[TEST][XPU][CI] disable xpu tests for nonexistent input norm kernels by microslaw · Pull Request #59568 · vllm-project/vllm

microslaw · 2026-10-01T07:57:39Z

Overview

#59195 introduced override of FusedMMInputNorm.enabled(), now it always returns True; This forces tests of FusedMMInputNorm on float based XPU kernels that do not exist. This pr disables those tests.

Claims

Stop testing missing kernels, keep the working uint8 path

Validation

python3 -m pytest -v -rs -p vllm._xpu_ops tests/model_executor/layers/test_mm_input_norm.py
tee /results/mmnorm/test_mm_input_norm.log
W1001 07:35:33.549000 200 torch/_opaque_base.py:6] torch._opaque_base is deprecated, use torch._custom_class_base instead
W1001 07:35:33.550000 200 torch/_library/opaque_object.py:288] register_opaque_type is deprecated, use register_custom_class instead
W1001 07:35:33.551000 200 torch/_library/opaque_object.py:211] typ='value' is deprecated, use typ='constant' instead
============================= test session starts ==============================
platform linux -- Python 3.12.3, pytest-9.1.1, pluggy-1.6.0 -- /opt/venv/bin/python3
cachedir: .pytest_cache
hypothesis profile 'default'
rootdir: /workspace/local
configfile: pyproject.toml
plugins: anyio-4.14.2, shard-0.1.2, asyncio-1.4.0, timeout-2.4.0, rerunfailures-16.6, schemathesis-4.24.3, forked-1.7.5, hypothesis-6.165.10, cov-7.1.0
asyncio: mode=Mode.STRICT, debug=False, asyncio_default_fixture_loop_scope=None, asyncio_default_test_loop_scope=function
collecting ... collected 29 items
Running 29 items in this shard: tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormModule::test_matches_reference_above_cudnn_grid_limit, tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormModule::test_identity_passthrough, tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormDtypes::test_dtype_combinations[in_dtype0-out_dtype0], tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormDtypes::test_dtype_combinations[in_dtype1-out_dtype1], tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormDtypes::test_dtype_combinations[in_dtype2-out_dtype2], tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormDtypes::test_dtype_combinations[in_dtype3-out_dtype3], tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormDtypes::test_dtype_combinations[in_dtype4-out_dtype4], tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormDtypes::test_dtype_combinations[in_dtype5-out_dtype5], tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormDtypes::test_dtype_combinations[in_dtype6-out_dtype6], tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormDtypes::test_dtype_combinations[in_dtype7-out_dtype7], tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormShapes::test_channel_variants[1], tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormShapes::test_channel_variants[3], tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormShapes::test_channel_variants[4], tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormShapes::test_block_boundaries[1], tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormShapes::test_block_boundaries[2048], tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormInputHandling::test_non_contiguous_input_matches_reference, tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormKernel::test_channel_boundary_crossing[1-1-1], tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormKernel::test_channel_boundary_crossing[1-3-1], tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormKernel::test_channel_boundary_crossing[3-3-7], tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormKernel::test_channel_boundary_crossing[2-8-63], tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormKernel::test_channel_boundary_crossing[4-3-1000], tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormKernel::test_non_contiguous_inputs_materialized, tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormKernel::test_rejects_mismatched_output_shape, tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormConstruction::test_input_norm_is_always_enabled, tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormConstruction::test_identity_config_buffers, tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormConstruction::test_non_identity_buffers, tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormConstruction::test_is_identity_params, tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormConstruction::test_buffers_initialized_on_default_device[image_mean0-image_std0-1.0], tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormConstruction::test_buffers_initialized_on_default_device[image_mean1-image_std1-0.00392156862745098]

tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormModule::test_matches_reference_above_cudnn_grid_limit SKIPPED [  3%]
tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormModule::test_identity_passthrough PASSED [  6%]
tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormDtypes::test_dtype_combinations[in_dtype0-out_dtype0] SKIPPED [ 10%]
tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormDtypes::test_dtype_combinations[in_dtype1-out_dtype1] SKIPPED [ 13%]
tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormDtypes::test_dtype_combinations[in_dtype2-out_dtype2] SKIPPED [ 17%]
tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormDtypes::test_dtype_combinations[in_dtype3-out_dtype3] SKIPPED [ 20%]
tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormDtypes::test_dtype_combinations[in_dtype4-out_dtype4] SKIPPED [ 24%]
tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormDtypes::test_dtype_combinations[in_dtype5-out_dtype5] SKIPPED [ 27%]
tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormDtypes::test_dtype_combinations[in_dtype6-out_dtype6] PASSED [ 31%]
tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormDtypes::test_dtype_combinations[in_dtype7-out_dtype7] PASSED [ 34%]
tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormShapes::test_channel_variants[1] SKIPPED [ 37%]
tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormShapes::test_channel_variants[3] SKIPPED [ 41%]
tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormShapes::test_channel_variants[4] SKIPPED [ 44%]
tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormShapes::test_block_boundaries[1] SKIPPED [ 48%]
tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormShapes::test_block_boundaries[2048] SKIPPED [ 51%]
tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormInputHandling::test_non_contiguous_input_matches_reference SKIPPED [ 55%]
tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormKernel::test_channel_boundary_crossing[1-1-1] PASSED [ 58%]
tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormKernel::test_channel_boundary_crossing[1-3-1] PASSED [ 62%]
tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormKernel::test_channel_boundary_crossing[3-3-7] PASSED [ 65%]
tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormKernel::test_channel_boundary_crossing[2-8-63] PASSED [ 68%]
tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormKernel::test_channel_boundary_crossing[4-3-1000] PASSED [ 72%]
tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormKernel::test_non_contiguous_inputs_materialized PASSED [ 75%]
tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormKernel::test_rejects_mismatched_output_shape PASSED [ 79%]
tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormConstruction::test_input_norm_is_always_enabled PASSED [ 82%]
tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormConstruction::test_identity_config_buffers PASSED [ 86%]
tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormConstruction::test_non_identity_buffers PASSED [ 89%]
tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormConstruction::test_is_identity_params PASSED [ 93%]
tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormConstruction::test_buffers_initialized_on_default_device[image_mean0-image_std0-1.0] PASSED [ 96%]
tests/model_executor/layers/test_mm_input_norm.py::TestFusedMMInputNormConstruction::test_buffers_initialized_on_default_device[image_mean1-image_std1-0.00392156862745098] PASSED [100%]

=========================== short test summary info ============================
SKIPPED [1] tests/model_executor/layers/test_mm_input_norm.py:101: XPU fused kernel only accepts uint8 input
SKIPPED [6] tests/model_executor/layers/test_mm_input_norm.py:139: XPU fused kernel only accepts uint8 input
SKIPPED [3] tests/model_executor/layers/test_mm_input_norm.py:166: XPU fused kernel only accepts uint8 input
SKIPPED [2] tests/model_executor/layers/test_mm_input_norm.py:180: XPU fused kernel only accepts uint8 input
SKIPPED [1] tests/model_executor/layers/test_mm_input_norm.py:196: XPU fused kernel only accepts uint8 input
======================= 16 passed, 13 skipped in 13.28s ========================
+ python3 -m pytest -v -rs tests/models/multimodal/generation/test_qwen2_vl.py -k embeddings_input
+ tee /results/mmnorm/test_qwen2_vl.log
============================= test session starts ==============================
platform linux -- Python 3.12.3, pytest-9.1.1, pluggy-1.6.0 -- /opt/venv/bin/python3
cachedir: .pytest_cache
hypothesis profile 'default'
rootdir: /workspace/local
configfile: pyproject.toml
plugins: anyio-4.14.2, shard-0.1.2, asyncio-1.4.0, timeout-2.4.0, rerunfailures-16.6, schemathesis-4.24.3, forked-1.7.5, hypothesis-6.165.10, cov-7.1.0
asyncio: mode=Mode.STRICT, debug=False, asyncio_default_fixture_loop_scope=None, asyncio_default_test_loop_scope=function
collecting ... collected 3 items
Running 3 items in this shard: tests/models/multimodal/generation/test_qwen2_vl.py::test_qwen2_vl_image_embeddings_input[10-128-half-Qwen/Qwen2-VL-2B-Instruct], tests/models/multimodal/generation/test_qwen2_vl.py::test_qwen2_vl_multiple_image_embeddings_input[10-128-half-Qwen/Qwen2-VL-2B-Instruct], tests/models/multimodal/generation/test_qwen2_vl.py::test_qwen2_vl_video_embeddings_input[10-128-half-Qwen/Qwen2-VL-2B-Instruct]

tests/models/multimodal/generation/test_qwen2_vl.py::test_qwen2_vl_image_embeddings_input[10-128-half-Qwen/Qwen2-VL-2B-Instruct] SKIPPED [ 33%]
tests/models/multimodal/generation/test_qwen2_vl.py::test_qwen2_vl_multiple_image_embeddings_input[10-128-half-Qwen/Qwen2-VL-2B-Instruct] SKIPPED [ 66%]
tests/models/multimodal/generation/test_qwen2_vl.py::test_qwen2_vl_video_embeddings_input[10-128-half-Qwen/Qwen2-VL-2B-Instruct] SKIPPED [100%]

=============================== warnings summary ===============================
<frozen importlib._bootstrap>:488
  <frozen importlib._bootstrap>:488: DeprecationWarning: builtin type SwigPyPacked has no __module__ attribute

<frozen importlib._bootstrap>:488
  <frozen importlib._bootstrap>:488: DeprecationWarning: builtin type SwigPyObject has no __module__ attribute

-- Docs: https://docs.pytest.org/en/stable/how-to/capture-warnings.html
=========================== short test summary info ============================
SKIPPED [1] tests/models/multimodal/generation/test_qwen2_vl.py:343: XPU fused input norm only accepts uint8 input
SKIPPED [1] tests/models/multimodal/generation/test_qwen2_vl.py:382: XPU fused input norm only accepts uint8 input
SKIPPED [1] tests/models/multimodal/generation/test_qwen2_vl.py:422: XPU fused input norm only accepts uint8 input
======================== 3 skipped, 2 warnings in 0.13s ========================

Details

Pull Request Checklist

I used vLLM's /pr-checklist skill. (Mandatory for agents, optional for humans).
AI assistance was used during the creation of this PR.
Design Fit: Minimizes impact on core components, reuses existing functionality, and justifies added complexity.
Testing and Validation: Validates the change and ensures any added tests are meaningful and reliable, with CI coverage or documented CI resource constraints and validation performed outside CI.
Code Quality and Style: Keeps code and comments clear and concise, and updates relevant documentation and examples.
Pull Request Contents: Includes a brief summary and relevant links, supports claims with evidence, explains root causes and implementation trade-offs, and follows the contributing guide.

claude

Claude Code Review

This pull request is from a fork — automated review is disabled. A repository maintainer can comment @claude review to run a one-time review.

microslaw · 2026-10-01T08:02:00Z

/pr-checklist

microslaw · 2026-10-01T08:05:12Z

@jikunshang could you take a look?

jbyczkow

LGTM

Signed-off-by: Milosz Grunwald <milosz.grunwald@intel.com>

microslaw · 2026-10-02T06:52:09Z

as per request from @jikunshang I've changed the implementation from python decorator to pytest --deselect ...

jikunshang · 2026-10-02T07:53:49Z

/ci run --allow-stale

github-actions · 2026-10-02T07:54:09Z

✅ Triggered Buildkite CI #92575 for commit 373712b9908d.

⚠️ This PR is 2 commits behind upstream main. Running CI at your own risk because --allow-stale was requested; outdated CI configuration may cause failures. Before merging, merge or rebase onto the latest main, then rerun /ci run on the latest PR commit.

Signed-off-by: Milosz Grunwald <milosz.grunwald@intel.com>

jikunshang · 2026-10-02T14:04:25Z

/ci run

github-actions · 2026-10-02T14:04:48Z

✅ Triggered Buildkite CI #92635 for commit 6543a8016dc1.

vllm-agent · 2026-10-02T14:52:40Z

CI selector (shadow): 0 test steps (0 jobs) instead of 0 (0 jobs)

Shadow mode: this changes nothing about what CI runs. It shows what the evidence-based selector would pick for this PR, next to today's rules. How it works.

Feedback welcome: reply here if it would skip a step this change needs, or runs something unrelated.

steps (jobs)	Today's rules	Selector	Would skip	Would add
NVIDIA, CPU and others	0 (0)	0 (0)	0 (0)	0 (0)
AMD mirrors	0 (0)	0 (0)	0 (0)	0 (0)

Selector would run (0)

none

Would skip (today's rules run them) (0)

none

Would add (today's rules do not run them) (0)

none

AMD mirrors: would skip (0)

none

AMD mirrors: would add (0)

none

_{2 changed files · base c91dccc080 · head 6543a8016d · Python record: build 92561 at 4056c8ac1f · kernel record: table 4056c8a (build 92561), map 4056c8a · not counted: 10 build steps, 5 A100 steps the generator no longer emits, 0 optional steps the selector would also run}

…kernels (vllm-project#59568)" This reverts commit 8c3b720. Signed-off-by: Yan Ma <yan.ma@intel.com>

microslaw requested review from AndreasKaratzas, DarkLight1337 and ywang96 as code owners October 1, 2026 07:57

claude Bot reviewed Oct 1, 2026

View reviewed changes

mergify Bot added multi-modality Related to multi-modality (#4194) qwen Related to Qwen models intel-gpu Related to Intel GPU labels Oct 1, 2026

jbyczkow approved these changes Oct 1, 2026

View reviewed changes

test: disable xpu ci tests for float input norm kernels

1c259e2

Signed-off-by: Milosz Grunwald <milosz.grunwald@intel.com>

microslaw force-pushed the microslaw/fix-xpu-multimodal-input-norm branch from 5f299f8 to 1c259e2 Compare October 2, 2026 06:47

microslaw requested review from Harry-Chen and khluu as code owners October 2, 2026 06:47

mergify Bot added the ci/build label Oct 2, 2026

jikunshang approved these changes Oct 2, 2026

View reviewed changes

test: disable qwen vl tests

efe65c7

Signed-off-by: Milosz Grunwald <milosz.grunwald@intel.com>

microslaw force-pushed the microslaw/fix-xpu-multimodal-input-norm branch from 373712b to efe65c7 Compare October 2, 2026 11:41

microslaw and others added 2 commits October 2, 2026 13:42

Merge branch 'main' into microslaw/fix-xpu-multimodal-input-norm

0d73e10

Merge branch 'main' into microslaw/fix-xpu-multimodal-input-norm

6543a80

jikunshang merged commit 8c3b720 into vllm-project:main Oct 3, 2026
20 checks passed

yma11 mentioned this pull request Oct 3, 2026

[XPU] fix fused_input_norm for MM #59865

Merged

6 tasks

Hotragn mentioned this pull request Oct 4, 2026

[Bugfix][Multimodal] Honor modality-scoped mm_processor_kwargs in every model that reads them #54527

Open

yma11 added a commit to yma11/vllm that referenced this pull request Oct 9, 2026

Revert "[TEST][XPU][CI] disable xpu tests for nonexistent input norm …

04da351

…kernels (vllm-project#59568)" This reverts commit 8c3b720. Signed-off-by: Yan Ma <yan.ma@intel.com>

Navigation Menu

Repository navigation

[TEST][XPU][CI] disable xpu tests for nonexistent input norm kernels - #59568

[TEST][XPU][CI] disable xpu tests for nonexistent input norm kernels#59568
jikunshang merged 4 commits into
vllm-project:mainfrom
microslaw:microslaw/fix-xpu-multimodal-input-norm

microslaw commented Oct 1, 2026 •

edited

Loading

Uh oh!

claude Bot left a comment

Uh oh!

microslaw commented Oct 1, 2026

Uh oh!

microslaw commented Oct 1, 2026

Uh oh!

jbyczkow left a comment

Uh oh!

microslaw commented Oct 2, 2026

Uh oh!

jikunshang commented Oct 2, 2026

Uh oh!

github-actions Bot commented Oct 2, 2026

Uh oh!

jikunshang commented Oct 2, 2026

Uh oh!

github-actions Bot commented Oct 2, 2026

Uh oh!

vllm-agent commented Oct 2, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Conversation

microslaw commented Oct 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Overview

Claims

Validation

Details

Uh oh!

claude Bot left a comment

Choose a reason for hiding this comment

Claude Code Review

Uh oh!

microslaw commented Oct 1, 2026

Uh oh!

microslaw commented Oct 1, 2026

Uh oh!

jbyczkow left a comment

Choose a reason for hiding this comment

Uh oh!

microslaw commented Oct 2, 2026

Uh oh!

jikunshang commented Oct 2, 2026

Uh oh!

github-actions Bot commented Oct 2, 2026

Uh oh!

jikunshang commented Oct 2, 2026

Uh oh!

github-actions Bot commented Oct 2, 2026

Uh oh!

vllm-agent commented Oct 2, 2026

CI selector (shadow): 0 test steps (0 jobs) instead of 0 (0 jobs)

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

microslaw commented Oct 1, 2026 •

edited

Loading