class light_pfp_autogen.active_learning.ActiveLearning(active_learning_config: ActiveLearningConfig)#

Bases: object

Interface for active learning in molecular dynamics (MD) simulations.

clean_up_models(keep_final: bool = True) → None#
initialize() → None#

Initialize the active learning process.
Steps:
1. Perform initial training if not already done.
2. Call _check_previous_iterations method to check the status of previous iterations,

and clean up possible conflicted files.

  1. Call _pre_sampling method to prepare for the first MD sampling.

print_md_statistics(start: int = 0, stop: Optional[int] = None) → None#

Prints the statistics of the MD sampling.

Parameters

  • start (int) – The starting iteration number. Default is 0.

  • stop (int) – The stopping iteration number. If None, it will be set to
    the current iteration. Default is None.

print_training_statistics(start: int = -1, stop: Optional[int] = None) → None#

Prints the energy forces and stress MAE of the models.

Parameters

  • start (int) – The starting iteration number. If -1, it will be set to the initial model.
    Default is -1.

  • stop (int) – The stopping iteration number. If None, it will be set to
    the current iteration. Default is None.

update() → None#

Train the new model with the collected data and prepare for the next MD sampling.
Also, the iteration count is incremented by 1.