Models

fyt.configs.models

AggregatedMetricResult

Bases: BaseModel

key property

Normalized string name of the metric (single rule: normalize_key).

AggregatedMetricsManagerInput

Bases: BaseModel

scalar_metrics_results property

The per-run metric results restricted to their scalar results.

collect_metric_values(metric_name)

Collect all values for a specific metric across all runs.

Parameters:

Name	Type	Description	Default
metric_name	MetricType | str	The type of metric to collect	required

Returns:

Type	Description
list[float]	A tuple of (values, seeds) where both lists have the same length.
list[int]	Only includes runs where the metric has a valid finite value.

get_unique_metric_types()

Get all unique scalar metric types from all runs.

Returns:

Type	Description
list[MetricType | str]	Sorted list of unique metric types (excluding confusion_matrix)

CategoricalImputationComponentConfig

Bases: PreProcessorComponentConfig

Configuration for a single categorical imputation component.

CategoricalTransformationComponentConfig

Bases: PreProcessorComponentConfig

Configuration for a single categorical transformation component.

CorrelationFilterConfig

Bases: BaseModel

Configuration for correlation-based feature filtering.

CrossValidationConfig

Bases: BaseModel

Configuration for cross-validation.

DataManagerConfig

Bases: YamlBaseModel

validate_exclude_columns()

Validate that exclude_columns does not contain id_column or target_column.

Returns:

Type	Description
DataManagerConfig	The validated config instance.

Raises:

Type	Description
ValueError	If exclude_columns contains the id_column, the target_column, or an explicitly provided categorical or numerical column.

validate_test_data_consistency()

Validate that test_data_path is provided if use_test_data is True.

Returns:

Type	Description
DataManagerConfig	The validated config instance.

Raises:

Type	Description
ValueError	If use_test_data is True but test_data_path is not provided.

DataSplit

Bases: BaseModel

Class for holding train-test split data.

FeatureImportanceConfig

Bases: BaseModel

Configuration for feature importance logging.

FeatureSelectionConfig

Bases: BaseModel

Configuration for feature selection.

FeatureSelectorConfig

Bases: YamlBaseModel

Configuration for the FeatureSelector.

MetricResultModel

Bases: BaseModel

key property

Normalized string name of the metric (single rule: normalize_key).

MetricResults

Bases: BaseModel

scalar_results property

The scalar metric results from the results list.

get_scalar_metric_value(metric_name)

Get the value of a specific scalar metric.

Parameters:

Name	Type	Description	Default
metric_name	MetricType | str	The type of metric to retrieve	required

Returns:

Type	Description
float | None	The metric value if found and valid, None otherwise

NumericalImputationComponentConfig

Bases: PreProcessorComponentConfig

Configuration for a single numerical imputation component.

validate_impute_zeros_with_strategy()

Validate that impute_zeros is only set when strategy is not 'passthrough'.

Returns:

Type	Description
'NumericalImputationComponentConfig'	The validated config instance.

Raises:

Type	Description
ValueError	If impute_zeros is True while strategy is 'passthrough'.

NumericalTransformationComponentConfig

Bases: PreProcessorComponentConfig

Configuration for a single numerical transformation component.

OptunaCategoricalParam

Bases: OptunaParamModel

Categorical parameter space definition for Optuna.

get_optuna_suggestion(trial)

Get the Optuna suggestion for this categorical parameter.

OptunaConfig

Bases: BaseModel

Configuration for Optuna hyperparameter tuning.

validate_direction_matches_metric()

Reject maximizing loss-like metrics (e.g. log_loss).

Returns:

Type	Description
OptunaConfig	The validated config instance.

Raises:

Type	Description
ValueError	If the optimization metric is a loss but direction is 'maximize'.

OptunaFloatParam

Bases: OptunaParamModel

Float parameter space definition for Optuna.

get_optuna_suggestion(trial)

Get the Optuna suggestion for this float parameter.

OptunaIntParam

Bases: OptunaParamModel

Integer parameter space definition for Optuna.

get_optuna_suggestion(trial)

Get the Optuna suggestion for this integer parameter.

OptunaParamModel

Bases: BaseModel

Parameter space definition for Optuna.

PreProcessorComponentConfig

Bases: BaseModel

Base configuration for a PreProcessor component.

validate_feature_specification()

Validate that features and exclude_features are not both specified.

Returns:

Type	Description
'PreProcessorComponentConfig'	The validated config instance.

Raises:

Type	Description
ValueError	If both 'features' and 'exclude_features' are specified.

validate_features(all_columns)

Validate the provided features with all the available ones.

Parameters:

Name	Type	Description	Default
all_columns	list[str]	List of all available columns.	required

Raises:

Type	Description
ValueError	If 'features' or 'exclude_features' contain duplicates or reference columns not present in the dataset.

PreProcessorConfig

Bases: YamlBaseModel

Configuration for the PreProcessor.

PredictCommandConfig

Bases: YamlBaseModel

Configuration for batch prediction with a saved pipeline artifact.

RunPipelineCommandConfig

Bases: YamlBaseModel

validate_task_consistency()

Validate that the model type and metrics match the task type.

Returns:

Type	Description
RunPipelineCommandConfig	The validated config instance.

Raises:

Type	Description
ValueError	If the model type or a configured metric is incompatible with the configured task.

validate_test_data_usage()

Validate DataManager and Trainer config consistency for test data usage.

ScalarMetricResult

Bases: MetricResultModel

A single numeric metric value (built-in or custom-registered).

TargetEncodingConfig

Bases: BaseModel

Configuration for target encoding.

TargetProcessorConfig

Bases: YamlBaseModel

Configuration for the TargetProcessor.

TrainerConfig

Bases: YamlBaseModel

Configuration for the Trainer.

validate_validation_size_and_cross_validation()

Warn when validation_size is set but cross-validation will ignore it.