Self-supervised Transformers for 10x more efficient trainingai.facebook.com2 points·halotrope··0 commentsOpen articleSaveView on HN