Fix executorch/extension/llm/custom_op/...

[mergennachin]

Reviewed By: rascani

Differential Revision: D90888544

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/16665

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[meta-codesync]

@mergennachin has exported this pull request. If you are a Meta employee, you can view the originating Diff in D90888544.

[github-actions]

This PR needs a release notes: label

If your change should be included in the release notes (i.e. would users of this library care about this change?), please use a label starting with release notes:. This helps us keep track and include your important work in the next release notes.

To add a label, you can comment to pytorchbot, for example
@pytorchbot label "release notes: none"

For more information, see
https://github.com/pytorch/pytorch/wiki/PyTorch-AutoLabel-Bot#why-categorize-for-release-notes-and-how-does-it-work.

[Copilot]

Pull request overview

This PR addresses test stability issues by adding threadpool configuration to prevent OMP errors and updating CUDA test requirements to specify minimum compute capability.

Changes:

• Added compute capability check for CUDA/Triton tests requiring CC >= 9.0
• Integrated threadpool reset workaround in test_sdpa_with_kv_cache.py to prevent OMP errors
• Updated TARGETS dependency to include portable_lib for test_sdpa_with_kv_cache

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.

File	Description
extension/llm/custom_ops/test_update_cross_attn_cache.py	Added CUDA compute capability check (CC >= 9.0) for Triton-based tests
extension/llm/custom_ops/test_sdpa_with_kv_cache.py	Added _unsafe_reset_threadpool call to prevent OMP threading errors
extension/llm/custom_ops/TARGETS	Added portable_lib dependency for test_sdpa_with_kv_cache to support threadpool import

---

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

[Copilot]

The TARGETS file is missing a corresponding dependency update for test_update_cross_attn_cache. While test_update_cross_attn_cache.py doesn't currently use _unsafe_reset_threadpool, the file is importing from the same module family and may benefit from the same OMP error workaround that was applied to test_sdpa_with_kv_cache and test_quantized_sdpa. Consider adding the dependency if similar threading issues arise in the future.

[meta-codesync]

@mergennachin has imported this pull request. If you are a Meta employee, you can view this in D90888544.

[Copilot]

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated no new comments.

---

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

[Copilot]

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.

---

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

[Copilot]

The compute capability check may raise an IndexError when CUDA is available but no GPU device is present. Consider adding a device count check or wrapping in a try-except block. For example, the condition should handle the case where torch.cuda.device_count() is 0.

Suggested change

	CUDA_CC_COMPATIBLE = CUDA_AVAILABLE and torch.cuda.get_device_capability()[0] >= 9
	CUDA_CC_COMPATIBLE = (
	CUDA_AVAILABLE
	and torch.cuda.device_count() > 0
	and torch.cuda.get_device_capability()[0] >= 9
	)