[TMLR] HyperCLIP: Prompt-Conditioned Image Encoders for Contrastive Vision-Language Pre-training - View it on GitHub