asimtools.asimmodules.benchmarking package

asimtools.asimmodules.benchmarking.parity module

Generates a parity plot and collects evaluation statistics comparing energy and/or forces and/or stress to existing values in the provided dataset. This asimmodule can work in parallel based on the number of cores specified.

Author: mkphuthi@github.com

asimtools.asimmodules.benchmarking.parity.calc_parity_data(subset: List, calculator: Dict, properties: Sequence = ('energy', 'forces', 'stress'), force_prob: float = 1.0) Dict[source]

Calculates parity data for each atoms instance in subset

Parameters:
  • subset (List) – List of atoms instances

  • calculator (Dict) – Calculator specification, see asimtools.calculators.load_calc()

  • properties (List, optional) – Properties to evaluate, choose from “energy”, “forces” and “stress”, defaults to (‘energy’, ‘forces’, ‘stress’)

  • force_prob (float, optional) – Fraction of forces to consider, may speed up sampling large structures by subsampling forces randomly, defaults to 1.0

Returns:

Dictionary with reference and predicted values for each property

Return type:

Dict

asimtools.asimmodules.benchmarking.parity.parity(images: Dict, calculator: Dict, force_prob: float = 1.0, nprocs: int = 1, unit: str = 'meV', index: str = ':', properties: Sequence = ('energy', 'forces', 'stress')) Dict[source]

Generates a parity plot and collects evaluation statistics comparing energy and/or forces and/or stress to existing values in the provided dataset

Parameters:
  • images (Dict) – Images specification, see asimtools.utils.get_images()

  • calculator (Dict) – Calculator specification, see asimtools.calculators.load_calc()

  • force_prob (float, optional) – Fraction of forces to consider in force parity, can be used for speeding up large structures by only subsampling randomly, defaults to 1.0

  • nprocs (int, optional) – Number of process to parallelize over, must be less than number of datapoints, defaults to 1

  • unit (str, optional) – Unit to plot, choose from meV, eV or kcal/mol, defaults to ‘meV’

  • index (int, optional) – index of data to use from full dataset, defaults to ‘:’

  • properties (Sequence, optional) – Properties to evaluate, choose from “energy”, “forces” and “stress”, defaults to (‘energy’, ‘forces’, ‘stress’)

Returns:

Dictionary of results

Return type:

Dict

asimtools.asimmodules.benchmarking.parity.rmse(yhat: Sequence, y: Sequence) float[source]

Calculate Root Mean Square Error between to sequences

Parameters:
  • yhat (Sequence) – Predicted values

  • y (Sequence) – True values

Returns:

RMSE

Return type:

float

Module contents