Update RVQ projection layers during training

lucidrains / audiolm-pytorch

Implementation of AudioLM, a SOTA Language Modeling Approach to Audio Generation out of Google Research, in Pytorch

MIT License

2.32k stars 249 forks source link

Closed ilya16 closed 7 months ago

ilya16 commented 7 months ago

The projection layers in RVQ / RLFQ / RFSQ (if present) were not updated during training.

lucidrains commented 7 months ago

🤦‍♂️ yes, thank you!