DMKD Learning Transferable Visual Models From Natural Language Supervision [240710] MobileCLIP: Fast Image-Text Models through Multi-Modal Reinforced Trainin [240724] Sigmod Loss for Language Image Pre-Training [240814]