Try creating homogeneous from csv files but got DGLHeteroGraph

[RtekiX]

nodes = pd.read_csv(’./nodes.csv’)

node_feat = th.tensor(nodes[[‘this is feature name list’]].to_numpy(), dtype=th.float32)

node_label = th.tensor(nodes[[‘label’]].to_numpy(), dtype=th.float32)

edges = pd.read_csv(’./edges.csv’)

src = th.tensor(edges[‘src’].to_numpy(), dtype=th.int32)

dst = th.tensor(edges[‘dst’].to_numpy(), dtype=th.int32)

weight = th.tensor(edges[‘weight’].to_numpy(), dtype=th.float32)

g = dgl.graph((src, dst))

g.ndata[‘x’] = node_feat

g.ndata[‘y’] = node_label

g.edata[‘w’] = weight


I tried creating a homogeneous DGLGraph from code above, but print(type(g)) got <class ‘dgl.heterograph.DGLHeteroGraph’>. Is this common?

nodes.csv format: feat1, feat2, …, featN, label

edges.csv format: src, dst, weight

dgl version: dgl-cuda11.3 0.8.1

[Rhett-Ying]

this is expected as Homogeneous is just a special case of Heterogenous and DGLHeteroGraph is used in the underlying implementation.

[RtekiX]

Thank you. I realized this point after calling DGLGraph.ishomogeneous.