API Reference / agent / tools / workflow /

evaluate.py

Model evaluation and reports.

Functions

Func

execute_evaluate

Line 9
execute_evaluate(**kwargs) -> Dict[str, Any]

Execute evaluation with support for timeseries and anomaly models.

Backs the tuiml_evaluate tool. Detects the model family (classifier, regressor, clusterer, timeseries, anomaly) and computes the appropriate metrics; the report stage additionally builds a formatted text report.

Parameters

model_id
str
Identifier of a trained model from tuiml_train (arrives via **kwargs, like all parameters below). One of model_id / model_path is required.
model_path
str = None
Explicit path to a serialized model file.
data
str
Dataset to evaluate on (dataset_id, file path, or built-in name).
stage
str = None
Optional stage: 'report' for a human-readable evaluation report, or 'metrics' / None for a plain metrics dict.
stage_kwargs
dict = None
Extra stage arguments (currently unused by the stages).
metrics
str or list = 'auto'
Metrics passed to model.evaluate on the standard path.

Returns

result
dict
On success: status ('success') and metrics; the report stage adds report (formatted text) and model_type; timeseries evaluation adds train_size, test_size and forecast_preview. On failure: status ('error'), error, error_type and optionally suggestion.