A Transformer is a sequence-to-sequence model architecture described in the paper Attention Is All You Need.