Updating torchvision tutorial using Transforms API v2

[vfdev-5]

Description

• Updating torchvision tutorial using Transforms API v2

TODO:

• fix exec error with COCO eval (Update coco_utils.py vision#7869)
• update _static/tv-training-code.py
• update _static/torchvision_finetuning_instance_segmentation.ipynb or remove as unused and outdated? => decided to skip this and opt for sphinx gallery way of doing things.

Checklist

• The issue that is being fixed is referred in the description (see above "Fixes #ISSUE_NUMBER")
• Only one issue is addressed in this pull request
• Labels from the issue that this PR is fixing are added to this pull request
• No unnecessary issues are included into this pull request.

cc @datumbox @nairbv @fmassa @NicolasHug @YosuaMichael

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/tutorials/2533

• 📄 Preview Python docs built from this PR

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

✅ No Failures

As of commit 615bf83 with merge base 827843f ():
💚 Looks good so far! There are no failures yet. 💚

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

[vfdev-5]

@svekars can please hint where the source of the notebook when we open it in colab: https://colab.research.google.com/github/pytorch/tutorials/blob/gh-pages/_downloads/torchvision_finetuning_instance_segmentation.ipynb ?

Seems like _static/torchvision_finetuning_instance_segmentation.ipynb is not used and looks very out-dated. We can remove it, maybe?

[NicolasHug]

Thanks a lot @vfdev-5 , I made a first pass but only have minor comments, this looks good

[NicolasHug]

Thanks @vfdev-5 , the changes LGTM

[vfdev-5]

@svekars we are planning to rewrite this tutorial as a python file (same as a lot of other tutorials, e.g. https://github.com/pytorch/tutorials/blob/main/intermediate_source/spatial_transformer_tutorial.py) and rely on sphinx gallery to render the content and ipynb file.
Do you have any thoughts / suggestions about that, also concerning CI part ? For this tutorial we need to download the dataset, model weights and run the training/evaluation. Locally the training on recent GPUs takes less than 5 minutes (we can also reduce the number of epochs).