[5/n] Migrate CUTLASS MLA, hadamard, awq, allspark and DSV3 fused a gemm to torch stable ABI (continued) (#42339)

Signed-off-by: Mikayla Gawarecki <mikaylagawarecki@gmail.com>
Co-authored-by: Mikayla Gawarecki <mikaylagawarecki@gmail.com>
This commit is contained in:
Chris Leonard
2026-05-13 07:24:39 +00:00
committed by GitHub
co-authored by Mikayla Gawarecki
parent 503697c9ce
commit 85b2fecab7
20 changed files with 525 additions and 407 deletions
+8 -4
View File
@@ -2,10 +2,14 @@
#ifndef _marlin_cuh
#define _marlin_cuh
#include <torch/all.h>
#include <ATen/cuda/CUDAContext.h>
#include <c10/cuda/CUDAGuard.h>
// These torch headers are only needed by non-stable callers (e.g. ops.cu).
// Guard them so that stable ABI targets can still include marlin.cuh
// for Vec, constants, and cp_async helpers without pulling in torch/all.h.
#ifndef TORCH_TARGET_VERSION
#include <torch/all.h>
#include <ATen/cuda/CUDAContext.h>
#include <c10/cuda/CUDAGuard.h>
#endif
#include <cuda.h>
#include <cuda_fp16.h>
#include <cuda_runtime.h>