Ahmadzei's picture
added 3 more tables for large emb model
5fa1a76
BeIT is trained to predict the visual tokens corresponding to the masked patches.