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

ggml : explicit assignment of deltas

Browse source
This commit is contained in:
Georgi Gerganov 2023-04-17 18:41:14 +03:00
parent 0f5ee9e13f
commit a8592fbc11
No known key found for this signature in database
GPG key ID: 449E073F9DC10735
1 changed files with 2 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

4
ggml.c
View file

@ -726,8 +726,8 @@ static void quantize_row_q4_0(const float * restrict x, void * restrict vy, int
const float id0 = d0 ? 1.0f/d0 : 0.0f; const float id0 = d0 ? 1.0f/d0 : 0.0f;
const float id1 = d1 ? 1.0f/d1 : 0.0f; const float id1 = d1 ? 1.0f/d1 : 0.0f;
y[i].d0 = d0; y[i].d0 = GGML_FP32_TO_FP16(d0);
y[i].d1 = d1; y[i].d1 = GGML_FP32_TO_FP16(d1);
for (int l = 0; l < 4; l++) { for (int l = 0; l < 4; l++) {
const float32x4_t v = vmulq_n_f32(srcv[l], id0); const float32x4_t v = vmulq_n_f32(srcv[l], id0);