Help with Key Error when Running Examples from Readme

[hohoCode]

From readme instructions, I tried the listed examples: python gnn_node.py --dataset rel-f1 --task driver-position , and also: python gnn_node.py --dataset rel-avito --task ad-ctr

My environment has fresh installations of latest relbench (v1.1)/pyg etc.

But both runs give me key error: "Tried to collect 'num_sampled_nodes' but did not find any occurrences of it in any node and/or edge type", as below.

../lib/python3.10/site-packages/relbench/modeling/utils.py:14: FutureWarning: casting datetime64[ns] values to int64 with .astype(...) is deprecated and will raise in a future version. Use .view(...) instead.
  unix_time = ser.astype("int64").values
  0%|                                                                                                                                                                                         | 0/10 [00:00<?, ?it/s]
Traceback (most recent call last):
  File "/home/yye/relbench/examples/gnn_node.py", line 193, in <module>
    train_loss = train()
  File "/home/yye/relbench/examples/gnn_node.py", line 133, in train
    pred = model(
  File ".../python/torch/2/0/dist/lib/python3.10/torch/nn/modules/module.py", line 1501, in _call_impl
    return forward_call(*args, **kwargs)
  File "/home/yye/relbench/examples/model.py", line 102, in forward
    batch.num_sampled_nodes_dict,
  File ".../python3.10/site-packages/torch_geometric/data/hetero_data.py", line 161, in __getattr__
    return self.collect(key[:-5])
  File ".../python3.10/site-packages/torch_geometric/data/hetero_data.py", line 565, in collect
    raise KeyError(f"Tried to collect '{key}' but did not find any "
KeyError: "Tried to collect 'num_sampled_nodes' but did not find any occurrences of it in any node and/or edge type"

[rishabh-ranjan]

I was unable to reproduce this error with a fresh installation. @weihua916 @rusty1s any ideas what might be going on?

[hohoCode]

What is your pyg version? Mine is the latest version.

[quang-truong]

Have you tried to set subgraph_type=directional for NeighborLoader? Also, the current baseline's GNN doesn't take num_sampled_nodes_dict and num_sampled_edges_dict as arguments, so I think you can safely remove it.

def forward(
        self,
        x_dict: Dict[NodeType, Tensor],
        edge_index_dict: Dict[NodeType, Tensor],
        num_sampled_nodes_dict: Optional[Dict[NodeType, List[int]]] = None,
        num_sampled_edges_dict: Optional[Dict[EdgeType, List[int]]] = None,
    ) -> Dict[NodeType, Tensor]:
        for _, (conv, norm_dict) in enumerate(zip(self.convs, self.norms)):
            x_dict = conv(x_dict, edge_index_dict)
            x_dict = {key: norm_dict[key](x) for key, x in x_dict.items()}
            x_dict = {key: x.relu() for key, x in x_dict.items()}

        return x_dict