Tensorboard implementation with Heterogenous Graph

[shahfasal]

Hi Team,
I am using code example to do link prediction.
I have implemented tensorboard functionality to write test and train loss. Now I was trying to add Tensorboard functionality on the model using tensorboard.add_graph

with torch.no_grad():
    model.encoder(train_data.x_dict, train_data.edge_index_dict)  
    writer.add_graph(model, [data.x_dict, data.edge_index_dict])

this throws the error

RuntimeError: Cannot create dict for key type '(str, str, str)', only int, float, complex, Tensor, device and string keys are supported

tensorboard expects tensor while as encoder is taking dictionary input... How can we resolve this to add our model to tensorboard?
Thanks
cc @rusty1s

[rusty1s]

I see. This is indeed a problem. You may be able to fix this by converting the edge type keys into a single string:

edge_index_dict = ['__'.join(edge_type) for edge_type, edge_index in train_data.edge_index_dict.items()]
model.encoder(train_data.x_dict, edge_index_dict)

and re-map to a tuple once again inside the model.