Nicolas Hug

Looks like this can be closed now that https://github.com/pytorch/vision/pull/8615 is merged. Thanks @lqrhy3 for the issue and @venkatram-dev for the PR!

Thanks for the report @cwestergren . Yeah, sadly the official MNIST mirror from Lecun is down. And the existing alternative is down too (well, it leads to downloading empty files)....

Thanks for the PR @larryliu0820 and sorry for the late reply, I was away last week. Support for `torch.export` is still an ongoing effort for the v2 transforms and we...

Thanks for the report @occipita . And sorry for the weird user experience. AFAICT this was introduced in https://github.com/pytorch/vision/pull/6943. I have no idea why it was done this way, this...

Hi @TonyCongqianWang , > I copy pasted the custom transform from your [tutorial page](https://pytorch.org/vision/stable/auto_examples/transforms/plot_custom_transforms.html#:~:text=How%20to%20write%20your%20own%20v2%20transforms%20Note,from%20torchvision%20import%20tv_tensors%20from%20torchvision.transforms%20import%20v2) and inserted it into the transform pipeline in your reference/detection/presets.py script. Can you share a minimal...

Thanks for the report @N00bcak , I agree this is insufficiently documented. I'm happy to review a PR to improve the docs.

Sorry folks, we don't own the conda packages anymore :/ If using `pip` works (which seems to be the case), then do that if you can.

Hi @jameszhouyi , torchvision doesn't support 3.13t yet

The failures are tracked in https://github.com/pytorch/hub/issues/317 > could be easily fixed Easy-fix PRs are more than welcome ;)

Hi @Heasterian , it looks like you are passing GPU tensors. This means JPEG will rely on `decode_jpeg()` and `encode_jpeg()`, which only support CUDA devices, not ROCm, becuase nvjpeg is...