flash_fwd_hdim192_fp16_sm80.cu 392 B

12345678910
  1. // Copyright (c) 2024, Tri Dao.
  2. // Splitting the different head dimensions to different files to speed up compilation.
  3. // This file is auto-generated. See "generate_kernels.py"
  4. #include "flash_fwd_launch_template.h"
  5. template<>
  6. void run_mha_fwd_<cutlass::half_t, 192, false>(Flash_fwd_params &params, cudaStream_t stream) {
  7. run_mha_fwd_hdim192<cutlass::half_t, false>(params, stream);
  8. }