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

llama-bench : add model sizes (#2771)

Browse source 
* llama-bench : add model sizes

* more compact markdown output

* back to GiB

* adjust column sizes
This commit is contained in:
slaren 2023-08-25 15:16:19 +02:00 • committed by GitHub
parent 12e2e33a97
commit 154725c543
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
3 changed files with 74 additions and 10 deletions
Download patch file Download diff file Expand all files Collapse all files

18
llama.cpp
View file

@ -5297,13 +5297,29 @@ int llama_model_n_embd(const struct llama_model * model) {
return model->hparams.n_embd;
}
int llama_model_type(const struct llama_model * model, char * buf, size_t buf_size) {
int llama_model_desc(const struct llama_model * model, char * buf, size_t buf_size) {
return snprintf(buf, buf_size, "%s %s %s",
model->name.c_str(),
llama_model_type_name(model->type),
llama_model_ftype_name(model->ftype).c_str());
}
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;
}
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;
}
int llama_model_quantize(
const char * fname_inp,
const char * fname_out,