why you check the range of inputs when calculate lpips？

[zealousfool]

I found this function, when calculate lpips metrics.
def _valid_img(img: Tensor, normalize: bool) -> bool:
"""Check that input is a valid image to the network."""
value_check = img.max() <= 1.0 and img.min() >= 0.0 if normalize else img.min() >= -1
return img.ndim == 4 and img.shape[1] == 3 and value_check # type: ignore[return-value]

I think a lot of models don't return the [-1,1] or [0,1] as image output, so why do we have to check the range of inputs when calculate lpips?