【Mindspore】ResizeArea涉及的infershape问题

算子有两个输入，第一个是原图像4维Tensor，名字是images；第二个是输出图像Tensor要求的高度和宽度，是一个1维Tensor，名字是size

现在我在c++侧写infershape，我想从size里拿到高度和宽度放到这里

 

输入：[batch, height, width, channels]
第2/3维根据输入`size`的取值确定，涉及动态shape
输出：[batch, new_height, new_width, channels]

infer shape思路参考
const int64_t max_length = 1024 * 1024 *1024;
const int64_t height_max_length = min(height *100, max_length);
const int64_t width_max_length = min(width *100, max_length);

ShapeVector shape_out = {batch, SHP_ANY, SHP_ANY, channels};
ShapeVector shape_min = {batch, 0, 0, channels};
ShapeVector shape_max = {batch, height_max_length, width_max_length, channels};

并且在底层代码实现中明确限制当超过max_length时打印错误日志