Fix deterministic issue when getting pipeline dtype and device #10696
Conversation
This file contains 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
dimitribarbot
commented
Jan 31, 2025
| @@ -1619,10 +1619,12 @@ def components(self) -> Dict[str, Any]: | |||
| k: getattr(self, k) for k in self.config.keys() if not k.startswith("_") and k not in optional_parameters | |||
| } | |||
|
|
|||
| if set(components.keys()) != expected_modules: | |||
| actual = sorted(set(components.keys())) | |||
| expected = sorted(expected_modules) | |||
There was a problem hiding this comment.
I'm not sure if using sorted is necessary here, since expected_modules comes from the _get_signature_keys call which already calls sorted internally.
However, I leave it that way because it allows to be completely decoupled from the _get_signature_keys function. And I don't think it will add a big performance overhead (sorting of n elements, n=~10, elements already sorted).
dimitribarbot
commented
Jan 31, 2025
| @@ -585,3 +585,104 @@ def test_video_to_video(self): | |||
| with io.StringIO() as stderr, contextlib.redirect_stderr(stderr): | |||
| _ = pipe(**inputs) | |||
| self.assertTrue(stderr.getvalue() == "", "Progress bar should be disabled") | |||
|
|
|||
|
|
|||
| @require_torch_gpu | |||
There was a problem hiding this comment.
I need this to test multiple devices across pipeline components, so that testing with distinct pipeline devices fails without the code changes. Otherwise, I guess only CPU can be available?
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Outdated suggestions cannot be applied.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?
This PR aims to fix a deterministic issue when getting the pipeline dtype or device across multiple runs of the same Python script.
Internally, the ordering issue comes from the use of
setin the_get_signature_keysfunction in thesrc/diffusers/pipelines/pipeline_utils.pyfile. As stated in the Python documentation,setdoes not guarantee that the elements will have the same order across multiple runs of a Python script. This is due to the hashing algorithm which is initialized with a random salt each time the python interpreter is started. More information can be found here (especially in the second note at the end of the section).Fixes #10671.
Before submitting
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.
@sayakpaul