Skip to content

GitLab

Switch branch/tag
Find file Normal viewHistoryPermalink
splitkv_mla.h 207 Bytes
Newer Older
Shengyu Liu's avatar
Reorganize files and add sparse prefill/decoding kernels on hopper  
Shengyu Liu committed
1
2
3
4 
#pragma once

#include "params.h"
Shengyu Liu's avatar
Multiple updates and refactorings (#150)  
Shengyu Liu committed
5 
namespace sm90::decode::sparse_fp8 {
Shengyu Liu's avatar
Reorganize files and add sparse prefill/decoding kernels on hopper  
Shengyu Liu committed
6
Shengyu Liu's avatar
Multiple updates and refactorings (#150)  
Shengyu Liu committed
7
8 
template<ModelType MODEL_TYPE, int NUM_HEADS>
void run_flash_splitkv_mla_fp8_sparse_kernel(const SparseAttnDecodeParams &params);
Shengyu Liu's avatar
Reorganize files and add sparse prefill/decoding kernels on hopper  
Shengyu Liu committed
9
10 

}
Shengyu Liu's avatar
Multiple updates and refactorings (#150)  
Shengyu Liu committed
11