apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-21 15:49:31 +08:00
Code Issues Actions 6 Packages Projects Releases Wiki Activity

[Iluvatar GPU] Optimize attention performance and fix moe load ckpt error (#3651)

Browse Source
This commit is contained in:
yzwu
2025-09-22 21:13:59 +08:00
committed by GitHub
parent 5532e8a323
commit 504461b6b5
17 changed files with 1344 additions and 363 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
custom_ops/gpu_ops/helper.h
View File

@@ -193,11 +193,13 @@ public:
typedef uint8_t data_t;
};
#ifndef PADDLE_WITH_COREX
template <> class PDTraits<paddle::DataType::FLOAT8_E4M3FN> {
public:
typedef __nv_fp8_e4m3 DataType;
typedef paddle::float8_e4m3fn data_t;
};
#endif
template <typename T, int Size> struct alignas(sizeof(T) * Size) AlignedVector {
T val[Size];