-
Couldn't load subscription status.
- Fork 25.7k
Fix AT_USE_JITERATOR checks #100464
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
Fix AT_USE_JITERATOR checks #100464
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/100464
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 065ee34: This comment was automatically generated by Dr. CI and updates every 15 minutes. |
`AT_USE_JITERATOR` evaluates to false when the definition isn't included, so these files were not using jiterator at all. [ghstack-poisoned]
`AT_USE_JITERATOR` evaluates to false when the definition isn't included, so these files were not using jiterator at all. ghstack-source-id: be1e006 Pull Request resolved: pytorch#100464
`AT_USE_JITERATOR` evaluates to false when the definition isn't included, so these files were not using jiterator at all. [ghstack-poisoned]
`AT_USE_JITERATOR` evaluates to false when the definition isn't included, so these files were not using jiterator at all. [ghstack-poisoned]
`AT_USE_JITERATOR` evaluates to false when the definition isn't included, so these files were not using jiterator at all. [ghstack-poisoned]
`AT_USE_JITERATOR` evaluates to false when the definition isn't included, so these files were not using jiterator at all. [ghstack-poisoned]
| namespace at::native { | ||
|
|
||
| CONSTEXPR_EXCEPT_WIN_CUDA char acos_name[] = "acos"; | ||
| CONSTEXPR_EXCEPT_WIN_CUDA char acos_name[] = "acos_impl"; |
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.
The _impl is necessary because otherwise it calls acos defined in the jiterator headers which fails because the template parameters are different. Same issue exists for all of these.
| auto common_dtype = iter.common_dtype(); | ||
| if (at::isComplexType(common_dtype)) { | ||
| #if AT_USE_JITERATOR | ||
| // Disabled due to accuracy issues |
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.
Some of these might just need tolerance bumps, but I didn't want to spend much time on this.
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.
Could you open an issue describing what would need doing and tag it as "good first issue"?
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.
Opened #100842
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.
Good catch!
Is there any way we could have a regression test for this (even if it's xfailed for now, given that all these are deactivated now).
Also cc @kshitij12345 as he probably wrote a few of these.
| auto common_dtype = iter.common_dtype(); | ||
| if (at::isComplexType(common_dtype)) { | ||
| #if AT_USE_JITERATOR | ||
| // Disabled due to accuracy issues |
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.
Could you open an issue describing what would need doing and tag it as "good first issue"?
|
@pytorchbot rebase |
|
@pytorchbot successfully started a rebase job. Check the current status here |
`AT_USE_JITERATOR` evaluates to false when the definition isn't included, so these files were not using jiterator at all. [ghstack-poisoned]
|
Successfully rebased |
|
@pytorchbot merge |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
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.
large/extreme values could have bugs in the implementation, libc++ might not handle those cases correctly (iirc we had to skip some mac tests for extreme values for the same reason)
Stack from ghstack (oldest at bottom):
AT_USE_JITERATORevaluates to false when the definition isn'tincluded, so these files were not using jiterator at all.