From_networkx return none

Questions
#1

I have a list of edge，I use nx.from_edgelist to create a networkx graph and the networkx graph was seemingly created correctlly，but then I use the from_networkx function in DGL to get a DGLGrpah but the result was None，and here is my code，was anything wrong？

g =  dgl.DGLGraph()
graph = nx.from_edgelist(edge_list)  
g.from_networkx(graph)

I print nx.info(graph) and the result was
Name:
Type: Graph
Number of nodes: 1915
Number of edges: 3086
Average degree: 3.2230

#2

I’m not sure what’s the problem. However you can directly add edges to DGLGraph by g.add_edges(src_list, dst_list) after g.add_nodes(num_nodes).

#3

I check the pytorch code, it seems that if I add an operation .to(self.deivce) on the DGLgraph，the result will be None. However，for a tensor called features, features.to(self.deivce) works fine. Does the operation violate some rules of DGL？

#4

g.to(device) is an inplace operation. (Maybe we should change this). Therefore it doesn’t return anything, but directly move g to devices.