Skip to content

Remove hotpatches that circumvent MAGMA bug#34357

Closed
vishwakftw wants to merge 4 commits intopytorch:masterfrom
vishwakftw:remove-hotfix-lu-det
Closed

Remove hotpatches that circumvent MAGMA bug#34357
vishwakftw wants to merge 4 commits intopytorch:masterfrom
vishwakftw:remove-hotfix-lu-det

Conversation

@vishwakftw
Copy link
Copy Markdown
Contributor

@vishwakftw vishwakftw commented Mar 6, 2020

Changelog:

  • The magma implementation of small singular square batch matrices had a bug that resulted in nan values in the LU factorization result. This has been fixed in MAGMA 2.5.2. This PR removes the existing patch that was a temporary workaround for this bug.

Test Plan:

  • Existing tests for det and lu should pass

@vishwakftw
Copy link
Copy Markdown
Contributor Author

vishwakftw commented Mar 6, 2020

cc: @seemethere @soumith

This PR requires that the CI images have the updated MAGMA binaries as well to pass tests.

@dr-ci
Copy link
Copy Markdown

dr-ci Bot commented Mar 6, 2020

💊 CircleCI build failures summary and remediations

As of commit 40a1f39 (more details on the Dr. CI page):


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


This comment was automatically generated by Dr. CI (expand for details).Follow this link to opt-out of these comments for your Pull Requests.

Please report bugs/suggestions on the GitHub issue tracker.

This comment has been revised 16 times.

@yf225 yf225 requested review from seemethere and yf225 March 7, 2020 01:59
Copy link
Copy Markdown
Contributor

@yf225 yf225 left a comment

Choose a reason for hiding this comment

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

Thanks a lot for the fix! @seemethere Wondering do you mind checking if our CI has MAGMA >= 2.5.2? Thanks!

@seemethere
Copy link
Copy Markdown
Member

CI should now be updated with the latest version of magma

@vishwakftw
Copy link
Copy Markdown
Contributor Author

@seemethere thank you for confirming!

Copy link
Copy Markdown
Contributor

@facebook-github-bot facebook-github-bot left a comment

Choose a reason for hiding this comment

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

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

@facebook-github-bot
Copy link
Copy Markdown
Contributor

@seemethere merged this pull request in cb06cb7.

@facebook-github-bot
Copy link
Copy Markdown
Contributor

@seemethere merged this pull request in cb06cb7.

@ezyang
Copy link
Copy Markdown
Contributor

ezyang commented Mar 13, 2020

We're still running CUDA 9 on master with the old version on MAGMA, which means this breaks master. You'll need to do something about that first before we can land this.

@vishwakftw
Copy link
Copy Markdown
Contributor Author

@seemethere I believe we don't have MAGMA binaries for CUDA 9 on Anaconda either.

facebook-github-bot pushed a commit that referenced this pull request Apr 30, 2020
Summary:
Changelog:
- The magma implementation of small singular square batch matrices had a bug that resulted in nan values in the LU factorization result. This has been fixed in MAGMA 2.5.2. This PR removes the existing patch that was a temporary workaround for this bug.
Pull Request resolved: #35973

Test Plan:
- Existing tests for det and lu should pass

This is a re-submit of #34357

Differential Revision: D21336552

Pulled By: seemethere

fbshipit-source-id: 9c3b350966913147f1d5811927f3cae10fe620f1
laurentdupin pushed a commit to laurentdupin/pytorch that referenced this pull request Apr 24, 2026
Summary:
Changelog:
- The magma implementation of small singular square batch matrices had a bug that resulted in nan values in the LU factorization result. This has been fixed in MAGMA 2.5.2. This PR removes the existing patch that was a temporary workaround for this bug.
Pull Request resolved: pytorch#34357

Test Plan: - Existing tests for det and lu should pass

Differential Revision: D20422879

Pulled By: seemethere

fbshipit-source-id: 8dd7a30b5c31fc5b844e0a11965efd46067e936a
laurentdupin pushed a commit to laurentdupin/pytorch that referenced this pull request Apr 24, 2026
Summary:
Changelog:
- The magma implementation of small singular square batch matrices had a bug that resulted in nan values in the LU factorization result. This has been fixed in MAGMA 2.5.2. This PR removes the existing patch that was a temporary workaround for this bug.
Pull Request resolved: pytorch#35973

Test Plan:
- Existing tests for det and lu should pass

This is a re-submit of pytorch#34357

Differential Revision: D21336552

Pulled By: seemethere

fbshipit-source-id: 9c3b350966913147f1d5811927f3cae10fe620f1
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.

7 participants