-
Notifications
You must be signed in to change notification settings - Fork 5.6k
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
Update and disable push_tests_mps
workflow
#10442
base: main
Are you sure you want to change the base?
Conversation
Does removing them help launch the tests? Could we maybe trigger it manually? |
I don't know if we can trigger it manually. Currently the tests launch on push to https://github.com/huggingface/diffusers/actions/runs/12588771861/job/35087350062#step:7:8211 Most pipeline tests seem to be failing and there are some warnings
It suggests some pipeline tests are failing because of it, but there are no tracebacks for individual test failures, probably due to the segfault. |
Sorry should have been clearer. What we could do to ensure the MPS stuff runs successfully:
Does this sound good? |
Nice seems like the tests are getting picked up? We can let them run and fix anything in a follow-up PR? WDYT? |
There are still some failures so far, I'll compare to one of the last runs to check if anything is fixed. I think priority is fixing the segfault but I see no indication of anything that could be causing it. We could separate the tests like models, pipelines, schedulers, might help break it down as there are a lot of |
Makes sense and okay for me to tackle that in this PR. Thanks! Let me know if I can help. |
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. |
@@ -220,7 +220,7 @@ def set_timesteps( | |||
self.num_inference_steps = num_inference_steps | |||
|
|||
step_ratio = self.config.num_train_timesteps // self.num_inference_steps | |||
timesteps = (np.arange(0, num_inference_steps) * step_ratio).round()[::-1].copy().astype(np.int64) | |||
timesteps = (np.arange(0, num_inference_steps) * step_ratio).round()[::-1].copy().astype(np.int32) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Won't it cause any numerical problems or qualitative problems?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
PR is now only workflow changes, we may want to just disable it until we get the new runner though.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
That is okay with me.
12fe529
to
1c7df8a
Compare
push_tests_mps
workflowpush_tests_mps
workflow
push_tests_mps
workflowpush_tests_mps
workflow
What does this PR do?
Updates
push_tests_mps
to separate tests by category and fix some environment issues. For now a lot of failures are due to a problem with GitHub's runners so we also disable the workflow.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