Commit Graph

  • d0a53aad48 Fix tensorboard log location Daniel Povey 2022-04-12 11:51:15 +08:00
  • 65818d16de Add more results Daniel Povey 2022-04-12 11:48:16 +08:00
  • 61c34a7465 Merge branch 'master' of https://github.com/luomingshuang/icefall luomingshuang 2022-04-12 10:28:02 +08:00
  • bdeff338c2
    Fix CI errors. (#310) Fangjun Kuang 2022-04-12 09:09:56 +08:00
  • 73c0f2aacd Fix CI errors. Fangjun Kuang 2022-04-12 08:50:35 +08:00
  • 6a425ed793 Keep the original tolerance Guanbo Wang 2022-04-11 20:12:00 -04:00
  • ba245aa60f Use default storage_type Guanbo Wang 2022-04-11 20:10:50 -04:00
  • 407998267c Remove lazy_load option Guanbo Wang 2022-04-11 18:12:54 -04:00
  • 36ec10c41a Merge remote-tracking branch 'upstream/master' into gigaspeech_recipe Guanbo Wang 2022-04-11 17:49:01 -04:00
  • 22f011e5ab Update results Guanbo Wang 2022-04-11 21:46:40 +00:00
  • f485b66d54 Decode with post-processing Guanbo Wang 2022-04-11 21:46:15 +00:00
  • 6d07cf9245 Decrease num_paths while CUDA OOM Guanbo Wang 2022-04-11 21:45:29 +00:00
  • 118e195004
    Update results for tedlium3 pruned RNN-T (#307) Mingshuang Luo 2022-04-11 22:19:26 +08:00
  • 93c60a9d30
    Code style check for librispeech pruned transducer stateless2 (#308) Mingshuang Luo 2022-04-11 22:15:18 +08:00
  • ef0b6df8f8
    Update train.py Mingshuang Luo 2022-04-11 21:19:45 +08:00
  • 1b854e5c44
    Update beam_search.py Mingshuang Luo 2022-04-11 21:14:02 +08:00
  • 8dcf9644cb
    Update train.py Mingshuang Luo 2022-04-11 21:10:17 +08:00
  • 7320e0ea89
    Update model.py Mingshuang Luo 2022-04-11 21:09:12 +08:00
  • ead822477c Fix rebase Daniel Povey 2022-04-11 21:01:13 +08:00
  • e8eb0b94d9 Updating RESULTS.md; fix in beam_search.py Daniel Povey 2022-04-11 20:56:11 +08:00
  • a92133ef96 Minor fixes pkufool 2022-04-11 15:41:45 +08:00
  • ddd8f9e15e Minor fixes pkufool 2022-04-11 15:40:14 +08:00
  • cc0d4ffa4f Add mix precision support pkufool 2022-04-11 15:27:24 +08:00
  • 92e90bd055
    Merge branch 'master' into code-style-check-for-librispeech-pruned-transducer-stateless2 Mingshuang Luo 2022-04-11 20:53:21 +08:00
  • 16c6e0207b check some files luomingshuang 2022-04-11 20:41:32 +08:00
  • fecceee216 check some files luomingshuang 2022-04-11 20:38:10 +08:00
  • 187d59d59b check some files luomingshuang 2022-04-11 20:37:19 +08:00
  • 05fd40ba68 check some files luomingshuang 2022-04-11 20:24:46 +08:00
  • 9b69f901ec
    Update RESULTS.md Mingshuang Luo 2022-04-11 19:39:48 +08:00
  • 016a40d384
    Update README.md Mingshuang Luo 2022-04-11 19:39:17 +08:00
  • aee80b2828 update for results luomingshuang 2022-04-11 19:35:39 +08:00
  • 8cb727e24a
    Tedlium3 pruned transducer stateless (#261) Mingshuang Luo 2022-04-11 17:08:53 +08:00
  • 7012fd65b5
    Support mix precision training on the reworked model (#305) Wei Kang 2022-04-11 16:49:54 +08:00
  • cd6a2d903b do some changes for pruned RNN-T luomingshuang 2022-04-11 16:25:12 +08:00
  • d95750278b Minor fixes pkufool 2022-04-11 15:56:45 +08:00
  • 04fc4e4f4a Minor fixes pkufool 2022-04-11 15:41:45 +08:00
  • 22474e9abe Minor fixes pkufool 2022-04-11 15:40:14 +08:00
  • 4ebe821769 Add mix precision support pkufool 2022-04-11 15:27:24 +08:00
  • b7e7629168
    Merge branch 'k2-fsa:master' into tedlium3-pruned-transducer-stateless-new Mingshuang Luo 2022-04-11 15:18:32 +08:00
  • 34aad74a2c
    Merge pull request #303 from danpovey/fix_docs Daniel Povey 2022-04-11 15:14:06 +08:00
  • 03c7c2613d Fix docs in optim.py Daniel Povey 2022-04-11 15:13:42 +08:00
  • 6eb6d9b4cd
    Merge pull request #288 from danpovey/reworked_model Daniel Povey 2022-04-11 15:03:08 +08:00
  • 5078332088 Fix adding learning rate to tensorboard Daniel Povey 2022-04-11 14:58:15 +08:00
  • 651745b220 minor fix doc in emformer.py yaozengwei 2022-04-11 12:42:47 +08:00
  • 6c1f9b5181 Add wenet ref in ConvolutionModule class. yaozengwei 2022-04-11 12:38:45 +08:00
  • d5f9d49e53 Modify beam search to be efficient with current joienr Daniel Povey 2022-04-11 12:35:29 +08:00
  • 1d74c5e596 Support causal convolution in emformer encoder layer. yaozengwei 2022-04-11 12:28:15 +08:00
  • 46d52dda10 Fix dir names Daniel Povey 2022-04-11 12:03:41 +08:00
  • 6f8a9e97ec
    Merge branch 'k2-fsa:master' into tedlium3-pruned-transducer-stateless-new Mingshuang Luo 2022-04-11 10:07:06 +08:00
  • f721a2fd7a
    Minor fixes for logging (#296) Wei Kang 2022-04-10 23:34:18 +08:00
  • 08473a17aa
    Modify init (#301) Zengwei Yao 2022-04-10 23:29:28 +08:00
  • 85db12caac Merge remote-tracking branch 'dan/streaming' into streaming-1 Fangjun Kuang 2022-04-10 23:27:24 +08:00
  • 7f73043219 Merge remote-tracking branch 'dan/master' into streaming Fangjun Kuang 2022-04-10 23:25:56 +08:00
  • 26cfb7644d Merge remote-tracking branch 'dan/master' into streaming-1 Fangjun Kuang 2022-04-10 23:17:45 +08:00
  • f16b759397 Add streaming feature extractor. Fangjun Kuang 2022-04-10 23:07:41 +08:00
  • a24eef8096 update conv_emformer_transducer/emformer.py. yaozengwei 2022-04-10 20:29:22 +08:00
  • 8129470586 first upload the conv_emformer_transducer recipe, integrating convolution module into emformer layers. yaozengwei 2022-04-10 20:24:20 +08:00
  • 962cf868c9 Fix import Daniel Povey 2022-04-10 15:31:46 +08:00
  • d1e4ae788d Refactor how learning rate is set. Daniel Povey 2022-04-10 15:25:27 +08:00
  • 82d58629ea Implement 2p version of learning rate schedule. Daniel Povey 2022-04-10 13:50:31 +08:00
  • da50525ca5 Make lrate rule more symmetric Daniel Povey 2022-04-10 13:25:40 +08:00
  • 4d41ee0caa Implement 2o schedule Daniel Povey 2022-04-09 18:37:03 +08:00
  • db72aee1f0 Set 2n rule.. Daniel Povey 2022-04-09 18:15:56 +08:00
  • 3e131891a2 update test_emformer.py yaozengwei 2022-04-08 20:43:54 +08:00
  • d58002c414 update emformer.py yaozengwei 2022-04-08 20:31:32 +08:00
  • 37f4f65ca3 do a change luomingshuang 2022-04-08 19:12:30 +08:00
  • d8907769ee add and update some files luomingshuang 2022-04-08 19:00:19 +08:00
  • 5242275b8a
    Merge branch 'k2-fsa:master' into wenetspeech-pruned-transducer-stateless-pinyin Mingshuang Luo 2022-04-08 17:25:03 +08:00
  • 549826bff8 exclude black check for icefall/__init__.py in pyproject.toml. yaozengwei 2022-04-08 16:59:21 +08:00
  • 0f8ee68af2 Fix bug Daniel Povey 2022-04-08 16:53:42 +08:00
  • f587cd527d Change exponential part of lrate to be epoch based Daniel Povey 2022-04-08 16:24:21 +08:00
  • 6ee32cf7af Set new scheduler Daniel Povey 2022-04-08 16:10:06 +08:00
  • 2c76ea7f9e make imports style consistent. yaozengwei 2022-04-08 14:43:03 +08:00
  • a1c57ce8b3 Merge remote-tracking branch 'k2-fsa/master' into modify_init yaozengwei 2022-04-08 14:34:53 +08:00
  • 78b8792d1d
    Fix potential bugs in PyTorch that exist in label_smoothing. (#300) Fangjun Kuang 2022-04-08 13:41:33 +08:00
  • 4d1a337ea1 update icefall/__init__.py yaozengwei 2022-04-08 12:11:38 +08:00
  • 6092b20b6a Fix potential bugs in PyTorch that exist in label_smoothing. Fangjun Kuang 2022-04-08 12:00:11 +08:00
  • 7c0070e6f6
    Display torch version in the training log. (#299) Fangjun Kuang 2022-04-08 11:39:54 +08:00
  • f56e54c1c1 Display torch version in the training log. Fangjun Kuang 2022-04-08 11:11:06 +08:00
  • 2d1b90f758 update the docs of Emformer class in emformer.py yaozengwei 2022-04-08 10:59:39 +08:00
  • 374eacdd5c First upload emformer_pruned_transducer_stateless recipe, refator emformer codes from torchaudio. yaozengwei 2022-04-07 21:32:59 +08:00
  • b7ecb3b620 Minor fix pkufool 2022-04-07 13:03:10 +08:00
  • 36726ca188 Minor fixes for logging pkufool 2022-04-07 13:01:02 +08:00
  • 3d2c261684 black Guanbo Wang 2022-04-06 21:05:42 -04:00
  • 3e5436e5d4
    Merge branch 'k2-fsa:master' into gigaspeech_recipe Wang, Guanbo 2022-04-06 20:57:51 -04:00
  • e56d32728d Update RESULTS Guanbo Wang 2022-04-06 20:53:47 -04:00
  • f857d5a9ea Added README Guanbo Wang 2022-04-06 20:30:09 -04:00
  • a4e1471d1d Decrease num_paths while CUDA OOM Guanbo Wang 2022-04-06 19:47:44 -04:00
  • d9addb7c43 update params in decode.py Guanbo Wang 2022-04-06 19:41:24 -04:00
  • 9a5340b21f Use bpe_500 Guanbo Wang 2022-04-06 19:06:54 -04:00
  • 64bb39bf17 Update params in train.py Guanbo Wang 2022-04-06 19:06:22 -04:00
  • 79211633ed Use BucketingSampler for valid and test dataloader Guanbo Wang 2022-04-06 19:05:45 -04:00
  • 3ddcc7939b Delete compute_fbank_gigaspeech.py Guanbo Wang 2022-04-06 18:57:53 -04:00
  • 61486a0f76 Remove initial_speed Daniel Povey 2022-04-06 13:17:26 +08:00
  • a41e93437c Change some defaults in LR-setting rule. Daniel Povey 2022-04-06 12:36:58 +08:00
  • ceeb95bcb8
    update icefall/__init__.py to import more common functions. (#294) Zengwei Yao 2022-04-06 11:55:29 +08:00
  • 8063577760 update icefall/__init__.py to import more common functions. yaozengwei 2022-04-06 11:30:26 +08:00
  • d5437ba623 Export stateless pruned transducer to jit script module pkufool 2022-04-06 10:31:28 +08:00
  • 9423b3899f Update emformer_pruned_transducer_stateless/emformer.py and upload emformer_pruned_transducer_stateless/test_emformer.py. yaozengwei 2022-04-04 22:16:46 +08:00
  • fe43c1349e First upload emformer_pruned_transducer_stateless/emformer.py, modified from torchaudio. yaozengwei 2022-04-02 21:14:24 +08:00