The Problem in the GGSNN-GMN Model

[liaoluoxing]

Hello, dear author.
I meet some problem in the Part 2 of GGSNN-GMN(binary_function_similarity/Models
/GGSNN-GMN/).

When i run the command "Example: run the validation on Dataset-1 using the model_checkpoint that we trained on Dataset-1", there are some errors. The following code is the error message:

Traceback (most recent call last):
  File "/code/gnn.py", line 219, in <module>
    main()
  File "/code/gnn.py", line 209, in main
    model_validate(config)
  File "/code/gnn.py", line 110, in model_validate
    gnn_model.model_validate()
  File "/code/core/gnn_model.py", line 289, in model_validate
    self._model_initialize(training_set, is_training=False)
  File "/code/core/gnn_model.py", line 144, in _model_initialize
    self._create_network(batch_generator, is_training=is_training)
  File "/code/core/gnn_model.py", line 119, in _create_network
    node_feature_dim = first_batch_graphs.node_features.shape[-1]
AttributeError: 'tuple' object has no attribute 'node_features'

I suspect that I did some things wrongly. If you can help me, I would be very appreciate. Thank you a lot.

[iDTer]

The same problem.

[iDTer]

core/gnn_model.py进行如下修改：

def _create_network(self, batch_generator, is_training):
"""Build the model and set _tensors, _placeholders and _model."""
# Automatically infer the node and edge features dim.

    _it = None
    if is_training and self._config['training']['mode'] == 'triplet':
        _it = batch_generator.triplets()
    else:
        _it = batch_generator.pairs()

    first_batch_graphs = None
    if is_training and self._config['training']['mode'] == 'pair':
        first_batch_graphs, _ = next(_it)
    else:
        first_batch_graphs = next(_it)

    # Set the feature dimensions
    # node_feature_dim = first_batch_graphs.node_features.shape[-1]
    # edge_feature_dim = first_batch_graphs.edge_features.shape[-1]

    # Set the feature dimensions. 
    if isinstance(first_batch_graphs, GraphData):
        node_feature_dim = first_batch_graphs.node_features.shape[-1]
        edge_feature_dim = first_batch_graphs.edge_features.shape[-1]
    elif isinstance(first_batch_graphs, tuple):
        node_feature_dim = first_batch_graphs[0].node_features.shape[-1]
        edge_feature_dim = first_batch_graphs[0].edge_features.shape[-1]
    else:
        raise TypeError("Unexpected type for first_batch_graphs")