lggvu
/

mmvap-avcocktail

voice-activity-projection

Model card Files Files and versions

MMVAP — AVCocktail

Audio-visual early-fusion Voice Activity Projection (VAP) turn-taking model.

Training data: Trained from scratch on the AVCocktail corpus.

Checkpoint: Epoch 5, val/loss=3.3068 (lowest val loss checkpoint; note the run's epoch 3 checkpoint was not retained on disk).

Files

hparams.yaml
checkpoint.ckpt

Source code

This checkpoint was produced with mm-turn-taking.

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Collection including lggvu/mmvap-avcocktail

MM-Turn-Taking-Candor-AVCocktail

Pretrained models related to Candor & AVCocktail datasets. • 11 items • Updated 2 days ago