[10c/n] Migrate MoE kernels to torch stable ABI (#44565)

Signed-off-by: Chris Leonard <chleonar@redhat.com>
Co-authored-by: Shengqi Chen <harry-chen@outlook.com>
This commit is contained in:
Chris Leonard
2026-06-10 23:02:26 -07:00
committed by GitHub
co-authored by Shengqi Chen
parent 2f2c5cf4f1
commit 6e64c1bab1
34 changed files with 1296 additions and 1098 deletions
-19
View File
@@ -130,25 +130,6 @@ TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, ops) {
"Tensor? qzeros_or_none, bool inplace) -> Tensor");
// conditionally compiled so impl registrations are in source file
#endif
#ifndef USE_ROCM
// Expert-specialization mxfp8 blockscaled grouped quantization (SM100+).
ops.def(
"mxfp8_experts_quant("
" Tensor input, Tensor problem_sizes, Tensor expert_offsets,"
" Tensor blockscale_offsets, Tensor! quant_output, Tensor! scale_factor)"
" -> ()");
// conditionally compiled so impl registration is in source file
// Expert-specialization mxfp8 blockscaled grouped GEMM (SM100+).
ops.def(
"cutlass_mxfp8_grouped_mm("
" Tensor a, Tensor b, Tensor sfa, Tensor sfb, Tensor! out,"
" Tensor problem_sizes, Tensor expert_offsets, Tensor blockscale_offsets)"
" -> ()");
// conditionally compiled so impl registration is in source file
#endif
}