model_errors

mbgdml._gdml.train.model_errors(model, dataset, is_valid=False, n_test=None, max_processes=None, use_torch=False)

Computes model errors for either validation or testing purposes.

Parameters:

• model (dict) – Trained GDML model.

• dataset (dict) – Dataset to test the model against.

• is_valid (bool, default: False) – Is this for model validation? Determines how we get the structure indices.

• n_test (int, default: None) – Number of desired testing indices. None will test against all available structures.

Returns:

• int – Number of structures predicted.

• numpy.ndarray – (ndim: 1) Energy errors or None.

• numpy.ndarray – (ndim: 3) Force errors.