ViTamin: Designing Scalable Vision Models in the Vision-Language Era | Read Paper on Bytez