search

lucidrains / e2-tts-pytorch

Implementation of E2-TTS, "Embarrassingly Easy Fully Non-Autoregressive Zero-Shot TTS", in Pytorch
MIT License
228 stars 21 forks source link

fix gradient accumulation #24

Closed SWivid closed 1 month ago

SWivid commented 1 month ago
  1. correctly pass in grad_accumulation_steps value to trainer
  2. fix sync issue, otherwise _unscale_gradient halfway if mixed precision used
lucidrains commented 1 month ago

thank you!