CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-28 08:31:25 +00:00
Code Issues Packages Projects Releases Wiki Activity

ggml-quants : fix make_qp_quants NANs and IQ1 assertion errors

Browse Source
This commit is contained in:
Francis Couture-Harpin
2025-08-17 11:36:30 -04:00
parent 19f4decae0
commit 184cdc6b27
1 changed files with 4 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
ggml/src/ggml-quants.c
View File

@@ -901,7 +901,7 @@ static float make_qp_quants(int n, int nmax, const float * GGML_RESTRICT x, uint
for (int i = 0; i < n; ++i) { for (int i = 0; i < n; ++i) {
max = MAX(max, x[i]); max = MAX(max, x[i]);
} }
if (!max) { // all zero if (max < GROUP_MAX_EPS) { // all zero
for (int i = 0; i < n; ++i) { L[i] = 0; } for (int i = 0; i < n; ++i) { L[i] = 0; }
return 0.f; return 0.f;
} }
@@ -966,7 +966,7 @@ static float make_qp_quants(int n, int nmax, const float * GGML_RESTRICT x, uint
break; break;
} }
} }
return sumlx/suml2; return suml2 > 0.0f ? sumlx / suml2 : 0.0f;
} }
static void quantize_row_q2_K_impl(const float * GGML_RESTRICT x, block_q2_K * GGML_RESTRICT y, int k, const float * GGML_RESTRICT quant_weights) { static void quantize_row_q2_K_impl(const float * GGML_RESTRICT x, block_q2_K * GGML_RESTRICT y, int k, const float * GGML_RESTRICT quant_weights) {
@@ -4266,7 +4266,7 @@ static void quantize_row_iq1_s_impl(const float * GGML_RESTRICT x, void * GGML_R
sumw[j+1] = sumw[j] + weight[i]; sumw[j+1] = sumw[j] + weight[i];
} }
} }
float best_score = -FLT_MIN, scale = max; float best_score = -FLT_MAX, scale = max;
int besti1 = -1, besti2 = -1, best_shift = 0; int besti1 = -1, besti2 = -1, best_shift = 0;
for (int i1 = 0; i1 <= block_size; ++i1) { for (int i1 = 0; i1 <= block_size; ++i1) {
for (int i2 = i1; i2 <= block_size; ++i2) { for (int i2 = i1; i2 <= block_size; ++i2) {
@@ -4442,7 +4442,7 @@ static void quantize_row_iq1_m_impl(const float * GGML_RESTRICT x, void * GGML_R
idx[2*j] = j; idx[2*j] = j;
} }
qsort(pairs, block_size, 2*sizeof(float), iq1_sort_helper); qsort(pairs, block_size, 2*sizeof(float), iq1_sort_helper);
float best_score = -FLT_MIN, scale = max; float best_score = -FLT_MAX, scale = max;
int besti1 = -1, besti2 = -1, best_k = -1; int besti1 = -1, besti2 = -1, best_k = -1;
// 0: +, + // 0: +, +
// 1: +, - // 1: +, -