mindspore_gl.HeterGraph

class mindspore_gl.HeterGraph

The heterogeneous Graph.

This is the class which should be annotated in construct function for GNNCell class.

property dst_idx

A list of tensor with shape \((N\_EDGES)\), represents the destination node index of COO edge matrix.

get_homo_graph(etype)

Get the specific nodes, edges for etype.

Parameters

etype (int) – The edge type.

Returns

List[Tensor], a homo graph.

Supported Platforms:

GPU

Examples

>>> import mindspore as ms
>>> from mindspore_gl import Graph, HeterGraph, HeterGraphField
>>> from mindspore_gl.nn import GNNCell
>>> n_nodes = [9, 2]
>>> n_edges = [11, 1]
>>> src_idx = [ms.Tensor([0, 2, 2, 3, 4, 5, 5, 6, 8, 8, 8], ms.int32), ms.Tensor([0], ms.int32)]
>>> dst_idx = [ms.Tensor([1, 0, 1, 5, 3, 4, 6, 4, 8, 8, 8], ms.int32), ms.Tensor([1], ms.int32)]
>>> heter_graph_field = HeterGraphField(src_idx, dst_idx, n_nodes, n_edges)
>>> node_feat = ms.Tensor([[1], [2], [1], [2], [0], [1], [2], [3], [1]], ms.float32)
...
>>> class TestSum(GNNCell):
...     def construct(self, x, g: Graph):
...         g.set_vertex_attr({"x": x})
...         for v in g.dst_vertex:
...             v.h = g.sum([u.x for u in v.innbs])
...         return [v.h for v in g.dst_vertex]
...
>>> class TestHeterGraph(GNNCell):
...     def __init__(self):
...         super().__init__()
...         self.sum = TestSum()
...
...     def construct(self, x, hg: HeterGraph):
...         return self.sum(x, *hg.get_homo_graph(0))
...
>>> ret = TestHeterGraph()(node_feat, *heter_graph_field.get_heter_graph()).asnumpy().tolist()
>>> print(ret)
    [[1.0], [2.0], [0.0], [0.0], [3.0], [2.0], [1.0], [0.0], [3.0]]

property n_edges

A list of integer, represent the edges count of the graph.

property n_nodes

A list of integer, represent the nodes count of the graph.

property src_idx

A list of tensor with shape \((N\_EDGES)\), represents the source node index of COO edge matrix.