diff --git a/dsv4/kernels/attention/fmha_sm100.cuh b/dsv4/kernels/attention/fmha_sm100.cuh index 06683854..7d8cc947 100644 --- a/dsv4/kernels/attention/fmha_sm100.cuh +++ b/dsv4/kernels/attention/fmha_sm100.cuh @@ -31,6 +31,7 @@ #include #include +#include // CUTLASS C++ includes (CUDA device code only) #if defined(__CUDA_ARCH__)