Convert NetworkX heterogeneous Multigraph into PyTorch Geometric data

Question

I have a heterogeneous graph with different node types and attributes stored as NetworkX Multigraph. I want to train a GNN model on this graph, therefore I need to convert it into PyTorch Geometric datatype.

I tried built-in function

from torch_geometric.utils.convert import from_networkx
data_PyG = from_networkx(G)

But it only works if G is networkx.Graph or networkx.DiGraph with the same set of attributes. Is there a way to automatically convert nx.Multigraph into HeteroData()?