int tengine_graph_init(void)
{
    struct options opt;   //options为garph的行时参数
    opt.num_thread = 4;
    opt.cluster = TENGINE_CLUSTER_ALL;  //TENGINE_CLUSTER_ALL表示使用所有的CPU核心
    opt.precision = TENGINE_MODE_FP32;  //TENGINE_MODE_FP32表示使用FP32的精度
    opt.affinity = 0;  //affinity表示是否使用CPU亲和性，0表示不使用,没有绑定线程到核心
    int dims[] = {1, 3, 640, 640};
    std::vector<float> input_data(3*640*640);

    if(init_tengine() != 0)
    {
       cout<<"tengine init failed"<<endl;
       return -1; 
    }
    cout<<get_tengine_version()<<endl;

    //构建计算图
    graph_t graph = create_graph(nullptr, "tengine", "yolov5s.tmfile");
    if(graph != nullptr)
        cout<<"graph build success"<<endl;
    else
    {
        cout<<"graph build faild"<<endl;
        return -1;
    }

    tensor_t input_tensor = get_graph_input_tensor(graph, 0, 0);
    if (input_tensor == nullptr)
    {
        cout<<"Get input tensor failed"<<endl;
        return -1;
    }

    if (set_tensor_shape(input_tensor, dims, 4) < 0)
    {
        cout<<"Set input tensor shape failed"<<endl;
        return -1;
    }

    if (set_tensor_buffer(input_tensor, input_data.data(), 3*640*640 * sizeof(float)) < 0)
    {
        cout<<"Set input tensor buffer failed"<<endl;
        return -1;
    }

    /* prerun graph, set work options(num_thread, cluster, precision) */
    if (prerun_graph_multithread(graph, opt) < 0)
    {
        cout<<"Prerun multithread graph failed"<<endl;
        return -1;
    }
    else
        cout<<"prerun graph success"<<endl;
    return 0;
}