Model

fyt.registries.model

ModelRegistry

Bases: ComponentRegistry[BaseEstimator]

Registry for machine learning model types.

Note

The ModelRegistry is designed to manage the creation of machine learning model instances based on string identifiers. It includes special handling for ensemble models like VotingClassifier and StackingClassifier.

In case you want to use a different ensemble model, you can register it with a custom factory function that handles the creation logic, similar to how VotingClassifier and StackingClassifier are handled in this registry.

create(key, **kwargs) classmethod

Create a model, handling ensemble types specially.

Parameters:

Name	Type	Description	Default
key	str | Enum	Model type identifier.	required
**kwargs	Any	Model parameters.	{}

Returns:

Type	Description
BaseEstimator	A scikit-learn compatible model instance.

Raises:

Type	Description
ValueError	If the model type is not registered.

create_estimators(estimators_list, random_state=None) classmethod

Create a list of estimators for ensemble models.

Duplicate model types are given unique names by suffixing the ordinal occurrence (e.g. random_forest, random_forest_2, ...).

Parameters:

Name	Type	Description	Default
estimators_list	list[Estimator]	List of estimator configurations.	required
random_state	int | None	Random state forwarded to each base estimator that accepts it, unless the estimator's own parameters already set it.	None

Returns:

Type	Description
list[tuple[str, BaseEstimator]]	List of (name, estimator) tuples with unique names.

Raises:

Type	Description
ValueError	If a nested ensemble model is specified.

create_seeded(model_type, params, random_state)

Create a model, injecting random_state when the estimator accepts it.

Parameters:

Name	Type	Description	Default
model_type	str | Enum	Model type identifier.	required
params	dict[str, Any]	Model parameters. An explicit random_state here takes precedence over the random_state argument.	required
random_state	int	Random state injected when params does not set one.	required

Returns:

Type	Description
BaseEstimator	A scikit-learn compatible model instance.

Raises:

Type	Description
TypeError	If random_state was explicitly set in params but the estimator does not accept it.