make datapoint methods private

[pmeier]

Reasons:

1. They should not have been public in the first place. Although they are undocumented, users can still do Image(...).crop(...) and might think this is normal way to do it.
2. They clash with the PyTorch core ops. As of now, this only happens for Datapoint.resize, which overwrites Tensor.resize. However, there might be more of these in the future.

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/vision/7733

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

❌ 1 New Failure

As of commit 58b8eba:

NEW FAILURE - The following job has failed:

• unittests-linux (3.8, linux.g5.4xlarge.nvidia.gpu, cuda, 11.8) / linux-job (gh)

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[pmeier]

This was missed in #7113. We technically don't need it, since both the dispatcher

vision/torchvision/transforms/v2/functional/_misc.py

Lines 65 to 70 in 08c9938

	elif isinstance(inpt, (datapoints.Image, datapoints.Video)):
	return inpt.normalize(mean=mean, std=std, inplace=inplace)
	else:
	raise TypeError(
	f"Input can either be a plain tensor or an `Image` or `Video` datapoint, " f"but got {type(inpt)} instead."
	)

and the transform

vision/torchvision/transforms/v2/_misc.py

Line 162 in 08c9938

_transformed_types = (datapoints.Image, is_simple_tensor, datapoints.Video)

filter on the datapoint type. However, this is the only dispatcher that does not have a method defined on the datapoint base class.

[NicolasHug]

Thanks Philip, LGTM if green

I guess now we need an "official" way of supporting user-defined Datapoints. But we can start thinking about that later.

[pmeier]

Superseded by #7747.