All issues

Search results

Can audio codecs designed for reconstruction (e.g., DAC) be used as upstream models in s3prl?

Status: Open.

#585 In s3prl/s3prl;

· liusanchun5-art opened on Mar 15, 2026

Upstream Trainable hugely reduce accuracy score for SID for both hubert and w2v2

Status: Open.

#583 In s3prl/s3prl;

· david-wei-01001 opened on Feb 14, 2026

Masking issue in AttentivePoolingModule

Status: Open.

#579 In s3prl/s3prl;

· kpeverson opened on Jan 29, 2026

Downstream checkpoints that lack reproducibility.

Status: Open.

#577 In s3prl/s3prl;

· hhhhhh45 opened on Jan 11, 2026

Support for Continued Pretraining of wav2vec2 on New Unlabeled Audio

Status: Open.

#576 In s3prl/s3prl;

· akbar20gh opened on Dec 19, 2025

Issues with deprecated api in torchaudio

Status: Open.

#575 In s3prl/s3prl;

· ftshijt opened on Dec 9, 2025

Leaderboard website is down

Status: Open.

#574 In s3prl/s3prl;

· mxmpl opened on Aug 20, 2025

Audio Deepfake Detection Task

Status: Open.

#573 In s3prl/s3prl;

· hashim19 opened on Jul 29, 2025

Inference for emotion is not implemented

Status: Open.

#572 In s3prl/s3prl;

· KansaiTraining opened on Jul 4, 2025

Supports for Numpy 2.0 and above

Status: Open.

#571 In s3prl/s3prl;

· ftshijt opened on May 29, 2025

extra tokens are not optimised by CTC loss but used as targets for model to predict

Status: Open.

#570 In s3prl/s3prl;

· wenjie-p opened on May 2, 2025

Use custom hubert as teacher to train a hubert distilled

Status: Open.

#569 In s3prl/s3prl;

· eloi-tavus opened on Apr 26, 2025