test_model_load.cpp

#include <iostream>
#include <fstream>
#include "llama-cpp.h"

int main() {
    std::cout << "=== 模型加载测试 ===" << std::endl;
    
    // 检查模型文件
    const char* model_path = "model/model_1.8B.gguf";
    std::ifstream file(model_path, std::ios::binary);
    if (!file) {
        std::cerr << "[ERROR] 无法打开模型文件: " << model_path << std::endl;
        return 1;
    }
    
    file.seekg(0, std::ios::end);
    size_t size = file.tellg();
    std::cout << "[OK] 模型文件大小: " << size << " bytes (" << size / (1024*1024) << " MB)" << std::endl;
    file.close();
    
    // 尝试加载模型
    std::cout << "正在加载模型..." << std::endl;
    
    ggml_backend_load_all();
    
    llama_model_params mparams = llama_model_default_params();
    llama_model* model = llama_model_load_from_file(model_path, mparams);
    
    if (model == nullptr) {
        std::cerr << "[ERROR] 模型加载失败!" << std::endl;
        std::cerr << "可能原因:" << std::endl;
        std::cerr << "1. 内存不足（需要8GB+）" << std::endl;
        std::cerr << "2. CPU不支持AVX2指令集" << std::endl;
        std::cerr << "3. 模型文件损坏" << std::endl;
        return 1;
    }
    
    std::cout << "[OK] 模型加载成功!" << std::endl;
    
    // 获取模型信息
    llama_context_params cparams = llama_context_default_params();
    llama_context* ctx = llama_new_context_with_model(model, cparams);
    
    if (ctx == nullptr) {
        std::cerr << "[WARNING] 无法创建上下文（可能内存不足）" << std::endl;
    } else {
        std::cout << "[OK] 上下文创建成功" << std::endl;
        llama_free(ctx);
    }
    
    llama_model_free(model);
    std::cout << "=== 测试完成 ===" << std::endl;
    
    return 0;
}