ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-01-26 17:20:01 +00:00
Code Issues Packages Projects Releases Wiki Activity

Fix q6_0 dequantize (#1005)

Browse Source 
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
This commit is contained in:
Kawrakow
2025-11-24 09:13:46 +01:00
committed by GitHub
parent 9a63e768ea
commit 0243356650
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
ggml/src/ggml-cuda/dequantize.cuh
View File

@@ -91,7 +91,7 @@ static __device__ __forceinline__ void dequantize_q6_0(const void * vx, const in
const dfloat d = x[ib].d;
const uint8_t h = x[ib].qh[iqs%8] >> 2*(iqs/8);
const uint8_t h = x[ib].qh[iqs%8] >> 4*(iqs/8);
v.x = ((x[ib].qs[iqs] & 0xf) | ((h & 0x3) << 4));
v.y = ((x[ib].qs[iqs] >> 4) | ((h & 0xc) << 2));