yolo cudaMemcpyAsync(output, buffers[outputIndex], output_size * sizeof(float), cudaMemcpyDeviceToH

yolov7 tensorrt预测时报错，

CHECK(cudaMemcpyAsync(output, buffers[outputIndex], output_size * sizeof(float), cudaMemcpyDeviceToHost, stream));

原因：模型的 input和outputname写错了

解决方法：onnx导出的input和output记下来，导出tensorrt也用。

        if opt.dynamic:
            dynamic_axes = {'input': {0: 'batch', 2: 'height', 3: 'width'},  # size(1,3,640,640)
             'output': {0: 'batch', 2: 'y', 3: 'x'}}

        torch.onnx.export(model, img, f, verbose=False, opset_version=11, input_names=['input'],
                          output_names=output_names,
                          dynamic_axes=dynamic_axes)

yolov7 c++推理部分：

#include <fstream>
#include <