diff --git a/dsv4/kernels/attention/fmha_sm100.cpp b/dsv4/kernels/attention/fmha_sm100.cpp
index a4188a8c..6c64144a 100644
--- a/dsv4/kernels/attention/fmha_sm100.cpp
+++ b/dsv4/kernels/attention/fmha_sm100.cpp
@@ -5,7 +5,6 @@
 #include "fmha_sm100.cuh"
 #include <ATen/ATen.h>
 #include <torch/extension.h>
-#include <cuda_runtime.h>
 
 namespace dsv4 {
 namespace kernels {
diff --git a/dsv4/kernels/attention/fmha_sm100.cuh b/dsv4/kernels/attention/fmha_sm100.cuh
index bd69d497..06683854 100644
--- a/dsv4/kernels/attention/fmha_sm100.cuh
+++ b/dsv4/kernels/attention/fmha_sm100.cuh
@@ -29,16 +29,18 @@
 
 #pragma once
 
-#include <cuda.h>
 #include <cuda_runtime.h>
-#include <cuda_fp8.h>
 #include <cuda_bf16.h>
+
+// CUTLASS C++ includes (CUDA device code only)
+#if defined(__CUDA_ARCH__)
 #include <cutlass/cutlass.h>
 #include <cute/arch/mma_sm100_umma.hpp>
 #include <cute/arch/copy_sm100.hpp>
 #include <cute/arch/copy_sm100_tma.hpp>
 #include <cute/arch/tmem_allocator_sm100.hpp>
 #include <cute/arch/cluster_sm100.hpp>
+#endif
 
 namespace dsv4 {
 namespace kernels {