DeepSpeed
632ab6b2
- make torch.compile more adaptive to old pytorch version
Go
Login via GitHub
Home
Pricing
FAQ
Install
Login
via GitHub
Commit
View On
GitHub
Commit
1 year ago
make torch.compile more adaptive to old pytorch version Signed-off-by: Guokai Ma <guokai.ma@intel.com>
References
#7648 - Put Muon optimizer momentum buffer on GPU
Author
delock
Parents
e02c0ec9
Loading