mirror of
https://github.com/ggml-org/llama.cpp.git
synced 2026-06-27 23:50:20 -05:00
a976ff081b
* tests: add end-to-end tests per model architecture * fixup for rebase * fix use-after-free in llama-model-loader.cpp * fix CI * fix WebGPU * fix CI * disable CI for macOS-latest-cmake-arm64 * use expert_weights_scale only if != 0.0f * comments
41 lines
1.1 KiB
C++
41 lines
1.1 KiB
C++
#pragma once
|
|
|
|
#include "gguf.h"
|
|
#include "llama.h"
|
|
#include "llama-arch.h"
|
|
|
|
#include <vector>
|
|
|
|
struct llama_model_saver {
|
|
struct gguf_context * gguf_ctx = nullptr;
|
|
const bool gguf_ctx_owned;
|
|
const struct llama_model * model;
|
|
const struct LLM_KV llm_kv;
|
|
|
|
llama_model_saver(const struct llama_model * model);
|
|
llama_model_saver(enum llm_arch arch, struct gguf_context * gguf_ctx);
|
|
~llama_model_saver();
|
|
|
|
void add_kv(enum llm_kv key, uint32_t value);
|
|
void add_kv(enum llm_kv key, int32_t value);
|
|
void add_kv(enum llm_kv key, float value);
|
|
void add_kv(enum llm_kv key, bool value);
|
|
void add_kv(enum llm_kv key, const char * value);
|
|
|
|
[[noreturn]]
|
|
void add_kv(enum llm_kv key, char value); // needed to make the template below compile
|
|
|
|
template <typename Container>
|
|
void add_kv(enum llm_kv key, const Container & value, bool per_layer = false);
|
|
|
|
void add_kv(enum llm_kv key, const std::vector<std::string> & value);
|
|
|
|
void add_tensor(const struct ggml_tensor * tensor);
|
|
|
|
void add_kv_from_model();
|
|
|
|
void add_tensors_from_model();
|
|
|
|
void save(const std::string & path_model);
|
|
};
|