DeepSpeed
d4e18950 - [COMPILE] workflow for deepspeed + torch.compile (#6570)

Commit
1 year ago
[COMPILE] workflow for deepspeed + torch.compile (#6570) We use simple model + deepspeed zero 3 + torch.compile and count graph break numbers to demonstrate current status of combing deepspeed + torch.compile. --------- Co-authored-by: Masahiro Tanaka <81312776+tohtana@users.noreply.github.com>
Author
Parents
Loading