pgl.heter_graph_wrapper module: Heterogenous Graph data holders for Paddle GNN.

This package provides interface to help building static computational graph for PaddlePaddle.

class pgl.heter_graph_wrapper.HeterGraphWrapper(name, edge_types, node_feat={}, edge_feat={}, **kwargs)

Bases: object

Implement a heterogeneous graph wrapper that creates a graph data holders that attributes and features in the heterogeneous graph. And we provide interface to_feed to help converting Graph data into feed_dict.

Parameters

• name – The heterogeneous graph data prefix

• node_feat – A dict of list of tuples that decribe the details of node feature tenosr. Each tuple mush be (name, shape, dtype) and the first dimension of the shape must be set unknown (-1 or None) or we can easily use HeterGraph.node_feat_info() to get the node_feat settings.

• edge_feat – A dict of list of tuples that decribe the details of edge feature tenosr. Each tuple mush be (name, shape, dtype) and the first dimension of the shape must be set unknown (-1 or None) or we can easily use HeterGraph.edge_feat_info() to get the edge_feat settings.

Examples

import paddle.fluid as fluid
import numpy as np
from pgl import heter_graph
from pgl import heter_graph_wrapper
num_nodes = 4
node_types = [(0, 'user'), (1, 'item'), (2, 'item'), (3, 'user')]
edges = {
    'edges_type1': [(0,1), (3,2)],
    'edges_type2': [(1,2), (3,1)],
}
node_feat = {'feature': np.random.randn(4, 16)}
edges_feat = {
    'edges_type1': {'h': np.random.randn(2, 16)},
    'edges_type2': {'h': np.random.randn(2, 16)},
}

g = heter_graph.HeterGraph(
                num_nodes=num_nodes,
                edges=edges,
                node_types=node_types,
                node_feat=node_feat,
                edge_feat=edges_feat)

gw = heter_graph_wrapper.HeterGraphWrapper(
                    name='heter_graph',
                    edge_types = g.edge_types_info(),
                    node_feat=g.node_feat_info(),
                    edge_feat=g.edge_feat_info())

to_feed(heterGraph, edge_types_list='__ALL__')

Convert the graph into feed_dict.

This function helps to convert graph data into feed dict for fluid.Excecutor to run the model.

Parameters

• heterGraph – the HeterGraph data object

• edge_types_list – the edge types list to be fed

Returns

A dictinary contains data holder names and its coresponding data.