INNER CODE UNIT · Python

get_dgl_graph

IBM/transition-amr-parser · src/ibm_neural_aligner/main.py:447

    def get_dgl_graph(self, amr):
        vocab = self.amr_tokenizer.vocab

        # init graph
        g = dgl.DGLGraph()

        # get tree
        tree = convert_amr_to_tree(amr)

        # add node structure
        node_ids = get_node_ids(amr)
        node_TO_idx = {k: i for i, k in enumerate(node_ids)}
        N = len(node_ids)
        g.add_nodes(N)

        # add node features
        node_labels = [amr.nodes[k] for k in node_ids]
        node_tokens = torch.tensor([vocab[tok] for tok in node_labels], dtype=torch.long)

View source record →