Molecular Dynamics

class apax.md.ase_calc.ASECalculator(model_dir: Path | list[Path], dr_threshold: float = 0.5, transformations: Callable = [], padding_factor: float = 1.5, **kwargs)

ASE Calculator for apax models. Always implements energy and force predictions. Stress predictions and corresponding uncertainties are added to implemented_properties based on whether the stress flag is set in the model config and whether a model ensemble is loaded.

batch_eval(atoms_list: list[Atoms], batch_size: int = 64, silent: bool = False) → list[Atoms]

Evaluate the model on a list of Atoms. This is preferable to assigning the calculator to each Atoms instance for 2 reasons: 1. Processing can be abtched, which is advantageous for larger datasets. 2. Inputs are padded so no recompilation is triggered when evaluating differently sized systems.

Parameters:

• atoms_list – List of Atoms to be evaluated.

• batch_size – Processing batch size. Does not affect results, only speed and memory requirements.

• silent – Whether or not to suppress progress bars.

Returns:

List of Atoms with labels predicted by the model.

Return type:

evaluated_atoms_list

calculate(atoms, properties=['energy'], system_changes=['positions', 'numbers', 'cell', 'pbc', 'initial_charges', 'initial_magmoms'])

Do the calculation.

properties: list of str

List of what needs to be calculated. Can be any combination of ‘energy’, ‘forces’, ‘stress’, ‘dipole’, ‘charges’, ‘magmom’ and ‘magmoms’.

system_changes: list of str

List of what has changed since last calculation. Can be any combination of these six: ‘positions’, ‘numbers’, ‘cell’, ‘pbc’, ‘initial_charges’ and ‘initial_magmoms’.

Subclasses need to implement this, but can ignore properties and system_changes if they want. Calculated properties should be inserted into results dictionary like shown in this dummy example:

self.results = {'energy': 0.0,
                'forces': np.zeros((len(atoms), 3)),
                'stress': np.zeros(6),
                'dipole': np.zeros(3),
                'charges': np.zeros(len(atoms)),
                'magmom': 0.0,
                'magmoms': np.zeros(len(atoms))}

The subclass implementation should first call this implementation to set the atoms attribute and create any missing directories.

implemented_properties: List[str] = ['energy', 'forces']

Properties calculator can handle (energy, forces, …)

class apax.md.function_transformations.FunctionTransformation

class apax.md.function_transformations.GaussianAcceleratedMolecularDynamics

Applies a boost potential to the system that pulls it towards a target energy. https://pubs.acs.org/doi/10.1021/acs.jctc.5b00436

Parameters:

• energy_target (float) – Target potential energy below which to apply the boost potential.

• spring_constant (float) – Spring constant of the boost potential.

class apax.md.function_transformations.GlobalCalibration(energy_factor: float, forces_factor: float)

Applies a global calibration to energy and force uncertainties. Energy ensemble predictions are rescaled according to EQ 7 in https://doi.org/10.1063/5.0036522

Parameters:

• energy_factor (float) – Global calibration factor by which to scale the energy uncertainty.

• forces_factor (float) – Global calibration factor by which to scale the force uncertainties.

class apax.md.function_transformations.UncertaintyDrivenDynamics

UDD requires an uncertainty aware model. It drives the dynamics towards higher uncertainty regions up to some maximum bias energy. https://doi.org/10.1038/s43588-023-00406-5

Parameters:

• height (float) – Maximum bias potential that can be applied

• width (float) – Width of the Gaussian bias.