#!/usr/bin/env python3
# Copyright (c) Facebook, Inc. and its affiliates.
#
# This source code is licensed under the MIT license found in the
# LICENSE file in the root directory of this source tree.
from abc import ABC, abstractmethod
from typing import Any, Dict, Optional, Set, Tuple
import numpy as np
import pandas as pd
from ax.core.experiment import Experiment
from ax.core.map_data import MapData
from ax.early_stopping.utils import align_partial_results
from ax.exceptions.core import UnsupportedError
from ax.utils.common.logger import get_logger
from ax.utils.common.typeutils import not_none
logger = get_logger(__name__)
[docs]class BaseEarlyStoppingStrategy(ABC):
"""Interface for heuristics that halt trials early, typically based on early
results from that trial."""
[docs] @abstractmethod
def should_stop_trials_early(
self,
trial_indices: Set[int],
experiment: Experiment,
**kwargs: Dict[str, Any],
) -> Dict[int, Optional[str]]:
"""Decide whether to complete trials before evaluation is fully concluded.
Typical examples include stopping a machine learning model's training, or
halting the gathering of samples before some planned number are collected.
Args:
trial_indices: Indices of candidate trials to stop early.
experiment: Experiment that contains the trials and other contextual data.
Returns:
A dictionary mapping trial indices that should be early stopped to
(optional) messages with the associated reason.
"""
pass # pragma: nocover
[docs]class PercentileEarlyStoppingStrategy(BaseEarlyStoppingStrategy):
"""Implements the strategy of stopping a trial if its performance
falls below that of other trials at the same step."""
def __init__(
self,
percentile_threshold: float = 50.0,
min_progression: float = 0.1,
min_curves: float = 5,
) -> None:
"""Construct a PercentileEarlyStoppingStrategy instance.
Args:
percentile_threshold: Falling below this threshold compared to other trials
at the same step will stop the run. Must be between 0.0 and 100.0.
e.g. if percentile_threshold=25.0, the bottom 25% of trials are stopped.
Note that "bottom" here is determined based on performance, not
absolute values; if `minimize` is False, then "bottom" actually refers
to the top trials in terms of metric value.
min_progression: Only stop trials if the latest progression value
(e.g. timestamp) is greater than this threshold. Prevents stopping
prematurely before enough data is gathered to make a decision.
min_curves: Minimum number of trial curves that need to be available to
make a stopping decision.
"""
self.percentile_threshold = percentile_threshold
self.min_progression = min_progression
self.min_curves = min_curves
[docs] def should_stop_trials_early(
self,
trial_indices: Set[int],
experiment: Experiment,
**kwargs: Dict[str, Any],
) -> Dict[int, Optional[str]]:
"""Stop a trial if its performance is in the bottom `percentile_threshold`
of the trials at the same step.
Args:
trial_indices: Indices of candidate trials to consider for early stopping.
experiment: Experiment that contains the trials and other contextual data.
Returns:
A dictionary mapping trial indices that should be early stopped to
(optional) messages with the associated reason. An empty dictionary
means no suggested updates to any trial's status.
"""
if experiment.optimization_config is None:
raise UnsupportedError( # pragma: no cover
"Experiment must have an optimization config in order to use an "
"early stopping strategy."
)
optimization_config = not_none(experiment.optimization_config)
objective_name = optimization_config.objective.metric.name
minimize = optimization_config.objective.minimize
data = experiment.lookup_data()
if data.df.empty:
logger.info(
"PercentileEarlyStoppingStrategy received empty data. "
"Not stopping any trials."
)
return {}
if not isinstance(data, MapData):
raise ValueError(
"PercentileEarlyStoppingStrategy expects MapData, but the "
f"data attached to experiment is of type {type(data)}."
)
map_keys = data.map_keys
if len(map_keys) > 1:
raise ValueError( # pragma: no cover
"PercentileEarlyStoppingStrategy expects MapData with a single "
"map key, but the data attached to the experiment has multiple: "
f"{data.map_keys}."
)
map_key = map_keys[0]
df = data.df
metric_to_aligned_means, _ = align_partial_results(
df=df,
progr_key=map_key,
metrics=[objective_name],
)
aligned_means = metric_to_aligned_means[objective_name]
decisions = {
trial_index: self.should_stop_trial_early(
trial_index=trial_index,
experiment=experiment,
df=aligned_means,
percentile_threshold=self.percentile_threshold,
map_key=map_key,
minimize=minimize,
)
for trial_index in trial_indices
}
return {
trial_index: reason
for trial_index, (should_stop, reason) in decisions.items()
if should_stop
}
[docs] def should_stop_trial_early(
self,
trial_index: int,
experiment: Experiment,
df: pd.DataFrame,
percentile_threshold: float,
map_key: str,
minimize: bool,
) -> Tuple[bool, Optional[str]]:
"""Stop a trial if its performance is in the bottom `percentile_threshold`
of the trials at the same step.
Args:
trial_index: Indices of candidate trial to stop early.
experiment: Experiment that contains the trials and other contextual data.
df: Dataframe of partial results after applying interpolation,
filtered to objective metric.
percentile_threshold: Falling below this threshold compared to other trials
at the same step will stop the run. Must be between 0.0 and 100.0.
e.g. if percentile_threshold=25.0, the bottom 25% of trials are stopped.
Note that "bottom" here is determined based on performance, not
absolute values; if `minimize` is False, then "bottom" actually refers
to the top trials in terms of metric value.
map_key: Name of the column of the dataset that indicates progression.
minimize: Whether objective value is being minimized.
Returns:
A tuple `(should_stop, reason)`, where `should_stop` is `True` iff the
trial should be stopped, and `reason` is an (optional) string providing
information on why the trial should or should not be stopped.
"""
logger.debug(f"Considering trial {trial_index} for early stopping.")
if trial_index not in df:
logger.info(
f"There is not yet any data associated with trial {trial_index}. "
"Not early stopping this trial."
)
return False, "No data available to make an early stopping decision."
trial_last_progression = not_none(df[trial_index].dropna()).index.max()
if trial_last_progression < self.min_progression:
reason = (
f"Most recent progression ({trial_last_progression}) is less than "
"the specified minimum progression for early stopping "
f"({self.min_progression}). "
)
logger.info(
f"Trial {trial_index}'s m{reason[1:]} Not early stopping this trial."
)
return False, reason
# dropna() here will exclude trials that have not made it to the
# last progression of the trial under consideration, and therefore
# can't be included in the comparison
data_at_last_progression = df.loc[trial_last_progression].dropna()
if len(data_at_last_progression) < self.min_curves:
logger.info(
f"The number of trials with data ({len(data_at_last_progression)}) "
f"at trial {trial_index}'s last progression ({trial_last_progression}) "
"is less than the specified minimum number for early stopping "
f"({self.min_curves}). Not early stopping this trial."
)
reason = (
f"Number of trials with data ({len(data_at_last_progression)}) at "
f"last progression ({trial_last_progression}) is less than the "
f"specified minimum number for early stopping ({self.min_curves})."
)
return False, reason
percentile_threshold = (
100.0 - self.percentile_threshold if minimize else self.percentile_threshold
)
percentile_value = np.percentile(data_at_last_progression, percentile_threshold)
trial_objective_value = data_at_last_progression[trial_index]
should_early_stop = (
trial_objective_value > percentile_value
if minimize
else trial_objective_value < percentile_value
)
comp = "better" if should_early_stop else "worse"
reason = (
f"Trial objective value {trial_objective_value} is {comp} than "
f"{percentile_threshold:.1f}-th percentile ({percentile_value}) "
"across comparable trials."
)
logger.debug(
f"Early stopping decision for {trial_index}: {should_early_stop}. "
f"Reason: {reason}"
)
return should_early_stop, reason
