flash-attention/hopper/instantiations/flash_bwd_hdim96_fp16_softcap_sm80.cu

// Copyright (c) 2024, Jay Shah, Ganesh Bikshandi, Ying Zhang, Vijay Thakkar, Pradeep Ramani, Tri Dao.
// Splitting the different template instantiations to different files to speed up compilation.
// This file is auto-generated. See "generate_kernels.py"

#include "flash_bwd_launch_template.h"

#ifndef FLASHATTENTION_DISABLE_SM8x
#ifndef FLASHATTENTION_DISABLE_HDIM96
template<>
void run_mha_bwd_<80, cutlass::half_t, 96, true>(Flash_bwd_params &params, cudaStream_t stream) {
    run_mha_bwd_hdim96<80, cutlass::half_t, true>(params, stream);
}
template<>
void run_mha_bwd_<86, cutlass::half_t, 96, true>(Flash_bwd_params &params, cudaStream_t stream) {
    run_mha_bwd_hdim96<86, cutlass::half_t, true>(params, stream);
}
#endif
#endif