From 154725c5436808e5c519685d0279e850596dbe62 Mon Sep 17 00:00:00 2001 From: slaren Date: Fri, 25 Aug 2023 15:16:19 +0200 Subject: llama-bench : add model sizes (#2771) * llama-bench : add model sizes * more compact markdown output * back to GiB * adjust column sizes --- llama.cpp | 18 +++++++++++++++++- 1 file changed, 17 insertions(+), 1 deletion(-) (limited to 'llama.cpp') diff --git a/llama.cpp b/llama.cpp index d12b6d1c..4529ac82 100644 --- a/llama.cpp +++ b/llama.cpp @@ -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, -- cgit v1.2.3