Official implementation of SFCLIP: Building Vision-Language Models on Solid Foundations with Masked Distillation - View it on GitHub