FlashMLA/csrc/flash_fwd_mla_fp16_sm90.cu
2025-02-24 01:58:53 -08:00

4 lines
147 B
Plaintext

#include "flash_fwd_mla_kernel.h"
template void run_mha_fwd_splitkv_mla<cutlass::half_t, 576>(Flash_fwd_mla_params &params, cudaStream_t stream);