Pytorch implementation of the paper "Self-Attention with Relative Position Representations"
For the entire Seq2Seq framework, you can refer to this repo.