gdmlModel#

class mbgdml.models.gdmlModel(model, comp_ids=None, criteria=None, for_predict=True)[source]#
Parameters:

  • model (str or dict) – Path to GDML npz model or dict.

  • comp_ids (iterable, default: None) – Model component IDs that relate entity IDs of a structure to a fragment label. This overrides any model comp_ids if they are present.

  • criteria (mbgdml.descriptors.Criteria, default: None) – Initialized descriptor criteria for accepting a structure based on a descriptor and cutoff.

  • for_predict (bool, default: True) – Unpack the model for use in predictors.

add_modifications(dset)[source]#

mbGDML-specific modifications of models.

Transfers information from data set to model.

Parameters:

dset (dict) – Dictionary of a mbGDML data set.

property code_version#

mbGDML version used to train the model.

Raises:

AttributeError – If accessing information when no model is loaded.

property md5#

Unique MD5 hash of model.

Type:

str

save(path)[source]#

Save model dict as npz file.

Changes in model attributions are not considered.

Parameters:

path (str) – Path to save the npz file.