The class torchvision.ops.RoIAlign outputs the result of roi_align based on the input feature map. In tradition, the output size is smaller than the input size when using roi_align. So my question is what will happen if the output size is larger than the input size. Will it just divided the input feature map into output_size[0] * output_size[1] equal parts and then do the same thing as it will do when the output size is smaller than the input feature map size?