Check Type Annotations for transforms/functional_pil and transforms/functional_tensor

[oke-aditya]

🚀 Feature

As spotted in #4234 (review)
There are some minor cases where type annotations do not match.

We should probably rectify them and ensure that the types annotations match.

P.S.
But I'm willing to leave for any new contributor! Otherwise I will complete it in the next week

cc @vfdev-5 @datumbox @pmeier I think this can be a good first issue for people awaiting to contribute 😄

[pmeier]

After some more thought, this is probably a torch.jit issue in the sense that Union is not supported. We could try to change the annotation for the tensor transform, but my guess is that the tests will fail.

[oke-aditya]

Probably no new contributor seems to take this So I will have go this week.

[lezwon]

hi @oke-aditya I see the type annotations do not match here too:

vision/torchvision/prototype/transforms/functional/_geometry.py

Lines 235 to 241 in 97385df

	def _center_crop_parse_output_size(output_size: List[int]) -> List[int]:
	if isinstance(output_size, numbers.Number):
	return [int(output_size), int(output_size)]
	elif isinstance(output_size, (tuple, list)) and len(output_size) == 1:
	return [output_size[0], output_size[0]]
	else:
	return list(output_size)

I could make a PR to fix this.

[datumbox]

@lezwon I believe this might be intentional to support JIT. :(

If it can be fixed without breaking the JIT-scriptability that's great. But if that's not possible, we should leave as-is to allow inference on mobile devices.