Commit Graph

  • 7600268539 fixing the CI test jinzr 2024-01-30 09:59:19 +08:00
  • a1177c979a minor updates jinzr 2024-01-30 09:48:57 +08:00
  • d5da6abb49 Merge branch 'master' into dev/multi-zh-streaming-decoding jinzr 2024-01-30 09:41:28 +08:00
  • 37b975cac9
    fixed a CI test for wenetspeech (#1476) zr_jin 2024-01-27 06:41:56 +08:00
  • 9644c1722a minor fixes jinzr 2024-01-27 03:21:26 +08:00
  • b9bbdfaadc Comply to issue #1149 jinzr 2024-01-27 03:18:07 +08:00
  • c606ef5e50 Update export-onnx.py jinzr 2024-01-27 02:09:16 +08:00
  • d9227665eb Update run-wenetspeech-pruned-transducer-stateless2.sh jinzr 2024-01-27 00:45:09 +08:00
  • 1c30847947
    Whisper Fine-tuning Recipe on Aishell1 (#1466) Yuekai Zhang 2024-01-27 00:32:30 +08:00
  • 8d39f9508b
    Fix torchscript export to use tokens.txt instead of lang_dir (#1475) Fangjun Kuang 2024-01-26 19:18:33 +08:00
  • 4109253f95 Fix export for stateless7_streaming Fangjun Kuang 2024-01-26 19:16:54 +08:00
  • d64cdf6d4f Fix export for stateless5 Fangjun Kuang 2024-01-26 17:35:32 +08:00
  • ed68914fe2 fix decoder for emformer_rnnt2 Fangjun Kuang 2024-01-26 16:58:44 +08:00
  • 283227c0c5 Fix export for pruned_stateless_emformer_rnnt2 Fangjun Kuang 2024-01-26 16:39:12 +08:00
  • 6c98fbc309
    Merge 752e16be1038211bada5d4f15eb4b59d3f6ae9f6 into c401a2646b347bf1fff0c2ce1a4ee13b0f482448 Erwan Zerhouni 2024-01-26 16:24:46 +08:00
  • 7a4d8c9c1d Use torch.jit.script for the LSTM transducer decoder model Fangjun Kuang 2024-01-26 16:07:16 +08:00
  • c401a2646b
    minor fix of zipformer/optim.py (#1474) Zengwei Yao 2024-01-26 15:50:11 +08:00
  • 951b537e40 minor fix of zipformer/optim.py yaozengwei 2024-01-26 15:39:58 +08:00
  • e1880b7413 fix exporting tal_csasr recipe Fangjun Kuang 2024-01-26 12:48:58 +08:00
  • 7cf3ea8b33 fix export for aidatatang_200zh Fangjun Kuang 2024-01-26 12:17:48 +08:00
  • 71ee509e7d Use tokens.txt to replace bpe.model Fangjun Kuang 2024-01-26 12:05:21 +08:00
  • c65b734838 Fix export for wenetspeech Fangjun Kuang 2024-01-26 11:59:42 +08:00
  • c72bba1e0b Fix export gigaspeech models Fangjun Kuang 2024-01-26 10:50:04 +08:00
  • 9c494a3329
    typos fixed (#1472) zr_jin 2024-01-25 18:41:43 +08:00
  • fd4ebf3bfe add manifest dir option root 2024-01-25 08:31:08 +00:00
  • 4170f44c35
    final polish zr_jin 2024-01-25 12:47:05 +08:00
  • 6ba1e63c41 Small change to avoid hardcoded change in make_kn_lm.py Xinyuan Li 2024-01-24 23:38:35 -05:00
  • eec59410f1 Fix style check issues Xinyuan Li 2024-01-24 11:45:16 -05:00
  • 5d94a19026 prepare for 1000h dataset Triplecq 2024-01-24 11:33:36 -05:00
  • d864da4d65 validation scripts Triplecq 2024-01-25 01:25:28 +09:00
  • 7047a579b8 Undo changes to util summary writer Xinyuan Li 2024-01-23 23:23:22 -05:00
  • 8dc1ca194d Use symlinks whenever possible Xinyuan Li 2024-01-23 23:21:37 -05:00
  • d725bad4fd Rename asr_datamodule to slu_datamodule Xinyuan Li 2024-01-23 21:39:40 -05:00
  • 9755157d1e typos fixed jinzr 2024-01-24 10:23:08 +08:00
  • fd9f7b466b Restructure recipe directories Xinyuan Li 2024-01-23 20:19:32 -05:00
  • 5e88f80b50 Remove tdnn architecture from fluent speech commands recipe Xinyuan Li 2024-01-23 20:14:39 -05:00
  • f35fa8aa8f add blank penalty in decoding script Triplecq 2024-01-23 17:10:10 -05:00
  • 559ed150bb
    Fix typo (#1471) Yifan Yang 2024-01-23 22:51:09 +08:00
  • 358292eb03
    Fix typo Yifan Yang 2024-01-23 22:45:27 +08:00
  • a8e9dc2488 all combinations of epochs and avgs Triplecq 2024-01-23 21:12:17 +09:00
  • 033a021b72
    Merge 90228632926c03cdf219e6c984e397f7404615a1 into ebe97a07b082f9bee4a18b5f8e54c453187a74bb Piotr Żelasko 2024-01-23 10:11:02 +01:00
  • ebe97a07b0
    Reworked README.md (#1470) zr_jin 2024-01-23 16:26:24 +08:00
  • 6b188fc733
    Update README.md zr_jin 2024-01-23 16:22:28 +08:00
  • 16012547d2 Update README.md jinzr 2024-01-23 16:01:39 +08:00
  • 2844bbf115 Update README.md jinzr 2024-01-23 15:52:10 +08:00
  • 8383a7cfd5 Update README.md jinzr 2024-01-23 15:45:47 +08:00
  • 70e70edc8a
    Merge branch 'master' into spgispeech_zipformer zr_jin 2024-01-23 10:22:43 +08:00
  • 46605eaef2 fix wrong order of token slice Yuekai Zhang 2024-01-22 16:24:46 +08:00
  • ab08201f6c remove model file Yuekai Zhang 2024-01-22 16:15:56 +08:00
  • 8d9ab308af fix lint root 2024-01-22 08:10:26 +00:00
  • b623c3be15 fix requirements Yuekai Zhang 2024-01-22 15:20:59 +08:00
  • bda48291db using monkey patch to replace models Yuekai Zhang 2024-01-22 14:41:14 +08:00
  • 5dfc3ed7f9
    Fix buffer size of DynamicBucketingSampler (#1468) Yifan Yang 2024-01-21 02:10:42 +08:00
  • d305c7cceb Implement recipe for Fluent Speech Commands dataset Xinyuan Li 2024-01-19 13:37:00 -05:00
  • 1ab3073a5a
    Fix for flake8 Yifan Yang 2024-01-19 17:57:22 +08:00
  • ae9ac81c55 update yifanyeung 2024-01-19 17:51:38 +08:00
  • 25c1670431 update yifanyeung 2024-01-19 17:48:00 +08:00
  • c2e769ffe1 update yifanyeung 2024-01-19 17:44:51 +08:00
  • 69730a7f4b Fix buffer size yifanyeung 2024-01-19 17:39:12 +08:00
  • eae650e342 Add changes Xinyuan Li 2024-01-17 12:21:09 -05:00
  • 84e4af93d7 add whisper fine-tuning results Yuekai Zhang 2024-01-17 16:17:32 +08:00
  • 4590da3a83 support decoding giga marcoyang 2024-01-17 10:17:30 +08:00
  • 7bdde9174c
    A Zipformer recipe with Byte-level BPE for Aishell-1 (#1464) zr_jin 2024-01-16 21:08:35 +08:00
  • c6669de38d doc_str fixed jinzr 2024-01-16 15:29:44 +08:00
  • d90a24b946 Delete onnx_check_bbpe.py jinzr 2024-01-16 15:18:22 +08:00
  • 34e290ac4c minor updates jinzr 2024-01-16 15:17:16 +08:00
  • 674390e63e re-org the bbpe recipe for aishell jinzr 2024-01-16 14:49:50 +08:00
  • ad94191055 set bpe_model as required jinzr 2024-01-16 11:40:11 +08:00
  • b63576ccd0 added scripts for testing pretrained models jinzr 2024-01-16 11:34:36 +08:00
  • d7f284a60a removed unused softlinks jinzr 2024-01-16 11:19:10 +08:00
  • 9669fa05a3 added vocab_size jinzr 2024-01-16 11:14:42 +08:00
  • 13541424a9 Update RESULTS.md jinzr 2024-01-16 10:59:27 +08:00
  • fa96660ac9 support finetune zipformer marcoyang 2024-01-16 10:19:51 +08:00
  • 057238c27e add methods to load gigaspeech cuts for finetune marcoyang 2024-01-16 10:19:13 +08:00
  • 557b35cefc clean codes Yuekai Zhang 2024-01-15 20:40:44 +08:00
  • eea46458c5 revert asr data module Yuekai Zhang 2024-01-15 19:59:48 +08:00
  • e883bb60d4 remove seamless for next PR Yuekai Zhang 2024-01-15 19:34:03 +08:00
  • ac53222054 add model saving Yuekai Zhang 2024-01-15 14:56:18 +08:00
  • 2ce09809cd support large-v3 Yuekai Zhang 2024-01-14 18:27:41 +08:00
  • fa7ad4dc72 update deepspeed model loading Yuekai Zhang 2024-01-12 17:29:24 +08:00
  • b6418acda2 support deepspeed to finetune large model Yuekai Zhang 2024-01-12 16:14:10 +08:00
  • 92895f774f clean up codes Yuekai Zhang 2024-01-11 16:45:05 +08:00
  • 98d11abedb remove padding to 30s, compute validation loss once Yuekai Zhang 2024-01-11 16:30:48 +08:00
  • 07cefa82a7 change scaleadam to adamw Yuekai Zhang 2024-01-11 14:49:27 +08:00
  • 8b832f168d update lhotse version Yuekai Zhang 2024-01-09 01:48:25 -08:00
  • 5bf3a9cfe0 using audio with any length Yuekai Zhang 2023-09-26 17:09:04 +08:00
  • 6c2cd5b4c3 support whisper ft Yuekai Zhang 2023-09-26 10:46:35 +08:00
  • bb1c4466e3 rename train, train2, add support to fine-tune embedding table Yuekai Zhang 2023-09-11 18:46:38 -07:00
  • d926585b10 fix loading Yuekai Zhang 2023-09-08 16:46:42 +08:00
  • 2a288fb9bf add custom tokenizer Yuekai Zhang 2023-09-08 16:40:17 +08:00
  • 22ee287312 add token files Yuekai Zhang 2023-09-08 16:09:15 +08:00
  • 7e387dd54b change vocab table Yuekai Zhang 2023-09-08 16:07:39 +08:00
  • 72e9a436b8 fix typo Yuekai Zhang 2023-09-07 22:45:17 -07:00
  • cc6432443d add decoding with avg model Yuekai Zhang 2023-09-07 20:18:14 +08:00
  • 5f399dc780 load checkpoint to decode Yuekai Zhang 2023-09-07 05:01:24 -07:00
  • e81545714a update decoding from checkpoint Yuekai Zhang 2023-09-07 17:39:37 +08:00
  • 0d6d8f9473 update fine-tuning lr Yuekai Zhang 2023-09-07 02:32:30 -07:00
  • cbc3852876 add fairseq2 require Yuekai Zhang 2023-09-07 01:21:43 -07:00
  • 3a7ad277ad add requirements Yuekai Zhang 2023-09-07 15:57:38 +08:00
  • 363c3f1f82 update finetuning codes Yuekai Zhang 2023-09-07 15:20:00 +08:00