halfway refactoring, wip adding other model types

This commit is contained in:
Concedo
2023-04-01 01:13:05 +08:00
parent 56949197fe
commit 6b86f5ea22
19 changed files with 13135 additions and 319 deletions
+10 -1
View File
@@ -1,3 +1,4 @@
#pragma once
#include "common.h"
#include <cassert>
@@ -14,8 +15,16 @@
#include "llama.h"
#include "ggml.h"
//return val: 0=fail, 1=(original ggml, alpaca), 2=(ggmf), 3=(ggjt)
enum FileFormat
{
FAIL=0,
GGML=1,
GGHF=2,
GGJT=3
};
int check_file_format(const std::string & fname);
FileFormat check_file_format(const std::string & fname);
std::vector<llama_token> legacy_llama_tokenize(struct llama_context * ctx, const std::string & text, bool add_bos);
static bool legacy_llama_model_load(const std::string & fname, llama_context & lctx, int n_ctx, int n_parts, ggml_type memory_type, bool vocab_only, llama_progress_callback progress_callback, void *progress_callback_user_data);