PSRoiAlign: SymInt support + meta-implem

[NicolasHug]

Towards #8055

cc @pmeier

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/vision/8058

• 📄 Preview Python docs built from this PR

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

❌ 3 New Failures, 29 Unrelated Failures

As of commit feed2d7 with merge base 8a0b491 ():

NEW FAILURES - The following jobs have failed:

• Prototype tests on Linux / unittests-prototype (3.8, linux.g5.4xlarge.nvidia.gpu, cuda, 11.8) / linux-job (gh)
• Tests / unittests-macos (3.10, macos-12) / macos-job (gh)
• Tests / unittests-windows (3.8, windows.4xlarge, cpu) / windows-job (gh)

FLAKY - The following jobs failed but were likely due to flakiness present on trunk:

• CMake / linux (linux.12xlarge, cpu) / linux-job (gh)
• CMake / linux (linux.g5.4xlarge.nvidia.gpu, cuda, 11.8) / linux-job (gh)
• CMake / macos (macos-12) / macos-job (gh)
• CMake / macos (macos-m1-12) / macos-job (gh)
• CMake / windows (windows.4xlarge, cpu) / windows-job (gh)
• CMake / windows (windows.g5.4xlarge.nvidia.gpu, cuda, 11.8) / windows-job (gh)
• Docs / build / linux-job (gh)
• Lint / c-source / linux-job (gh)
• Lint / python-source-and-configs / linux-job (gh)
• Lint / python-types / linux-job (gh)
• Prototype tests on Linux / unittests-prototype (3.10, linux.12xlarge, cpu) / linux-job (gh)
• Prototype tests on Linux / unittests-prototype (3.11, linux.12xlarge, cpu) / linux-job (gh)
• Prototype tests on Linux / unittests-prototype (3.8, linux.12xlarge, cpu) / linux-job (gh)
• Prototype tests on Linux / unittests-prototype (3.9, linux.12xlarge, cpu) / linux-job (gh)
• Tests / onnx / linux-job (gh)
• Tests / unittests-extended / linux-job (gh)
• Tests / unittests-linux (3.10, linux.12xlarge, cpu) / linux-job (gh)
• Tests / unittests-linux (3.11, linux.12xlarge, cpu) / linux-job (gh)
• Tests / unittests-linux (3.8, linux.12xlarge, cpu) / linux-job (gh)
• Tests / unittests-linux (3.8, linux.g5.4xlarge.nvidia.gpu, cuda, 11.8) / linux-job (gh)
• Tests / unittests-linux (3.9, linux.12xlarge, cpu) / linux-job (gh)
• Tests / unittests-macos (3.11, macos-12) / macos-job (gh)
• Tests / unittests-macos (3.8, macos-12) / macos-job (gh)
• Tests / unittests-macos (3.8, macos-m1-12) / macos-job (gh)
• Tests / unittests-macos (3.9, macos-12) / macos-job (gh)
• Tests / unittests-windows (3.10, windows.4xlarge, cpu) / windows-job (gh)
• Tests / unittests-windows (3.11, windows.4xlarge, cpu) / windows-job (gh)
• Tests / unittests-windows (3.8, windows.g5.4xlarge.nvidia.gpu, cuda, 11.8) / windows-job (gh)
• Tests / unittests-windows (3.9, windows.4xlarge, cpu) / windows-job (gh)

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

[NicolasHug]

CC @zou3519

I just opened pytorch/pytorch#111797 which I believe could be a fix for the problem I'm facing here:

This test (and a bunch of others)

vision/test/test_ops.py

Lines 186 to 189 in 3fb88b3

	@pytest.mark.parametrize("seed", range(10))
	@pytest.mark.parametrize("device", cpu_and_cuda_and_mps())
	@pytest.mark.parametrize("contiguous", (True, False))
	def test_backward(self, seed, device, contiguous, deterministic=False):

is parametrized over cpu, CUDA and MPS, and fails on MPS.

I'm trying to xfail the MPS parametrization as above (lines 10-12), but optests complains with:

E RuntimeError: In failures dict, got test name 'TestPSRoIAlign.test_autograd_registration__test_backward[True-mps-0]'. We parsed this as running test 'test_autograd_registration' on 'test_backward[True-mps-0]', but test_backward[True-mps-0] does not exist on the TestCase 'TestPSRoIAlign]. Maybe you need to change the test name?

The problem is, if I replace TestPSRoIAlign.test_autograd_registration__test_backward[True-mps-0] with TestPSRoIAlign.test_autograd_registration__test_backward in line 10, then I'm getting and "unexpected success":

E torch.testing._internal.optests.generate_tests.OpCheckError: generate_opcheck_tests: Unexpected success for operator torchvision::ps_roi_align on test TestPSRoIAlign.test_autograd_registration__test_backward. This may mean that you have fixed this test failure. Please rerun the test with PYTORCH_OPCHECK_ACCEPT=1 to automatically update the test runner or manually remove the expected failure in the failure dict at /home/nicolashug/dev/vision/test/optests_failures_dict.jsonFor more details, see https://docs.google.com/document/d/1Pj5HRZvdOq3xpFpbEjUZp2hBovhy7Wnxw14m6lF2154/edit

[zou3519]

@NicolasHug can the tests be run under unittest, or are they pytest only?

[NicolasHug]

It's pytest only

[NicolasHug]

The MPS checks are going to fail until pytorch/pytorch#112251 is resovled on torch core's side. I don't know how long it will take, so I'm going to remove the opchecks from this PR (and from the other ones) so that we can still have symint / torch.compile support on these ops (albeit temporarily untested)

[vfdev-5]

Stamp