5fa1a76
1
ViTMAE has a similar pretraining objective, except it must predict the pixels instead of visual tokens.
