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

Add llama_init_ggml c api.

Browse source
This commit is contained in:
Yaohui Liu 2023-05-20 11:42:28 +08:00
parent 484f6e9438
commit e78a971859
No known key found for this signature in database
GPG key ID: E86D01E1809BD23E
3 changed files with 16 additions and 6 deletions
Download patch file Download diff file Expand all files Collapse all files

7
examples/quantize/quantize.cpp
View file

@ -52,6 +52,13 @@ int main(int argc, char ** argv) {
return 1; return 1;
} }
// needed to initialize f16 tables
{
struct ggml_init_params params = { 0, NULL, false };
struct ggml_context * ctx = ggml_init(params);
ggml_free(ctx);
}
// parse command line arguments // parse command line arguments
const std::string fname_inp = argv[1]; const std::string fname_inp = argv[1];
std::string fname_out; std::string fname_out;

11
llama.cpp
View file

@ -2188,6 +2188,11 @@ struct llama_context * llama_init_from_file(
return ctx; return ctx;
} }
void llama_init_ggml(struct ggml_init_params params) {
struct ggml_context * ctx = ggml_init(params);
ggml_free(ctx);
}
void llama_free(struct llama_context * ctx) { void llama_free(struct llama_context * ctx) {
delete ctx; delete ctx;
} }
@ -2198,12 +2203,6 @@ int llama_model_quantize(
enum llama_ftype ftype, enum llama_ftype ftype,
int nthread) { int nthread) {
try { try {
// needed to initialize f16 tables
{
struct ggml_init_params params = { 0, NULL, false };
struct ggml_context * ctx = ggml_init(params);
ggml_free(ctx);
}
llama_model_quantize_internal(fname_inp, fname_out, ftype, nthread); llama_model_quantize_internal(fname_inp, fname_out, ftype, nthread);
return 0; return 0;
} catch (const std::string & err) { } catch (const std::string & err) {

4
llama.h
View file

@ -97,6 +97,10 @@ extern "C" {
const char * path_model, const char * path_model,
struct llama_context_params params); struct llama_context_params params);
// Init the ggml context (it won't return a context ptr because it will free
// the ctx after initialing it).
LLAMA_API void llama_init_ggml(struct ggml_init_params params);
// Frees all allocated memory // Frees all allocated memory
LLAMA_API void llama_free(struct llama_context * ctx); LLAMA_API void llama_free(struct llama_context * ctx);