// Copyright (c) 2023, Tri Dao.
// Splitting the different head dimensions to different files to speed up
// compilation. This file is auto-generated. See "generate_kernels.py"

#include "flash_fwd_launch_template.h"

template <>
void run_mha_fwd_<cutlass::half_t, 64, false>(Flash_fwd_params& params,
                                              cudaStream_t stream) {
  run_mha_fwd_hdim64<cutlass::half_t, false>(params, stream);
}