Hi, TY for the interesting work. Would you be able to provide the zeroshot accuracies for the CLIP models you trained (E.G. ViT-B-32)?