File size: 929 Bytes
eb8ddce |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 |
// Copyright (c) 2024, Tri Dao.
// Splitting the different head dimensions to different files to speed up compilation.
#include "flash_fwd_combine_launch_template.h"
template void run_mha_fwd_combine_<float, float, 64>(Flash_fwd_params ¶ms, cudaStream_t stream, bool enable_pdl);
template void run_mha_fwd_combine_<float, float, 128>(Flash_fwd_params ¶ms, cudaStream_t stream, bool enable_pdl);
template void run_mha_fwd_combine_<cutlass::half_t, float, 64>(Flash_fwd_params ¶ms, cudaStream_t stream, bool enable_pdl);
template void run_mha_fwd_combine_<cutlass::half_t, float, 128>(Flash_fwd_params ¶ms, cudaStream_t stream, bool enable_pdl);
template void run_mha_fwd_combine_<cutlass::bfloat16_t, float, 64>(Flash_fwd_params ¶ms, cudaStream_t stream, bool enable_pdl);
template void run_mha_fwd_combine_<cutlass::bfloat16_t, float, 128>(Flash_fwd_params ¶ms, cudaStream_t stream, bool enable_pdl);
|