Mirrors/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-25 05:48:47 +01:00
Code Issues Packages Projects Releases Wiki Activity

Guard against all weights in a super-block being zero (#3010)

Browse Source 
* Guard against all weights in a super-block being zero

* Also guard against extremely small weights

Closes #2982 

---------

Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
This commit is contained in:
Kawrakow 2023-09-05 09:55:33 +02:00 committed by GitHub
parent 35938ee3b0
commit d59bd97065
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 7 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
k_quants.c
View File

@ -83,7 +83,7 @@ static float make_qx_quants(int n, int nmax, const float * restrict x, int8_t *
float ax = fabsf(x[i]); float ax = fabsf(x[i]);
if (ax > amax) { amax = ax; max = x[i]; } if (ax > amax) { amax = ax; max = x[i]; }
} }
if (!amax) { // all zero if (amax < 1e-30f) { // all zero
for (int i = 0; i < n; ++i) { for (int i = 0; i < n; ++i) {
L[i] = 0; L[i] = 0;
} }
@ -1086,6 +1086,12 @@ void quantize_row_q6_K_reference(const float * restrict x, block_q6_K * restrict
} }
if (!max_abs_scale) {
memset(&y[i], 0, sizeof(block_q6_K));
y[i].d = ggml_fp32_to_fp16(0.f);
continue;
}
float iscale = -128.f/max_scale; float iscale = -128.f/max_scale;
y[i].d = ggml_fp32_to_fp16(1/iscale); y[i].d = ggml_fp32_to_fp16(1/iscale);
for (int ib = 0; ib < QK_K/16; ++ib) { for (int ib = 0; ib < QK_K/16; ++ib) {