Closed rainmaker712 closed 3 years ago
"A fully featured implementation of Routing Transformer. The paper proposes using k-nearest neighbors to route similar queries / keys into the same cluster for attention."
It seems like they use k-means not knn.
It might be my misunderstanding but FYI.
fixed!
lucidrains
commented
3 years ago
"A fully featured implementation of Routing Transformer. The paper proposes using k-nearest neighbors to route similar queries / keys into the same cluster for attention."
It seems like they use k-means not knn.
It might be my misunderstanding but FYI.