fix stride check error in fused_qk_norm_group_quant

[yzhou103]

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 2637 --add-label <label>

[Copilot]

Pull request overview

This PR updates fused_qk_rmsnorm_group_quant to correctly support non-contiguous (but last-dimension-contiguous) Q/K/residual inputs by replacing overly strict contiguity checks with stride-based validation, and it fixes scale indexing for transpose_scale by passing explicit row/column strides into the kernel.

Changes:

• Relax input/output layout requirements from fully contiguous to stride(1)==1 with sane leading-dimension stride checks.
• Make q_out_scale writes use explicit (row_stride, col_stride) so transpose_scale is handled via correct addressing rather than shape-only assumptions.
• Update the op test to generate realistic strided Q/K/residual views via torch.split, exercising the new stride checks.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
csrc/kernels/fused_qk_rmsnorm_group_quant.cu	Replaces .is_contiguous() gating with stride-based checks and threads scale strides into the kernel to fix transpose_scale addressing.
op_tests/test_fused_qk_rmsnorm_group_quant.py	Constructs strided Q/K/residual inputs (via split-views) and adjusts scale buffer allocation for transpose_scale testing.

---

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