[Vulkan] Match ATen special values in pow.Tensor_Scalar - #23251

Draft

mergennachin wants to merge 8 commits into

gh/mergennachin/34/headfrom

gh/mergennachin/35/head

Commits on Sep 29, 2026

[INITIAL] Recreate the Vulkan transformer and conformance stack with ghstack
mergennachin
committed
0dfae81
[UPDATE] Add FP16 exponent and intermediate-rounding regression tests
mergennachin
committed
a4d0f14

Commits on Oct 3, 2026

[UPDATE] Rebase remaining Vulkan stack onto main after parts 1–3 landed
mergennachin
committed
1f058df
[UPDATE] Cover exact FP16 halfway rounding in texture reductions
mergennachin
committed
6280b7c
[UPDATE] Match arg-reduction partitioning to the buffer-only last-dimension kernel; rebase after PR 23243 landed
mergennachin
committed
443b4c4
[UPDATE] Add exact FP16 zero/subnormal and overflow-threshold coverage
mergennachin
committed
71fb388
[UPDATE] Exercise int32 reduction shaders directly beyond the removed clamp
mergennachin
committed
cfa1852
[UPDATE] Support any.dim without keepdim using texture reduction and GPU repacking
mergennachin
committed
c0a14d0