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

kompute : remove Q6_K from list of supported quant types

Browse source
This commit is contained in:
Cebtenzzre 2023-10-04 16:19:19 -04:00 • committed by cebtenzzre
parent 24a4a5956a
commit 3d850db767
1 changed files with 1 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

3
llama.cpp
View file

@ -6498,8 +6498,7 @@ struct llama_context * llama_new_context_with_model(
&& (model->ftype == LLAMA_FTYPE_ALL_F32
|| model->ftype == LLAMA_FTYPE_MOSTLY_F16
|| model->ftype == LLAMA_FTYPE_MOSTLY_Q4_0
|| model->ftype == LLAMA_FTYPE_MOSTLY_Q4_1
|| model->ftype == LLAMA_FTYPE_MOSTLY_Q6_K)) {
|| model->ftype == LLAMA_FTYPE_MOSTLY_Q4_1)) {
// this allocates all Vulkan resources and memory buffers
ctx->ctx_kompute = ggml_vk_init();