AMRBART (large-sized model)
AMRBART model is continually pre-trained on the English text and AMR Graphs based on the BART model. It was introduced in the paper: Graph Pre-training for AMR Parsing and Generation by bai et al. and first released in this repository.