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

Free model gpu buffers on exit

Browse source
This commit is contained in:
0cc4m 2023-12-16 14:49:03 +01:00
parent cd34b87de0
commit c05883f3a9
1 changed files with 3 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

3
llama.cpp
View file

@ -1451,6 +1451,9 @@ struct llama_model {
ggml_cl_free_data(tensors_by_name[i].second); ggml_cl_free_data(tensors_by_name[i].second);
} }
#elif defined(GGML_USE_VULKAN) #elif defined(GGML_USE_VULKAN)
for (size_t i = 0; i < tensors_by_name.size(); ++i) {
ggml_vk_free_data(tensors_by_name[i].second);
}
ggml_vk_cleanup(); ggml_vk_cleanup();
#endif #endif
} }