Position Encoding of EVA-2

baaivision / EVA

EVA Series: Visual Representation Fantasies from BAAI

MIT License

2.31k stars 167 forks source link

Open vateye opened 1 year ago

vateye commented 1 year ago

Hi, I found the EVA-2 adopts both learnable positional embedding and rotary embedding at the same time. Why do you choose such design? Thanks.

Jiaqi-Chen-00 commented 1 year ago

Hi, how do I modify this rotary embedding when my image input shape is not H == W?