Source code for smac.intensifier.abstract_intensifier

from __future__ import annotations

from abc import abstractmethod
from typing import Any, Callable, Iterator, Optional

import time

import numpy as np
from ConfigSpace import Configuration

from smac.runhistory import TrialInfo, TrialInfoIntent, TrialValue
from smac.runhistory.runhistory import RunHistory
from smac.scenario import Scenario
from smac.stats import Stats
from smac.utils.logging import format_array, get_logger

__copyright__ = "Copyright 2022, automl.org"
__license__ = "3-clause BSD"

logger = get_logger(__name__)


[docs]class AbstractIntensifier: """Base class for all racing methods. The intensification is designed to output a TrialInfo object with enough information to run a given configuration (for example, the trial info contains the instance/seed pair, as well as the associated resources). A worker can execute this TrialInfo object and produce a TrialValue object with the execution results. Each intensifier process the TrialValue object and updates its internal state in preparation for the next iteration. Parameters ---------- scenario : Scenario min_config_calls : int, defaults to 1 Minimum number of trials per config (summed over all calls to intensify). max_config_calls : int, defaults to 2000 Maximum number of trials per config (summed over all calls to intensify). min_challenger : int, defaults to 1 Minimal number of challengers to be considered (even if time_bound is exhausted earlier). seed : int | None, defaults to none """ def __init__( self, scenario: Scenario, min_config_calls: int = 1, max_config_calls: int = 2000, min_challenger: int = 1, seed: int | None = None, ): if seed is None: seed = scenario.seed self._scenario = scenario self._seed = seed self._rng = np.random.RandomState(seed) self._deterministic = scenario.deterministic self._min_config_calls = min_config_calls self._max_config_calls = max_config_calls self._min_challenger = min_challenger self._stats: Stats | None = None # Set the instances self._instances: list[str | None] if scenario.instances is None: # We need to include None here to tell whether None instance was evaluated or not self._instances = [None] else: # Removing duplicates here # Fun fact: When using a set here, it always includes randomness self._instances = list(dict.fromkeys(scenario.instances)) # General attributes self._num_trials = 0 # Number of trials done in an iteration so far self._challenger_id = 0 self._repeat_configs = False # Repeating configurations is discouraged for parallel trials. self._iteration_done = False # Marks the end of an iteration. self._target_function_time = 0.0 @property def repeat_configs(self) -> bool: """Whether configs should be repeated or not.""" return self._repeat_configs @property def iteration_done(self) -> bool: """Whether an iteration is done or not.""" return self._iteration_done @property def num_trials(self) -> int: """How many trials have been done in an iteration so far.""" return self._num_trials @property @abstractmethod def uses_seeds(self) -> bool: """If the intensifier needs to make use of seeds.""" raise NotImplementedError @property @abstractmethod def uses_budgets(self) -> bool: """If the intensifier needs to make use of budgets.""" raise NotImplementedError @property @abstractmethod def uses_instances(self) -> bool: """If the intensifier needs to make use of instances.""" raise NotImplementedError
[docs] @abstractmethod def get_target_function_seeds(self) -> list[int]: """Which seeds are used to call the target function.""" raise NotImplementedError
[docs] @abstractmethod def get_target_function_budgets(self) -> list[float | None]: """Which budgets are used to call the target function.""" raise NotImplementedError
[docs] @abstractmethod def get_target_function_instances(self) -> list[str | None]: """Which instances are used to call the target function.""" raise NotImplementedError
@property def meta(self) -> dict[str, Any]: """Returns the meta data of the created object.""" return { "name": self.__class__.__name__, "min_config_calls": self._min_config_calls, "max_config_calls": self._max_config_calls, "min_challenger": self._min_challenger, "seed": self._seed, }
[docs] def get_next_trial( self, challengers: list[Configuration] | None, incumbent: Configuration, get_next_configurations: Callable[[], Iterator[Configuration]] | None, runhistory: RunHistory, repeat_configs: bool = True, n_workers: int = 1, ) -> tuple[TrialInfoIntent, TrialInfo]: """Abstract method for choosing the next challenger. If no more challengers are available, the method should issue a SKIP via RunInfoIntent.SKIP, so that a new iteration can sample new configurations. Parameters ---------- challengers : list[Configuration] | None Promising configurations. incumbent : Configuration Incumbent configuration. get_next_configurations : Callable[[], Iterator[Configuration]] | None, defaults to none Function that generates next configurations to use for racing. runhistory : RunHistory repeat_configs : bool, defaults to true If false, an evaluated configuration will not be generated again. n_workers : int, optional, defaults to 1 The maximum number of workers available. Returns ------- TrialInfoIntent Indicator of how to consume the TrialInfo object. TrialInfo An object that encapsulates necessary information of the trial. """ raise NotImplementedError()
[docs] def process_results( self, trial_info: TrialInfo, trial_value: TrialValue, incumbent: Configuration | None, runhistory: RunHistory, time_bound: float, log_trajectory: bool = True, ) -> tuple[Configuration, float | list[float]]: """The intensifier stage will be updated based on the results/status of a configuration execution. Also, a incumbent will be determined. Parameters ---------- trial_info : TrialInfo trial_value: TrialValue incumbent : Configuration | None Best configuration seen so far. runhistory : RunHistory time_bound : float Time [sec] available to perform intensify. log_trajectory: bool Whether to log changes of incumbents in the trajectory. Returns ------- incumbent: Configuration Current (maybe new) incumbent configuration. incumbent_costs: float | list[float] Empirical cost(s) of the incumbent configuration. """ raise NotImplementedError()
def _next_challenger( self, challengers: list[Configuration] | None, get_next_configurations: Callable[[], Iterator[Configuration]] | None, runhistory: RunHistory, repeat_configs: bool = True, ) -> Configuration | None: """Retuns the next challenger to use in intensification. If challenger is none, then the optimizer will be used to generate the next challenger. Parameters ---------- challengers : list[Configuration] | None Promising configurations to evaluate next. get_next_configurations : Callable[[], Iterator[Configuration]] | None, defaults to none Function that generates next configurations to use for racing. runhistory : RunHistory repeat_configs : bool, defaults to true If false, an evaluated configuration will not be generated again. Returns ------- configuration : Configuration | None Next challenger to use. If no challenger was found, none is returned. """ start_time = time.time() chall_gen: Iterator[Optional[Configuration]] if challengers: # iterate over challengers provided logger.debug("Using provied challengers.") chall_gen = (c for c in challengers) elif get_next_configurations: # generating challengers on-the-fly if optimizer is given logger.debug("Generating new challenger from optimizer.") chall_gen = get_next_configurations() else: raise ValueError("No configurations (function) provided. Can not generate challenger!") logger.debug("Time spend to select next challenger: %.4f" % (time.time() - start_time)) # Select challenger from the generators assert chall_gen is not None for challenger in chall_gen: # Repetitions allowed if repeat_configs: return challenger used_configs = runhistory.get_configs() # set(runhistory.get_configs()) # Otherwise, select only a unique challenger if challenger not in used_configs: return challenger logger.debug("No valid challenger was generated!") return None def _compare_configs( self, incumbent: Configuration, challenger: Configuration, runhistory: RunHistory, log_trajectory: bool = True, ) -> Configuration | None: """Compare two configuration wrt the runhistory and return the one which performs better (or None if the decision is not safe). Decision strategy to return x as being better than y: * x has at least as many trials as y. * x performs better than y on the intersection of trials on x and y. Note ---- Implicit assumption: Challenger was evaluated on the same instance-seed pairs as incumbent. Returns ------- configuration : Configuration | None The better configuration. If the decision is not sure, none is returned. """ inc_trials = runhistory.get_trials(incumbent, only_max_observed_budget=True) chall_trials = runhistory.get_trials(challenger, only_max_observed_budget=True) to_compare_trials = set(inc_trials).intersection(chall_trials) # Performance on challenger trials, the challenger only becomes incumbent # if it dominates the incumbent chal_perf = runhistory.average_cost(challenger, to_compare_trials, normalize=True) inc_perf = runhistory.average_cost(incumbent, to_compare_trials, normalize=True) assert type(chal_perf) == float assert type(inc_perf) == float # Line 15 if np.any(chal_perf > inc_perf) and len(chall_trials) >= self._min_config_calls: chal_perf_format = format_array(chal_perf) inc_perf_format = format_array(inc_perf) # Incumbent beats challenger logger.debug( f"Incumbent ({inc_perf_format}) is better than challenger " f"({chal_perf_format}) on {len(chall_trials)} trials." ) return incumbent # Line 16 # This statement is true if both incumbent trials and challenger trials are the same if not set(inc_trials) - set(chall_trials): # No plateau walks if np.any(chal_perf >= inc_perf): chal_perf_format = format_array(chal_perf) inc_perf_format = format_array(inc_perf) logger.debug( f"Incumbent ({inc_perf_format}) is at least as good as the " f"challenger ({chal_perf_format}) on {len(chall_trials)} trials." ) assert self._stats if log_trajectory and self._stats.incumbent_changed == 0: self._stats.add_incumbent(cost=chal_perf, incumbent=incumbent) return incumbent # Challenger is better than incumbent and has at least the same trials as incumbent. # -> Change incumbent n_samples = len(chall_trials) chal_perf_format = format_array(chal_perf) inc_perf_format = format_array(inc_perf) logger.info( f"Challenger ({chal_perf_format}) is better than incumbent ({inc_perf_format}) " f"on {n_samples} trials." ) self._log_incumbent_changes(incumbent, challenger) if log_trajectory: assert self._stats self._stats.add_incumbent(cost=chal_perf, incumbent=challenger) return challenger # Undecided return None def _log_incumbent_changes( self, incumbent: Configuration | None, challenger: Configuration | None, ) -> None: if incumbent is None or challenger is None: return params = sorted([(param, incumbent[param], challenger[param]) for param in challenger.keys()]) logger.info("Changes in incumbent:") for param in params: if param[1] != param[2]: logger.info("--- %s: %r -> %r" % param) else: logger.debug("--- %s remains unchanged: %r", param[0], param[1])