xtts deepspeed init with dtype=torch.float32 even though model=self.gpt_inference.half() #3662
nguyenquocthao
started this conversation in
General
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Path: TTS/tts/layers/xtts/gpt.py
Shouldn't dtype = torch.half?
Beta Was this translation helpful? Give feedback.
All reactions