Development version

You are reading the latest (dev) docs. For the stable version, see 0.13.

MLflow API

from sklearn_genetic.mlflow_log import MLflowConfig

Requires MLflow: pip install sklearn-genetic-opt[mlflow].

MLflowConfig

Configures MLflow logging for a GASearchCV or GAFeatureSelectionCV run. Pass an instance to the log_config parameter of the search estimator.

Each call to fit creates:

• One parent run for the full search (with run_name and tags)
• One child run per evaluated candidate (with parameters and CV score)

MLflowConfig(
    tracking_uri,
    experiment,
    run_name,
    save_models=False,
    registry_uri=None,
    tags=None,
)

Parameters

Parameter	Type	Default	Description
tracking_uri	str	—	Address of the MLflow tracking server (e.g., "http://localhost:5000")
experiment	str	—	Name of the MLflow experiment. Created if it doesn't exist
run_name	str	—	Name for the parent run (stored as a mlflow.runName tag)
save_models	bool	False	Log the fitted estimator as an artifact for each candidate run
registry_uri	str	None	Address of the MLflow model registry server
tags	dict	None	Dictionary of tag name → value applied to the parent run

Usage

from sklearn_genetic.mlflow_log import MLflowConfig

mlflow_config = MLflowConfig(
    tracking_uri="http://localhost:5000",
    experiment="my-experiment",
    run_name="RF hyperparameter search",
    save_models=True,
    tags={"team": "ml", "dataset": "breast_cancer"},
)

search = GASearchCV(
    estimator=your_estimator,
    param_grid=your_param_grid,
    cv=your_cv,
    scoring="roc_auc",
    evolution_config=...,
    log_config=mlflow_config,  # <-- pass here
)

search.fit(X_train, y_train)

See Also

• MLflow Integration Guide — full tutorial with setup instructions
• Callbacks — combine MLflow logging with early stopping