Graph2Vec fit method by karateclub

Question

I am struggling with running the graph2vec module by karateclub (or any other provider of a similar one) on my networkx Graph G. Graph2Vec was introduced in this paper and promises to build quite successful embeddings.

So far, my Code looks like that:

from karateclub import Graph2Vec

graph2vec_model = Graph2Vec(
    dimensions=2
)

graph2vec.fit(**G)

I already tried to apply the fit function to multiple versions of my Graph, with one or two asterisks inside the parentheses.

• Error message with one *: TypeError: fit() takes 2 positional arguments but 8513 were given (as my Graph has 8513 nodes)
• Error message with two **: TypeError: karateclub.graph_embedding.graph2vec.Graph2Vec.fit() argument after ** must be a mapping, not MultiDiGraph
• Error message after trying .fit(*[G]) or .fit(G):

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
Input In [75], in <cell line: 1>()
----> 1 graph2vec.fit(*[G])

File /opt/homebrew/lib/python3.9/site-packages/karateclub/graph_embedding/graph2vec.py:66, in Graph2Vec.fit(self, graphs)
     59 """
     60 Fitting a Graph2Vec model.
     61 
     62 Arg types:
     63     * **graphs** *(List of NetworkX graphs)* - The graphs to be embedded.
     64 """
     65 self._set_seed()
---> 66 graphs = self._check_graphs(graphs)
     67 documents = [
     68     WeisfeilerLehmanHashing(
     69         graph, self.wl_iterations, self.attributed, self.erase_base_features
     70     )
     71     for graph in graphs
     72 ]
     73 documents = [
     74     TaggedDocument(words=doc.get_graph_features(), tags=[str(i)])
     75     for i, doc in enumerate(documents)
     76 ]

File /opt/homebrew/lib/python3.9/site-packages/karateclub/estimator.py:64, in Estimator._check_graphs(self, graphs)
     62 def _check_graphs(self, graphs: List[nx.classes.graph.Graph]):
     63     """Check the Karate Club assumptions for a list of graphs."""
---> 64     graphs = [self._check_graph(graph) for graph in graphs]
     66     return graphs

File /opt/homebrew/lib/python3.9/site-packages/karateclub/estimator.py:64, in <listcomp>(.0)
     62 def _check_graphs(self, graphs: List[nx.classes.graph.Graph]):
     63     """Check the Karate Club assumptions for a list of graphs."""
---> 64     graphs = [self._check_graph(graph) for graph in graphs]
     66     return graphs

File /opt/homebrew/lib/python3.9/site-packages/karateclub/estimator.py:57, in Estimator._check_graph(self, graph)
     55 def _check_graph(self, graph: nx.classes.graph.Graph) -> nx.classes.graph.Graph:
     56     """Check the Karate Club assumptions about the graph."""
---> 57     self._check_indexing(graph)
     58     graph = self._ensure_integrity(graph)
     60     return graph

File /opt/homebrew/lib/python3.9/site-packages/karateclub/estimator.py:50, in Estimator._check_indexing(graph)
     47 @staticmethod
     48 def _check_indexing(graph: nx.classes.graph.Graph):
     49     """Checking the consecutive numeric indexing."""
---> 50     numeric_indices = [index for index in range(graph.number_of_nodes())]
     51     node_indices = sorted([node for node in graph.nodes()])
     53     assert numeric_indices == node_indices, "The node indexing is wrong."

AttributeError: 'int' object has no attribute 'number_of_nodes'

Why, and especially where does my Graph get converted to a int object? Using number_of_nodes() on my normal graph G works just fine.

Sorry for this messy documentation, could anyone help me out here? How do I need to use the fit method or how do I need to transform my graph to fit in?

Thanks a lot