int8 support #488
Merged
int8 support #488
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
@mitchellnw looks low risk, ready to merge? |
|
@rwightman yep, ready to merge! |
rwightman
reviewed
Apr 16, 2023
src/training/main.py
Outdated
| @@ -277,6 +287,8 @@ def main(args): | |||
| if args.ddp_static_graph: | |||
| # this doesn't exist in older PyTorch, arg only added if enabled | |||
| ddp_args['static_graph'] = True | |||
| if args.use_bnb_linear is not None: | |||
There was a problem hiding this comment.
couldn't this be put right after replace_linear() in the previous if use_bnb_linear block?
There was a problem hiding this comment.
yep good point that simplifies -- i'll make the change and run a quick test to make sure all good.
There was a problem hiding this comment.
fixed, thanks @rwightman!
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR introduces beta support for int8 training and inference. You can enable int8 training with
--use-bnb-linear SwitchBackLinearGlobalor--use-bnb-linear SwitchBackLinearGlobalMemEfficient. For CLIP VIT-Huge this should currently correspond to a 10% training speedup with negligable accuracy difference.More speedups coming when the attention layer is refactored so that linear layers may be replaced there, too. However, that will require more changes so best to get this merged first.