CLIP for Language-Image Representation