Use DeepStack with PyTorch

[Etoye]

Hey, thanks a lot for the cool library. I am looking forward to using it for experiments with ensembling my models, but with a PyTorch as main training library. After going through code and especially deepstack.base.Member could notice a Member class. What format of train_probs, val_probs, and other parameters must be? Will be great and useful, probably not only for me, if you could include non-keras example and show everyone proper utilization of the DeepStack. Thanks!