Wav2Vec-Aug: Improved self-supervised training with limited data