transformers
e61eed44 - [Trainer] use deepspeed SP process group when Accelerate doesn’t build a mesh (#43799)

Commit
97 days ago
[Trainer] use deepspeed SP process group when Accelerate doesn’t build a mesh (#43799) * use deepspeed SP process group when Accelerate doesn’t build a mesh * use only when SP is enabled * switch order of check --------- Co-authored-by: Marc Sun <57196510+SunMarc@users.noreply.github.com>
Author
Parents
Loading