whether the training of EVA involves masking text (caption) token?

baaivision / EVA

EVA Series: Visual Representation Fantasies from BAAI

MIT License

2.31k stars 167 forks source link

whether the training of EVA involves masking text (caption) token? #135

Closed leyangjin closed 9 months ago

leyangjin commented 10 months ago

I am new to this area. Just want to check that whether the training of EVA model involves masking text (caption) token, or the training of EVA model only involves masking image patches. Thank you so much for your help.

Quan-Sun commented 10 months ago

@leyangjin only masking image patches.