Skip to content

fix unfold for meta tensors#75717

Closed
bdhirsh wants to merge 7 commits intogh/bdhirsh/204/basefrom
gh/bdhirsh/204/head
Closed

fix unfold for meta tensors#75717
bdhirsh wants to merge 7 commits intogh/bdhirsh/204/basefrom
gh/bdhirsh/204/head

Conversation

@bdhirsh
Copy link
Collaborator

@bdhirsh bdhirsh commented Apr 13, 2022

Not sure when this happened, but unfold should be CompositeExplicitAutograd so it works for meta tensors.

Stack from ghstack:

Differential Revision: D35705379

[ghstack-poisoned]
@facebook-github-bot
Copy link
Contributor

facebook-github-bot commented Apr 13, 2022

🔗 Helpful links

💊 CI failures summary and remediations

As of commit 1774ea9 (more details on the Dr. CI page):


💚 💚 Looks good so far! There are no failures yet. 💚 💚


This comment was automatically generated by Dr. CI (expand for details).

Please report bugs/suggestions to the (internal) Dr. CI Users group.

Click here to manually regenerate this comment.

bdhirsh added a commit that referenced this pull request Apr 13, 2022
ghstack-source-id: f8eac05
Pull Request resolved: #75717
bdhirsh added 2 commits April 14, 2022 13:37
Not sure when this happened, but `unfold` should be `CompositeExplicitAutograd` so it works for meta tensors.




[ghstack-poisoned]
Not sure when this happened, but `unfold` should be `CompositeExplicitAutograd` so it works for meta tensors.




[ghstack-poisoned]
Not sure when this happened, but `unfold` should be `CompositeExplicitAutograd` so it works for meta tensors.




[ghstack-poisoned]
Not sure when this happened, but `unfold` should be `CompositeExplicitAutograd` so it works for meta tensors.




[ghstack-poisoned]
@bdhirsh
Copy link
Collaborator Author

bdhirsh commented Apr 17, 2022

It looks like there's a problem with XLA's as_strided lowering that causes tests to break (since marking unfold as CompositeExplicitAutograd causes it to decompose into an as_strided call for XLA).

I'm just gonna add Meta directly instead, just to cut through.

Not sure when this happened, but `unfold` should be `CompositeExplicitAutograd` so it works for meta tensors.




[ghstack-poisoned]
@bdhirsh
Copy link
Collaborator Author

bdhirsh commented Apr 17, 2022

@bdhirsh has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

@facebook-github-bot
Copy link
Contributor

@pytorchbot merge this

(Initiating merge automatically since Phabricator Diff has merged)

@github-actions
Copy link
Contributor

Hey @bdhirsh.
You've committed this PR, but it does not have both a 'release notes: ...' and 'topics: ...' label. Please add one of each to the PR. The 'release notes: ...' label should represent the part of PyTorch that this PR changes (fx, autograd, distributed, etc) and the 'topics: ...' label should represent the kind of PR it is (not user facing, new feature, bug fix, perf improvement, etc). The list of valid labels can be found here for the 'release notes: ...' and here for the 'topics: ...'.
For changes that are 'topic: not user facing' there is no need for a release notes label.

facebook-github-bot pushed a commit that referenced this pull request Apr 18, 2022
Summary:
Pull Request resolved: #75717

Not sure when this happened, but `unfold` should be `CompositeExplicitAutograd` so it works for meta tensors.

Test Plan: Imported from OSS

Reviewed By: zhxchen17

Differential Revision: D35705379

Pulled By: bdhirsh

fbshipit-source-id: a566db935fe42cabdc5eb6a025bae9a60ddd2f25
malfet pushed a commit that referenced this pull request Apr 20, 2022
Pull Request resolved: #75717

Approved by: https://github.com/ezyang

(cherry picked from commit 7be1b29)
@facebook-github-bot facebook-github-bot deleted the gh/bdhirsh/204/head branch April 22, 2022 14:17
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants