Bug in hetero_link_pred.py walkthrough

[sperry304]

Hi there, I'm trying to run the code here and encountered the following issue at two points in the script - when trying to run the initial model.encoder call and then again when trying to train the model. Everything else in this script and the companion CSV loader script seems to work great. Any ideas?

Here's the error message:
https://pastebin.com/AySBvnrG

I did forget one other error that pops up in the torch.bincount call:
RuntimeError: bincount only supports 1-d non-negative integral inputs.

Here's the shape of the final training data object:
HeteroData(
user={ num_nodes=610 },
movie={ x=[9742, 404] },
(user, rates, movie)={
edge_index=[2, 85712],
edge_label=[85712, 1],
edge_label_index=[2, 85712]
},
(movie, rev_rates, user)={ edge_index=[2, 85712] }
)

NOTE: The error seems to be caused by something from the load CSV script, not this one. Sorry and please feel free to delete.

[rusty1s]

Mh, this works fine for me on PyG master. Which version are you using?

[sperry304]

Mh, this works fine for me on PyG master. Which version are you using?

You're right, I was using some slightly differently formatted data files from the load CSV script. This can go ahead and be closed :)