Inference Using Models Trained with Anvil

Inference functions for trained models.

openadmet.models.inference.inference.load_anvil_model_and_metadata(model_dir)[source]

Load the Anvil model from the specified path.

Parameters:

model_dir (Union[str, Path]) – Path to the directory containing the trained model and its metadata.

Returns:

A tuple containing the loaded model, feature object, metadata, and data specification.

Return type:

tuple

openadmet.models.inference.inference.predict(input_path: str, input_col: str, model_dir: str | Path | list[Union[str, pathlib._local.Path]], write_csv: bool = False, output_csv: str = None, debug: bool = False, accelerator: str = 'gpu', log: bool = True, aq_fxn_args: dict | None = None, **kwargs)[source]

Predict using a trained model.

Parameters:

  • input_path (Union[str, Path, pd.DataFrame]) – Path to the input data file (CSV or SDF or parquet) or a pandas DataFrame.

  • input_col (str) – Name of the column containing SMILES strings.

  • model_dir (Union[str, Path, list[Union[str, Path]]]) – Path(s) to the directory(ies) containing the trained model(s) and their metadata.

  • write_csv (bool, optional) – Whether to write the output to a CSV file. Default is False.

  • output_csv (str, optional) – Path to the output CSV file. If None, defaults to ‘predictions.csv’ in the current directory. Default is None.

  • debug (bool, optional) – Whether to enable debug logging. Default is False.

  • accelerator (str, optional) – Accelerator to use for prediction (‘cpu’ or ‘gpu’). Default is ‘gpu’.

  • log (bool, optional) – Whether to enable logging. Default is True.

  • aq_fxn_args (dict, optional) – Dictionary of acquisition function names and their arguments to compute additional metrics. Default is None.

  • **kwargs – Additional keyword arguments.

Returns:

DataFrame containing the input data along with prediction results.

Return type:

pd.DataFrame