root/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-24 10:24:35 +00:00
Code Issues Actions 5 Packages Projects Releases Wiki Activity

Disable BLAS altogether - the bug is not just for qunatized mat mul

Browse Source
This commit is contained in:
Georgi Gerganov 2023-03-24 23:47:06 +02:00
parent b3f460e941
commit 8520fc310e
No known key found for this signature in database
GPG Key ID: 449E073F9DC10735
1 changed files with 3 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
ggml.c
View File

@ -5854,6 +5854,9 @@ static bool ggml_compute_forward_mul_mat_use_blas(
const int ne0 = dst->ne[0]; const int ne0 = dst->ne[0];
const int ne1 = dst->ne[1]; const int ne1 = dst->ne[1];
// TMP: disable BLAS for now there is definitely a bug
return false;
// TODO: find the optimal values for these // TODO: find the optimal values for these
if (ggml_is_contiguous(src0) && if (ggml_is_contiguous(src0) &&
ggml_is_contiguous(src1) && ((ne0 >= 32 && ne1 >= 32 && ne10 >= 32))) { ggml_is_contiguous(src1) && ((ne0 >= 32 && ne1 >= 32 && ne10 >= 32))) {