QGNN_Train_Terminal.txt

This is an example showing terminal output when training the QGNN.

Going in order, and ignoring warnings:

The printed dictionary is your gnn_train.yaml file.

Then, CPU or GPU is printed, depending on you choice of device.

Next, an example graph is printed to make sure your choice of cuts is what you want.

Then, the layers of your network are printed. Note the Quantum Torch Layer from Pennylane in the edge and node networks.

Finally, every 10 graphs, a progress report will print.



(acorn) [03:21] mcamp-notebook-1:~/acorn/examples/pennylane/standalone_gnn $ python3 qgnn_run.py gnn_train.yaml qgnn_testtt
WARNING:root:FRNN is not available, install it at https://github.com/murnanedaniel/FRNN. Using PyG radius instead.
WARNING:root:FRNN is not available, as no GPU is available
WARNING:root:cuDF not found, using pandas instead
{'stage': 'edge_classifier', 'model': 'InteractionGNN', 'input_dir': '/home/mcamp/acorn/examples/pennylane/module_map/', 'stage_dir': '/home/mcamp/acorn/examples/Example_1/gnn/', 'project': 'CF_Example_1_GNN', 'gpus': 1, 'nodes': 1, 'data_split': [80, 10, 10], 'num_workers': [0, 0, 0], 'weighting': [{'weight': 2.0, 'conditions': {'y': True, 'pt': [1000, inf], 'nhits': [3, inf], 'primary': True, 'pdgId': ['not_in', [11, -11]]}}], 'hard_cuts': {'pt': [5000, inf]}, 'edge_cut': 0.5, 'undirected': True, 'node_features': ['r', 'phi', 'z'], 'node_scales': [1000, 3.14, 1000], 'hidden': 128, 'n_graph_iters': 8, 'nb_node_layer': 2, 'nb_edge_layer': 2, 'emb_channels': 0, 'layernorm': True, 'batchnorm': False, 'aggregation': ['sum', 'mean', 'max', 'std'], 'hidden_activation': 'SiLU', 'output_activation': 'Tanh', 'warmup': 5, 'lr': 0.001, 'factor': 0.9, 'patience': 10, 'max_epochs': 100} 

cpu 

Data(eta_angle_1=[245066], norm_x_1=[245066], norm_y_1=[245066], x=[245066], region=[245066], hit_id=[245066], y=[2392], cluster_x_2=[245066], norm_x_2=[245066], r=[245066], module_id=[245066], phi_angle_1=[245066], norm_y_2=[245066], phi_angle_2=[245066], cluster_z_1=[245066], norm_z_2=[245066], eta_angle_2=[245066], eta=[245066], cluster_y_1=[245066], cluster_z_2=[245066], cluster_y_2=[245066], cluster_x_1=[245066], phi=[245066], norm_z_1=[245066], z=[245066], track_edges=[2, 842], pdgId=[842], radius=[842], nhits=[842], particle_id=[842], pt=[842], primary=[842], config=[1], event_id='000000101', edge_index=[2, 2392], truth_map=[842], weights=[2392]) 

InteractionGNN(
  (aggregation): MultiAggregation([
    SumAggregation(),
    MeanAggregation(),
    MaxAggregation(),
    StdAggregation(),
  ], mode=cat)
  (node_encoder): Sequential(
    (0): Linear(in_features=3, out_features=128, bias=True)
    (1): LayerNorm((128,), eps=1e-05, elementwise_affine=False)
    (2): SiLU()
    (3): Linear(in_features=128, out_features=128, bias=True)
    (4): LayerNorm((128,), eps=1e-05, elementwise_affine=False)
    (5): Tanh()
  )
  (edge_encoder): Sequential(
    (0): Linear(in_features=256, out_features=128, bias=True)
    (1): LayerNorm((128,), eps=1e-05, elementwise_affine=False)
    (2): SiLU()
    (3): Linear(in_features=128, out_features=128, bias=True)
    (4): LayerNorm((128,), eps=1e-05, elementwise_affine=False)
    (5): Tanh()
  )
  (edge_network): Sequential(
    (0): Linear(in_features=384, out_features=128, bias=True)
    (1): LayerNorm((128,), eps=1e-05, elementwise_affine=False)
    (2): SiLU()
    (3): Linear(in_features=128, out_features=4, bias=True)
    (4): <Quantum Torch Layer: func=qnode>
    (5): Linear(in_features=4, out_features=128, bias=True)
    (6): LayerNorm((128,), eps=1e-05, elementwise_affine=False)
    (7): Tanh()
  )
  (node_network): Sequential(
    (0): Linear(in_features=1152, out_features=128, bias=True)
    (1): LayerNorm((128,), eps=1e-05, elementwise_affine=False)
    (2): SiLU()
    (3): Linear(in_features=128, out_features=4, bias=True)
    (4): <Quantum Torch Layer: func=qnode>
    (5): Linear(in_features=4, out_features=128, bias=True)
    (6): LayerNorm((128,), eps=1e-05, elementwise_affine=False)
    (7): Tanh()
  )
  (output_edge_classifier): Sequential(
    (0): Linear(in_features=384, out_features=128, bias=True)
    (1): LayerNorm((128,), eps=1e-05, elementwise_affine=False)
    (2): SiLU()
    (3): Linear(in_features=128, out_features=128, bias=True)
    (4): LayerNorm((128,), eps=1e-05, elementwise_affine=False)
    (5): SiLU()
    (6): Linear(in_features=128, out_features=1, bias=True)
  )
) 

epoch 0, graph 1, loss 0.9598536491394043, forward time 8.306082725524902, back time 21.870222330093384,