 del, 4376 sub ]
+2024-01-15 08:44:46,915 INFO [decode.py:389] Wrote detailed error stats to whisper/exp_large_v2/errs-valid-beam-search-epoch-999-avg-1.txt
+2024-01-15 08:44:46,919 INFO [decode.py:403] 
+For valid, CER of different settings are:
+beam-search	2.3	best for valid
+
+2024-01-15 08:44:46,933 INFO [decode.py:490] Done!

exp_large_v2/tensorboard/events.out.tfevents.1705052086.5753425.551322.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:67d021d40e6ef662ec0c09bb8f2a088bab30f5c84eaf89d4d5206bb6ab8b8074
+size 408

exp_large_v2/tensorboard/events.out.tfevents.1705053323.5753425.631090.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6079c1bcc1019b466052f83ae5562605da010d2cfd412b6aa930c243c773a32d
+size 408

exp_large_v2/tensorboard/events.out.tfevents.1705053424.5753425.637962.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e80eba0e1fd54f692f6dffe79c0961df5bc02efb363bf37167a38cf9c3e07ba
+size 50182

exp_large_v3/epoch-5-avg-3.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:457f63500291d6dd86ddb9d24be0f1ed224c025b7b9e565aaca442c8afbb07d6
+size 6174412832

exp_large_v3/log/cer-summary-test-epoch-5-avg-3.txt

@@ -0,0 +1,2 @@
+settings	CER
+beam-search	2.84

exp_large_v3/log/log-decode-epoch-5-avg-3-2024-01-15-05-54-57

@@ -0,0 +1,20 @@
+2024-01-15 05:54:57,924 INFO [decode.py:373] Decoding started
+2024-01-15 05:54:57,926 INFO [decode.py:374] {'subsampling_factor': 4, 'feature_dim': 80, 'nhead': 4, 'attention_dim': 512, 'num_encoder_layers': 12, 'num_decoder_layers': 6, 'vgg_frontend': False, 'use_feat_batchnorm': True, 'search_beam': 20, 'output_beam': 7, 'min_active_states': 30, 'max_active_states': 10000, 'use_double_scores': True, 'env_info': {'k2-version': '1.24.3', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': '279b0c87015a615b81b147251814d737a548f397', 'k2-git-date': 'Wed May 24 22:24:09 2023', 'lhotse-version': '1.20.0.dev+git.0089643.clean', 'torch-version': '2.0.1+cu118', 'torch-cuda-available': True, 'torch-cuda-version': '11.8', 'python-version': '3.1', 'icefall-git-branch': None, 'icefall-git-sha1': None, 'icefall-git-date': None, 'icefall-path': '/workspace/icefall', 'k2-path': '/usr/local/lib/python3.10/dist-packages/k2/__init__.py', 'lhotse-path': '/usr/local/lib/python3.10/dist-packages/lhotse/__init__.py', 'hostname': '5767730', 'IP address': '0.88.2.50'}, 'epoch': 5, 'avg': 3, 'method': 'beam-search', 'beam_size': 10, 'exp_dir': PosixPath('whisper/exp_large_v3_rerun'), 'model_name': 'large-v3', 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 50, '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': False, 'suffix': 'epoch-5-avg-3', 'decoding_options': DecodingOptions(task='transcribe', language='zh', temperature=0.0, sample_len=None, best_of=None, beam_size=10, patience=None, length_penalty=None, prompt=None, prefix=None, suppress_tokens='-1', suppress_blank=True, without_timestamps=True, max_initial_timestamp=1.0, fp16=True), 'cleaner': <whisper.normalizers.basic.BasicTextNormalizer object at 0x7f7124bfe8c0>, 'normalizer': <tn.chinese.normalizer.Normalizer object at 0x7f7124bfe8f0>}
+2024-01-15 05:54:57,926 INFO [decode.py:380] device: cuda
+2024-01-15 05:55:27,262 INFO [decode.py:416] Number of model parameters: 1541570560
+2024-01-15 05:55:27,262 INFO [asr_datamodule.py:380] About to get test cuts
+2024-01-15 05:55:27,263 INFO [asr_datamodule.py:345] About to create test dataset
+2024-01-15 05:55:30,840 INFO [decode.py:306] batch 0/?, cuts processed until now is 7
+2024-01-15 05:59:45,032 INFO [decode.py:306] batch 100/?, cuts processed until now is 869
+2024-01-15 06:03:55,695 INFO [decode.py:306] batch 200/?, cuts processed until now is 1781
+2024-01-15 06:08:06,556 INFO [decode.py:306] batch 300/?, cuts processed until now is 2716
+2024-01-15 06:12:21,508 INFO [decode.py:306] batch 400/?, cuts processed until now is 3623
+2024-01-15 06:16:34,661 INFO [decode.py:306] batch 500/?, cuts processed until now is 4480
+2024-01-15 06:19:28,079 INFO [decode.py:323] The transcripts are stored in whisper/exp_large_v3_rerun/recogs-test-beam-search-epoch-5-avg-3.txt
+2024-01-15 06:19:28,189 INFO [utils.py:564] [test-beam-search] %WER 2.84% [2077 / 73194, 51 ins, 73 del, 1953 sub ]
+2024-01-15 06:19:28,374 INFO [decode.py:339] Wrote detailed error stats to whisper/exp_large_v3_rerun/errs-test-beam-search-epoch-5-avg-3.txt
+2024-01-15 06:19:28,378 INFO [decode.py:353] 
+For test, CER of different settings are:
+beam-search	2.84	best for test
+
+2024-01-15 06:19:28,381 INFO [decode.py:437] Done!

exp_large_v3/tensorboard/events.out.tfevents.1705288270.5765802.289785.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4732578885e0a0479b03ec4a1c54e069d4ab2ee8411e816cd0f9c7082c6e461b
+size 189

exp_large_v3/tensorboard/events.out.tfevents.1705288650.5765802.291955.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6419e624adc564c2aa2bdc18030169556dcb7a265016c6acaeb94ee583ecbfb9
+size 19552

exp_medium/epoch-10-avg-4.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e84887c07117b714d69b7f73ad0551f0b7cc427206dfd568c33a96200fc136c
+size 3055780238

exp_medium/log/cer-summary-test-epoch-10-avg-4.txt

@@ -0,0 +1,2 @@
+settings	CER
+beam-search	3.27

exp_medium/log/log-decode-epoch-10-avg-4-2024-01-11-09-25-43

@@ -0,0 +1,20 @@
+2024-01-11 09:25:44,243 INFO [decode.py:368] Decoding started
+2024-01-11 09:25:44,245 INFO [decode.py:369] {'subsampling_factor': 4, 'feature_dim': 80, 'nhead': 4, 'attention_dim': 512, 'num_encoder_layers': 12, 'num_decoder_layers': 6, 'vgg_frontend': False, 'use_feat_batchnorm': True, 'search_beam': 20, 'output_beam': 7, 'min_active_states': 30, 'max_active_states': 10000, 'use_double_scores': True, 'env_info': {'k2-version': '1.24.3', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': '279b0c87015a615b81b147251814d737a548f397', 'k2-git-date': 'Wed May 24 22:24:09 2023', 'lhotse-version': '1.18.0', 'torch-version': '2.0.1+cu118', 'torch-cuda-available': True, 'torch-cuda-version': '11.8', 'python-version': '3.1', 'icefall-git-branch': None, 'icefall-git-sha1': None, 'icefall-git-date': None, 'icefall-path': '/workspace/icefall', 'k2-path': '/usr/local/lib/python3.10/dist-packages/k2/__init__.py', 'lhotse-path': '/usr/local/lib/python3.10/dist-packages/lhotse/__init__.py', 'hostname': '5724415', 'IP address': '0.87.88.255'}, 'epoch': 10, 'avg': 4, 'method': 'beam-search-beam10', 'exp_dir': PosixPath('whisper/exp_medium'), 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 100, '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, 'suffix': 'epoch-10-avg-4', 'decoding_options': DecodingOptions(task='transcribe', language='zh', temperature=0.0, sample_len=None, best_of=None, beam_size=10, patience=None, length_penalty=None, prompt=None, prefix=None, suppress_tokens='-1', suppress_blank=True, without_timestamps=True, max_initial_timestamp=1.0, fp16=True), 'cleaner': <whisper.normalizers.basic.BasicTextNormalizer object at 0x7f715c82ff40>, 'normalizer': <tn.chinese.normalizer.Normalizer object at 0x7f715c82feb0>}
+2024-01-11 09:25:44,245 INFO [decode.py:375] device: cuda
+2024-01-11 09:25:52,381 INFO [decode.py:384] Calculating the averaged model over epoch range from 6 (excluded) to 10
+2024-01-11 09:26:13,996 INFO [decode.py:401] Number of model parameters: 762321920
+2024-01-11 09:26:13,996 INFO [asr_datamodule.py:376] About to get test cuts
+2024-01-11 09:26:13,998 INFO [asr_datamodule.py:341] About to create test dataset
+2024-01-11 09:26:28,119 INFO [decode.py:300] batch 0/?, cuts processed until now is 16
+2024-01-11 09:42:50,889 INFO [decode.py:300] batch 100/?, cuts processed until now is 1774
+2024-01-11 09:59:25,591 INFO [decode.py:300] batch 200/?, cuts processed until now is 3634
+2024-01-11 10:16:14,174 INFO [decode.py:300] batch 300/?, cuts processed until now is 5509
+2024-01-11 10:31:55,967 INFO [decode.py:300] batch 400/?, cuts processed until now is 7130
+2024-01-11 10:32:20,787 INFO [decode.py:317] The transcripts are stored in whisper/exp_medium/recogs-test-beam-search-epoch-10-avg-4.txt
+2024-01-11 10:32:20,943 INFO [utils.py:564] [test-beam-search] %WER 3.27% [3422 / 104765, 75 ins, 144 del, 3203 sub ]
+2024-01-11 10:32:21,206 INFO [decode.py:333] Wrote detailed error stats to whisper/exp_medium/errs-test-beam-search-epoch-10-avg-4.txt
+2024-01-11 10:32:21,210 INFO [decode.py:347] 
+For test, CER of different settings are:
+beam-search	3.27	best for test
+
+2024-01-11 10:32:21,215 INFO [decode.py:422] Done!

exp_medium/tensorboard/events.out.tfevents.1704956878.5730418.703.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3f71d4985192eada735f37dd818f62a748d12b48afb0a70c4ab96dd7bca8dbf2
+size 103281