Skip to content

Fix bug in dglu xfilter math - #568

Merged
Anerudhan merged 1 commit into
NVIDIA:developfrom
vasunvidia:vrengasamy/dglu_xfilter_fix
Aug 20, 2026
Merged

Fix bug in dglu xfilter math#568
Anerudhan merged 1 commit into
NVIDIA:developfrom
vasunvidia:vrengasamy/dglu_xfilter_fix

Conversation

@vasunvidia

@vasunvidia vasunvidia commented Aug 13, 2026

Copy link
Copy Markdown
Contributor

Before submitting

  • I agree to license this contribution under the terms of LICENSE.txt.
  • I ran pre-commit run and committed any formatting changes.
  • I added GitHub labels: one cat-*, one or more mod-*, and one orig-* (see label list).

Affected area

Summary

Why

Related issues

API and compatibility impact

Testing

Summary by CodeRabbit

  • Bug Fixes
    • Corrected gradient handling for clipped activation values in grouped GEMM operations.
    • Gradients are now preserved only within the supported ranges for both vectorized and scalar execution paths, improving consistency and numerical correctness.

@coderabbitai

coderabbitai Bot commented Aug 13, 2026

Copy link
Copy Markdown
Contributor

Review Change Stack

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info
⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 2ce42381-2bec-4b5a-bdaa-dbeae13173c7

📥 Commits

Reviewing files that changed from the base of the PR and between c622373 and a39a71e.

📒 Files selected for processing (1)
  • python/cudnn/gemm/cutedsl/grouped/dglu/moe_blockscaled_grouped_gemm_dglu_rubin.py

📝 Walkthrough

Walkthrough

The DGeGLU vectorized and scalar gradient paths now use binary clipping masks. x1 gradients are removed above 7. x2 gradients are preserved only within the inclusive range [-7, 7].

Changes

DGeGLU gradient filtering

Layer / File(s) Summary
Apply clipping masks
python/cudnn/gemm/cutedsl/grouped/dglu/moe_blockscaled_grouped_gemm_dglu_rubin.py
The vectorized and scalar paths now use binary validity masks instead of clipped activation values when filtering x1 and x2 gradients.

Estimated code review effort: 2 (Simple) | ~10 minutes

Mergeability Score: ⚪ Minimal · up to a39a7

This localized change adjusts the dglu xfilter math behavior without any identified merge-blocking risk; it is merge-ready after normal checks and review.

Suggested reviewers: anerudhan

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name Status Explanation Resolution
Description check ⚠️ Warning The description contains only the template and omits the summary, rationale, affected area, compatibility impact, and testing details. Complete the required sections with the affected area, change summary, rationale, compatibility impact, related issues, and exact testing results.
✅ Passed checks (4 passed)
Check name Status Explanation
Title check ✅ Passed The title clearly and concisely describes the main change: fixing the dglu xfilter math bug.
Docstring Coverage ✅ Passed No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check ✅ Passed Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check ✅ Passed Check skipped because no linked issues were found for this pull request.
✨ Finishing Touches
🧪 Generate unit tests (beta)
  • Create PR with unit tests

Comment @coderabbitai help to get the list of available commands.

@Anerudhan
Anerudhan self-requested a review August 17, 2026 04:27
@Anerudhan Anerudhan added orig-nv-eng Reported or requested by NVIDIA engineering. mod-cutedsl CuTeDSL kernels, generated kernels, examples, or related integration work. cat-bugfix labels Aug 17, 2026
@Anerudhan Anerudhan added this to the Frontend 1.28.0 milestone Aug 17, 2026
@Anerudhan

Copy link
Copy Markdown
Collaborator

@cudnn-ci-bot run oss

@cudnn-ci-bot

Copy link
Copy Markdown

🚀 Running mirror pipeline

Branch: cudnn-gh/pr-568-a39a71e
Pipeline: 63045557
Targets: oss

@Anerudhan
Anerudhan merged commit 2aab0cd into NVIDIA:develop Aug 20, 2026
1 check passed
@Anerudhan Anerudhan mentioned this pull request Sep 2, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

cat-bugfix mod-cutedsl CuTeDSL kernels, generated kernels, examples, or related integration work. orig-nv-eng Reported or requested by NVIDIA engineering.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants