Skip to content

fix missing attribute in Rubin grouped dGLU - #625

Merged
sraman-rgb merged 1 commit into
NVIDIA:developfrom
filaretov:hfilaretov/fix-moe-blockscaled-grouped-gemm-dglu-rubin
Aug 17, 2026
Merged

fix missing attribute in Rubin grouped dGLU#625
sraman-rgb merged 1 commit into
NVIDIA:developfrom
filaretov:hfilaretov/fix-moe-blockscaled-grouped-gemm-dglu-rubin

Conversation

@filaretov

@filaretov filaretov commented Aug 17, 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

  • Python API or bindings

Summary

A variable wasn't assigned to an attribute, so we hit a missing attribute later in the code.

Traceback snippet:

15:   File "/usr/local/lib/python3.12/dist-packages/cudnn/gemm/cutedsl/grouped/dglu/moe_blockscaled_grouped_gemm_dglu_rubin.py", line 2126, in kernel
15:     if cutlass.const_expr(self.use_single_group_runtime_offsets):
15:                           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
15: AttributeError: 'BlockScaledMoEGroupedGemmDgluKernel' object has no attribute 'use_single_group_runtime_offsets'

Related PRs

Related to #588 and #590.

Testing

I ran our failing training runs with this patch and they worked.

Summary by CodeRabbit

  • Refactor
    • Preserved the single-group runtime offsets configuration for use during kernel execution.

Signed-off-by: Hristo Filaretov <hfilaretov@nvidia.com>
@coderabbitai

coderabbitai Bot commented Aug 17, 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: 809cdb6b-128d-4f6a-a1dc-be49119839d3

📥 Commits

Reviewing files that changed from the base of the PR and between 491805e and 09b93c6.

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

Included review availability: Your plan includes up to 12 reviews per rolling hour; 11 remain after this review.


📝 Walkthrough

Walkthrough

The constructor now stores use_single_group_runtime_offsets on the kernel instance for later kernel execution.

Changes

Runtime offset configuration

Layer / File(s) Summary
Constructor configuration
python/cudnn/gemm/cutedsl/grouped/dglu/moe_blockscaled_grouped_gemm_dglu_rubin.py
The constructor preserves use_single_group_runtime_offsets as an instance attribute.

Estimated code review effort: 1 (Trivial) | ~2 minutes

Merge Risk: ⚪ Minimal · up to 09b93

This change assigns the missing runtime attribute in the Rubin grouped dGLU path, preventing the reported attribute error. No actionable merge-blocking risk remains beyond normal checks and review.

Suggested reviewers: sraman-rgb

🚥 Pre-merge checks | ✅ 5
✅ Passed checks (5 passed)
Check name Status Explanation
Title check ✅ Passed The title clearly identifies the missing attribute fix in the Rubin grouped dGLU implementation.
Description check ✅ Passed The description includes the required checklist, affected area, summary, related work, and testing, but omits separate Why and API compatibility sections.
Docstring Coverage ✅ Passed Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.
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.

@filaretov

Copy link
Copy Markdown
Contributor Author

I can't seem to add labels, but these are relevant: orig-nv-eng, cat-bugfix, and mod-cutedsl.

@sraman-rgb

Copy link
Copy Markdown
Collaborator

@cudnn-ci-bot run oss

@sraman-rgb sraman-rgb self-assigned this Aug 17, 2026
@cudnn-ci-bot

Copy link
Copy Markdown

🚀 Running mirror pipeline

Branch: cudnn-gh/pr-625-09b93c6
Pipeline: 63126268
Targets: oss

@sraman-rgb
sraman-rgb merged commit 6c39f8b into NVIDIA:develop Aug 17, 2026
1 check passed
hxbai pushed a commit to hxbai/cudnn-frontend that referenced this pull request Aug 18, 2026
Signed-off-by: Hristo Filaretov <hfilaretov@nvidia.com>
@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

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants