diff --git a/src/device/reduce_kernel.h b/src/device/reduce_kernel.h
index d36dfe5a7..af927a924 100644
--- a/src/device/reduce_kernel.h
+++ b/src/device/reduce_kernel.h
@@ -260,7 +260,7 @@ struct Apply_Cast<float ,__nv_bfloat16, /*EltPerPack=*/2> {
   template<> \
   struct Apply_Cast<A, B, EltPerPack> { \
     __device__ __forceinline__ static BytePack<sizeof(B)*EltPerPack> cast(BytePack<sizeof(A)*EltPerPack> a) { \
-      return toPack(VecB(fromPack<VecA>(a))); \
+      return toPack(static_cast<VecB>(fromPack<VecA>(a))); \
     } \
   }; \
   template<> \