matt/ollama37
1
0
Fork 0
mirror of https://github.com/dogkeeper886/ollama37.git synced 2025-12-11 00:07:07 +00:00
Code Issues Packages Projects Releases Wiki Activity

update llama.cpp submodule to 66c1968f7 (#2618)

Browse Source
This commit is contained in:
Jeffrey Morgan
2024-02-20 17:42:31 -05:00
committed by GitHub
parent ace2cdf1c6
commit 4613a080e7
6 changed files with 39 additions and 130 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
llm/ext_server/ext_server.cpp
View File

@@ -80,7 +80,7 @@ void llama_server_init(ext_server_params *sparams, ext_server_resp_t *err) {
params.main_gpu = sparams->main_gpu;
params.use_mlock = sparams->use_mlock;
params.use_mmap = sparams->use_mmap;
params.numa = sparams->numa;
params.numa = (ggml_numa_strategy)sparams->numa;
params.embedding = sparams->embedding;
if (sparams->model != NULL) {
params.model = sparams->model;
@@ -111,7 +111,8 @@ void llama_server_init(ext_server_params *sparams, ext_server_resp_t *err) {
}
#endif
llama_backend_init(params.numa);
llama_backend_init();
llama_numa_init(params.numa);
// load the model
if (!llama->load_model(params)) {

2
llm/ext_server/ext_server.h
View File

@@ -41,7 +41,7 @@ typedef struct ext_server_params {
int32_t main_gpu; // the GPU that is used for scratch and small tensors
bool use_mlock; // force system to keep model in RAM
bool use_mmap; // use mmap if possible
bool numa; // attempt optimizations that help on some NUMA systems
int numa; // attempt optimizations that help on some NUMA systems
bool embedding; // get only sentence embedding
ext_server_lora_adapter_t *lora_adapters;
char *mmproj;