flash_fwd_mla_fp16_gfx936.cu

#include "flash_fwd_mla_kernel.h"

template void run_mha_fwd_splitkv_mla<cutlass::half_t, 576>(Flash_fwd_mla_params &params, const std::string& kv_cache_dtype, cudaStream_t stream, bool is_q_nope_pe = false);