cleanup ported v2 tests

[pmeier]

This does couple of things:

1. Enable make_image to return a fake channels last image. For PyTorch core only 4D tensors can be channels last, but most of the time we test 3D ones. Setting memory_format=torch.channels_last without batch dims, will now return the memory laid out correctly.
2. Fix the naming of the test helper to create datapoints.BoundingBoxes. We currently have

   • vision/test/common_utils.py

     Line 644 in e0e6f7e

     def make_bounding_box(

     
     This is what we should use in terms of functionality

   • vision/test/common_utils.py

     Line 719 in e0e6f7e

     make_bounding_boxes = from_loaders(make_bounding_box_loaders)

     
     This is the legacy function that creates an iterator over multiple datapoints.BoundingBoxes.
3. Remove some obsolete helper functions and tests from test/test_transforms_v2_functional.py
4. Port the remaining correctness tests for F.resize from test/test_transforms_v2_functional.py to test/test_transforms_v2_refactored.py
5. Refactor the affine bounding box helper to also accept expand=True and datapoints.BoundingBoxes directly, rather than a torch.Tensor and explicit metadata.
6. Use the result from 5. to implement reference tests for F.resize and F.rotate

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/vision/7723

• 📄 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 37b1458 with merge base 4cba51c ():

NEW FAILURE - The following job has failed:

• pytorch/vision / wheel-py3_8-cpu-aarch64 (gh)

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

[pmeier]

Superseded by a few smaller PRs.