LgbModelConfig

LgbModelConfig

Configuration container for LightGBM models in the factrainer framework.

This class encapsulates all necessary components for training and prediction with LightGBM models.

Warnings

Do not instantiate this class directly by calling LgbModelConfig(...). Use the factory method LgbModelConfig.create instead.

Attributes:

Name	Type	Description
learner	LgbLearner	Component responsible for training LightGBM models.
predictor	LgbPredictor	Component responsible for making predictions with trained models.
train_config	LgbTrainConfig	Configuration for training parameters (params, num_boost_round, etc.).
pred_config	LgbPredictConfig	Configuration for prediction parameters (iteration range, output type, etc.).

See Also

LgbModelConfig.create : Factory method for creating configurations. LgbTrainConfig : Configuration for training parameters. LgbPredictConfig : Configuration for prediction parameters. SingleModelContainer : For training a single model. CvModelContainer : For cross-validation workflows.

Examples:

>>> from factrainer.lightgbm import LgbModelConfig, LgbTrainConfig
>>> # Create configuration with default prediction settings
>>> train_config = LgbTrainConfig(
...     params={"objective": "regression", "metric": "rmse"}, num_boost_round=100
... )
>>> model_config = LgbModelConfig.create(train_config)
>>> # Create configuration with custom prediction settings
>>> from factrainer.lightgbm import LgbPredictConfig
>>> pred_config = LgbPredictConfig(raw_score=True)
>>> model_config = LgbModelConfig.create(train_config, pred_config)

Attributes

learner instance-attribute

learner: LgbLearner

predictor instance-attribute

predictor: LgbPredictor

train_config instance-attribute

train_config: LgbTrainConfig

pred_config instance-attribute

pred_config: LgbPredictConfig

Functions

create classmethod

create(
    train_config: LgbTrainConfig,
    pred_config: LgbPredictConfig | None = None,
) -> Self

Create a new LgbModelConfig instance.

Parameters:

Name	Type	Description	Default
train_config	LgbTrainConfig	Configuration for training parameters including LightGBM params, number of boosting rounds, callbacks, etc.	required
pred_config	LgbPredictConfig | None	Configuration for prediction parameters. If None, uses default LgbPredictConfig() which performs standard prediction.	None

Returns:

Type	Description
LgbModelConfig	A configuration instance ready for use with SingleModelContainer or CvModelContainer.