Is your feature request related to a problem? Please describe.
For digital pathology training pipeline to use torch.Tensor, there are still two transform that are not supporting this input types RandRotate90 and RandZoomD. I know that there have been efforts on this front, so I am wondering if these transforms are in the pipeline? If so, is it possible to prioritize them? If not, is there anything that holding us from doing that besides time?