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

prevent gpu offload if kquant is selected with clblast for now

Browse source
This commit is contained in:
Concedo 2023-06-12 21:57:31 +08:00
parent 215edf420b
commit 120851df53
1 changed files with 5 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

6
llama.cpp
View file

@ -1077,7 +1077,11 @@ static void llama_model_load_internal(
#if defined(GGML_USE_CLBLAST) #if defined(GGML_USE_CLBLAST)
if (file_version == LLAMA_FILE_VERSION_GGJT_V3) { if (file_version == LLAMA_FILE_VERSION_GGJT_V3) {
if (hparams.ftype >= LLAMA_FTYPE_MOSTLY_Q2_K && hparams.ftype <= LLAMA_FTYPE_MOSTLY_Q6_K) { if (hparams.ftype >= LLAMA_FTYPE_MOSTLY_Q2_K && hparams.ftype <= LLAMA_FTYPE_MOSTLY_Q6_K) {
printf("\n===\nK-Quants are currently not supported with CLBlast!!!\nPlease select a q4_0, q4_0, q5_0 or q5_1 format instead!\n=====\n"); if(n_gpu_layers>0)
{
n_gpu_layers = 0;
printf("\n===\nCLBlast cannot offload layers for K-Quants!\nPlease select a q4_0, q4_0, q5_0 or q5_1 format instead!\n=====\n");
}
} }
} }
#endif #endif