Tuner

class pyatf.Tuner

Represents a tuner, that can be customized via its member setter functions.

Tuner()

Default constructor.

tuning_parameters(*tps: TP)

Sets program’s tuning parameters.

search_technique(search_technique: SearchTechnique | SearchTechnique1D)

Sets the search technique for exploration.

verbosity(verbosity: int = 2)

Verbosity level of log messages. 0: no output. 1: progress bars for search space generation and exploration. 2: level 1 plus search space size and best configuration. 3: level 2 plus log of all tested configurations.

log_file(log_file: str)

Sets path to logfile.

tune(cost_function: CostFunction, abort_condition: AbortCondition | None = None)

Tunes cost_function until abort_condition is met.

Default abort condition: explore full search space.

make_step(cost_function: CostFunction)

Make one tuning step using cost_function.

get_tuning_data() → TuningData

Returns the tuning data object.

class pyatf.TP

TP(name, range, constraint)

Specifies a tuning parameter by its name, range, and constraint.

Parameters:

• name – Tuning parameter’s name as string

• range – Either: 1) pyatf.Interval(min, max) which is an interval of values between min and max (both including); intervals may have as optional argument a step_size and function generator (for using values generator(min), ..., generator(max); or 2) pyatf.Set(*values).

class pyatf.tuning_data.Configuration

Configuration of tuning parameters (name-value pairs).

class pyatf.tuning_data.TuningData

Tuning data object.

tuning_parameters(read-only)

constrained_search_space_size(read-only)

unconstrained_search_space_size(read-only)

search_space_generation_ns(read-only)

search_technique(read-only)

abort_condition(read-only)

tuning_start_timestamp(read-only)

terminated_early(read-only)

history(read-only)

improvement_history(read-only)

number_of_evaluated_configurations(read-only)

number_of_evaluated_valid_configurations(read-only)

number_of_evaluated_invalid_configurations(read-only)

total_tuning_duration()

configuration_of_min_cost()

search_space_coordinates_of_min_cost()

search_space_index_of_min_cost()

timestamp_of_min_cost()

duration_to_min_cost()

evaluations_to_min_cost()

valid_evaluations_to_min_cost()