vrr/kvcache-ai-ktransformers
2
0
Fork 0
mirror of https://github.com/kvcache-ai/ktransformers.git synced 2026-05-04 14:41:18 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

update kt-kernel

Browse source
This commit is contained in:
ovowei 2025-11-03 15:19:52 +08:00
parent 1a925769d9
commit f854d03bd7
119 changed files with 4459 additions and 6368 deletions
Download patch file Download diff file Expand all files Collapse all files

30
kt-kernel/operators/moe_kernel/api/mat_kernel.h Normal file
View file

@ -0,0 +1,30 @@
#pragma once
#include <cstddef>
#include <cstdint>
#include <type_traits>
#include "common.h"
using GemmFn = void (*)(const KERNEL_CBLAS_LAYOUT layout, const KERNEL_CBLAS_TRANSPOSE transa,
const KERNEL_CBLAS_TRANSPOSE transb, const KERNEL_CBLAS_OFFSET offsetc, const size_t m,
const size_t n, const size_t k, const float alpha, const void* a, const size_t lda,
const int8_t oa, const void* b, const size_t ldb, const int8_t ob, const float beta, int32_t* c,
const size_t ldc, const int32_t* oc);
struct MatKernelSelection {
GemmFn fn;
int divide_elements_size;
};
MatKernelSelection select_kernel_for_int4(MatKernelVariant variant);
MatKernelSelection select_kernel_for_int8(MatKernelVariant variant);
template <typename T>
MatKernelSelection select_mat_kernel(MatKernelVariant variant) {
if constexpr (std::is_same_v<typename T::dt, int4_2_t>) {
return select_kernel_for_int4(variant);
} else {
return select_kernel_for_int8(variant);
}
}