how to code the following graph structure ?

which kind of data structure is suitable in this case for both forward inference and backward propagation ?

Do edges belong to different types? If not, DGLGraph naturally supports multigraphs.

@mufeili yes, the edges are of different arithmetic operations.

However, does the multigraph have support for both forward inference and backward propagation operations ?

If the edges belong to different types and you want to distinguish the types, you will need to construct a heterogeneous graph.

What do you mean by “support for both forward inference and backward propagation operations”? If you simply want to use it with a graph neural network, then yes.