def is_supported(self, model: BaseInferenceEngine,
training_data: ASEDataStore, X: Tuple[Atoms]):
# Check total count
if training_data.count() < self.min_entries:
return False
# Get the training points
train_X, _ = training_data.get_all_data()
if not training_data.convert_to_pmg:
train_X = [AseAtomsAdaptor.get_molecule(x) for x in train_X]
train_X = self.cm.fit_transform(train_X)
# Make a distance computer
# TODO (wardlt): Here is where we could benefit from checking if training set is updated
nn = NearestNeighbors(n_neighbors=self.k,
n_jobs=self.n_jobs,
metric=self.metric).fit(train_X)
# Get the distance
strc = AseAtomsAdaptor.get_molecule(X[0])
features = self.cm.transform([strc])
dists, _ = nn.kneighbors(features)
return np.mean(dists, axis=1)[0] < self.threshold
def molToMolecule(filename):
"""
:code:`molToMolecule` can help us read the .mol file and return a pymatgen Molecule object.
:param filename: The filename (relative path) that we want.
:type filename: python string object
:returns: A pymatgen Molecule object
"""
from pymatgen.io.ase import AseAtomsAdaptor
from ase.io import read
mol = read(filename, format='mol')
return AseAtomsAdaptor.get_molecule(mol)
def infer(self, X: Atoms) -> float:
# Convert to pymatgen format needed by matminer
strc = AseAtomsAdaptor.get_molecule(X[0])
return self.model.predict([strc])[0]
