Contrastive Learning of Auditory Representations