search

Doubiiu / CodeTalker

[CVPR 2023] CodeTalker: Speech-Driven 3D Facial Animation with Discrete Motion Prior
MIT License
515 stars 57 forks source link

stage2 training does not converge #53

Open RujingD opened 1 year ago

RujingD commented 1 year ago

stage1 is convergent, stage2 training loss does not decrease