CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-03 09:22:01 +00:00
Code Issues Packages Projects Releases Wiki Activity

Disable BLAS branch in mul_mat - seems there is a bug

Browse Source
This commit is contained in:
Georgi Gerganov
2023-03-24 23:39:17 +02:00
parent 04c6f5ed6f
commit b3f460e941
1 changed files with 5 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
ggml.c
View File

@@ -5858,11 +5858,11 @@ static bool ggml_compute_forward_mul_mat_use_blas(
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))) {
//// disable BLAS for Q4_0 and Q4_1 // disable BLAS for Q4_0 and Q4_1
//// looks like there is no benefit and we only waste a lot of memory // there is a bug that has to be fixed before enabling
//if (src0->type == GGML_TYPE_Q4_0 || src0->type == GGML_TYPE_Q4_1) { if (src0->type == GGML_TYPE_Q4_0 || src0->type == GGML_TYPE_Q4_1) {
// return false; return false;
//} }
//printf("BLAS: %d %d %d %d %d\n", ne0, ne1, ne10, ne00, ne01); //printf("BLAS: %d %d %d %d %d\n", ne0, ne1, ne10, ne00, ne01);
return true; return true;