Commit Graph

  • 5aafbb970e
    SPGISpeech recipe (#334) Desh Raj 2022-05-16 08:52:14 -04:00
  • 2f2934a115 fix train.py Daniel Povey 2022-05-16 19:41:16 +08:00
  • 59478b1ef3 Update tests. Fangjun Kuang 2022-05-16 19:22:21 +08:00
  • d61c8aa3bc Hopefully this finishes the full orthogonalization. Daniel Povey 2022-05-16 19:18:12 +08:00
  • 896993714b Add tests to ensure that the model is torch scriptable. Fangjun Kuang 2022-05-16 19:12:34 +08:00
  • 96a544fb69 Various fixes to support torch script. Fangjun Kuang 2022-05-16 18:53:56 +08:00
  • 67f916e599 Draft towards 2nd orthogonalization Daniel Povey 2022-05-16 16:16:12 +08:00
  • 8aeaf1421a Rename orthogonalize to diagonalize Daniel Povey 2022-05-16 12:43:47 +08:00
  • 9859e33c06 Remove optimizer.reset() which is not supported for Eve Daniel Povey 2022-05-16 12:13:36 +08:00
  • d58debdb3f Use Eve with orthogonalization Daniel Povey 2022-05-16 10:34:25 +08:00
  • 995371ad95 Move train.py changes to the right dir Daniel Povey 2022-05-15 22:17:08 +08:00
  • cee5396058 Orthogonalize every 2k iters Daniel Povey 2022-05-15 21:50:40 +08:00
  • f94ad976c6 fix decode.py yaozengwei 2022-05-15 16:25:34 +08:00
  • 1c80a5b29a minor fix of doc yaozengwei 2022-05-15 14:20:17 +08:00
  • 6aded4d6df add remaining files yaozengwei 2022-05-15 12:12:25 +08:00
  • fe0719d8ba minor fix yaozengwei 2022-05-15 12:11:05 +08:00
  • 648a0b37d5 add Emformer module yaozengwei 2022-05-14 23:18:16 +08:00
  • 943cb9d5a3 add EmformerEncoderLayer module yaozengwei 2022-05-14 13:10:44 +08:00
  • 3838b84313 add conv module yaozengwei 2022-05-13 22:23:39 +08:00
  • 5dc5f8305a add emformer attention module yaozengwei 2022-05-13 17:07:40 +08:00
  • a0c7095e42 add relative position encoding yaozengwei 2022-05-12 21:27:32 +08:00
  • a9dccdc33f
    Streaming merge (#369) Zengwei Yao 2022-05-15 21:08:30 +08:00
  • 3e82387717 minor fix yaozengwei 2022-05-15 21:00:55 +08:00
  • 58d476c134 minor fix yaozengwei 2022-05-15 20:56:17 +08:00
  • 8f23d9eb41 modify .flake8 yaozengwei 2022-05-15 20:53:25 +08:00
  • d83daf750e Merge branch 'master' into streaming_merge yaozengwei 2022-05-15 20:26:16 +08:00
  • ed30271715 add CER numbers Desh Raj 2022-05-15 08:08:49 -04:00
  • c9d84aeb5c Merge remote-tracking branch 'k2-fsa/master' yaozengwei 2022-05-15 18:02:27 +08:00
  • 2d8b07ef3a delete two dir yaozengwei 2022-05-15 17:10:47 +08:00
  • 9828b8f628 fix decode.py yaozengwei 2022-05-15 16:25:34 +08:00
  • bb32556f9e Add and test reset() function Daniel Povey 2022-05-15 16:20:10 +08:00
  • a1dc020270 train scalars slower Daniel Povey 2022-05-15 15:44:56 +08:00
  • 0989aec741 Replace Eve optimizer with Abel. Daniel Povey 2022-05-15 14:33:08 +08:00
  • 6f7860a0a6
    Fix GitHub CI for decoding GigaSpeech dev/test datasets (#366) Fangjun Kuang 2022-05-15 14:25:35 +08:00
  • 5e3bf4ce5a Remove some debug code Daniel Povey 2022-05-15 14:22:12 +08:00
  • 27558858a3 minor fix of doc yaozengwei 2022-05-15 14:20:17 +08:00
  • 6306f24430 Some changes to algorithm; more diagnostics printing Daniel Povey 2022-05-15 14:12:23 +08:00
  • 67c402a369 Add some debugging/diagnostic code Daniel Povey 2022-05-15 13:28:00 +08:00
  • 6670cac531 Fix GitHub CI for decoding GigaSpeech dev/test datasets Fangjun Kuang 2022-05-15 13:21:18 +08:00
  • 9630f9a3ba
    Update GigaSpeech reults (#364) Guanbo Wang 2022-05-15 00:57:40 -04:00
  • a0e7a6d745 Merge remote-tracking branch 'upstream/master' into gigaspeech_update_results Guanbo Wang 2022-05-15 00:39:57 -04:00
  • 44bf91437b Update README.md Guanbo Wang 2022-05-15 00:39:39 -04:00
  • 2142fca8f8 add remaining files yaozengwei 2022-05-15 12:12:25 +08:00
  • 20a23d13ce minor fix yaozengwei 2022-05-15 12:11:05 +08:00
  • 747960677e Prob. 1st working version of Abel Daniel Povey 2022-05-15 10:13:06 +08:00
  • 3b5acd16e9 Update results Guanbo Wang 2022-05-14 20:27:32 -04:00
  • e8f9c39382 Update export.py Guanbo Wang 2022-05-14 19:56:52 -04:00
  • 8b60d43ead add Emformer module yaozengwei 2022-05-14 23:18:16 +08:00
  • 4fc1638959 pre commit hook Desh Raj 2022-05-14 10:41:06 -04:00
  • f23dd43719
    Update results for libri+giga multi dataset setup. (#363) Fangjun Kuang 2022-05-14 21:45:39 +08:00
  • ef98646ed0 More fixes. Fangjun Kuang 2022-05-14 21:32:20 +08:00
  • 9ffc77a0f2 Use an n-gram LM to rescore the lattice from fast_beam_search. Fangjun Kuang 2022-05-14 20:54:04 +08:00
  • b265a5c875 add EmformerEncoderLayer module yaozengwei 2022-05-14 13:10:44 +08:00
  • 1f7de1c123 Typo fixes. Fangjun Kuang 2022-05-14 11:31:08 +08:00
  • a006d6494f Fix CI. Fangjun Kuang 2022-05-14 09:05:11 +08:00
  • ad3fb63ad6 Merge remote-tracking branch 'dan/master' into update-giga-libri-results Fangjun Kuang 2022-05-14 08:58:11 +08:00
  • 2d7096dfc6
    Decode gigaspeech in GitHub actions (#362) Fangjun Kuang 2022-05-14 08:53:22 +08:00
  • 633741d3b1 Update decode.py Guanbo Wang 2022-05-13 19:02:23 -04:00
  • 02b4b469a2 remove change in librispeech Desh Raj 2022-05-13 14:03:38 -04:00
  • a2fb1859db pre commit hooks Desh Raj 2022-05-13 13:58:20 -04:00
  • 2ce48a2c21 Update decode.py and train.py to use periodically averaged models. Fangjun Kuang 2022-05-13 23:22:30 +08:00
  • b44b3a77f4 Minor fixes. Fangjun Kuang 2022-05-13 22:52:44 +08:00
  • a1c6bae5d6 Init pruned_transducer_stateless4b as copy of 4 Daniel Povey 2022-05-13 22:48:17 +08:00
  • 993b36e0b8 Merge remote-tracking branch 'upstream/master' into knowledge_base_1b_L2_ng Daniel Povey 2022-05-13 22:47:20 +08:00
  • 56974a900d Update results for libri+giga multi dataset setup. Fangjun Kuang 2022-05-13 22:28:27 +08:00
  • 3360dc5afc add conv module yaozengwei 2022-05-13 22:23:39 +08:00
  • e0536c9aee Merge branch 'master' of https://github.com/k2-fsa/icefall into spgi Desh Raj 2022-05-13 10:23:04 -04:00
  • 63dcc1d3f4 remove duplicate files Desh Raj 2022-05-13 10:15:02 -04:00
  • d4a8648a0c remove unused scripts and soft link common scripts Desh Raj 2022-05-13 10:11:37 -04:00
  • 2381ba544d add pretrained model to HF Desh Raj 2022-05-13 10:00:43 -04:00
  • 7b786ce0b9 Remove random combiner. Fangjun Kuang 2022-05-13 18:53:22 +08:00
  • 14d91c4645 Minor fixes. Fangjun Kuang 2022-05-13 18:22:45 +08:00
  • be5a17b47c Typo fixes. Fangjun Kuang 2022-05-13 18:14:18 +08:00
  • fc5c2f04b1 Add CI for pruned_transducer_stateless5 Fangjun Kuang 2022-05-13 18:10:55 +08:00
  • a613e85900 Update results. Fangjun Kuang 2022-05-13 18:04:03 +08:00
  • 2cfb2f58f0 add emformer attention module yaozengwei 2022-05-13 17:07:40 +08:00
  • ba87eb4461 minor fixes Fangjun Kuang 2022-05-13 15:25:11 +08:00
  • ecbcb25532 Minor fixes. Fangjun Kuang 2022-05-13 15:09:29 +08:00
  • fdedae2b53 Minor fixes. Fangjun Kuang 2022-05-13 14:15:22 +08:00
  • a13dcff870 minor fixes Fangjun Kuang 2022-05-13 13:56:53 +08:00
  • 0f180b3ce2
    Validate that there are no OOV tokens in BPE-based lexicons. (#359) Fangjun Kuang 2022-05-13 14:00:35 +08:00
  • e780dccb00 Typo fixes. Fangjun Kuang 2022-05-13 13:58:05 +08:00
  • cce3a7c280 Begin to add CI for gigaspeech. Fangjun Kuang 2022-05-13 13:40:40 +08:00
  • e30e042c39
    Update decoding script for gigaspeech and remove duplicate files. (#361) Fangjun Kuang 2022-05-13 13:03:16 +08:00
  • ba99671fba Update decoding script for gigaspeech and remove duplicate files. Fangjun Kuang 2022-05-13 12:55:06 +08:00
  • 44f4aa5f66 Try to resolve merge issues etc Daniel Povey 2022-05-13 11:32:23 +08:00
  • 4f933f5413 Merge changes from knowledge_base_1bfast; fix nheads 4->8 Daniel Povey 2022-05-13 11:26:18 +08:00
  • 48a6a9a549
    GigaSpeech RNN-T experiments (#318) Guanbo Wang 2022-05-12 23:03:26 -04:00
  • 7b7acdf369
    Support --iter in export.py (#360) Fangjun Kuang 2022-05-13 10:51:44 +08:00
  • 197a3be9b8 typo Guanbo Wang 2022-05-12 22:47:13 -04:00
  • 9f0d5ccd5f Support --iter in export.py Fangjun Kuang 2022-05-13 10:47:01 +08:00
  • 8965a5e7f3 Update RESULTS.md Guanbo Wang 2022-05-12 22:27:37 -04:00
  • 2cf0d51e89 add tensorboard Desh Raj 2022-05-12 22:14:35 -04:00
  • 72d38950d1 add results Desh Raj 2022-05-12 22:07:21 -04:00
  • f62f8fba20 remove conformer ctc; minor fixes in RNN-T Desh Raj 2022-05-12 22:02:08 -04:00
  • c2583ab1e0 Validate that there are no OOV tokens in BPE-based lexicons. Fangjun Kuang 2022-05-13 08:47:08 +08:00
  • 994b8a7716 Merge remote-tracking branch 'dan/master' into deeper-conformer Fangjun Kuang 2022-05-13 07:41:50 +08:00
  • d2b41a000b Rename to avoid conflicts. Fangjun Kuang 2022-05-13 07:41:38 +08:00
  • ebdb97f615 Update readme. Fangjun Kuang 2022-05-13 07:39:58 +08:00
  • aeb8986e35
    Ignore padding frames during RNN-T decoding. (#358) Fangjun Kuang 2022-05-13 07:39:14 +08:00