-
Notifications
You must be signed in to change notification settings - Fork 6.5k
Speed up the peft lora unload #5741
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
|
The documentation is not available anymore as the PR was closed or merged. |
Co-Authored-By: Younes Belkada <[email protected]>
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks Sourab for making this work with the new PR on PEFT. I have some minor comments, no blockers.
For context, the mentioned PR is now merged and this change is necessary to work with PEFT main. We ran a test on OPT-350m, where we saw 15x speedup in unloading with the new changes.
Co-Authored-By: Benjamin Bossan <[email protected]>
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks a lot for this and for taking care of the backward compatibility! left one suggestion above
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM, thanks.
|
Thanks! |
* Update peft_utils.py * fix bug * make the util backwards compatible. Co-Authored-By: Younes Belkada <[email protected]> * fix import issue * refactor the backward compatibilty condition * rename the conditional variable * address comments Co-Authored-By: Benjamin Bossan <[email protected]> * address comment --------- Co-authored-by: Younes Belkada <[email protected]> Co-authored-by: Benjamin Bossan <[email protected]>
* Update peft_utils.py * fix bug * make the util backwards compatible. Co-Authored-By: Younes Belkada <[email protected]> * fix import issue * refactor the backward compatibilty condition * rename the conditional variable * address comments Co-Authored-By: Benjamin Bossan <[email protected]> * address comment --------- Co-authored-by: Younes Belkada <[email protected]> Co-authored-by: Benjamin Bossan <[email protected]>
What does this PR do?