blackboxopt.optimization_loops.utils

EvaluationFunctionError (ValueError)

Raised on errors originating from the user defined evaluation function.

Source code in blackboxopt/optimization_loops/utils.py

class EvaluationFunctionError(ValueError):
    """Raised on errors originating from the user defined evaluation function."""

    def __init__(self, evaluation_specification: EvaluationSpecification):
        self.message = (
            "An error occurred when attempting to call the user specified evaluation "
            "function with the specification below. Please check the cause of this "
            "exception in the output further up for the original stacktrace.\n"
            f"{evaluation_specification}"
        )
        self.evaluation_specification = evaluation_specification

evaluation_function_wrapper(evaluation_function, evaluation_specification, objectives, catch_exceptions_from_evaluation_function, logger)

Wrapper for evaluation functions. The evaluation result returned by the evaluation function is checked to contain all relevant objectives. An empty evaluation with a stacktrace is reported to the optimizer in case an unhandled Exception occurrs during the evaluation function call when catch_exceptions_from_evaluation_function is set to True, otherwise an EvaluationFunctionError is raised based on the original exception.

Source code in blackboxopt/optimization_loops/utils.py

def evaluation_function_wrapper(
    evaluation_function: Callable[[EvaluationSpecification], Evaluation],
    evaluation_specification: EvaluationSpecification,
    objectives: List[Objective],
    catch_exceptions_from_evaluation_function: bool,
    logger: logging.Logger,
) -> Evaluation:
    """Wrapper for evaluation functions. The evaluation result returned by the
    evaluation function is checked to contain all relevant objectives. An empty
    evaluation with a stacktrace is reported to the optimizer in case an unhandled
    Exception occurrs during the evaluation function call when
    `catch_exceptions_from_evaluation_function` is set to `True`, otherwise an
    `EvaluationFunctionError` is raised based on the original exception.
    """
    try:
        evaluation = evaluation_function(evaluation_specification)
    except Exception as e:
        if not catch_exceptions_from_evaluation_function:
            raise EvaluationFunctionError(evaluation_specification) from e

        stacktrace = traceback.format_exc()

        logger.warning(
            "Reporting FAILURE due to unhandled error in evaluation function. See "
            + "DEBUG log level output or evaluation.stacktrace for details. "
            + "Alternatively, disable automated exception handling by setting "
            + "catch_exceptions_from_evaluation_function=False to exit on errors."
        )
        logger.debug(stacktrace)

        evaluation = evaluation_specification.create_evaluation(
            stacktrace=stacktrace, objectives={o.name: None for o in objectives}
        )

    raise_on_unknown_or_incomplete(
        exception=ObjectivesError,
        known=[o.name for o in objectives],
        reported=evaluation.objectives.keys(),
    )

    return evaluation

init_max_evaluations_with_limit_logging(timeout_s, logger, max_evaluations=None)

[summary]

Parameters:

Name	Type	Description	Default
timeout_s	float	[description]	required
logger	Logger	[description]	required
max_evaluations	int	[description]	None

Returns:

Type	Description
float	[description]

Source code in blackboxopt/optimization_loops/utils.py

def init_max_evaluations_with_limit_logging(
    timeout_s: float, logger: logging.Logger, max_evaluations: int = None
) -> float:
    """[summary]

    Args:
        timeout_s: [description]
        logger: [description]
        max_evaluations: [description]

    Returns:
        [description]
    """
    if max_evaluations:
        logger.info(
            "Starting optimization run. Stops when complete or "
            + f"{max_evaluations} evaluations reached."
        )
        return float(max_evaluations)

    if timeout_s == float("inf"):
        logger.info("Starting optimization run. Stops when complete.")
    else:
        timeout_pretty = datetime.timedelta(seconds=timeout_s)
        logger.info(
            "Starting optimization run. Stops when complete or "
            + f"{timeout_pretty} passed."
        )

    return float("inf")