Skip to content

Conversation

@swolchok
Copy link
Contributor

Partial fix for #7748.

[ghstack-poisoned]
@swolchok
Copy link
Contributor Author

swolchok commented Jan 21, 2025

Stack from ghstack (oldest at bottom):

@pytorch-bot
Copy link

pytorch-bot bot commented Jan 21, 2025

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/7800

Note: Links to docs will display an error until the docs builds have been completed.

❌ 1 New Failure

As of commit f538e6c with merge base 948fba6 (image):

NEW FAILURE - The following job has failed:

This comment was automatically generated by Dr. CI and updates every 15 minutes.

@facebook-github-bot facebook-github-bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Jan 21, 2025
swolchok added a commit that referenced this pull request Jan 21, 2025
Partial fix for #7748.

ghstack-source-id: 81b8638
ghstack-comment-id: 2605573512
Pull Request resolved: #7800
@swolchok swolchok added the release notes: ops & kernels Changes to the opset and any new / changed kernel implementations label Jan 21, 2025
out,
l3,
1e-2,
executorch::runtime::testing::internal::kDefaultBFloat16Atol);
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why do you need to pass the default atol here?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

macros don't have default arguments

return internal::kDefaultHalfAtol;
}
if (t == ScalarType::BFloat16) {
return internal::kDefaultBFloat16Atol;
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is great, I wasn't aware of these defaults. Do these play well with the EXPECT_TENSOR_CLOSE_WITH_TOL macro?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

yes, though atol matters a lot less to me than rtol in all these diffs

Copy link
Contributor

@manuelcandales manuelcandales Jan 23, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

see my comment on #7807 , I think you should also introduce kDefaultHalfRtol and kDefaultBFloat16Rtol.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

replied; 1e-2 is not needed for all ops

@swolchok swolchok merged commit f5ba5bc into main Jan 23, 2025
44 of 47 checks passed
@swolchok swolchok deleted the gh/swolchok/156/head branch January 23, 2025 17:59
YIWENX14 pushed a commit that referenced this pull request Jan 28, 2025
zonglinpeng pushed a commit to zonglinpeng/executorch that referenced this pull request Jan 30, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. release notes: ops & kernels Changes to the opset and any new / changed kernel implementations

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants