Some question about ATRank

[crazygirlfym]

It seems that you atrank is different from the one described in paper. For example, bilinear attention is used in paper, but scale-dot attention here. Vanilla attention in paper, but multi-head attention here.

[lizy124]

Query comes from the decoding layer, key and value comes from the encoding layer and is called vanilla attention (which the paper didn't say), the most basic attention. Query, key and value are all from the encoding layer called self attention.