feature: CpuOffload pre_forward don't attempt to move if already on device #3695
Conversation
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
…y on that the device. This is more noticiable in large step itterations on diffusion loops when the pre_froward can get called many times
sam-noumena
approved these changes
Jul 26, 2025
Contributor
|
cc @SunMarc |
SunMarc
approved these changes
Aug 6, 2025
Member
SunMarc
left a comment
SunMarc
left a comment
There was a problem hiding this comment.
Thanks ! I'll merge this now for the release
| if self.prev_module_hook is not None: | ||
| self.prev_module_hook.offload() | ||
| clear_device_cache() | ||
| if self.prev_module_hook is not None and isinstance(self.prev_module_hook, UserCpuOffloadHook): |
Member
There was a problem hiding this comment.
it should be fine to not check for isinstance(self.prev_module_hook, UserCpuOffloadHook) as we expect it to be UserCpuOffloadHook in the docstring.
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
Member
|
@bot /style |
Contributor
|
Style bot fixed some files and pushed the changes. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 participants
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Outdated suggestions cannot be applied.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.
What does this PR do?
Added optimisation to not attempt to move devices if already on that the device. This is more noticeable in large step iterations on diffusion loops when the pre_forward can get called many times
Possibly the there is a better way than using next() and noticed it's possibly to set the prev_module_hook to invalid object so added an additional check. Glad for any feedback to improve. Or perhaps there's an even better way.
Fixes # (issue)
Issue was not as bad as I initially thought. But from speaking with maintainers there was some smaller gains found. Mainly noticeable in large loops
huggingface/diffusers#11872
Before submitting
Pull Request section?
to it if that's the case.
documentation guidelines, and
here are tips on formatting docstrings.
Who can review?
Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.