ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-05-01 03:41:53 +00:00
Code Issues Packages Projects Releases Wiki Activity

Fix #300 (#301)

Browse Source 
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
This commit is contained in:
Kawrakow
2025-04-01 08:29:25 +02:00
committed by GitHub
parent 3c3825d7f6
commit ba3030c9c3
1 changed files with 8 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
ggml/src/iqk/iqk_mul_mat.cpp
View File

@@ -15514,8 +15514,13 @@ struct HelperQ8KVR8 : public BaseHelper<step> {
template <int D, int step>
struct HelperQ40 final : public BaseHelper<step> {
using Base = BaseHelper<step>;
#if defined __AVX2__
using block_q8 = block_q8_2;
constexpr static int block_size_q = QK8_2;
#else
using block_q8 = block_q8_0;
constexpr static int block_size_q = QK8_0;
#endif
HelperQ40(const char * data, int stride) : Base(data, stride) {}
// Needed for v * softmax(k * q)
@@ -15558,8 +15563,8 @@ struct HelperQ40 final : public BaseHelper<step> {
template <int D, int step>
struct HelperQ41 final : public BaseHelper<step> {
using Base = BaseHelper<step>;
using block_q8 = block_q8_1;
constexpr static int block_size_q = QK8_1;
using block_q8 = block_q8_2;
constexpr static int block_size_q = QK8_2;
HelperQ41(const char * data, int stride) : Base(data, stride) {}
// Needed for v * softmax(k * q)
@@ -16414,7 +16419,7 @@ struct FlashQKfp32 {
#ifdef __aarch64__
MAKE_FUNCS(mul_mat_qX_0_q8_0<DequantizerQ40, nq);
#else
MAKE_FUNCS(mul_mat_qX_1_q8_2_T<Q4_0_Unpacker, nq);
MAKE_FUNCS(mul_mat_qX_1_q8_2_T<Q4_0_1_Unpacker, nq);
#endif
}
else if constexpr (std::is_same_v<KHelper, HelperQ41<D, k_step>>) {