Added wrap_dataset_for_transforms_v2 into datasets and handled beta w…

[vfdev-5]

Description:

• Added wrap_dataset_for_transforms_v2 into datasets and handled beta warning
• Added tests

Usage:

python -c "import torchvision; torchvision.disable_beta_transforms_warning(); from torchvision.datasets import wrap_dataset_for_transforms_v2"
> 

python -c "from torchvision.datasets import wrap_dataset_for_transforms_v2"
> 
/vision/torchvision/transforms/v2/__init__.py:54: UserWarning: The torchvision.datapoints and torchvision.transforms.v2 namespaces are still Beta. While we will try our best to maintain backward compatibility, some APIs or behaviors might change without a deprecation cycle. To help us improve these new features, please provide your feedback here: https://github.com/pytorch/vision/issues/6753.You can silence this warning by calling torchvision.disable_beta_transform_warning().
  warnings.warn(_BETA_TRANSFORMS_WARNING)
/vision/torchvision/datapoints/__init__.py:14: UserWarning: The torchvision.datapoints and torchvision.transforms.v2 namespaces are still Beta. While we will try our best to maintain backward compatibility, some APIs or behaviors might change without a deprecation cycle. To help us improve these new features, please provide your feedback here: https://github.com/pytorch/vision/issues/6753.You can silence this warning by calling torchvision.disable_beta_transform_warning().
  warnings.warn(_BETA_TRANSFORMS_WARNING)

python -c "from torchvision.datasets import CIFAR10"
> 

cc @bjuncek @pmeier

[NicolasHug]

This test is failing, but I'd suggest removing both tests.
They can interfere with the import and warnings in ways that we may not predict properly, and this may lead to surprising or undesired behaviour (like masking a warning it shouldn't mask).

I agree we should test this, it can be done with a strategy similar to the one proposed in #7269, but we should probably leave that for next week

[pmeier]

Just a few comments on the tests. Module level __getattr__ for the import looks fine to me to resolve the circular dependency.

[pmeier]

Test are failing though: https://github.com/pytorch/vision/actions/runs/4202967731/jobs/7291773255#step:10:33509

Same for the lint job: https://app.circleci.com/pipelines/github/pytorch/vision/23383/workflows/5bcb63e6-c258-47a8-8f66-cd09db2a43bb/jobs/1834284

[NicolasHug]

Thanks a lot for the PR @vfdev-5 . Some minor comment but the behaviour LGTM. We should also remove wrap_dataset_for_transforms_v2 from datapoints.__init__.py before merging

[pmeier]

LGTM is CI is green. Thanks Victor!

[NicolasHug]

Thanks a lot @vfdev-5 , still LGTM!

[github-actions]

Hey @vfdev-5!

You merged this PR, but no labels were added. The list of valid labels is available at https://github.com/pytorch/vision/blob/main/.github/process_commit.py