vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Fix MacOS Sonoma model quantization (#4052)

Browse source 
Co-authored-by: Jared Van Bortel <jared@nomic.ai>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
This commit is contained in:
Michael Potter 2023-11-14 09:34:41 -08:00 • committed by GitHub
parent 36eed0c42c
commit 6bb4908a17
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
3 changed files with 19 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

5
ggml-quants.c
View file

@ -1368,7 +1368,12 @@ static float make_qkx2_quants(int n, int nmax, const float * restrict x, const f
float max = x[0];
float sum_w = weights[0];
float sum_x = sum_w * x[0];
#ifdef HAVE_BUGGY_APPLE_LINKER
// use 'volatile' to prevent unroll and work around a bug in Apple ld64 1015.7
for (volatile int i = 1; i < n; ++i) {
#else
for (int i = 1; i < n; ++i) {
#endif
if (x[i] < min) min = x[i];
if (x[i] > max) max = x[i];
float w = weights[i];