All pull requests

Pull requests

• Fix VAD.save_boundaries crash on no-speech audio and missing last segment in samples

  #3117·tanishqbaweja opened Oct 29, 2026·7 of 13 tasks·

  ·

• fix(distributed): initialise the DDP group on the active accelerator

  #3116·li-lizhe opened Oct 24, 2026Contributor·

  ·

• Normalize smoothed KL mean by non-padding targets

  #3115·kocean9-freedom opened Oct 23, 2026·7 of 11 tasks·

  ·

• Fix HF lobes (e.g. Whisper) silently keeping random weights when loading SpeechBrain checkpoints

  #3114·Echolonius opened Oct 22, 2026·7 tasks done·

  ·
  1

• Propagate non-timeout errors from parallel metric evaluation

  #3112·kocean9-freedom opened Oct 22, 2026·7 of 12 tasks·

  ·

• Respect positive_label in binary metric counts

  #3111·akshitkhokhani opened Oct 21, 2026·

  ·

• Ignore data generated under tests/samples by recipe tests

  #3110·quetion667 opened Oct 21, 2026·4 of 7 tasks·

  ·
  1

• Fix InputNormalization padding masks for nondefault length dimensions

  #3109·mottopanikeiku opened Oct 18, 2026·7 of 13 tasks·

  ·

• Keep dtype and device in TextEncoder.prepend_bos_index / append_eos_index for tensor input

  #3108·sungbin1015 opened Oct 16, 2026Contributor·7 of 13 tasks·

  ·

• Keep ECAPA-TDNN pooling masks in the input dtype

  #3107·Cassian433 opened Oct 15, 2026·7 tasks done·

  ·
  1

• Fix wer_summary reporting WER=0.0 when all scored references are empty

  #3102·Virtuoso461 opened Oct 12, 2026·

  ·

• Exclude padding when relative loss lengths round above an integer

  #3098·antonsoo opened Oct 9, 2026·7 of 13 tasks·

  ·

• Preserve batches when a distributed worker gets only one

  #3097·antonsoo opened Oct 9, 2026·7 of 13 tasks·

  ·

• Add formant estimation to vocal features

  #3096·pplantinga opened Oct 6, 2026Collaborator·

  ·

• Fix optimizer selection in SENSE recipe

  #3094·MaryemBouziane opened Oct 6, 2026Contributor·

  ·

• Use reference length as the denominator of WeightedErrorRateStats

  #3089·MohammadHijjawi97 opened Sep 28, 2026·7 of 13 tasks·

  ·

• Fix DNS composite_eval.py always raising TypeError (eval_composite missing sample_rate)

  #3079·Anai-Guo opened Aug 31, 2026·

  ·

• Fix BinaryMetricStats DER denominator

  #3078·Kayvan-Zahiri opened Aug 30, 2026Contributor·7 tasks done·

  ·

• fix(inference): make DiffWaveVocoder.forward and PIQAudioInterpreter.forward callable

  #3077·Anai-Guo opened Aug 29, 2026·6 of 7 tasks·

  ·

• Make RelPosEncXL able to encode the sign of the relative position

  #3073·yentur opened Aug 9, 2026·7 tasks done·

  ·

• [Code from Samsung AI Cambridge] BESTOW streaming for speech-to-text translation

  #3059·TParcollet opened Jun 3, 2026Collaborator·

  ·

• Bump gitpython from 3.1.37 to 3.1.47 in /recipes/BinauralWSJ0Mix

  dependenciespython
  #3051·dependabot[bot] opened Apr 26, 2026Bot·

  ·

• Add ECAPA2 to VoxCeleb

  #3039·othman-istaiteh opened Mar 4, 2026·7 of 13 tasks·

  ·
  19

• Optimize minDCF memory footprint

  #3037·othman-istaiteh opened Feb 27, 2026·6 of 13 tasks·

  ·

• fix spectrogram drop mask staring position

  #3036·gfdb opened Feb 17, 2026Contributor·13 tasks·

  ·
  3