Skip to content

fix: batch norm issue encountered in RAFT #3758

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
zewenli98 merged 2 commits into from
Aug 21, 2025
Merged

fix: batch norm issue encountered in RAFT #3758

zewenli98 merged 2 commits into from
Aug 21, 2025

Conversation

@zewenli98
Copy link
Collaborator

@zewenli98 zewenli98 commented Aug 7, 2025

Description

Pytorch decomposes torch.ops.aten._native_batch_norm_legit.no_stats into a bunch of aten ops, which makes it way slow. See the issue below for details.

Fixes #3731

Type of change

  • Bug fix (non-breaking change which fixes an issue)

Checklist:

  • My code follows the style guidelines of this project (You can use the linters)
  • I have performed a self-review of my own code
  • I have commented my code, particularly in hard-to-understand areas and hacks
  • I have made corresponding changes to the documentation
  • I have added tests to verify my fix or my feature
  • New and existing unit tests pass locally with my changes
  • I have added the relevant labels to my PR in so that relevant reviewers are notified

@zewenli98 zewenli98 self-assigned this Aug 7, 2025
@meta-cla meta-cla bot added the cla signed label Aug 7, 2025
@github-actions github-actions bot added component: lowering Issues re: The lowering / preprocessing passes component: conversion Issues re: Conversion stage component: converters Issues re: Specific op converters component: api [Python] Issues re: Python API component: dynamo Issues relating to the `torch.compile` or `torch._dynamo.export` paths labels Aug 7, 2025
@zewenli98 zewenli98 mentioned this pull request Aug 7, 2025
Closed
narendasan
narendasan reviewed Aug 7, 2025
View reviewed changes
Copy link
Collaborator

@narendasan narendasan left a comment

Choose a reason for hiding this comment

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

looks fine, is there a reason we use numpy instead of torch?

@zewenli98
Copy link
Collaborator Author

zewenli98 commented Aug 7, 2025 •
edited
Loading

looks fine, is there a reason we use numpy instead of torch?

because it would be FakeTensor

@zewenli98 zewenli98 requested a review from narendasan August 9, 2025 00:00
narendasan
narendasan approved these changes Aug 12, 2025
View reviewed changes
Copy link
Collaborator

@narendasan narendasan left a comment

Choose a reason for hiding this comment

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

Lgtm

@zewenli98 zewenli98 merged commit 844cfd4 into main Aug 21, 2025
81 of 83 checks passed
@zewenli98 zewenli98 deleted the fix_raft branch January 12, 2026 21:12
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@narendasan narendasan narendasan approved these changes

@peri044 peri044 Awaiting requested review from peri044

@cehongwang cehongwang Awaiting requested review from cehongwang

Assignees

@zewenli98 zewenli98

Labels

cla signed component: api [Python] Issues re: Python API component: conversion Issues re: Conversion stage component: converters Issues re: Specific op converters component: dynamo Issues relating to the `torch.compile` or `torch._dynamo.export` paths component: lowering Issues re: The lowering / preprocessing passes

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

🐛 [Bug] perf gap reduce on RAFT

3 participants

@zewenli98 @narendasan