Experiment reporter

fyt.core.experiment_reporter

Experiment reporting — logs metrics, artifacts, and plots to an experiment tracker.

Extracted from Trainer to follow the Single Responsibility Principle. All logging is delegated to a BaseExperimentLogger implementation.

ExperimentReporter

Handles all experiment logging, artifact creation, and plot generation.

Delegates actual logging operations to a BaseExperimentLogger implementation.

Parameters:

Name	Type	Description	Default
experiment_logger	'BaseExperimentLogger'	Logger backend (e.g. MlflowLogger).	required
feature_importance_config	FeatureImportanceConfig	Configuration for feature importance computation.	required

experiment_logger property

The underlying experiment logger.

report(metrics, data_split, model, model_type, model_params, best_params=None, optuna_results=None, optimization_metric=None, optimization_direction=None)

Log a complete experiment run: params, metrics, model, and artifacts.

Parameters:

Name	Type	Description	Default
metrics	MetricResults	Computed evaluation metrics.	required
data_split	DataSplit	Data split used for evaluation.	required
model	Any	Trained model instance.	required
model_type	str	Model type identifier.	required
model_params	dict[str, Any]	Base model parameters.	required
best_params	dict[str, Any] | None	Best parameters from hyperparameter optimization.	None
optuna_results	DataFrame | None	DataFrame of Optuna trial results.	None
optimization_metric	str | None	Normalized name of the tuned metric (the column holding trial scores in optuna_results).	None
optimization_direction	str | None	"maximize" or "minimize".	None