[Test] Port roi_align (CUDA) to stable ABI.

[adabeyta]

Ports roi_align CUDA/ROCm kernels to the stable ABI.

Notes

• Requires [Stable ABI] Move fastAtomicAdd to torch/headeronly. pytorch#191501.
• Determinism layer change: The CUDA backward's at::globalContext().alertNotDeterministic call can't be made from a stable kernel, so it moves to the Python backward as torch._prims_common.alert_not_deterministic, gated to grad_output.device.type == "cuda". The CPU backward remains unchanged .

Stable-ABI audit (torch-abi-audit)

Ran torch-abi-audit on a CUDA build. The counts don't move with this layer since the CUDA kernel's unstable references are all shared with the remaining legacy ops.

    Package: torchvision
      Torch ABI:   UNSTABLE
      CPython ABI: n/a
      Bundled libs: 3
      -- bundled libs --
        [UNSTABLE] [not-abi3] _C.so            (stable_shim=0,  unstable=103)
        [STABLE  ] [not-abi3] _C_stable.so     (stable_shim=70, unstable=0)
        [STABLE  ] [not-abi3] image_stable.so  (stable_shim=76, unstable=0)

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/vision/9573

• 📄 Preview Python docs built from this PR

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

❗ 1 Active SEVs

There are 1 currently active SEVs. If your PR is affected, please view them below:

• [ALERT ONLY] - System Maintenance

❌ 24 New Failures, 1 Unrelated Failure, 1 Unclassified Failure

As of commit bf0a70a with merge base 0b23e88 ():

NEW FAILURES - The following jobs have failed:

• Build Aarch64 Linux Wheels / pytorch/vision / build-wheel-py3_10-cuda-aarch6412_6-aarch64 (gh)
  RuntimeError: Error compiling objects for extension
• Build Aarch64 Linux Wheels / pytorch/vision / build-wheel-py3_10-cuda-aarch6413_0-aarch64 (gh)
  API rate limit exceeded
• Build Aarch64 Linux Wheels / pytorch/vision / build-wheel-py3_10-cuda-aarch6413_2-aarch64 (gh)
  RuntimeError: Error compiling objects for extension
• Build Aarch64 Linux Wheels / pytorch/vision / upload / upload-wheel-py3_10-cuda-aarch6412_6-aarch64 (gh)
  Unable to download artifact(s): Artifact not found
• Build Aarch64 Linux Wheels / pytorch/vision / upload / upload-wheel-py3_10-cuda-aarch6413_0-aarch64 (gh)
  Unable to download artifact(s): Artifact not found
• Build Aarch64 Linux Wheels / pytorch/vision / upload / upload-wheel-py3_10-cuda-aarch6413_2-aarch64 (gh)
  Unable to download artifact(s): Artifact not found
• Build Linux Wheels / pytorch/vision / build-manywheel-py3_10-cuda12_6 (gh)
• Build Linux Wheels / pytorch/vision / build-manywheel-py3_10-cuda13_0 (gh)
• Build Linux Wheels / pytorch/vision / build-manywheel-py3_10-cuda13_2 (gh)
• Build Linux Wheels / pytorch/vision / build-manywheel-py3_10-rocm7_1 (gh)
  RuntimeError: Error compiling objects for extension
• Build Linux Wheels / pytorch/vision / build-manywheel-py3_10-rocm7_2 (gh)
  RuntimeError: Error compiling objects for extension
• Build Linux Wheels / pytorch/vision / upload / upload-manywheel-py3_10-cuda12_6 (gh)
  Unable to download artifact(s): Artifact not found
• Build Linux Wheels / pytorch/vision / upload / upload-manywheel-py3_10-cuda13_0 (gh)
  API rate limit exceeded
• Build Linux Wheels / pytorch/vision / upload / upload-manywheel-py3_10-cuda13_2 (gh)
  Unable to download artifact(s): Artifact not found
• Build Linux Wheels / pytorch/vision / upload / upload-manywheel-py3_10-rocm7_1 (gh)
  API rate limit exceeded
• Build Linux Wheels / pytorch/vision / upload / upload-manywheel-py3_10-rocm7_2 (gh)
  API rate limit exceeded
• Build Windows Wheels / pytorch/vision / build-wheel-py3_10-cuda12_6 (gh)
  C:\actions-runner\_work\vision\vision\pytorch\vision\torchvision\csrc\ops\cuda\roi_align_kernel.cu(9): fatal error C1083: Cannot open include file: 'torch/headeronly/cuda/AtomicAdd.h': No such file or directory
• Build Windows Wheels / pytorch/vision / build-wheel-py3_10-cuda13_0 (gh)
• Build Windows Wheels / pytorch/vision / build-wheel-py3_10-cuda13_2 (gh)
• Build Windows Wheels / pytorch/vision / upload / upload-wheel-py3_10-cuda12_6 (gh)
  API rate limit exceeded
• Build Windows Wheels / pytorch/vision / upload / upload-wheel-py3_10-cuda13_0 (gh)
  API rate limit exceeded
• Build Windows Wheels / pytorch/vision / upload / upload-wheel-py3_10-cuda13_2 (gh)
  API rate limit exceeded
• CMake / linux (linux.g5.4xlarge.nvidia.gpu, cuda, 12.6) / linux-job (gh)
  RuntimeError: Command docker exec -t 7c4143be7e820a6dc7a240f67b1eeb1a76c3c22137ed654080387f5e3e7f414b /exec failed with exit code 1
• Tests / unittests-linux (3.10, linux.g5.4xlarge.nvidia.gpu, cuda, 12.8) / linux-job (gh)
  RuntimeError: Command docker exec -t 884e56216a6d6b1402e6596aa2261f0be846fb9a529d9851f44102919a0a8d39 /exec failed with exit code 1

UNCLASSIFIED FAILURE - DrCI could not classify the following job because the workflow did not run on the merge base. The failure may be pre-existing on trunk or introduced by this PR:

• .github/workflows/pr-labels.yml (gh) (this job did not run on the merge base, so DrCI cannot tell whether the failure is pre-existing)

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

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

• Tests / unittests-rocm (3.10, linux.rocm.gpu.gfx942.1, rocm, 7.1) / linux-job (gh) (trunk failure)

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