Vulkan: support zeros padding and align_corners=false in grid_sampler_2d

[msluszniak]

Summary

grid_sampler_2d was gated to a single configuration, the one RIFE's WarpModule uses: bilinear, border padding, align_corners=True. Everything else was rejected by check_grid_sampler_2d_node and fell back to CPU, including torch's own defaults (padding_mode="zeros", align_corners=False), which is what the deformable attention in DETR-derived detectors emits.

Changes

padding_mode and align_corners become specialization constants, so the four supported combinations share one shader variant and each pipeline still compiles with the branches folded away. Reflection padding and nearest/bicubic interpolation remain unsupported and still fall back.

The two padding paths are not interchangeable: border clamps the sample coordinate, which also pins the interpolation weights at the edge, while zeros must not clamp, because the weights stay as computed and each out-of-range corner contributes a zero value instead.

Test

test_vulkan_backend_grid_sampler_2d covers all four combinations with a grid deliberately overshooting to +/-1.6, so the zeros and border paths actually diverge; an in-range grid passes even with the padding branch broken.

Also checked end to end on a Mali-G76 against eager PyTorch, max abs error 5.96e-08 (float32 eps) for each of the four configurations.

On an RF-DETR keypoint model this moves all 8 grid_sampler_2d calls off the host, taking the graph from 7 delegate blobs and 11 host calls to 2 and 3.

cc @SS-JIA @manuelcandales @digantdesai @cbilgin

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/22803

• 📄 Preview Python docs built from this PR

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

⚠️ 15 Awaiting Approval

As of commit 3433a4f with merge base 14dab10 ():

AWAITING APPROVAL - The following workflows need approval before CI can run:

• Build documentation (gh)
• Build Presets (gh)
• Lint (gh)
• periodic (gh)
• pull (gh)
• Test ARM Backend (gh)
• Test CoreML Backend (gh)
• Test Cortex-M Backend (gh)
• Test NXP Backend (gh)
• Test QNN Backend (gh)
• Test QNN Windows MSVC build (gh)
• Test Vulkan Backend (gh)
• Test Vulkan Backend (specialized runners) (gh)
• Test WebGPU Backend (gh)
• Test XNNPACK Backend (gh)

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

[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.