Are Pre-Trained Convolutions Better Than Pre-Trained Transformers? (2021)

2 points| fzliu | 27 days ago |arxiv.org