pix2seq icon indicating copy to clipboard operation
pix2seq copied to clipboard

Question about inference

Open SY-Xuan opened this issue 3 years ago • 2 comments

During inference, the token (5-th) that is corresponded to the object class may be classified to the coordinates. In the other hand, the token that is corresponded to the coordinates still has chance to be classified to the class of the object. How to deal with such situation? Thanks a lot.

SY-Xuan avatar Jul 15 '22 07:07 SY-Xuan

you can offset logits before sampling if you want to disable certain predictions (e.g. set -1e9 for logits that correspond to class/coordinates). but we find that just free sampling is fine.

chentingpc avatar Jul 16 '22 17:07 chentingpc

Thanks for your kindly replying!

SY-Xuan avatar Jul 17 '22 11:07 SY-Xuan