experiments/codes/utils/data.py [391:415]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
                elem_edges = meta_graph["edges"]
                # populate edge ids
                for elem in elem_edges:
                    if elem[0] not in node2id:
                        node2id[elem[0]] = len(node2id)
                    if elem[1] not in node2id:
                        node2id[elem[1]] = len(node2id)
                edge_mapping = torch.zeros(
                    (len(self.label2id), len(node2id) + len(elem_edges))
                ).long()
                num_nodes = len(node2id)
                edge_ct = num_nodes
                edge_indicator = [0 for ni in range(num_nodes)]
                for ei, elem in enumerate(elem_edges):
                    edges.append([node2id[elem[0]], num_nodes + ei])
                    edges.append([num_nodes + ei, node2id[elem[1]]])
                    edge_mapping[self.get_label2id(elem[2])][num_nodes + ei] = 1
                    edge_ct += 1
                    # we are adding 1 to the edge indicator to keep the first position common for nodes
                    edge_indicator.append(self.get_label2id(elem[2]) + 1)
                x = torch.arange(edge_ct).unsqueeze(1)
                # torch.nn.init.xavier_uniform_(x, gain=1.414)
                edge_index = list(zip(*edges))
                edge_index = torch.LongTensor(edge_index)  # 2 x num_edges
                if edge_index.dim() != 2:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



graphlog/dataset.py [211:235]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
                elem_edges = meta_graph["edges"]
                # populate edge ids
                for elem in elem_edges:
                    if elem[0] not in node2id:
                        node2id[elem[0]] = len(node2id)
                    if elem[1] not in node2id:
                        node2id[elem[1]] = len(node2id)
                edge_mapping = torch.zeros(
                    (len(self.label2id), len(node2id) + len(elem_edges))
                ).long()
                num_nodes = len(node2id)
                edge_ct = num_nodes
                edge_indicator = [0 for ni in range(num_nodes)]
                for ei, elem in enumerate(elem_edges):
                    edges.append([node2id[elem[0]], num_nodes + ei])
                    edges.append([num_nodes + ei, node2id[elem[1]]])
                    edge_mapping[self.get_label2id(elem[2])][num_nodes + ei] = 1
                    edge_ct += 1
                    # NOTE: We are adding 1 to the edge indicator to keep the first position common for nodes
                    edge_indicator.append(self.get_label2id(elem[2]) + 1)
                x = torch.arange(edge_ct).unsqueeze(1)
                edge_index = list(zip(*edges))
                edge_index = torch.LongTensor(edge_index)  # type: ignore
                # 2 x num_edges
                if edge_index.dim() != 2:  # type: ignore
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -