-
Notifications
You must be signed in to change notification settings - Fork 22k
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
adjust thresholds for gluon_inception_v3, beit_base_patch16_224, phli… #127664
base: main
Are you sure you want to change the base?
adjust thresholds for gluon_inception_v3, beit_base_patch16_224, phli… #127664
Conversation
We may need to triage beit_base_patch16_224 and phlippe_resnet models further, as the debug print PR seems to suggest a real regression (see: #126692 (comment)) |
…hreshold for phlippe_resnet as it's fixed by cudnnv9
@pytorchbot rebase |
@pytorchbot started a rebase job onto refs/remotes/origin/viable/strict. Check the current status here |
Rebase failed due to Command
Raised by https://github.com/pytorch/pytorch/actions/runs/9507897428 |
@pytorchbot rebase |
@pytorchbot started a rebase job onto refs/remotes/origin/viable/strict. Check the current status here |
Rebase failed due to Command
Raised by https://github.com/pytorch/pytorch/actions/runs/9570550855 |
…inductor_tolerance_fixes
Looks like this PR hasn't been updated in a while so we're going to go ahead and mark this as |
Compiling the error logs for accuracy failures from #126692, we get
So out of the four failing models, gluon_inception_v3 and beit_base_patch16_224 are 0.01151, and 0.01333 respectively where the tolerance is 0.01, so a very minor mismatch. Similarly phlippe_resnet is 0.00102, where the tolerance is at 0.001.
This PR adjusts the thresholds for these three models that are failing with minor accuracy mismatch by bumping them up to the next tolerance requirement, and then removing the xfail from the tests.
cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @chenyang78 @kadeng @chauhang @amjames @nWEIdia @eqy