RAT: Bridging RNN Efficiency and Attention Accuracy via Chunk-based Sequence Modeling

Open in new window