Philip Meier

From `torchvision` perspective there is no change needed. The old links work were fully re-instated so older versions of `torchvision` should work as well.

> * on the functionals, passing `fill=0` or `fill=any_color` may result in different results between PIL and Tensors (passing `fill=None` should result in consistent results). Whether those differences are within...

Since we have another report in #8083, do we want to tackle this? IMO, we should just align the two branches https://github.com/pytorch/vision/blob/f69eee6108cd047ac8b62a2992244e9ab3c105e1/torchvision/transforms/v2/functional/_geometry.py#L588-L594 with something like ```python bool_mask = mask <...

I've removed @datumbox from the list above as per his offline request.

We discussed this offline and the consensus is that we can have a go at it for the constructors.

I wrestled with this the last few days and the result is somewhat sobering. Imagine we currently have a transform like ```py import collections.abc from typing import * import torch...

Unfortunately, I wouldn't get my hopes up:  Messaged them multiple times without a response ...

Just dropping in to say that it would be really useful to have something like that. My use case is that part of our CI is unable to solve a...

Seems like a real failure: https://hud.pytorch.org/pr/97010#12167701793

Maybe not. The failure first appears in #96125 and at that point in the stack, the functionality hasn't changed at all. It is then gone in #96332 which fixes something...