Is there a way to use one hot encoded targets in Pytorch for Mixup/Cutmix?

Question

So as far I as know regularization techniques like mixup and cutmix need one hot encoded targets, Pytorch loss functions seem to accept the labels directly, so I wrote a quick c