Neural network

In this section, you can find the information of neural network in SIMPLE-NN.

Running mode

• train
• train_list
• valid_list
• test
• test_list
• add_NNP_ref
• ref_list
• train_atomic_E
• test_atomic_E
• use_force
• use_stress
• shuffle_dataloader

Network

• nodes
• acti_func
• double_precision
• weight_initializer
• dropout
• use_scale
• use_pca
• use_atomic_weights
• weight_modifier

Optimization

• optimizer
• batch_size
• full_batch
• total_epoch
• learning_rate
• decay_rate
• l2_regularization

Loss function

• loss_scale
• E_loss_type
• F_loss_type
• energy_coeff
• force_coeff
• stress_coeff

Logging & saving

• show_interval
• save_interval
• energy_criteria
• force_criteria
• stress_criteria
• print_structure_rmse

Continue

• continue
• clear_prev_status
• clear_prev_optimizer
• start_epoch

Parallelism

• use_gpu
• GPU_number
• inter_op_threads
• intra_op_threads
• subprocesses