ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-03-10 14:00:08 +00:00
Code Issues Packages Projects Releases Wiki Activity

Remove the scales, they are not needed

Browse Source
This commit is contained in:
Iwan Kawrakow
2025-06-12 13:17:27 +03:00
parent 5432108e9c
commit 8ba852a2f3
1 changed files with 0 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
ggml/src/iqk/iqk_gemm_kquants.cpp
View File

@@ -755,9 +755,7 @@ struct DequantizerQ4K_AVX2 final : public BaseDequantizer<block_q4_K> {
inline void prepare(int i, int j) {
bits.prepare(x[i].qs, j);
}
Q4Bits_AVX2 bits;
Scales8K s8k;
};
struct DequantizerQ5K_AVX2 final : public BaseDequantizer<block_q5_K> {
@@ -777,7 +775,6 @@ struct DequantizerQ5K_AVX2 final : public BaseDequantizer<block_q5_K> {
const __m256i mh = _mm256_set1_epi8(0x10);
Q4Bits_AVX2 bits;
__m256i hbits;
Scales8K s8k;
};
template <typename Dequantizer, int nrc_y>