Move Shape Registration back to CPP #3340
Closed
Conversation
This file contains hidden or 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
|
This pull request was exported from Phabricator. Differential Revision: D65615437 |
✅ Deploy Preview for pytorch-fbgemm-docs ready!
To edit notification comments on pull requests, go to your Netlify site configuration. |
Summary: X-link: facebookresearch/FBGEMM#434 In D64147797, I attempted to modernize our custom function shape registration using torch's register_fake, which has a bit better support than cpp meta registration. Although I think the motivation still is good, this diff introduced a significant number of complexities due to python being more brittle. Ultimately, I think it was a failure and made fbgemm custom ops harder to integrate broadly. In this diff, I revert custom operator shape registration to using CPP. We still are a bit improved over our previous state as we dropped the catch-all registration that prevented custom op compatibility with torch.export. Reviewed By: sryap, jianyuh Differential Revision: D65615437
jwfromm
force-pushed
the
export-D65615437
branch
from
b095888 to
ba6a7e7
Compare
|
This pull request was exported from Phabricator. Differential Revision: D65615437 |
|
This pull request has been merged in f79508e. |
q10
pushed a commit
to q10/FBGEMM
that referenced
this pull request
Apr 10, 2025
Summary: X-link: pytorch#3340 Pull Request resolved: facebookresearch/FBGEMM#434 In D64147797, I attempted to modernize our custom function shape registration using torch's register_fake, which has a bit better support than cpp meta registration. Although I think the motivation still is good, this diff introduced a significant number of complexities due to python being more brittle. Ultimately, I think it was a failure and made fbgemm custom ops harder to integrate broadly. In this diff, I revert custom operator shape registration to using CPP. We still are a bit improved over our previous state as we dropped the catch-all registration that prevented custom op compatibility with torch.export. Reviewed By: sryap, jianyuh Differential Revision: D65615437 fbshipit-source-id: 24bf5eee9125ee975b6921028705439b42a809b5
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.
Summary:
In D64147797, I attempted to modernize our custom function shape registration using torch's register_fake, which has a bit better support than cpp meta registration. Although I think the motivation still is good, this diff introduced a significant number of complexities due to python being more brittle.
Ultimately, I think it was a failure and made fbgemm custom ops harder to integrate broadly. In this diff, I revert custom operator shape registration to using CPP. We still are a bit improved over our previous state as we dropped the catch-all registration that prevented custom op compatibility with torch.export.
Differential Revision: D65615437