Use hipMemcpy rather than hipMemcpyAsync

Signed-off-by: wenjh <wenjh@sugon.com>

transformer_engine/common/gemm/rocm_gemm.cu

@@ -1610,10 +1610,10 @@ void hipblaslt_goupedgemm(std::vector<const Tensor*>& inputA, std::vector<const
   // Copy them to device memory
   // hipblaslt_ext::UserArguments* d_userArgs;
   // NVTE_CHECK_CUDA(hipMallocAsync(&d_userArgs, m.size() * sizeof(hipblaslt_ext::UserArguments), stream));
-  NVTE_CHECK_CUDA(hipMemcpyAsync(d_userArgs,
+  NVTE_CHECK_CUDA(hipMemcpy(d_userArgs,
                             userArgs,
                             m.size() * sizeof(hipblaslt_ext::UserArguments),
-                            hipMemcpyHostToDevice, stream));
+                            hipMemcpyHostToDevice));
   NVTE_CHECK_HIPBLASLT(groupedgemm.run(d_userArgs, stream));
   // NVTE_CHECK_HIPBLASLT(groupedgemm.initialize(heuristicResult[0].algo, workspace, false, stream));
   // NVTE_CHECK_HIPBLASLT(groupedgemm.run(stream));