Mark CV-CUDA tests with needs_cuda

[zy1git]

Mark CV-CUDA tests with needs_cuda by:
marks=( pytest.mark.skipif(not CVCUDA_AVAILABLE, reason="CVCUDA is not available"), pytest.mark.needs_cuda, ),

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/vision/9305

• 📄 Preview Python docs built from this PR

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

✅ You can merge normally! (2 Unrelated Failures)

As of commit 8a2c4c7 with merge base 96e7797 ():

BROKEN TRUNK - The following jobs failed but were present on the merge base:

👉 Rebase onto the `viable/strict` branch to avoid these failures

• Build Linux Wheels / pytorch/vision / build-manywheel-py3_10-xpu (gh) (trunk failure)
  ##[error]The operation was canceled.
• Build Linux Wheels / pytorch/vision / upload / upload-manywheel-py3_10-xpu (gh) (trunk failure)
  Unable to download artifact(s): Artifact not found for name: pytorch_vision__3.10_xpu_x86_64

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

[NicolasHug]

We'll need to use these two marks on every single CV-CUDA test, so instead of duplicating this everywhere, let's simply define a global mark on top of the file as

CV_CUDA_TEST = (
                    pytest.mark.skipif(not CVCUDA_AVAILABLE, reason="CVCUDA is not available"),
                    pytest.mark.needs_cuda,
 )

and then use marks=CV_CUDA_TEST for all of these.

Let's also update the previous test classes from @AntoineSimoulin:

vision/test/test_transforms_v2.py

Lines 6859 to 6861 in 96e7797

	@pytest.mark.skipif(not CVCUDA_AVAILABLE, reason="test requires CVCUDA")
	@needs_cuda
	class TestCVDUDAToTensor:

and

vision/test/test_transforms_v2.py

Lines 6797 to 6799 in 96e7797

	@pytest.mark.skipif(not CVCUDA_AVAILABLE, reason="test requires CVCUDA")
	@needs_cuda
	class TestToCVCUDATensor:

to use this mark as well.

I suspect we might be able to also remove the CVCUDA_AVAILABLE global variable now (not sure).

[AntoineSimoulin]

Maybe we can create a @needs_cvcuda decorator (similar to @needs_cuda)? We will also have to modify the pytest_configure function to add this decorator. This way we can identify easily tests requiring cvcuda?

[justincdavis]

This significantly reduces the boilerplate!

I have gone ahead and updated the branch implementing CV-CUDA for the ToDtype transform and I will preemptively update the others as well.

As a note, I defined CV_CUDA_TEST as a list so we can use it with the pytestmark class attribute for TestCVCUDAToTensor and TestToCVCUDATensor. Example here