Scaling Language-Image Pre-training via Masking - 42Papers