dgl._ffi.base.DGLError: Expect number of features to match number of nodes (len(u)). Got 943 and 1682 instead

[Tillzz]

I wrote and modified the code according to my senior’s code, and this problem occurred and i dont know how to resolve it, i wanna get some help. thank you very much

    hg_user = dgl.heterograph({('user', 'um', 'movie'): (mvsu_u, mvsu_m),
                               ('movie', 'mu', 'user'): (mvsu_m, mvsu_u),
                               ('user', 'uu', 'user'): (mvsu_u, mvsu_u),
                               ('movie', 'mm', 'movie'): (mvsu_m, mvsu_m),
                               # ('user', 'umgmu', 'user'): (umgmu_u0, umgmu_u1),  # relation
                               ('movie', 'mg', 'genre'): (mvsg_m, mvsg_g),
                               ('genre', 'gm', 'movie'): (mvsg_g, mvsg_m)
                               })

    user_mp_info = [[['um', 'mu', 'um'], ['mu']],
                    [[-1, -1, -1], [movie_features.shape[1], user_features.shape[1]]]]

[BarclayII]

Is this all the code you have? This looks like an error thrown at node feature assignment and I couldn’t see it in your code snippet. Could you give us a reproducible example and the complete error stack trace?

[Tillzz]

thanks a lot for your help, and after deeply checked code，found that i just need one type of node in new_g, but I dont know how to solve. there is the HIN_Embedding_Layer class below.

[BarclayII]

@Tillzz

found that i just need one type of node in new_g,

What do you mean by “need one type of node in new_g”? Do you mean that the node type doesn’t exist? When creating the graph, you could specify the number of nodes for each type in num_nodes_dict argument in the dgl.heterograph() function.