spainn.calculator

class calculator.SPaiNNulator(atom_types: numpy.ndarray | torch.Tensor | str | None = None, modelpath: List[str] | str = '', cutoff: float = 10.0, properties: List[str] | None = None, n_states: Dict[str, int] | None = None, thresholds: Dict[str, float] | None = None, nac_key: str = 'nacs')[source]

Interface between SHARC and SchNetPack 2.0

Parameters:

  • atom_types (atomic charges or string of atoms)

  • modelpath (path(s) to trained model(s) or folder(s) with 'best_inference_model') – for adaptive sampling

  • cutoff (cutoff value)

  • properties (list of properties returned to SHARC)

  • n_states (dictionary of calculated states)

  • thresholds (dictionary of threshold values)

Examples

You can use this calculator to perform predictions of properties using a trained NN model.

Here we show, how to predict the energies for a target molecule. First we import all necessary modules followed by the creation of an ase.Atoms object of the target molecule.

>>> import os, sys
>>> import numpy as np
>>> import ase
>>>
>>> symbols = 'CNHHHH'
>>> positions = np.array(
>>>     [[ 0.0000,  0.0000,  0.0000 ],
>>>      [ 2.4321,  0.0000,  0.0000 ],
>>>      [-1.0111,  1.7951,  0.0000 ],
>>>      [ 3.4373,  1.6202, -0.2566 ],
>>>      [ 3.4373, -1.6202,  0.2566 ],
>>>      [-1.0111, -1.7951,  0.0000 ]]
>>> )
>>> # create ase Atoms object
>>> target_mol = ase.Atoms(symbols=symbols, positions=positions)

Next, we define the calculator NacCalculator used to predict the energy of the target molecule and perform the prediction.

>>> from schnetpack.transform import MatScipyNeighborList
>>> from spainn.interface import NacCalculator
>>>
>>> calc = NacCalculator(
>>>     model_file=os.path.join(os.getcwd(), 'train', 'best_model'),
>>>     neighbor_list=MatScipyNeighborList(cutoff=10.0)
>>> )
>>> target_mol.calc = calc
>>> # make prediction
>>> pred = target_mol.get_properties(['energy'])
calculate(sharc_coords: numpy.ndarray | torch.Tensor) → Dict[str, List[numpy.ndarray]][source]

Calculate properties from new positions.

If multiple models are used, the average values between the two predictions with the lowest NAC MAE will be returned

Parameters:

sharc_coords (Coordinates from SHARC simulation)

get_qm(spainn_output: List[numpy.ndarray]) → Dict[str, List[numpy.ndarray]][source]

Calculate QM string for SHARC with predictions from model.