Why positional encoding is added to different role in encoder and decoder.

facebookresearch / detr

End-to-End Object Detection with Transformers

Apache License 2.0

13.08k stars 2.37k forks source link

Open Kami-chanw opened 8 months ago

Kami-chanw commented 8 months ago

In figure above, positional encoding is add to K Q, but why not add to V? Besides, positional encoding is add to K (but not include Q and V)?

psc1453 commented 7 months ago

Same question