palma.base package

Submodules

palma.base.engine module

class palma.base.engine.BaseOptimizer(engine_parameters: dict)

Bases: object

Attributes:

allow_splitter

engine_parameters

estimator_

optimizer

transformer_

Methods

allowing_splitter
optimize

property allow_splitter

allowing_splitter(splitter)

property engine_parameters: Dict

abstract property estimator_: None

abstract optimize(X: DataFrame, y: Series, splitter: ValidationStrategy | None = None) → None

abstract property optimizer: None

abstract property transformer_: None

class palma.base.engine.FlamlOptimizer(problem: str, engine_parameters: dict)

Bases: BaseOptimizer

Attributes:

allow_splitter

engine_parameters

estimator_

optimizer

transformer_

Methods

allowing_splitter
optimize

property allow_splitter

property estimator_: BaseEstimator

optimize(X: DataFrame, y: DataFrame, splitter: ValidationStrategy | None = None) → None

property optimizer: AutoML

property transformer_: flaml.data.DataTransformer

palma.base.model module

class palma.base.model.ModelEvaluation(estimator)

Bases: object

Attributes:

components

id

unfit_estimator

Methods

add
fit

add(component, name=None)

property components

fit(project: Project)

property id: str

property unfit_estimator

palma.base.model_selection module

class palma.base.model_selection.ModelSelector(engine: str | BaseOptimizer, engine_parameters: Dict)

Bases: object

Wrapper to optimizers selecting the best model for a Project.

The optimization can be launched with the start method. Once the optimization is done, the best model can be accessed as the best_model_ attribute.

Parameters:

- engine (str): Currently accepted values are “FlamlOptimizer” or

“AutoSklearnOptimizer” (the latter is deprecatted).

- engine_parameters (dict): parameters passed to the engine.

Attributes:

run_id

Methods

- start(project: Project): look for best model

property run_id: str

start(project: Project)

palma.base.project module

class palma.base.project.Project(project_name: str, problem: str)

Bases: object

Represents a machine learning project with various components and logging capabilities.

Parameters:

project_name (str): The name of the project.

problem (str): The description of the machine learning problem.

Accepted values: “classification” or “regression”.

Attributes:

base_index (List[int]): List of base indices for the project.

components (dict): Dictionary containing project components.

date (datetime): The date and time when the project was created.

project_id (str): Unique identifier for the project.

is_started (bool): Indicates whether the project has been started.

problem (str): Description of the machine learning problem.

validation_strategy (ValidationStrategy): The validation strategy used in the project.

project_name (str): The name of the project.

study_name (str): The randomly generated study name.

X (pd.DataFrame): The feature data for the project.

y (pd.Series): The target variable for the project.

Methods

add(component: Component) -> None: Adds a component to the project.
start(X: pd.DataFrame, y: pd.Series, splitter, X_test=None, y_test=None, groups=None, **kwargs) -> None:	Starts the project with the specified data and validation strategy.

property X: DataFrame

add(component: Component) → None

property components: dict

property date: datetime

property is_started: bool

property problem: str

property project_id: str

property project_name: str

start(X: DataFrame, y: Series, splitter, X_test=None, y_test=None, groups=None, **kwargs) → None

property study_name: str

property validation_strategy: ValidationStrategy

property y: Series

palma.base.splitting_strategy module

class palma.base.splitting_strategy.ValidationStrategy(splitter: BaseShuffleSplit | BaseCrossValidator | List[tuple] | List[str], **kwargs)

Bases: object

Validation strategy for a machine learning project.

Parameters:

- splitter (Union[BaseShuffleSplit, BaseCrossValidator, List[tuple], List[str]]): The data splitting strategy.

Attributes:

- test_index (np.ndarray): Index array for the test set.

- train_index (np.ndarray): Index array for the training set.

- indexes_val (list): List of indexes for validation sets.

- indexes_train_test (list): List containing tuples of training and test indexes.

- id: Unique identifier for the validation strategy.

- splitter: The data splitting strategy.

Methods

- __call__(X: pd.DataFrame, y: pd.Series, X_test: pd.DataFrame = None, y_test: pd.Series = None, groups=None, **kwargs):

Applies the validation strategy to the provided data.

property groups

property id

property indexes_train_test: list

property indexes_val: list

property splitter

property test_index: ndarray

property train_index: ndarray

Module contents