Skip to content

Source transform for HF RoPE in static attention #12500

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.

Sign up for GitHub

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

Jump to bottom
Merged
merged 1 commit into from
Jul 16, 2025
Merged

Source transform for HF RoPE in static attention #12500

merged 1 commit into from
Jul 16, 2025
+61 −41

Conversation

sxu
Copy link
Contributor

@sxu sxu commented Jul 15, 2025 •
edited
Loading

Summary: LLama style RoPE shuffles/unshuffles the embedding dimension which is not as efficient as HF style RoPE. Add a source transform on static attention so HF style can be used.

Differential Revision: D78353775

@sxu sxu requested review from lucylq and jackzhxng as code owners July 15, 2025 17:27
@pytorch-bot PyTorch Bot
Copy link

pytorch-bot bot commented Jul 15, 2025 •
edited
Loading

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/12500

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

❌ 1 New Failure

As of commit 463f369 with merge base a8070ec (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 Jul 15, 2025
@facebook-github-bot
Copy link
Contributor

This pull request was exported from Phabricator. Differential Revision: D78353775

@github-actions GitHub Actions
Copy link

This PR needs a release notes: label

If your change should be included in the release notes (i.e. would users of this library care about this change?), please use a label starting with release notes:. This helps us keep track and include your important work in the next release notes.

To add a label, you can comment to pytorchbot, for example
@pytorchbot label "release notes: none"

For more information, see
https://github.com/pytorch/pytorch/wiki/PyTorch-AutoLabel-Bot#why-categorize-for-release-notes-and-how-does-it-work.

@sxu sxu requested review from limintang and billmguo July 15, 2025 17:28
@sxu @facebook-github-bot
Source transform for HF RoPE in static attention (pytorch#12500)
463f369 
Summary:

LLama style RoPE shuffles/unshuffles the embedding dimension which is not as efficient as HF style RoPE. Add a source transform on static attention so HF style can be used.

Reviewed By: billmguo

Differential Revision: D78353775
@sxu sxu force-pushed the export-D78353775 branch from 341a047 to 463f369 Compare July 15, 2025 18:06
@facebook-github-bot
Copy link
Contributor

This pull request was exported from Phabricator. Differential Revision: D78353775

@facebook-github-bot facebook-github-bot merged commit 9e05d89 into pytorch:main Jul 16, 2025
103 of 106 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@billmguo billmguo billmguo approved these changes

@lucylq lucylq Awaiting requested review from lucylq lucylq is a code owner

@jackzhxng jackzhxng Awaiting requested review from jackzhxng jackzhxng is a code owner

@limintang limintang Awaiting requested review from limintang

Assignees
No one assigned
Labels
CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. fb-exported
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
3 participants
@sxu @facebook-github-bot @billmguo