transformers
add53e25
- don't use no_sync when deepspeed doesn't support it for certain zero stages (#35157)
Go
Login via GitHub
Home
Pricing
FAQ
Install
Login
via GitHub
Commit
View On
GitHub
Commit
1 year ago
don't use no_sync when deepspeed doesn't support it for certain zero stages (#35157) * don't use no_sync when deepspeed doesn't support it for certain zero stages * chore: lint * fix no_sync context for deepspeed across all zero types * chore: lint
References
#35157 - don't use no_sync when deepspeed doesn't support it for certain zero stages
Author
winglian
Parents
7237b3ec
Loading