Transforms on GPU and multi GPU training with PyTorch Lightning

[YerePhy]

Hi,

In PyTorch Lightning official documentation they discourage the use of to or cuda call in the code. However, in order to perform GPU augmenations with MONAI, it is necessary to use monai.transforms.ToDevice which is nothing but a wrapped to call. Is it safe to use GPU data augmentation with MONAI + PyTorch Lightning in multi GPU training, could someone give an explanation?