Implementation of attention block in Manet #717
QueLastimaSenor
started this conversation in
General
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hi! I've got a question about implementation of PAB in Manet(https://arxiv.org/pdf/2009.02130.pdf, i suppose it's a kernel attention in the article).


I can't find any resemblance between pab and attention kernel.
It seems like you've written just an attention mechanism. Am i wrong?
Beta Was this translation helpful? Give feedback.
All reactions