Skip to content

test(activation): skip fp4 silu_and_mul_quant on non-gfx950 archs - #4925

Open
zjin-lcf wants to merge 1 commit into
ROCm:mainfrom
zjin-lcf:fix/activation-fp4-skip-gfx942
Open

test(activation): skip fp4 silu_and_mul_quant on non-gfx950 archs#4925
zjin-lcf wants to merge 1 commit into
ROCm:mainfrom
zjin-lcf:fix/activation-fp4-skip-gfx942

Conversation

@zjin-lcf

Copy link
Copy Markdown

FP4 (e2m1) output relies on CDNA4-only MFMA and is compiled in only for gfx950/gfx1250 (guarded by #if defined(__Float4_e2m1fn_x2) in csrc/kernels/activation_kernels.cu). On gfx942 (MI300A/MI300X) the fp4 branch is compiled out, so the dispatch hits AITER_CHECK(false) and aborts the whole test process. Gate the fp4 case on the runtime gfx (matching the mxfp4 skip idiom in test_moe_ep.py) so it is skipped cleanly, letting the remaining activation tests run.

FP4 (e2m1) output relies on CDNA4-only MFMA and is compiled in only for
gfx950/gfx1250 (guarded by #if defined(__Float4_e2m1fn_x2) in
csrc/kernels/activation_kernels.cu). On gfx942 (MI300A/MI300X) the fp4
branch is compiled out, so the dispatch hits AITER_CHECK(false) and
aborts the whole test process. Gate the fp4 case on the runtime gfx
(matching the mxfp4 skip idiom in test_moe_ep.py) so it is skipped
cleanly, letting the remaining activation tests run.

Co-authored-by: Cursor <cursoragent@cursor.com>
@zjin-lcf
zjin-lcf requested a review from a team August 23, 2026 03:34
@github-actions

Copy link
Copy Markdown
Contributor

🏷️ CI Guide

Runs automatically on every PR:

  • ✅ Pre-checks (submodule verification, code formatting)
  • ✅ Aiter op tests (gfx942 + gfx950)
  • ✅ Triton tests on MI35X (only when aiter/ops/triton/** or related paths are changed)

Extended tests (opt-in via labels):

Label Tests
ci:gfx1250-ffm-triton Run the five-shard gfx1250 FFM Triton test suite
ci:triton-300x Run an additional Triton test job on MI300X in PRs; main branch always runs both MI35X and MI300X
ci:sglang SGLang integration tests: DeepSeek-R1-MXFP4 accuracy, Qwen 3.5 accuracy
ci:atom ATOM benchmark: DeepSeek-R1-0528, GPT-OSS-120B
ci:atom_full ATOM accuracy suite for PR and main models from ATOM models_accuracy.json
ci:vllm vLLM benchmark: GPT-OSS-120B, DeepSeek-R1-0528, Kimi-K2.5
ci:all All standard extended tests (excludes ci:atom_full)

Only add ci:atom_full for FlyDSL or Triton upgrades.
Add labels via the sidebar or gh pr edit 4925 --add-label <label>

PR title tags:
Component tags ([Triton/Gluon], [HIP], [CK], [ASM], ...) are added to the PR title automatically from the changed files and re-synced on every push — change-type tags like [fix]/[Perf] and op tags like [MLA] are left untouched. Add the no-auto-title label to opt this PR out of title tagging.

@zufayu
zufayu requested a review from yzhou103 August 24, 2026 01:43
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.

1 participant