1111GGGGGGGGGGGGGGGGGGG

/home/ubuntu/.conda/envs/comformer1/bin/python /home/ubuntu/sophia/code/ComFormer/comformer/scripts/train_jarvis.py 
INFO: Pandarallel will run on 16 workers.
INFO: Pandarallel will use Memory file system to transfer data between the main process and workers.
name='iComformer' conv_layers=4 edge_layers=1 atom_input_features=92 edge_features=256 triplet_input_features=256 node_features=256 fc_layers=1 fc_features=256 output_features=1 node_layer_head=1 edge_layer_head=1 nn_based=False link='identity' zero_inflated=False use_angle=False angle_lattice=False classification=False
{'dataset': 'dft_3d', 'target': 'optb88vdw_total_energy', 'epochs': 500, 'batch_size': 64, 'weight_decay': 1e-05, 'learning_rate': 0.001, 'criterion': 'mse', 'optimizer': 'adamw', 'scheduler': 'onecycle', 'save_dataloader': False, 'pin_memory': False, 'write_predictions': True, 'num_workers': 4, 'classification_threshold': None, 'atom_features': 'cgcnn', 'model': {'name': 'iComformer', 'use_angle': False}, 'cutoff': 4.0, 'max_neighbors': 25, 'matrix_input': False, 'pyg_input': True, 'use_lattice': True, 'use_angle': False, 'neighbor_strategy': 'k-nearest', 'output_dir': 'yourpath-jarvis-optb88vdw_total_energy'}
config:
{'atom_features': 'cgcnn',
 'batch_size': 64,
 'classification_threshold': None,
 'criterion': 'mse',
 'cutoff': 4.0,
 'dataset': 'dft_3d',
 'distributed': False,
 'epochs': 500,
 'filename': 'sample',
 'id_tag': 'jid',
 'keep_data_order': False,
 'learning_rate': 0.001,
 'log_tensorboard': False,
 'matrix_input': False,
 'max_neighbors': 25,
 'model': {'angle_lattice': False,
           'atom_input_features': 92,
           'classification': False,
           'conv_layers': 4,
           'edge_features': 256,
           'edge_layer_head': 1,
           'edge_layers': 1,
           'fc_features': 256,
           'fc_layers': 1,
           'link': 'identity',
           'name': 'iComformer',
           'nn_based': False,
           'node_features': 256,
           'node_layer_head': 1,
           'output_features': 1,
           'triplet_input_features': 256,
           'use_angle': False,
           'zero_inflated': False},
 'n_early_stopping': None,
 'n_test': None,
 'n_train': None,
 'n_val': None,
 'neighbor_strategy': 'k-nearest',
 'num_workers': 4,
 'optimizer': 'adamw',
 'output_dir': 'yourpath-jarvis-optb88vdw_total_energy',
 'pin_memory': False,
 'progress': True,
 'pyg_input': True,
 'random_seed': 123,
 'save_dataloader': False,
 'scheduler': 'onecycle',
 'standard_scalar_and_pca': False,
 'store_outputs': True,
 'target': 'optb88vdw_total_energy',
 'target_multiplication_factor': None,
 'test_ratio': 0.1,
 'train_ratio': 0.8,
 'use_angle': False,
 'use_canonize': True,
 'use_lattice': True,
 'val_ratio': 0.1,
 'version': 'NA',
 'warmup_steps': 2000,
 'weight_decay': 1e-05,
 'write_checkpoint': True,
 'write_predictions': True}
Obtaining 3D dataset 55k ...
Reference:https://www.nature.com/articles/s41524-020-00440-1
Loading the zipfile...
fatal: not a git repository (or any of the parent directories): .git
Loading completed.
data range 3.3935 -10.50027
./saved_data/train_datatest_graph_angle.pkl
graphs not saved11111111111111111111111111111111111111
mean -3.214447 std 2.131234
normalize using training mean -3.214447 and std 2.131210
warning: could not load CGCNN features for 103
Setting it to max atomic number available here, 103
warning: could not load CGCNN features for 101
Setting it to max atomic number available here, 103
warning: could not load CGCNN features for 102
Setting it to max atomic number available here, 103
data range 2.53237 -10.32721
./saved_data/val_datatest_graph_angle.pkl
graphs not saved11111111111111111111111111111111111111
mean -3.247577 std 2.129600
normalize using training mean -3.214447 and std 2.131210
data range 2.51899 -10.49825
./saved_data/test_datatest_graph_angle.pkl
graphs not saved11111111111111111111111111111111111111
mean -3.173224 std 2.135519
normalize using training mean -3.214447 and std 2.131210
n_train: 44578
n_val: 5572
n_test: 5572
std train: 2.131210441953068
I am using the correct version of matformer
I am using the correct version of matformer
I am using the correct version of matformer
I am using the correct version of matformer
I am using the invariant version of EPCNet
config:
{'angle_lattice': False,
 'atom_input_features': 92,
 'classification': False,
 'conv_layers': 4,
 'edge_features': 256,
 'edge_layer_head': 1,
 'edge_layers': 1,
 'fc_features': 256,
 'fc_layers': 1,
 'link': 'identity',
 'name': 'iComformer',
 'nn_based': False,
 'node_features': 256,
 'node_layer_head': 1,
 'output_features': 1,
 'triplet_input_features': 256,
 'use_angle': False,
 'zero_inflated': False}
使用使用使用使用使用使用使用使用使用使用使用使用使用使用使用使用使用使用使用使用
使用warmup_steps= 2000
使用config.epochs= 500
使用steps_per_epoch= 696
使用pct_start= 0.005747126436781609
使用warmup_steps= 2000
使用config.epochs= 500
使用steps_per_epoch= 696
Total size: 18.967243194580078
Total trainable parameter number 4966497