Fix : RecursionError when using pipe.enable_sequential_cpu_offload()#9
Open
JVSCHANDRADITHYA wants to merge 2 commits intoNetflix:mainfrom
Open
Fix : RecursionError when using pipe.enable_sequential_cpu_offload()#9JVSCHANDRADITHYA wants to merge 2 commits intoNetflix:mainfrom
JVSCHANDRADITHYA wants to merge 2 commits intoNetflix:mainfrom
Conversation
Author
|
Hi @sam-motamed and @ttchengab, Wanted to check if this PR aligns with the intended approach... Will make changes if needed. |
This file contains hidden or 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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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 fixes Issue #8
This PR fixes a RecursionError that occurs when using
enable_sequential_cpu_offload()with the FP8 wrapper.Problem
The FP8 wrapper modifies module dtype inside the forward pass using
.to()When combined with
acceleratehooks, this leads to recursive forward calls and a crash.Fix
.to()calls from forwardResult
Tested with:
autocast_model_forward()to avoid ImportErrors (None Found)Validation
ATTACHMENTS :