Fix group topk dispatch for glm5

[junhaha666]

Motivation

Technical Details

Test Plan

Test Result

Submission Checklist

• Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.

[github-actions]

🏷️ CI Guide

Runs automatically on every PR:

• ✅ Pre-checks (submodule verification, code formatting)
• ✅ Aiter op tests (gfx942 + gfx950)
• ✅ Triton tests (only when aiter/ops/triton/** or related paths are changed)

Extended tests (opt-in via labels):

Label	Tests
ci:triton-355	Run Triton tests on MI355 in addition to MI325
ci:sglang	SGLang integration tests
ci:atom	ATOM benchmark (DeepSeek-R1 + GPT-OSS)
ci:vllm	vLLM benchmark
ci:all	All of the above

Add labels via the sidebar or gh pr edit 2611 --add-label <label>

[Copilot]

Pull request overview

Updates the grouped top‑k dispatch logic to avoid using moe_fused_gate for configurations with large experts-per-group (notably affecting GLM5), and adjusts the associated benchmark test to skip the moe_fused_gate comparison in those cases.

Changes:

• Update biased_grouped_topk dispatch to route to the HIP implementation when experts-per-group exceeds 32.
• Gate the moe_fused_gate (“sglang”) performance/correctness comparison in the benchmark test to only run when experts-per-group is ≤ 32.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 4 comments.

File	Description
op_tests/test_moeTopkSoftmax.py	Skips moe_fused_gate comparison when experts-per-group exceeds the supported threshold.
aiter/ops/topk.py	Adjusts dispatch logic to avoid moe_fused_gate when experts-per-group is too large.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.