All pull requests

Pull requests

• Add neural contextual biasing recipe for the Zipformer transducer (LibriSpeech)

  #2111·mahsa7823 opened Oct 7, 2026Contributor·3 tasks·

  ·
  9

• Add experimental Japanese Common Voice phone KWS recipe and evaluation

  #2110·mitsu-h opened Sep 27, 2026·

  ·
  11

• Fix ami/libricss SURT decode.py never forwarding batch to decode_one_batch

  #2108·Anai-Guo opened Aug 30, 2026·

  ·
  1

• Fix gigaspeech pruned_transducer_stateless2 train.py crashing on the first batch

  #2107·Anai-Guo opened Aug 30, 2026·

  ·
  1

• Fix iwslt22_ta/ST train.py requiring an 'sp' model the recipe never builds

  #2106·Anai-Guo opened Aug 29, 2026·

  ·
  1

• Fix europarl_st/lcma_srt passing kwargs that setup_dist and write_error_stats do not accept

  #2105·Anai-Guo opened Aug 29, 2026·

  ·
  3

• Update gigaspeech result

  #2093·yfyeung opened Jun 28, 2026Collaborator·

  ·
  5

• CommonVoice FR - Preprocessing stripped every lowercase character in the model

  #2085·Kizyow opened Jun 2, 2026·

  ·
  3

• Zapformer preview

  #2082·danpovey opened May 13, 2026Collaborator·

  ·
  57

• adding --use-adapter and --adapter-dim to zipformer_adapter decode.py

  #2079·Elquintas opened Apr 15, 2026·

  ·
  10

• egs/vctk/TTS: Add --local-data-dir option to skip VCTK download

  #2078·drakempham opened Apr 8, 2026·

  ·
  20

• feat(zipformer): Add multi-node DDP training support via torchrun/SLURM

  #2067·Houss3m opened Feb 8, 2026·

  ·
  14

• [not_for_merge_yet] fixing github workflows with automatic tests

  #2061·KarelVesely84 opened Jan 20, 2026Contributor·

  ·
  9

• fixed sampling rate issue on compute_fbank_librispeech.py

  #2060·Prithwin55 opened Jan 20, 2026·

  ·
  10

• [Not for merge] Contrastive Language-Speech Pre-training (CLSP) Recipe

  #2053·yfyeung opened Dec 30, 2025Collaborator·8 tasks done·

  ·
  1

• minimum length of encoder output for CTC training

  #2051·KarelVesely84 opened Dec 19, 2025Contributor·

  ·
  10

• Bilingual Japanese / English zipformer recipe (multi_ja_en) and MLS English recipe (mls_english)

  #2015·kinanmartin opened Aug 28, 2025·

  ·
  79

• Adding Common Voice Preprocessing for Farsi

  #1997·msghik opened Aug 2, 2025·

  ·
  2

• Add export-onnx.py for fluent_speech_commands

  #1992·csukuangfj opened Jul 16, 2025Collaborator·

  ·
  6

• Support using different musan augmentations for the same audio.

  #1975·csukuangfj opened Jul 1, 2025Collaborator·

  ·
  5

• Add Librispeech recipe example for parallel augmentation

  #1959·csukuangfj opened May 30, 2025Collaborator·

  ·

• [Not for merge] Support Zipformer encoder for LLM based ASR

  #1944·yfyeung opened May 15, 2025Collaborator·5 tasks done·

  ·
  4

• Support multiple parallel augmentations.

  #1942·csukuangfj opened May 12, 2025Collaborator·

  ·
  1

• [Not For Merge] implement qwen-omni like speech2speech model training recipe

  #1935·yuekaizhang opened Apr 29, 2025Collaborator·

  ·

• potential fix to k2SSL hubert_ce encoder forward function

  #1901·frankyoujian opened Apr 1, 2025Contributor·

  ·