Mirrors/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2025-01-13 13:52:22 +01:00
Code Issues Packages Projects Releases Wiki Activity

llama : save number of parameters and the size in llama_model (#10286)

Browse Source 
fixes #10285
This commit is contained in:
FirstTimeEZ 2024-11-16 13:42:13 +13:00 committed by GitHub
parent 74d73dc85c
commit 89e4caaaf0
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 18 additions and 13 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

31
src/llama.cpp
View File

@ -2907,9 +2907,15 @@ struct llama_model {
// for quantize-stats only
std::vector<std::pair<std::string, struct ggml_tensor *>> tensors_by_name;
int64_t t_load_us = 0;
int64_t t_load_us = 0;
int64_t t_start_us = 0;
// total number of parameters in the model
uint64_t n_elements = 0;
// total size of all the tensors in the model in bytes
size_t n_bytes = 0;
// keep track of loaded lora adapters
std::set<struct llama_lora_adapter *> lora_adapters;
@ -4275,8 +4281,8 @@ struct llama_model_loader {
int n_tensors = 0;
int n_created = 0;
int64_t n_elements = 0;
size_t n_bytes = 0;
uint64_t n_elements = 0;
size_t n_bytes = 0;
bool use_mmap = false;
bool check_tensors;
@ -5344,6 +5350,11 @@ static const char * llama_model_vocab_type_name(enum llama_vocab_type type){
}
}
static void llm_load_stats(llama_model_loader & ml, llama_model & model) {
model.n_elements = ml.n_elements;
model.n_bytes = ml.n_bytes;
}
static void llm_load_arch(llama_model_loader & ml, llama_model & model) {
model.arch = ml.get_arch();
if (model.arch == LLM_ARCH_UNKNOWN) {
@ -9256,6 +9267,7 @@ static int llama_model_load(const std::string & fname, llama_model & model, llam
throw std::runtime_error("error loading model vocabulary: " + std::string(e.what()));
}
llm_load_stats(ml, model);
llm_load_print_meta(ml, model);
if (model.vocab.type != LLAMA_VOCAB_TYPE_NONE &&
@ -18601,6 +18613,7 @@ static void llama_model_quantize_internal(const std::string & fname_inp, const s
llama_model model;
llm_load_arch(ml, model);
llm_load_hparams(ml, model);
llm_load_stats(ml, model);
struct quantize_state_internal qs(model, params);
@ -19953,19 +19966,11 @@ int32_t llama_model_desc(const struct llama_model * model, char * buf, size_t bu
}
uint64_t llama_model_size(const struct llama_model * model) {
uint64_t size = 0;
for (const auto & it : model->tensors_by_name) {
size += ggml_nbytes(it.second);
}
return size;
return model->n_bytes;
}
uint64_t llama_model_n_params(const struct llama_model * model) {
uint64_t nparams = 0;
for (const auto & it : model->tensors_by_name) {
nparams += ggml_nelements(it.second);
}
return nparams;
return model->n_elements;
}
struct ggml_tensor * llama_get_model_tensor(struct llama_model * model, const char * name) {