But RNNs are losing its edge over sequence modeling against fully attentional techniques like Transformer, at least it seems to be the consensus right now, Tran

But RNNs are losing its edge over sequence modeling against fully attentional techniques like Transformer, at least it seems to be the consensus right now, Transformer, when done right, offers better performance.