gncs
Where is n_node generated?
Hello
I am trying to run the template dataset for training the ISO17 dataset, and everything works fine but it occur the following issue when I am running the training command provided on the github page
"(graphdg-venv) popzq@GPUSrv:~/桌面/graphdg-master$ python3 ./scripts/run.py --train_path=iso17_split-0_train.pkl --test_path=iso17_split-0_test.pkl
2023-09-15 20:46:36.714 INFO: {
"augmentation": 1,
"batch_size": 32,
"decoder_depth": 2,
"decoder_width": 50,
"edge_core_depth": 2,
"edge_core_width": 50,
"edge_encoder_depth": 2,
"edge_encoder_width": 50,
"hidden_feats_edges": 10,
"hidden_feats_nodes": 10,
"latent_dims": 1,
"load_model": false,
"log_dir": "logs",
"log_level": "INFO",
"loss_threshold": 10,
"max_num_epochs": 100,
"model_dir": "models",
"name": "graphdg",
"node_core_depth": 2,
"node_core_width": 50,
"node_encoder_depth": 2,
"node_encoder_width": 50,
"num_core_models": 3,
"num_samples": 100,
"results_dir": "results",
"seed": 0,
"test_path": "iso17_split-0_test.pkl",
"train_path": "iso17_split-0_train.pkl",
"valid_fraction": 0.2,
"xyz_dir": "xyz"
}
2023-09-15 20:46:36.714 INFO: Loading training and test datasets
2023-09-15 21:05:56.443 INFO: Dataset sizes: 364256 train, 30 test
2023-09-15 21:08:46.546 INFO: Graph input dimensions: GraphFeatureDimensions(edges=21, nodes=12, globals=0)
graph_dims: GraphFeatureDimensions(edges=21, nodes=12, globals=0)
config: {'name': 'graphdg', 'seed': 0, 'log_dir': 'logs', 'model_dir': 'models', 'results_dir': 'results', 'xyz_dir': 'xyz', 'train_path': 'iso17_split-0_train.pkl', 'test_path': 'iso17_split-0_test.pkl', 'valid_fraction': 0.2, 'num_samples': 100, 'augmentation': 1, 'hidden_feats_edges': 10, 'hidden_feats_nodes': 10, 'latent_dims': 1, 'node_encoder_width': 50, 'node_encoder_depth': 2, 'edge_encoder_width': 50, 'edge_encoder_depth': 2, 'node_core_width': 50, 'node_core_depth': 2, 'edge_core_width': 50, 'edge_core_depth': 2, 'num_core_models': 3, 'decoder_width': 50, 'decoder_depth': 2, 'batch_size': 32, 'max_num_epochs': 100, 'loss_threshold': 10, 'load_model': False, 'log_level': 'INFO'}
hidden_dims: GraphFeatureDimensions(edges=22, nodes=12, globals=0)
node_encoder: <graphdg.mlp.MLP object at 0x7fa31f6c3b90>
edge_encoder: <graphdg.mlp.MLP object at 0x7fa31f6c3b90>
hidden_dims: GraphFeatureDimensions(edges=21, nodes=13, globals=0)
node_encoder: <graphdg.mlp.MLP object at 0x7fa31fb69710>
edge_encoder: <graphdg.mlp.MLP object at 0x7fa31fb69710>
p_hidden_dims: GraphFeatureDimensions(edges=31, nodes=23, globals=0)
p_in_dims: GraphFeatureDimensions(edges=21, nodes=13, globals=0)
graph_dims: GraphFeatureDimensions(edges=21, nodes=12, globals=0)
latent_dims 1
q_gnn <graphdg.graph.GraphNet object at 0x7fa31f880910>
p_gnn <graphdg.graph.GraphNet object at 0x7fa31f86e050>
Traceback (most recent call last):
File "./scripts/run.py", line 258, in n_node cannot be None")
ValueError: Field n_node cannot be None
"
Would you mind giving any clue where the n_node is generated and how to make sure it is not none?
