pqagent.pqagent module

class pqagent.pqagent.PqAgent(config_path: str = None, name: str = 'default')

Bases: object

compare_models()

eval(y_pred: DataFrame, y_true: DataFrame) → dict

evaluate_model(dataset: DataSet, model_to_use: str) → dict

Evaluates the performance of a model on a given dataset. This method applies the selected model to the provided dataset, generates predictions, and computes evaluation metrics. It returns both the calculated metrics and the predictions for further analysis.

Parameters:

• dataset (DataSet) – The dataset on which the model’s performance is to be evaluated. Should contain the data and corresponding labels.

• model_to_use – A string specifying the model to be used for predictions.

Returns:

A dictionary containing evaluation metrics and a DataFrame with the generated predictions.

Return type:

tuple(dict, pd.DataFrame)

classmethod from_file(filepath: str) → PqAgent

property get_models: list

load_model_from_checkpoint(model_name: str, trial_id: str = None, checkpoint_path: str = None) → None

models: Dict[str, PqModel]

predict(data: DataSet | DataFrame, model_to_use: str, get_comparison: bool = False) → DataFrame

preprocessor: Preprocessor

reset_ray()

retrain(dataset: DataSet, base_model: str, strategy: Callable, model_name: str | None = None, update_config_path: str = None) → None

save(file_path: str = None, folder_path: str = None) → None

Save the model and preprocessor to a specified file or folder.

Parameters: - file_path (str, optional): The path to save the file directly. - folder_path (str, optional): The folder path to save the file with a default naming convention.

Raises: - ValueError: If neither file_path nor folder_path is provided.

train_model(data: DataSet, model_name: str = 'base', train_val_split_feature_groups: list = None, save_train_data: bool = False) → None

static transform_fc_layers_args(config)

Converts the ‘args’ in ‘fc_layers’ from strings to tuples. This modifies the config in place if ‘fc_layers’ is present.

update_config(config_path: str)