C++上用onnxruntime进行onnx模型推理时，出现了内存报错

目前正在学习c++的onnxruntime部分，所以用了别人博客的代码，把代码的一些地方进行修改后发现在session.Run这一步一直出现内存报错，我已经把输入图像的维度都修改到了生成onnx模型时的维度了，也没有进行GPU的推理，实在想不出是哪里错了，也给输入进行一维化了，采用的是（NCHW）的格式。不知道到底哪里错了，特来请教各位！！代码如下：

#include <iostream>
#include <opencv2/highgui.hpp>
#include <opencv2/imgproc.hpp>
#include <opencv2/core.hpp>
#include <opencv2/imgcodecs.hpp>
#include <chrono>
#include <string>

//onnxruntime
#include <onnxruntime_cxx_api.h>
#include <onnxruntime_c_api.h>


using namespace std;

int main(int argc, char** argv)
{
    //输入网络的维度
    static constexpr const int width = 640;
    static constexpr const int height = 480;
    static constexpr const int channel = 1;
    std::array<int64_t, 4> input_shape_{ 1, 1, 480, 640 };

    //-------------------------------------------------------------onnxruntime-------------------------------------------------
    //图片和模型位置
#ifdef _WIN32
    const wchar_t* model_path = L"C:\\Users\\dell\\PycharmProjects\\cyh_torch\\SuperGlue\\Onnx\\CYH_SuperPoint_0.5.onnx";
#else
    const char* model_path = "C:\\Users\\dell\\PycharmProjects\\cyh_torch\\SuperGlue\\Onnx\\CYH_SuperPoint_0.5.onnx";
#endif
    cv::Mat imgSource = cv::imread("C:/Users/dell/Desktop/CYH/SuperGlue/SuperGlue_load/1.png", 0);

    Ort::Env env(OrtLoggingLevel::ORT_LOGGING_LEVEL_WARNING, "Detection");
    Ort::SessionOptions session_options;
    //CUDA加速开启
    //OrtSessionOptionsAppendExecutionProvider_CUDA(session_options, 0);
    session_options.SetGraphOptimizationLevel(GraphOptimizationLevel::ORT_ENABLE_ALL);

    Ort::AllocatorWithDefaultOptions allocator;
    //加载ONNX模型
    Ort::Session session(env, model_path, session_options);
    //获取输入输出的维度
    std::vector<int64_t> input_dims = session.GetInputTypeInfo(0).GetTensorTypeAndShapeInfo().GetShape();
    std::vector<int64_t> output_dims = session.GetOutputTypeInfo(0).GetTensorTypeAndShapeInfo().GetShape();

    /*
    session.GetOutputName(1, allocator);
    session.GetInputName(1, allocator);
    //输出模型输入节点的数量
    size_t num_input_nodes = session.GetInputCount();
    size_t num_output_nodes = session.GetOutputCount();
    */

    std::vector<const char*> input_node_names = { "image" };
    std::vector<const char*> output_node_names = { "scores", "keypoints", "descriptors" };
    input_dims[0] = output_dims[0] = 1;//batch size = 1

    std::vector<Ort::Value> input_tensors;
    auto memory_info = Ort::MemoryInfo::CreateCpu(OrtAllocatorType::OrtArenaAllocator, OrtMemType::OrtMemTypeDefault);

    //将图像存储到uchar数组中，BGR--->RGB
    std::array<uchar, 1 * 480 * 640> input_image_{};
    uchar* input = input_image_.data();
    for (int i = 0; i < imgSource.rows; i++) {
        for (int j = 0; j < imgSource.cols; j++) {
                //NCHW 格式
                input[i * imgSource.cols + j] = imgSource.ptr<uchar>(i)[j] / 255.0;
        }
    }

    input_tensors.push_back(Ort::Value::CreateTensor<uchar>(
        memory_info, input, input_image_.size(), input_shape_.data(), input_shape_.size()));
    //不知道输入维度时
    //input_tensors.push_back(Ort::Value::CreateTensor<uchar>(
    //        memory_info, input, input_image_.size(), input_dims.data(), input_dims.size()));

    chrono::steady_clock::time_point t1 = chrono::steady_clock::now();

    std::vector<Ort::Value> output_tensors;
    output_tensors = session.Run(Ort::RunOptions{ nullptr },
            input_node_names.data(), //输入节点名
            input_tensors.data(),     //input tensors
            input_tensors.size(),     //1
            output_node_names.data(), //输出节点名
            output_node_names.size()); //4
    chrono::steady_clock::time_point t2 = chrono::steady_clock::now();

    chrono::duration<double> delay_time = chrono::duration_cast<chrono::duration<double>>(t2 - t1); //milliseconds 毫秒

    cout << "前向传播平均耗时:" << delay_time.count() * 1000 / 100.0 << "ms" << endl;

    return 0;

}

报错的位置如下：

img