Local Token Mixing as a Requirement of Linear Recurrent Blocks

the need the short convolution served did not go away, it became implicit — the distinction between deleting a component and absorbing it is the lesson worth keeping

This Concept is waiting for its first lesson!

the need the short convolution served did not go away, it became implicit — the distinction between deleting a component and absorbing it is the lesson worth keeping

Are you a teacher? Sign in to start contributing.