gt4sd.training_pipelines.torchdrug.dataset module

Summary

Classes:

TorchDrugDataset

A generic TorchDrug dataset class that can be fed with custom data

Reference

class TorchDrugDataset(file_path, target_fields, smiles_field='smiles', verbose=1, lazy=False, transform=None, node_feature='default', edge_feature='default', graph_feature=None, with_hydrogen=False, kekulize=False)

Bases: MoleculeDataset

A generic TorchDrug dataset class that can be fed with custom data

__init__(file_path, target_fields, smiles_field='smiles', verbose=1, lazy=False, transform=None, node_feature='default', edge_feature='default', graph_feature=None, with_hydrogen=False, kekulize=False)

Constructor of TorchDrugDataset.

Parameters

• file_path (str) – The path to the .csv file containing the data.

• target_fields (str) – The columns containing the property to be optimized.

• smiles_field (str) – The column name containing the SMILES. Defaults to ‘smiles’.

• verbose (int) – output verbose level. Defaults to 1.

• lazy (Optional[bool, None]) – If yes, molecules are processed in the dataloader. This is faster for setup, but slower at training time. Defaults to False.

• transform (Optional[Callable, None]) – Optional data transformation function. Defaults to None.

• node_feature (Union[str, List[str], None]) – Node features to extract. Defaults to ‘default’.

• edge_feature (Union[str, List[str], None]) – Edge features to extract. Defaults to ‘default’.

• graph_feature (Union[str, List[str], None]) – Graph features to extract. Defaults to None.

• with_hydrogen (Optional[bool, None]) – Whether hydrogens are stored in molecular graph. Defaults to False.

• kekulize (Optional[bool, None]) – Whether aromatic bonds are converted to single/double bonds. Defaults to False.

__annotations__ = {}

__doc__ = 'A generic TorchDrug dataset class that can be fed with custom data'

__module__ = 'gt4sd.training_pipelines.torchdrug.dataset'

__parameters__ = ()

config_dict()

classmethod load_config_dict(config)