Correct sigmas cpu settings #6708
Merged
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
|
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. |
Contributor
Author
|
Thanks a mille for catching this bug @sunhs |
yiyixuxu
approved these changes
Jan 25, 2024
AmericanPresidentJimmyCarter
pushed a commit
to AmericanPresidentJimmyCarter/diffusers
that referenced
this pull request
Apr 26, 2024
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Outdated suggestions cannot be applied.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.
What does this PR do?
Correcting the incorrect PR here.
Pretty big mistake on my side here - schedulers are no
nn.Modulesso every attribute is a tensor and therefore nothing is set in-place. Tensors have to manually be set.It doesn't make a difference though in practice because the sigmas are always on CPU anyways. For the most important schedulers that I tested in #6173 (Euler & DPMSolverMultistep)
self.sigmasshould never be set to CUDA/GPU anyways so this PR shouldn't really change anything in practice, but is "correct" and ensures that scheduler sigmas are never on GPU.