flash_fwd_split_hdim224_bf16_sm80.cu 342 B

1234567
  1. // Copyright (c) 2023, 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 void run_mha_fwd_splitkv_dispatch<cutlass::bfloat16_t, 224, false>(Flash_fwd_params &params, cudaStream_t stream);