Active

mlptrain.training.active.train(mlp: MLPotential, method_name: str, selection_method: SelectionMethod = <mlptrain.training.selection.AbsDiffE object>, max_active_time: float = 1000, n_configs_iter: int = 10, temp: float = 300.0, max_e_threshold: Optional[float] = None, max_active_iters: int = 50, n_init_configs: int = 10, init_configs: Optional['mlptrain.ConfigurationSet'] = None, fix_init_config: bool = False, bbond_energy: Optional[dict] = None, fbond_energy: Optional[dict] = None, init_active_temp: Optional[float] = None, min_active_iters: int = 1, bias_start_iter: int = 0, restart_iter: Optional[int] = None, inherit_metad_bias: bool = False, constraints: Optional[List] = None, bias: mlptrain.Bias | mlptrain.PlumedBias | None = None, md_program: str = 'ASE', pbc: bool = False, box_size: Optional[list] = None, keep_al_trajs: bool = False) → None

Train a system using active learning, by propagating dynamics using ML driven molecular dynamics (MD) and adding configurations based on some selection criteria. Loop looks something like

Generate configurations -> train a MLP -> run MLP-MD -> frames

^ | |________ calc true _________

Active learning will loop until either:

  1. the iteration > max_active_iters

  2. no configurations are found to add

dftb, orca, gpaw

Parameters:

  • selection_method – Method used to select active learnt configurations

  • max_active_time – (float) Maximum propagation time in the active learning loop. Default = 1 ps

  • n_configs_iter – (int) Number of configurations to generate per active learning cycle

  • temp – (float) Temperature in K to propagate active learning at - higher is better for stability but requires more training

  • max_e_threshold – (float) Maximum relative energy threshold for configurations to be added to the training data

  • max_active_iters – (int) Maximum number of active learning iterations to perform. Will break if we hit the early stopping criteria

  • n_init_configs – (int) Number of initial configurations to generate, will be ignored if init_configs is not None

  • init_configs – (gt.ConfigurationSet) A set of configurations from which to start the active learning from

  • fix_init_config –

    (bool) Always start from the same initial configuration for the active learning loop. If False then:

    1. The structure with the lowest energy (true

    energy) is used

    1. If constraints and/or biases are attached, then

    the structure with the lowest biased energy (true energy + bias energy) is used

    1. If using the option to inherit metadynamics

    bias, then the structure with the lowest inherited biased energy (true energy + inherited bias energy) is used.

  • bbond_energy – (dict | None) Additional energy to add to a breaking bond. e.g. bbond_energy={(0, 1), 0.1} Adds 0.1 eV to the ‘bond’ between atoms 0 and 1 as velocities shared between the atoms in the breaking bond direction

  • fbond_energy – (dict | None) As bbond_energy but in the direction to form a bond

  • init_active_temp – (float | None) Initial temperature for velocities in the ‘active’ MD search for configurations

  • min_active_iters – (int) Minimum number of active iterations to perform

  • bias_start_iter – (int) Iteration index at which the bias starts to be applied. If the bias is PlumedBias, then UPPER_WALLS and LOWER_WALLS are still applied from iteration 0

  • restart_iter – (int | None) Iteration index at which to restart active learning

  • inherit_metad_bias – (bool) If True metadynamics bias is inherited from a previous iteration to the next during active learning

  • constraints – (List) List of ASE contraints to use in the dynamics during active learning

  • bias – (mlptrain.Bias | mlptrain.PlumedBias) Bias to add during MD simulations, useful for exploring under-explored regions in the dynamics

  • md_program – (str) ‘ASE’ or ‘OpenMM’

  • pbc – (bool) If True, MLP-MD propagates with periodic boundary conditions. The solvent should be therefore placed in a box and not a sphere. The training data are still treated as clusters in electronic structure computations.

  • box_size – (List | None) Size of the box where MLP-MD propagated.

  • keep_al_trajs – (bool) If True, MLP-MD trajectories generated during AL phase are saved into new folder.