Deep Contrastive Voice Recognition in Pytorch