Skip to content
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

Avoid double wrapping of all accelerate.prepare objects #1555

Merged
merged 9 commits into from
Jun 7, 2023
Merged

Conversation

muellerzr
Copy link
Collaborator

Solves double nested wrapping issues linked as a side-effect to huggingface/transformers#24050.

Also resets step to 0 during free_memory

@muellerzr muellerzr requested a review from sgugger June 7, 2023 17:19
@HuggingFaceDocBuilderDev
Copy link

HuggingFaceDocBuilderDev commented Jun 7, 2023

The documentation is not available anymore as the PR was closed or merged.

Copy link
Collaborator

@sgugger sgugger left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks!

@muellerzr muellerzr merged commit 65dd4f2 into main Jun 7, 2023
@muellerzr muellerzr deleted the free-memory branch June 7, 2023 17:37
sgugger pushed a commit that referenced this pull request Jun 7, 2023
* Add step reset to free memory

* Check if not Accelerated Optimizer

* Continue

* Another try

* Check the rest

* Try with just check on init

* Change logic based on review

* Update

* Oops very big logic issue!
@Yura52
Copy link
Contributor

Yura52 commented Jun 7, 2023

@muellerzr @sgugger Hi! I have zero familiarity with the code base, but is it intentional that the added setattr() call sets the "_is_accelerate_prepared" attribute, while the used getattr() calls access the different "_accelerate_prepared" (note missing "_is" suffix) attribute?

@muellerzr
Copy link
Collaborator Author

Nope, it is not! Would you like to put in a fix @Yura52? Otherwise I can do so quickly :)

@Yura52
Copy link
Contributor

Yura52 commented Jun 7, 2023

Nope, it is not! Would you like to put in a fix @Yura52? Otherwise I can do so quickly :)

Here is the PR: #1560 @muellerzr

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants