#!/usr/bin/env python3
# Copyright (c) Meta Platforms, Inc. and affiliates.
#
# This source code is licensed under the MIT license found in the
# LICENSE file in the root directory of this source tree.
# pyre-strict
from __future__ import annotations
from collections.abc import Callable, Generator, Iterable, Mapping
from copy import deepcopy
from dataclasses import dataclass
from datetime import datetime
from enum import IntEnum
from logging import LoggerAdapter
from time import sleep
from typing import Any, cast, NamedTuple
import ax.service.utils.early_stopping as early_stopping_utils
from ax.adapter.adapter_utils import get_fixed_features_from_experiment
from ax.adapter.base import Adapter
from ax.core.base_trial import BaseTrial
from ax.core.experiment import Experiment
from ax.core.generator_run import GeneratorRun
from ax.core.metric import Metric, MetricFetchE, MetricFetchResult
from ax.core.multi_type_experiment import (
filter_trials_by_type,
get_trial_indices_for_statuses,
MultiTypeExperiment,
)
from ax.core.runner import Runner
from ax.core.trial import Trial
from ax.core.trial_status import TrialStatus
from ax.core.utils import get_pending_observation_features_based_on_trial_status
from ax.exceptions.core import (
AxError,
DataRequiredError,
OptimizationComplete,
UnsupportedError,
UserInputError,
)
from ax.exceptions.generation_strategy import (
AxGenerationException,
MaxParallelismReachedException,
OptimizationConfigRequired,
)
from ax.generation_strategy.generation_strategy import GenerationStrategy
from ax.service.utils.analysis_base import AnalysisBase
from ax.service.utils.best_point import derelativize_opt_config, is_row_feasible
from ax.service.utils.best_point_mixin import BestPointMixin
from ax.service.utils.orchestrator_options import OrchestratorOptions, TrialType
from ax.service.utils.with_db_settings_base import DBSettings, WithDBSettingsBase
from ax.utils.common.constants import Keys
from ax.utils.common.executils import retry_on_exception
from ax.utils.common.logger import (
build_file_handler,
get_logger,
make_indices_str,
set_ax_logger_levels,
)
from ax.utils.common.timeutils import current_timestamp_in_millis
from pyre_extensions import assert_is_instance, none_throws
NOT_IMPLEMENTED_IN_BASE_CLASS_MSG = """ \
This method is not implemented in the base `Orchestrator` class. \
If this functionality is desired, specify the method in the \
Orchestrator subclass.
"""
GS_TYPE_MSG = "This optimization run uses a '{gs_name}' generation strategy."
OPTIMIZATION_COMPLETION_MSG = """Optimization completed with total of {num_trials} \
trials attached to the underlying Ax experiment '{experiment_name}'.
"""
FAILURE_EXCEEDED_MSG = (
"Failure rate exceeds the tolerated trial failure rate of {f_rate} (at least "
"{n_failed} out of first {n_ran} trials failed or were abandoned). Checks are "
"triggered both at the end of an optimization and if at least {min_failed} trials "
"have either failed, or have been abandoned, potentially automatically due to "
"issues with the trial."
)
EXPECTED_STAGED_MSG = (
"Expected all trials to be in status {expected} after running or staging, "
"found {t_idx_to_status}."
)
# Wait time b/w reports will not exceed 15 mins.
MAX_SECONDS_BETWEEN_REPORTS = 900
[docs]
class OptimizationResult(NamedTuple): # TODO[T61776778]
pass # TBD
[docs]
class OrchestratorInternalError(AxError):
"""Error that indicates an error within the `Orchestrator` logic."""
pass
[docs]
class FailureRateExceededError(AxError):
"""Error that indicates the optimization was aborted due to excessive
failure rate.
"""
pass
[docs]
class StatusQuoInfeasibleError(AxError):
"""Error that indicates the status-quo arm is infeasible
(i.e. violates outcome constraints)."""
pass
NO_RETRY_EXCEPTIONS: tuple[type[Exception], ...] = (
cast(type[Exception], OrchestratorInternalError),
cast(type[Exception], NotImplementedError),
cast(type[Exception], UnsupportedError),
)
[docs]
class OutputPriority(IntEnum):
"""Priority of a message. Messages with higher priority will be shown first, and
messages with the same priority will be sorted alphabetically."""
NOTSET = 0
DEBUG = 10
INFO = 20
TOPLINE = 30
WARNING = 40
ERROR = 50
[docs]
@dataclass
class MessageOutput:
"""Message to be shown in the output of the orchestrator."""
text: str
priority: OutputPriority | int
def __str__(self) -> str:
return self.text
def __repr__(self) -> str:
return f"MessageOutput(text={self.text}, priority={self.priority})"
[docs]
def append(self, text: str) -> None:
"""Append text to the text of an existing message."""
self.text += text
[docs]
class Orchestrator(AnalysisBase, BestPointMixin):
"""Closed-loop manager class for Ax optimization.
Attributes:
experiment: Experiment, in which results of the optimization
will be recorded.
generation_strategy: Generation strategy for the optimization,
describes models that will be used in optimization.
options: `OrchestratorOptions` for this Orchestrator instance.
db_settings: Settings for saving and reloading the underlying experiment
to a database. Expected to be of type
ax.storage.sqa_store.structs.DBSettings and require SQLAlchemy.
_skip_experiment_save: If True, Orchestrator will not re-save the
experiment passed to it. **Use only if the experiment had just
"""
experiment: Experiment
generation_strategy: GenerationStrategy
# pyre-fixme[24]: Generic type `LoggerAdapter` expects 1 type parameter.
logger: LoggerAdapter
# Mapping of form {short string identifier -> message to show in reported
# results}. This is a mapping and not a list to allow for changing of
# some optimization messages throughout the course of the optimization
# (e.g. progress report of the optimization).
markdown_messages: dict[str, MessageOutput]
# Number of trials that existed on the Orchestrator's experiment before
# the Orchestrator instantiation with that experiment.
_num_preexisting_trials: int
# Number of trials remaining to be scheduled during run_trials_and_yield_results.
# Saved as a property so that it can be accessed after optimization is complex (ex.
# for global stopping saving calculation).
_num_remaining_requested_trials: int = 0
# Total number of MetricFetchEs encountered during the course of optimization. Note
# this is different from and may be greater than the number of trials that have
# been marked either FAILED or ABANDONED due to metric fetching errors.
_num_metric_fetch_e_encountered: int = 0
# Number of trials that have been marked either FAILED or ABANDONED due to
# MetricFetchE being encountered during _fetch_and_process_trials_data_results
_num_trials_bad_due_to_err: int = 0
# Keeps track of whether the allowed failure rate has been exceeded during
# the optimization. If true, allows any pending trials to finish and raises
# an error through self._complete_optimization.
_failure_rate_has_been_exceeded: bool = False
# Timestamp of last optimization start time (milliseconds since Unix epoch);
# recorded in each `run_n_trials`.
_latest_optimization_start_timestamp: int | None = None
# Timeout setting for current optimization.
_timeout_hours: float | None = None
# Timestamp of when the last deployed trial started running.
_latest_trial_start_timestamp: float | None = None
# Will be set to `True` if generation strategy signals that the optimization
# is complete, in which case the optimization should gracefully exit early.
_optimization_complete: bool = False
# This will disable the global stopping strategy. It is useful in some
# applications where the user wants to run the optimization loop to exhaust
# the declared number of trials.
__ignore_global_stopping_strategy: bool = False
def __init__(
self,
experiment: Experiment,
generation_strategy: GenerationStrategy,
options: OrchestratorOptions,
db_settings: DBSettings | None = None,
_skip_experiment_save: bool = False,
) -> None:
self.experiment = experiment
# Set up logger with an optional filepath handler. Note: we set the
# logger before setting options since that can trigger errors.
self._set_logger(options=options)
self.options = options
# NOTE: Parallelism schedule is embedded in the generation
# strategy, as `GenerationStep.max_parallelism`.
self.generation_strategy = generation_strategy
if not isinstance(experiment, Experiment):
raise TypeError(f"{experiment} is not an Ax experiment.")
if not isinstance(generation_strategy, GenerationStrategy):
raise TypeError(f"{generation_strategy} is not a generation strategy.")
# Initialize storage layer for the orchestrator.
super().__init__(
db_settings=db_settings,
logging_level=self.options.logging_level,
suppress_all_errors=self.options.suppress_storage_errors_after_retries,
)
# Validate experiment and GS; ensure that experiment has immutable
# search space and opt. config to avoid storing their copies on each
# generator run.
self._validate_remaining_trials(experiment=experiment)
if self.options.enforce_immutable_search_space_and_opt_config:
self._enforce_immutable_search_space_and_opt_config()
if self.db_settings_set and not _skip_experiment_save:
self._maybe_save_experiment_and_generation_strategy(
experiment=experiment, generation_strategy=generation_strategy
)
# Number of trials that existed on experiment before this orchestrator.
self._num_preexisting_trials = len(experiment.trials)
# Whether to log the reason why no trials were generated next time
# we prepare new trials for deployment. Used to avoid spamming logs
# when trials are not generated for the same reason multiple times in
# a row.
self._log_next_no_trials_reason = True
self.markdown_messages = {
"Generation strategy": MessageOutput(
text=GS_TYPE_MSG.format(gs_name=generation_strategy.name),
priority=OutputPriority.DEBUG,
),
}
[docs]
@classmethod
def get_default_db_settings(cls) -> DBSettings:
raise NotImplementedError(
"Base `Orchestrator` does not specify default `DBSettings`. "
"DBSettings are required to leverage SQL storage functionality "
"and can be specified as argument to `Orchestrator` constructor or "
"via `get_default_db_settings` implementation on given orchestrator."
)
[docs]
@classmethod
def from_stored_experiment(
cls,
experiment_name: str,
options: OrchestratorOptions,
db_settings: DBSettings | None = None,
generation_strategy: GenerationStrategy | None = None,
reduced_state: bool = True,
**kwargs: Any,
) -> Orchestrator:
"""Create a ``Orchestrator`` with a previously stored experiment, which
the Orchestrator should resume.
Args:
experiment_name: Experiment to load and resume.
options: ``OrchestratorOptions``, with which to set up the new orchestrator.
db_settings: Optional ``DBSettings``, which to use for reloading the
experiment; also passed as ``db_settings`` argument to the
Orchestrator constructor.
generation_strategy: Generation strategy to use to provide candidates
for the resumed optimization. Provide this argument only if
the experiment does not already have a generation strategy
associated with it.
kwargs: Kwargs to pass through to the ``Orchestrator`` constructor.
"""
dbs = WithDBSettingsBase(
db_settings=db_settings or cls.get_default_db_settings()
)
exp, gs = dbs._load_experiment_and_generation_strategy(
experiment_name=experiment_name,
reduced_state=reduced_state,
)
if db_settings:
kwargs = {**kwargs, "db_settings": db_settings}
if not exp:
raise ValueError(f"Experiment {experiment_name} not found.")
if not gs and not generation_strategy:
raise ValueError(
f"Experiment {experiment_name} did not have a generation "
"strategy associated with it in the database, so a new "
"generation strategy must be provided as argument to "
"`Orchestrator.from_stored_experiment`."
)
if gs and generation_strategy and gs != generation_strategy:
# NOTE: In the future we may want to allow overriding of GS,
# in which case we can add a flag to this function and allow
# the override with warning.
raise UnsupportedError(
"Experiment was associated with generation strategy "
f"{gs.name} in DB, but a new generation strategy "
f"{generation_strategy.name} was provided. To use "
"the generation strategy currently in DB, do not "
"specify the `geneneration_strategy` kwarg."
)
orchestrator = cls(
experiment=exp,
generation_strategy=none_throws(generation_strategy or gs),
options=options,
# No need to resave the experiment we just reloaded.
_skip_experiment_save=True,
# NOTE: `kwargs` can include `db_settings` if those were
# provided to this function.
**kwargs,
)
return orchestrator
@property
def options(self) -> OrchestratorOptions:
"""Orchestrator options."""
return self._options # pyre-ignore [16]
@options.setter
def options(self, options: OrchestratorOptions) -> None:
"""Set Orchestrator options."""
self._validate_options(options=options)
self._options = options
# validate runners and metrics since validate_metrics is an option
self._validate_runner_and_implemented_metrics(experiment=self.experiment)
@property
def trial_type(self) -> str | None:
"""Trial type for the experiment this Orchestrator is running.
This returns None if the experiment is not a MultitypeExperiment
Returns:
Trial type for the experiment this Orchestrator is running if the
experiment is a MultiTypeExperiment and None otherwise.
"""
if isinstance(self.experiment, MultiTypeExperiment):
return self.options.mt_experiment_trial_type
return None
@property
def running_trials(self) -> list[BaseTrial]:
"""Currently running trials.
Note: if the experiment is a MultiTypeExperiment, then this will
only fetch trials of type `orchestrator.trial_type`.
Returns:
List of trials that are currently running.
"""
return filter_trials_by_type(
trials=self.experiment.trials_by_status[TrialStatus.RUNNING],
trial_type=self.trial_type,
)
@property
def trials(self) -> list[BaseTrial]:
"""All trials.
Note: if the experiment is a MultiTypeExperiment, then this will
only fetch trials of type `orchestrator.trial_type`.
Returns:
List of trials that are currently running.
"""
return filter_trials_by_type(
trials=list(self.experiment.trials.values()), trial_type=self.trial_type
)
@property
def running_trial_indices(self) -> set[int]:
"""Currently running trials.
Returns:
List of trials that are currently running.
"""
return get_trial_indices_for_statuses(
experiment=self.experiment,
statuses={TrialStatus.RUNNING},
trial_type=self.trial_type,
)
@property
def failed_abandoned_trial_indices(self) -> set[int]:
"""Failed or abandoned trials.
Note: if the experiment is a MultiTypeExperiment, then this will
only fetch trials of type `orchestrator.trial_type`.
Returns:
List of trials that are currently running.
"""
return get_trial_indices_for_statuses(
experiment=self.experiment,
statuses={TrialStatus.ABANDONED, TrialStatus.FAILED},
trial_type=self.trial_type,
)
@property
def pending_trials(self) -> list[BaseTrial]:
"""Running or staged trials on the experiment this Orchestrator is
running.
Note: if the experiment is a MultiTypeExperiment, then this will
only fetch trials of type `orchestrator.trial_type`.
Returns:
List of trials that are currently running or staged.
"""
staged_trials = filter_trials_by_type(
trials=self.experiment.trials_by_status[TrialStatus.STAGED],
trial_type=self.trial_type,
)
return self.running_trials + staged_trials
@property
def candidate_trials(self) -> list[BaseTrial]:
"""Candidate trials on the experiment this Orchestrator is running.
Note: if the experiment is a MultiTypeExperiment, then this will
only fetch trials of type `orchestrator.trial_type`.
Returns:
List of trials that are currently candidates.
"""
return filter_trials_by_type(
trials=self.experiment.trials_by_status[TrialStatus.CANDIDATE],
trial_type=self.trial_type,
)
@property
def trials_expecting_data(self) -> list[BaseTrial]:
"""Trials expecting data.
Note: if the experiment is a MultiTypeExperiment, then this will
only fetch trials of type `orchestrator.trial_type`.
"""
trials = []
for trial in self.experiment.trials.values():
if trial.status.expecting_data:
if self.trial_type is None or trial.trial_type == self.trial_type:
trials.append(trial)
return trials
@property
def runner(self) -> Runner:
"""``Runner`` specified on the experiment associated with this ``Orchestrator``
instance.
"""
if self.trial_type is not None:
runner = assert_is_instance(
self.experiment, MultiTypeExperiment
).runner_for_trial_type(trial_type=none_throws(self.trial_type))
else:
runner = self.experiment.runner
if runner is None:
raise UnsupportedError(
"`Orchestrator` requires that experiment specifies a `Runner`."
)
return runner
def __repr__(self) -> str:
"""Short user-friendly string representation."""
if not hasattr(self, "experiment"):
# Experiment, generation strategy, etc. attributes have not
# yet been set.
return f"{self.__class__.__name__}"
return (
f"{self.__class__.__name__}(experiment={self.experiment}, "
f"generation_strategy={self.generation_strategy}, options="
f"{self.options})"
)
# ---------- Methods below should generally not be modified in subclasses! ---------
# ---------- I. Methods that are often called outside the `Orchestrator`. ---------
[docs]
def generate_candidates(
self,
num_trials: int = 1,
reduce_state_generator_runs: bool = False,
) -> tuple[list[BaseTrial], Exception | None]:
"""Fetch the latest data and generate new candidate trials.
Args:
num_trials: Number of candidate trials to generate.
reduce_state_generator_runs: Flag to determine
whether to save model state for every generator run (default)
or to only save model state on the final generator run of each
batch.
Returns:
List of trials, empty if generation is not possible.
"""
# Trigger TTL check to ensure expired trials are marked as stale
self.experiment.trials
new_trials, err = self._get_next_trials(
num_trials=num_trials, n=self.options.batch_size
)
if len(new_trials) > 0:
new_generator_runs = [gr for t in new_trials for gr in t.generator_runs]
self._save_or_update_trials_and_generation_strategy_if_possible(
experiment=self.experiment,
trials=new_trials + self.experiment.trials_by_status[TrialStatus.STALE],
generation_strategy=self.generation_strategy,
new_generator_runs=new_generator_runs,
reduce_state_generator_runs=reduce_state_generator_runs,
)
return new_trials, err
[docs]
def run_n_trials(
self,
max_trials: int,
ignore_global_stopping_strategy: bool = False,
timeout_hours: float | None = None,
idle_callback: Callable[[Orchestrator], None] | None = None,
) -> OptimizationResult:
"""Run up to ``max_trials`` trials; will run all ``max_trials`` unless
completion criterion is reached. For base ``Orchestrator``, completion criterion
is reaching total number of trials set in ``OrchestratorOptions``, so if that
option is not specified, this function will run exactly ``max_trials`` trials
always.
Args:
max_trials: Maximum number of trials to run.
ignore_global_stopping_strategy: If set, Orchestrator will skip the global
stopping strategy in ``should_consider_optimization_complete``.
timeout_hours: Limit on length of ths optimization; if reached, the
optimization will abort even if completon criterion is not yet reached.
idle_callback: Callable that takes a Orchestrator instance as an argument to
deliver information while the trials are still running. Any output of
`idle_callback` will not be returned, so `idle_callback` must expose
information in some other way. For example, it could print something
about the state of the Orchestrator or underlying experiment to STDOUT,
write something to a database, or modify a Plotly figure or other object
in place. `ax.service.utils.report_utils.get_figure_and_callback` is a
helper function for generating a callback that will update a Plotly
figure.
Example:
>>> trials_info = {"n_completed": None}
>>>
>>> def write_n_trials(orchestrator:Orchestrator) -> None:
... trials_info["n_completed"] = len(orchestrator.experiment.trials)
>>>
>>> orchestrator.run_n_trials(
... max_trials=3, idle_callback=write_n_trials
... )
>>> print(trials_info["n_completed"])
3
"""
self.poll_and_process_results()
for _ in self.run_trials_and_yield_results(
max_trials=max_trials,
ignore_global_stopping_strategy=ignore_global_stopping_strategy,
timeout_hours=timeout_hours,
idle_callback=idle_callback,
):
pass
return self.summarize_final_result()
[docs]
def run_all_trials(
self,
timeout_hours: float | None = None,
idle_callback: Callable[[Orchestrator], None] | None = None,
) -> OptimizationResult:
"""Run all trials until ``should_consider_optimization_complete`` yields
true (by default, ``should_consider_optimization_complete`` will yield true when
reaching the ``num_trials`` setting, passed to Orchestrator on instantiation as
part of ``OrchestratorOptions``).
NOTE: This function is available only when ``OrchestratorOptions.num_trials`` is
specified.
Args:
timeout_hours: Limit on length of ths optimization; if reached, the
optimization will abort even if completon criterion is not yet reached.
idle_callback: Callable that takes a Orchestrator instance as an argument to
deliver information while the trials are still running. Any output of
`idle_callback` will not be returned, so `idle_callback` must expose
information in some other way. For example, it could print something
about the state of the Orchestrator or underlying experiment to STDOUT,
write something to a database, or modify a Plotly figure or other object
in place. `ax.service.utils.report_utils.get_figure_and_callback` is a
helper function for generating a callback that will update a Plotly
figure.
Example:
>>> trials_info = {"n_completed": None}
>>>
>>> def write_n_trials(orchestrator:Orchestrator) -> None:
... trials_info["n_completed"] = len(orchestrator.experiment.trials)
>>>
>>> orchestrator.run_all_trials(
... timeout_hours=0.1, idle_callback=write_n_trials
... )
>>> print(trials_info["n_completed"])
"""
if self.options.total_trials is None:
# NOTE: Capping on number of trials will likely be needed as fallback
# for most stopping criteria, so we ensure `num_trials` is specified.
raise ValueError(
"Please either specify `num_trials` in `OrchestratorOptions` input "
"to the `Orchestrator` or use `run_n_trials` instead of "
"`run_all_trials`."
)
return self.run_n_trials(
max_trials=none_throws(self.options.total_trials),
timeout_hours=timeout_hours,
idle_callback=idle_callback,
)
[docs]
def run_trials_and_yield_results(
self,
max_trials: int,
ignore_global_stopping_strategy: bool = False,
timeout_hours: float | None = None,
idle_callback: Callable[[Orchestrator], None] | None = None,
) -> Generator[dict[str, Any], None, None]:
"""Make continuous calls to `run` and `process_results` to run up to
``max_trials`` trials, until completion criterion is reached. This is the 'main'
method of a ``Orchestrator``.
Args:
max_trials: Maximum number of trials to run in this generator. The
generator will run trials until a completion criterion is reached,
a completion signal is received from the generation strategy, or
``max_trials`` trials have been run (whichever happens first).
ignore_global_stopping_strategy: If set, Orchestrator will skip the global
stopping strategy in ``should_consider_optimization_complete``.
timeout_hours: Maximum number of hours, for which
to run the optimization. This function will abort after running
for `timeout_hours` even if stopping criterion has not been reached.
If set to `None`, no optimization timeout will be applied.
idle_callback: Callable that takes a Orchestrator instance as an argument to
deliver information while the trials are still running. Any output of
`idle_callback` will not be returned, so `idle_callback` must expose
information in some other way. For example, it could print something
about the state of the Orchestrator or underlying experiment to STDOUT,
write something to a database, or modify a Plotly figure or other object
in place. `ax.service.utils.report_utils.get_figure_and_callback` is a
helper function for generating a callback that will update a Plotly
figure.
"""
if max_trials < 0:
raise ValueError(f"Expected `max_trials` >= 0, got {max_trials}.")
if timeout_hours is not None:
if timeout_hours < 0:
raise UserInputError(
f"Expected `timeout_hours` >= 0, got {timeout_hours}."
)
self._latest_optimization_start_timestamp = current_timestamp_in_millis()
self.__ignore_global_stopping_strategy = ignore_global_stopping_strategy
n_initial_candidate_trials = len(self.candidate_trials)
if n_initial_candidate_trials == 0 and max_trials < 0:
raise UserInputError(f"Expected `max_trials` >= 0, got {max_trials}.")
# trials are pre-existing only if they do not still require running
n_existing = len(self.trials) - n_initial_candidate_trials
# Until completion criterion is reached or `max_trials` is scheduled,
# schedule new trials and poll existing ones in a loop.
self._num_remaining_requested_trials = max_trials
while (
self._num_remaining_requested_trials > 0
and not self.should_consider_optimization_complete()[0]
):
if self.should_abort_optimization(timeout_hours=timeout_hours):
yield self._abort_optimization(num_preexisting_trials=n_existing)
return
# Run new trial evaluations until `run` returns `False`, which
# means that there was a reason not to run more evaluations yet.
# Also check that `max_trials` is not reached to not exceed it.
n_remaining_to_generate = self._num_remaining_requested_trials - len(
self.candidate_trials
)
while self._num_remaining_requested_trials > 0 and self.run(
max_new_trials=n_remaining_to_generate,
timeout_hours=timeout_hours,
):
# Not checking `should_abort_optimization` on every trial for perf.
# reasons.
n_already_run_by_orchestrator = (
len(self.trials) - n_existing - len(self.candidate_trials)
)
self._num_remaining_requested_trials = (
max_trials - n_already_run_by_orchestrator
)
n_remaining_to_generate = self._num_remaining_requested_trials - len(
self.candidate_trials
)
# this is safeguard in case no trial statuses have been updated, and
# wait_for_running_trials=False, in which case we do not want to continue
# to loop and poll
report_results = self._check_exit_status_and_report_results(
n_existing=n_existing, idle_callback=idle_callback, force_refit=False
)
if report_results is None:
return
else:
yield report_results
# When done scheduling, wait for the remaining trials to finish running
# (unless optimization is aborting, in which case stop right away).
if self.running_trials:
self.logger.info(
"Done submitting trials, waiting for remaining "
f"{len(self.running_trials)} running trials..."
)
while self.running_trials:
if self.should_abort_optimization(timeout_hours=timeout_hours):
yield self._abort_optimization(num_preexisting_trials=n_existing)
return
report_results = self._check_exit_status_and_report_results(
n_existing=n_existing, idle_callback=idle_callback, force_refit=True
)
if report_results is None:
return
else:
yield report_results
yield self._complete_optimization(
num_preexisting_trials=n_existing, idle_callback=idle_callback
)
return
# -------- II. Methods that are typically called within the `Orchestrator`. -------
[docs]
@retry_on_exception(retries=3, no_retry_on_exception_types=NO_RETRY_EXCEPTIONS)
def run_trials(
self,
new_trials: list[BaseTrial],
existing_trials: list[BaseTrial],
stage_only: bool = False,
) -> dict[int, dict[str, Any]]:
"""Deployment function, runs a single evaluation for each of the
given trials.
Override default implementation on the ``Runner`` if its desirable to deploy
trials in bulk.
NOTE: the `retry_on_exception` decorator applied to this function should also
be applied to its subclassing override if one is provided and retry behavior
is desired.
Args:
existing_trials: Iterable of trials to be deployed, each containing arms
with parameterizations to be evaluated. Can be a ``Trial``
if contains only one arm or a ``BatchTrial`` if contains
multiple arms.
new_trials: Iterable of trials to be deployed, each containing arms
with parameterizations to be evaluated. Can be a ``Trial``
if contains only one arm or a ``BatchTrial`` if contains
multiple arms. ``GeneratorRun``-s from the new trials will be saved to
the ``GenerationStrategy.generator_runs`` as well as on the trials.
Returns:
Dict of trial index to the run metadata of that trial from the deployment
process.
"""
if stage_only and not self.runner.staging_required:
raise UnsupportedError(
"`Orchestrator.run_trials(stage_only=True)` applies only to runners "
"that require staging trials before they can be run. "
f"Current runner: {self.runner}."
)
all_trials = [*existing_trials, *new_trials]
if not stage_only and self.runner.staging_required:
if not all(t.status == TrialStatus.STAGED for t in all_trials):
raise AxError(
EXPECTED_STAGED_MSG.format(
t_idx_to_status=[(t.index, t.status) for t in all_trials],
expected=TrialStatus.STAGED,
)
)
idcs_str = make_indices_str(indices=(t.index for t in all_trials))
self.logger.info(f"{'Stag' if stage_only else 'Runn'}ing trials {idcs_str}...")
# TODO: Add optional timeout between retries of `run_trial(s)`.
metadata = self.runner.run_multiple(trials=all_trials)
self.logger.debug(f"{'Staged' if stage_only else 'Ran'} trials {idcs_str}.")
if self.options.debug_log_run_metadata:
self.logger.debug(f"Run metadata: {metadata}.")
self._latest_trial_start_timestamp = current_timestamp_in_millis()
self._update_and_save_trials(
existing_trials=existing_trials,
new_trials=new_trials,
metadata=metadata,
stage_only=stage_only,
)
self._log_next_no_trials_reason = True
return metadata
[docs]
@retry_on_exception(retries=3, no_retry_on_exception_types=NO_RETRY_EXCEPTIONS)
def poll_trial_status(
self, poll_all_trial_statuses: bool = False
) -> dict[TrialStatus, set[int]]:
"""Polling function, checks the status of any non-terminal trials
and returns their indices as a mapping from TrialStatus to a list of indices.
NOTE: Does not need to handle waiting between polling while trials
are running; that logic is handled in ``orchestrator.poll``, which calls
this function.
Returns:
A dictionary mapping TrialStatus to a list of trial indices that have
the respective status at the time of the polling. This does not need to
include trials that at the time of polling already have a terminal
(ABANDONED, FAILED, COMPLETED) status (but it may).
"""
trials = (
list(self.experiment.trials.values())
if poll_all_trial_statuses
else self.pending_trials
)
trials = filter_trials_by_type(trials=trials, trial_type=self.trial_type)
if len(trials) == 0:
return {}
return self.runner.poll_trial_status(trials=trials)
[docs]
def wait_for_completed_trials_and_report_results(
self,
idle_callback: Callable[[Orchestrator], None] | None = None,
force_refit: bool = False,
) -> dict[str, Any]:
"""Continuously poll for successful trials, with limited exponential
backoff, and process the results. Stop once at least one successful
trial has been found. This function can be overridden to a different
waiting function as needed; it must call `poll_and_process_results`
to ensure that trials that completed their evaluation are appropriately
marked as 'COMPLETED' in Ax.
Args:
idle_callback: Callable that takes a Orchestrator instance as an argument to
deliver information while the trials are still running. Any output of
`idle_callback` will not be returned, so `idle_callback` must expose
information in some other way. For example, it could print something
about the state of the Orchestrator or underlying experiment to STDOUT,
write something to a database, or modify a Plotly figure or other object
in place. `ax.service.utils.report_utils.get_figure_and_callback` is a
helper function for generating a callback that will update a Plotly
figure.
force_refit: Whether to force a refit of the model during report_results.
Returns:
Results of the optimization so far, represented as a
dict. The contents of the dict depend on the implementation of
`report_results` in the given `Orchestrator` subclass.
"""
if self.options.init_seconds_between_polls is None:
raise ValueError(
"Default `wait_for_completed_trials_and_report_results` in base "
"`Orchestrator` relies on non-null `init_seconds_between_polls` "
"`Orchestrator` option."
)
seconds_between_polls = self.options.init_seconds_between_polls
backoff_factor = self.options.seconds_between_polls_backoff_factor
total_seconds_elapsed = 0
while len(self.pending_trials) > 0 and not self.poll_and_process_results():
if total_seconds_elapsed > MAX_SECONDS_BETWEEN_REPORTS:
break # If maximum wait time reached, check the stopping
# criterion again and and re-attempt scheduling more trials.
if idle_callback is not None:
try:
idle_callback(self)
except Exception as e:
self.logger.warning(
f"Exception raised in ``idle_callback``: {e}. "
"Continuing to poll for completed trials."
)
log_seconds = (
int(seconds_between_polls)
if seconds_between_polls > 2
else seconds_between_polls
)
self.logger.info(
f"Waiting for completed trials (for {log_seconds} sec, "
f"currently running trials: {len(self.running_trials)})."
)
sleep(seconds_between_polls)
total_seconds_elapsed += seconds_between_polls
seconds_between_polls *= backoff_factor
if idle_callback is not None:
idle_callback(self)
return self.report_results(force_refit=force_refit)
[docs]
def should_consider_optimization_complete(self) -> tuple[bool, str]:
"""Whether this Orchestrator should consider this optimization complete and not
run more trials (and conclude the optimization via ``_complete_optimization``).
NOTE: An optimization is considered complete when a generation strategy signaled
completion or when the ``should_consider_optimization_complete`` method on this
Orchestrator evaluates to ``True``. ``should_consider_optimization_complete``
method is also responsible for checking global_stopping_strategy's decision as
well. Alongside the stop decision, this function returns a string describing the
reason for stopping the optimization.
"""
if self._optimization_complete:
return True, ""
if len(self.pending_trials) == 0 and self._get_max_pending_trials() == 0:
return (
True,
"All pending trials have completed and max_pending_trials is zero.",
)
should_stop, message = self._should_stop_due_to_global_stopping_strategy()
if not should_stop:
if self.options.total_trials is None:
return False, ""
should_stop, message = self._should_stop_due_to_total_trials()
if should_stop:
self.logger.info(
f"Completing the optimization: {message}. "
f"`should_consider_optimization_complete` "
f"is `True`, not running more trials."
)
return should_stop, message
[docs]
def should_abort_optimization(self, timeout_hours: float | None = None) -> bool:
"""Checks whether this Orchestrator has reached some intertuption / abort
criterion, such as an overall optimization timeout, tolerated failure rate, etc.
"""
# If failure rate has been exceeded, log a warning and make sure we are not
# scheduling additional trials. Raises an exception after pending trials have
# completed, but does not abort the optimization immediately.
self.error_if_failure_rate_exceeded()
# If the status-quo arm (if specified) is infeasible,
# and if Options.terminate_if_status_quo_infeasible is True,
# raise an exception and abort the optimization.
if (
self.options.terminate_if_status_quo_infeasible
and self.experiment.status_quo
):
self._error_if_status_quo_infeasible()
# if optimization is timed out, return True, else return False
latest_optimization_start_timestamp = self._latest_optimization_start_timestamp
timeout_in_millis = (
timeout_hours * 60 * 60 * 1000 if timeout_hours is not None else None
)
timed_out = False
if (
latest_optimization_start_timestamp is not None
and timeout_in_millis is not None
):
time_elapsed_in_millis = (
current_timestamp_in_millis() - latest_optimization_start_timestamp
)
timed_out = time_elapsed_in_millis >= timeout_in_millis
if timed_out:
self.logger.error(
"Optimization timed out (timeout hours: " f"{timeout_hours})!"
)
return timed_out
[docs]
def report_results(self, force_refit: bool = False) -> dict[str, Any]:
"""Optional user-defined function for reporting intermediate
and final optimization results (e.g. make some API call, write to some
other db). This function is called whenever new results are available during
the optimization.
Args:
force_refit: Whether to force the implementation of this method to
refit the model on generation strategy before using it to produce
results to report (e.g. if using model to visualize data).
Returns:
An optional dictionary with any relevant data about optimization.
"""
# TODO[T61776778]: add utility to get best trial from arbitrary exp.
return {}
[docs]
def summarize_final_result(self) -> OptimizationResult:
"""Get some summary of result: which trial did best, what
were the metric values, what were encountered failures, etc.
"""
return OptimizationResult()
def _check_if_failure_rate_exceeded(self, force_check: bool = False) -> bool:
"""Checks if the failure rate (set in Orchestrator options) has been exceeded at
any point during the optimization.
NOTE: Both FAILED and ABANDONED trial statuses count towards the failure rate.
Args:
force_check: Indicates whether to force a failure-rate check
regardless of the number of trials that have been executed. If False
(default), the check will be skipped if the optimization has fewer than
five failed trials. If True, the check will be performed unless there
are 0 failures.
Effect on state:
If the failure rate has been exceeded, a warning is logged and the private
attribute `_failure_rate_has_been_exceeded` is set to True, which causes the
`_get_max_pending_trials` to return zero, so that no further trials are
scheduled and an error is raised at the end of the optimization.
Returns:
Boolean representing whether the failure rate has been exceeded.
"""
if self._failure_rate_has_been_exceeded:
return True
num_bad_in_orchestrator = self._num_bad_in_orchestrator()
# skip check if 0 failures
if num_bad_in_orchestrator == 0:
return False
# skip check if fewer than min_failed_trials_for_failure_rate_check failures
# unless force_check is True
if (
num_bad_in_orchestrator
< self.options.min_failed_trials_for_failure_rate_check
and not force_check
):
return False
num_ran_in_orchestrator = self._num_ran_in_orchestrator()
failure_rate_exceeded = (
num_bad_in_orchestrator / num_ran_in_orchestrator
) > self.options.tolerated_trial_failure_rate
if failure_rate_exceeded:
if self._num_trials_bad_due_to_err > num_bad_in_orchestrator / 2:
self.logger.warning(
"MetricFetchE INFO: Sweep aborted due to an exceeded error rate, "
"which was primarily caused by failure to fetch metrics. Please "
"check if anything could cause your metrics to be flaky or "
"broken."
)
# NOTE: this private attribute causes `_get_max_pending_trials` to
# return zero, which causes no further trials to be scheduled.
self._failure_rate_has_been_exceeded = True
return True
return False
[docs]
def error_if_failure_rate_exceeded(self, force_check: bool = False) -> None:
"""Raises an exception if the failure rate (set in Orchestrator options) has
been exceeded at any point during the optimization.
NOTE: Both FAILED and ABANDONED trial statuses count towards the failure rate.
Args:
force_check: Indicates whether to force a failure-rate check
regardless of the number of trials that have been executed. If False
(default), the check will be skipped if the optimization has fewer than
five failed trials. If True, the check will be performed unless there
are 0 failures.
"""
if self._check_if_failure_rate_exceeded(force_check=force_check):
raise self._get_failure_rate_exceeded_error(
num_bad_in_orchestrator=self._num_bad_in_orchestrator(),
num_ran_in_orchestrator=self._num_ran_in_orchestrator(),
)
def _error_if_status_quo_infeasible(self) -> None:
"""Raises an exception if the status-quo arm is infeasible and the
`terminate_if_status_quo_infeasible` option is set to True.
"""
status_quo_arm_name = none_throws(self.experiment.status_quo).name
# Find status_quo arm if it has reached terminal status
status_quo_trial = [
trial
for trial in self.experiment.trials.values()
if trial.status.is_terminal
and isinstance(trial, Trial)
and trial.arm is not None
and trial.arm.name == status_quo_arm_name
]
if not status_quo_trial:
# status_quo trial hasn't completed yet
return
try:
data_df = self.experiment.lookup_data().df
status_quo_df = data_df[data_df["arm_name"] == status_quo_arm_name]
if status_quo_df.empty:
return
except Exception as e:
self.logger.warning(
f"Could not fetch data to check status-quo arm feasibility: {e}"
)
return
# Check if the status-quo arm violates outcome constraints
if (
self.experiment.optimization_config is not None
and len(none_throws(self.experiment.optimization_config).all_constraints)
> 0
):
optimization_config = none_throws(self.experiment.optimization_config)
try:
if any(oc.relative for oc in optimization_config.all_constraints):
optimization_config = derelativize_opt_config(
optimization_config=optimization_config,
experiment=self.experiment,
)
feasibility_series = is_row_feasible(
df=status_quo_df,
optimization_config=optimization_config,
undetermined_value=None,
)
is_infeasible = any(
feasibility is False for feasibility in feasibility_series
)
if is_infeasible:
constraint_descriptions = [
f"{c.metric.name} {c.op.name} {c.bound}"
for c in optimization_config.outcome_constraints
]
error_msg = (
f"Status-quo arm '{status_quo_arm_name}' is infeasible. "
f"It violates one or more outcome constraints:\n"
+ "\n".join(f" - {desc}" for desc in constraint_descriptions)
)
raise StatusQuoInfeasibleError(error_msg)
except StatusQuoInfeasibleError as e:
raise e
except Exception as e:
self.logger.warning(
f"Status-quo arm feasibility calculation failed with error: {e}"
)
def _check_exit_status_and_report_results(
self,
n_existing: int,
idle_callback: Callable[[Orchestrator], None] | None,
force_refit: bool,
) -> dict[str, Any] | None:
if not self.options.wait_for_running_trials:
return None
return self.wait_for_completed_trials_and_report_results(
idle_callback, force_refit=True
)
[docs]
def run(self, max_new_trials: int, timeout_hours: float | None = None) -> bool:
"""Schedules trial evaluation(s) if stopping criterion is not triggered,
maximum parallelism is not currently reached, and capacity allows.
Logs any failures / issues.
Args:
max_new_trials: Maximum number of new trials this function should generate
and run (useful when generating and running trials in batches). Note
that this function might also re-deploy existing ``CANDIDATE`` trials
that failed to deploy before, which will not count against this number.
timeout_hours: Maximum number of hours, for which
to run the optimization. This function will abort after running
for `timeout_hours` even if stopping criterion has not been reached.
If set to `None`, no optimization timeout will be applied.
Returns:
Boolean representing success status.
"""
optimization_complete, _ = self.should_consider_optimization_complete()
if optimization_complete:
return False
if self.should_abort_optimization(timeout_hours=timeout_hours):
self.logger.info(
"`should_abort_optimization` is `True`, not running more trials."
)
return False
# Check if capacity allows for running new evaluations and generate as many
# trials as possible, limited by capacity and model requirements.
self._sleep_if_too_early_to_poll()
existing_trials, new_trials = self._prepare_trials(
max_new_trials=max_new_trials
)
if not existing_trials and not new_trials:
# Unable to gen. new run due to max parallelism limit or need for data
# or unable to run trials due to lack of capacity.
if self._optimization_complete:
return False
if len(self.pending_trials) < 1:
raise OrchestratorInternalError(
"No trials are running but model requires more data. This is an "
"invalid state of the Orchestrator, as no more trials can be added "
"but also no more data is expected as there are no running trials. "
"This should be investigated."
)
self._log_next_no_trials_reason = False
return False # Nothing to run.
if existing_trials:
idcs = sorted(t.index for t in existing_trials)
self.logger.debug(f"Will run pre-existing candidate trials: {idcs}.")
# TODO: Add optional timeout between retries of `run_trial(s)`.
self.run_trials(existing_trials=existing_trials, new_trials=new_trials)
return True
[docs]
def poll_and_process_results(self, poll_all_trial_statuses: bool = False) -> bool:
"""Takes the following actions:
1. Poll trial runs for their statuses
2. Find trials to fetch data for
3. Apply new trial statuses
4. Fetch data
5. Early-stop trials where possible
6. Save modified trials, having either new statuses or new data
Returns:
A boolean representing whether any trial evaluations completed
or have been marked as failed or abandoned, changing the number of
currently running trials.
"""
self._sleep_if_too_early_to_poll()
# POLL TRIAL STATUSES
new_status_to_trial_idcs = self.poll_trial_status(
poll_all_trial_statuses=poll_all_trial_statuses
)
trial_indices_with_updated_data_or_status = set()
# GET TRIALS TO FETCH DATA FOR
# This must be done before updating the trial statuses, so we can differentiate
# newly and previously completed trials.
trial_indices_to_fetch = self._get_trial_indices_to_fetch(
new_status_to_trial_idcs=new_status_to_trial_idcs
)
# UPDATE TRIAL STATUSES
trial_indices_with_updated_statuses = self._apply_new_trial_statuses(
new_status_to_trial_idcs=new_status_to_trial_idcs,
)
updated_any_trial_status = len(trial_indices_with_updated_statuses) > 0
trial_indices_with_updated_data_or_status.update(
trial_indices_with_updated_statuses
)
# FETCH DATA FOR TRIALS EXPECTING DATA
trial_indices_with_new_data = (
self._fetch_data_and_return_trial_indices_with_new_data(
trial_idcs=trial_indices_to_fetch,
)
)
trial_indices_with_updated_data_or_status.update(trial_indices_with_new_data)
# EARLY STOP TRIALS
stop_trial_info = early_stopping_utils.should_stop_trials_early(
early_stopping_strategy=self.options.early_stopping_strategy,
trial_indices=self.running_trial_indices,
experiment=self.experiment,
current_node=self.generation_strategy._curr,
)
self.experiment.stop_trial_runs(
trials=[self.experiment.trials[trial_idx] for trial_idx in stop_trial_info],
reasons=list(stop_trial_info.values()),
)
if len(stop_trial_info) > 0:
trial_indices_with_updated_data_or_status.update(set(stop_trial_info))
updated_any_trial_status = True
# UPDATE TRIALS IN DB
if (
len(trial_indices_with_updated_data_or_status) > 0
): # Only save if there were updates.
self.logger.debug(
f"Updating {len(trial_indices_with_updated_data_or_status)} "
"trials in DB."
)
self._save_or_update_trials_in_db_if_possible(
experiment=self.experiment,
trials=[
self.experiment.trials[i]
for i in trial_indices_with_updated_data_or_status
],
)
return updated_any_trial_status
# ------------------------- III. Protected helpers. -----------------------
def _fetch_data_and_return_trial_indices_with_new_data(
self, trial_idcs: set[int]
) -> set[int]:
"""Fetch data for any trials on the experiment that are expecting new data.
Args:
trial_idcs: A set of trial indices to fetch data for.
Returns:
Set of trial indices that were updated with new data. We're not asserting
that the new data is different than the old data, but may want to
in the future.
"""
if len(trial_idcs) > 0:
results = self._fetch_and_process_trials_data_results(
trial_indices=trial_idcs,
)
return {
i
for i, results_by_metric_name in results.items()
for r in results_by_metric_name.values()
if r.is_ok()
}
return set()
def _num_bad_in_orchestrator(self) -> int:
"""Returns the number of trials that have failed or been abandoned in the
orchestrator.
"""
# We only count failed trials with indices that came after the preexisting
# trials on experiment before Orchestrator use.
return sum(
1
for f in self.failed_abandoned_trial_indices
if f >= self._num_preexisting_trials
)
def _num_ran_in_orchestrator(self) -> int:
"""Returns the number of trials that have been run by the orchestrator."""
return len(self.experiment.trials) - self._num_preexisting_trials
def _apply_new_trial_statuses(
self, new_status_to_trial_idcs: dict[TrialStatus, set[int]]
) -> set[int]:
"""Apply new trial statuses to the experiment according to poll results.
Args:
new_status_to_trial_idcs: Changes to be applied to trial statuses from
poll_trial_status.
Returns:
Set of trial indices that were updated with new statuses.
"""
updated_trial_indices = set()
for status, trial_idcs in new_status_to_trial_idcs.items():
if status.is_candidate or status.is_deployed:
# No need to consider candidate, staged or running trials here (none of
# these trials should actually be candidates, but we can filter on that)
continue
if len(trial_idcs) > 0:
idcs = make_indices_str(indices=trial_idcs)
self.logger.info(f"Retrieved {status.name} trials: {idcs}.")
# Update trial statuses and record which trials were updated.
trials = self.experiment.get_trials_by_indices(trial_idcs)
updated_trial_indices.update(trial_idcs)
for trial in trials:
if status.is_failed or status.is_abandoned:
try:
reason = self.runner.poll_exception(trial)
trial.mark_as(status=status, unsafe=True, reason=reason)
except NotImplementedError:
# Some runners do not implement poll_failure_reason, so
# we fall back to marking the without a reason.
trial.mark_as(status=status, unsafe=True)
else:
trial.mark_as(status=status, unsafe=True)
return updated_trial_indices
def _identify_trial_indices_to_fetch(
self,
old_status_to_trial_idcs: Mapping[TrialStatus, set[int]],
new_status_to_trial_idcs: Mapping[TrialStatus, set[int]],
) -> set[int]:
"""
Identify trial indices to fetch data for based on changes in trial statuses.
Args:
old_status_to_trial_idcs: Mapping of old trial statuses
to their corresponding trial indices.
new_status_to_trial_idcs: Mapping of new trial statuses
to their corresponding trial indices.
Returns:
Set of trial indices to fetch data for.
"""
# Get newly completed trials
prev_completed_trial_idcs = old_status_to_trial_idcs.get(
TrialStatus.COMPLETED, set()
) | old_status_to_trial_idcs.get(TrialStatus.EARLY_STOPPED, set())
newly_completed = (
new_status_to_trial_idcs.get(TrialStatus.COMPLETED, set())
- prev_completed_trial_idcs
)
idcs = make_indices_str(indices=newly_completed)
if newly_completed:
self.logger.debug(f"Will fetch data for newly completed trials: {idcs}.")
else:
self.logger.debug("No newly completed trials; not fetching data for any.")
# Get running trials with metrics available while running
running_trial_indices_with_metrics = set()
if any(
m.is_available_while_running() for m in self.experiment.metrics.values()
):
running_trial_indices_with_metrics = new_status_to_trial_idcs.get(
TrialStatus.RUNNING, set()
) | old_status_to_trial_idcs.get(TrialStatus.RUNNING, set())
for status, indices in new_status_to_trial_idcs.items():
if status.is_terminal and indices:
running_trial_indices_with_metrics -= indices
if running_trial_indices_with_metrics:
idcs = make_indices_str(indices=running_trial_indices_with_metrics)
self.logger.debug(
f"Will fetch data for trials: {idcs} because some metrics "
"on experiment are available while trials are running."
)
# Get previously completed trials with new data after completion
recently_completed_trial_indices = self._get_recently_completed_trial_indices()
if len(recently_completed_trial_indices) > 0:
idcs = make_indices_str(indices=recently_completed_trial_indices)
self.logger.debug(
f"Will fetch data for trials: {idcs} because some metrics "
"on experiment have new data after completion."
)
# Combine all trial indices to fetch data for
trial_indices_to_fetch = (
newly_completed
| running_trial_indices_with_metrics
| recently_completed_trial_indices
)
return trial_indices_to_fetch
def _get_trial_indices_to_fetch(
self, new_status_to_trial_idcs: Mapping[TrialStatus, set[int]]
) -> set[int]:
"""Get trial indices to fetch data for the experiment given
`new_status_to_trial_idcs` and metric properties. This should include:
- newly completed trials
- running trials if the experiment has metrics available while running
- previously completed (or early stopped) trials if the experiment
has metrics with new data after completion which finished recently
Args:
new_status_to_trial_idcs: Changes about to be applied to trial statuses.
Returns:
Set of trial indices to fetch data for.
"""
old_status_to_trial_idcs = {status: set() for status in TrialStatus}
for trial in self.trials:
old_status_to_trial_idcs[trial.status].add(trial.index)
return self._identify_trial_indices_to_fetch(
old_status_to_trial_idcs=old_status_to_trial_idcs,
new_status_to_trial_idcs=new_status_to_trial_idcs,
)
def _get_recently_completed_trial_indices(self) -> set[int]:
"""Get trials that have completed within the max period specified by metrics."""
if len(self.experiment.metrics) == 0:
return set()
max_period = max(
m.period_of_new_data_after_trial_completion()
for m in self.experiment.metrics.values()
)
return {
t.index
for t in self.trials_expecting_data
if t.time_completed is not None
and datetime.now() - none_throws(t.time_completed) < max_period
}
def _process_completed_trials(self, newly_completed: set[int]) -> None:
# Fetch the data for newly completed trials; this will cache the data
# for all metrics. By pre-caching the data now, we remove the need to
# fetch it during candidate generation.
idcs = make_indices_str(indices=newly_completed)
self.logger.debug(f"Fetching data for trials: {idcs}.")
self._fetch_and_process_trials_data_results(
trial_indices=newly_completed,
)
def _abort_optimization(self, num_preexisting_trials: int) -> dict[str, Any]:
"""Conclude optimization without waiting for anymore running trials and
return results so far via `report_results`.
"""
self._record_optimization_complete_message()
return self.report_results(force_refit=True)
def _complete_optimization(
self,
num_preexisting_trials: int,
idle_callback: Callable[[Orchestrator], None] | None = None,
) -> dict[str, Any]:
"""Conclude optimization with waiting for anymore running trials and
return final results via `wait_for_completed_trials_and_report_results`.
"""
self._record_optimization_complete_message()
res = self.wait_for_completed_trials_and_report_results(
idle_callback=idle_callback, force_refit=True
)
# Raise an error if the failure rate exceeds tolerance at the
# end of the optimization.
self.error_if_failure_rate_exceeded(force_check=True)
self._warn_if_non_terminal_trials()
return res
def _validate_options(self, options: OrchestratorOptions) -> None:
"""Validates `OrchestratorOptions` for compatibility with given
`Orchestrator` class.
"""
if not (0.0 <= options.tolerated_trial_failure_rate < 1.0):
raise ValueError("`tolerated_trial_failure_rate` must be in [0, 1).")
if options.early_stopping_strategy is not None and options.validate_metrics:
if not any(
m.is_available_while_running() for m in self.experiment.metrics.values()
):
raise ValueError(
"Can only specify an early stopping strategy if at least one "
"metric is marked as `is_available_while_running`. Otherwise, we "
"will be unable to fetch intermediate results with which to "
"evaluate early stopping criteria."
)
if isinstance(self.experiment, MultiTypeExperiment):
if options.mt_experiment_trial_type is None:
raise UserInputError(
"Must specify `mt_experiment_trial_type` for MultiTypeExperiment."
)
if not self.experiment.supports_trial_type(
options.mt_experiment_trial_type
):
raise ValueError(
"Experiment does not support trial type "
f"{options.mt_experiment_trial_type}."
)
elif options.mt_experiment_trial_type is not None:
raise UserInputError(
"`mt_experiment_trial_type` must be None unless the experiment is a "
"MultiTypeExperiment."
)
def _get_max_pending_trials(self) -> int:
"""Returns the maximum number of pending trials specified in the options, or
zero, if the failure rate limit has been exceeded at any point during the
optimization.
"""
if self._failure_rate_has_been_exceeded:
return 0
return self.options.max_pending_trials
def _prepare_trials(
self, max_new_trials: int
) -> tuple[list[BaseTrial], list[BaseTrial]]:
"""Prepares one trial or multiple trials for deployment, based on
whether `run_trials_in_batches` is set to `True` in this Orchestrator's
options.
NOTE: If running trials in batches, exact number of trials run at once
is determined by available capacity and generation strategy's
requirement for more data and parallelism limitation.
Args:
max_new_trials: Maximum number of new trials to generate.
Returns:
Two lists of trials:
- list of existing candidate trials whose deployment was attempted
but failed before (empty if there were no such trials),
- list of new candidate trials that were created in the course of
this function (empty if no new trials were generated).
"""
# 1. Determine available capacity for running trials.
capacity = self.runner.poll_available_capacity()
if capacity != -1 and capacity < 1: # -1 indicates unlimited capacity.
self.logger.debug("There is no capacity to run any trials.")
return [], []
# 2. Determine actual number of trials to run based on capacity,
# limit on pending trials and limit on total trials.
n = capacity if self.options.run_trials_in_batches else 1
total_trials = self.options.total_trials
max_pending_trials = self._get_max_pending_trials()
num_pending_trials = len(self.pending_trials)
max_pending_upper_bound = max_pending_trials - num_pending_trials
if max_pending_upper_bound < 1:
self.logger.debug(
f"`max_pending_trials={max_pending_trials}` and {num_pending_trials} "
"trials are currently pending; not initiating any additional trials."
)
return [], []
n = max_pending_upper_bound if n == -1 else min(max_pending_upper_bound, n)
if total_trials is not None:
left_in_total = total_trials - len(self.trials_expecting_data)
n = min(n, left_in_total)
existing_candidate_trials = self.candidate_trials[:n]
n_new = min(n - len(existing_candidate_trials), max_new_trials)
new_trials, _err = (
self._get_next_trials(num_trials=n_new, n=self.options.batch_size)
if n_new > 0
else (
[],
None,
)
)
return existing_candidate_trials, new_trials
def _get_next_trials(
self, num_trials: int = 1, n: int | None = None
) -> tuple[list[BaseTrial], Exception | None]:
"""Produce up to `num_trials` new generator runs from the underlying
generation strategy and create new trials with them. Logs errors
encountered during generation.
NOTE: Fewer than `num_trials` trials may be produced if generation
strategy runs into its parallelism limit or needs more data to proceed.
Returns:
List of trials, empty if generation is not possible.
"""
try:
generator_runs = self._gen_new_trials_from_generation_strategy(
num_trials=num_trials, n=n
)
except OptimizationComplete as err:
completion_str = f"Optimization complete: {err}"
self.logger.info(completion_str)
self.markdown_messages["Optimization complete"] = MessageOutput(
text=completion_str,
priority=OutputPriority.DEBUG,
)
self._optimization_complete = True
return [], err
except DataRequiredError as err:
# TODO[T62606107]: consider adding a `more_data_required` property to
# check to generation strategy to avoid running into this exception.
if self._log_next_no_trials_reason:
self.logger.info(
"Generated all trials that can be generated currently. "
"Model requires more data to generate more trials."
)
self.logger.debug(f"Message from generation strategy: {err}")
return [], err
except MaxParallelismReachedException as err:
# TODO[T62606107]: consider adding a `step_max_parallelism_reached`
# check to generation strategy to avoid running into this exception.
if self._log_next_no_trials_reason:
self.logger.info(
"Generated all trials that can be generated currently. "
"Max parallelism currently reached."
)
self.logger.debug(f"Message from generation strategy: {err}")
return [], err
except AxGenerationException as err:
if self._log_next_no_trials_reason:
self.logger.info(
"Generated all trials that can be generated currently. "
"`generation_strategy` encountered an error "
f"{err}."
)
self.logger.debug(f"Message from generation strategy: {err}")
return [], err
except OptimizationConfigRequired as err:
if self._log_next_no_trials_reason:
self.logger.info(
"Generated all trials that can be generated currently. "
"`generation_strategy` requires an optimization config "
"to be set before generating more trials."
)
self.logger.debug(f"Message from generation strategy: {err}")
return [], err
if self.options.trial_type == TrialType.TRIAL and any(
len(generator_run_list[0].arms) > 1 or len(generator_run_list) > 1
for generator_run_list in generator_runs
):
raise OrchestratorInternalError(
"Generation strategy produced multiple arms when only one was expected."
)
trials = []
for generator_run_list in generator_runs:
if self.options.trial_type == TrialType.BATCH_TRIAL:
trial = self.experiment.new_batch_trial(
generator_runs=generator_run_list,
ttl_seconds=self.options.ttl_seconds_for_trials,
trial_type=self.trial_type,
should_add_status_quo_arm=self.options.status_quo_weight > 0,
)
else:
trial = self.experiment.new_trial(
generator_run=generator_run_list[0],
ttl_seconds=self.options.ttl_seconds_for_trials,
trial_type=self.trial_type,
)
trials.append(trial)
return trials, None
def _gen_new_trials_from_generation_strategy(
self,
num_trials: int,
n: int | None = None,
) -> list[list[GeneratorRun]]:
"""Generates a list ``GeneratorRun``s of length of ``num_trials`` using the
``_gen_multiple`` method of the Orchestrator's ``generation_strategy``, taking
into account any ``pending`` observations.
"""
self.generation_strategy.experiment = self.experiment
# For ``BatchTrial`-s, we generate trials using the new method that can
# produce GRs for multiple trials, with multiple nodes. But we don't yet
# want to enable that functionality for single-arm use cases of the
# ``Orchestrator``, as it's still in development.
if self.options.trial_type == TrialType.BATCH_TRIAL:
grs = self.generation_strategy.gen(
experiment=self.experiment,
num_trials=num_trials,
n=n,
)
return grs
else:
assert self.options.trial_type == TrialType.TRIAL # Sanity check.
pending = get_pending_observation_features_based_on_trial_status(
experiment=self.experiment
)
grs = self.generation_strategy.gen(
experiment=self.experiment,
num_trials=num_trials,
n=1,
pending_observations=pending,
fixed_features=get_fixed_features_from_experiment(
experiment=self.experiment
),
)
return grs
# TODO: pass self.trial_type to GS.gen for multi-type experiments
def _update_and_save_trials(
self,
existing_trials: list[BaseTrial],
new_trials: list[BaseTrial],
metadata: dict[int, dict[str, Any]],
stage_only: bool,
reduce_state_generator_runs: bool = False,
) -> None:
"""Updates trials with new run metadata and status; saves updates to DB.
Args:
exiting_trials: Trials that existed on this experiment during the
previous call to this function (these are trials, deployment of
which has already been attempted but failed, so we are
re-attempting it; these trials are already saved in DB if using
storage functionality).
new_trials: Trials that were newly created (these trials are not
yet saved in the DB if using storage functionality).
metadata: Run metadata for the trials, from `orchestrator.run_trials`.
Format is {trial index -> trial run metadata}. Trials present in
the metadata dict will be considered `RUNNING`, and the rest of
trials in `existing_trials` or `new_trials` (that are not present
in `metadata`) will be left as `CANDIDATE`.
reduce_state_generator_runs: Flag to determine
whether to save model state for every generator run (default)
or to only save model state on the final generator run of each
batch.
"""
def _process_trial(trial: BaseTrial) -> None:
if trial.index in metadata:
trial.update_run_metadata(metadata=metadata[trial.index])
try:
if stage_only:
trial.mark_staged()
else:
trial.mark_running(no_runner_required=True)
except ValueError as e:
self.logger.warning(
"Unable to mark trial as "
f"{'STAGED' if stage_only else 'RUNNING'} "
f"due to the following error:\n{str(e)}"
)
else:
self.logger.debug(
f"Trial {trial.index} did not deploy, status: {trial.status}."
)
new_generator_runs = []
for trial in existing_trials:
_process_trial(trial)
for trial in new_trials:
new_generator_runs.extend(trial.generator_runs)
_process_trial(trial)
self._save_or_update_trials_and_generation_strategy_if_possible(
experiment=self.experiment,
trials=[*existing_trials, *new_trials],
generation_strategy=self.generation_strategy,
new_generator_runs=new_generator_runs,
reduce_state_generator_runs=reduce_state_generator_runs,
)
def _sleep_if_too_early_to_poll(self) -> None:
"""Wait to query for capacity unless there has been enough time since last
scheduling.
"""
if self._latest_trial_start_timestamp is not None:
seconds_since_run_trial = round(
(
current_timestamp_in_millis()
- none_throws(self._latest_trial_start_timestamp)
)
/ 1000
)
if seconds_since_run_trial < self.options.min_seconds_before_poll:
sleep_duration_seconds = (
self.options.min_seconds_before_poll - seconds_since_run_trial
)
self.logger.debug(
f"Too early to poll, sleeping for {sleep_duration_seconds} seconds"
)
sleep(sleep_duration_seconds)
def _set_logger(self, options: OrchestratorOptions) -> None:
"""Set up the logger with appropriate logging levels."""
cls_name = self.__class__.__name__
logger = get_logger(name=f"{__name__}.{cls_name}@{hex(id(self))}")
set_ax_logger_levels(level=options.logging_level)
if options.log_filepath is not None:
handler = build_file_handler(
filepath=none_throws(options.log_filepath),
level=options.logging_level,
)
logger.addHandler(handler)
self.logger = LoggerAdapter(logger, extra={"output_name": cls_name})
def _validate_remaining_trials(self, experiment: Experiment) -> None:
"""Check how many trials are remaining in `total_trials` given the trials
already on experiment and make sure that there will be trials for the
Orchestrator to run.
"""
if not experiment.trials or not self.options.total_trials:
return
total_trials = none_throws(self.options.total_trials)
preexisting = len(experiment.trials)
msg = (
f"{experiment} already has {preexisting} trials associated with it. "
f"Total trials setting for this Orchestrator is {total_trials}, so "
)
if preexisting >= total_trials:
self.logger.warning(
msg + "no more trials would be run in this Orchestrator if "
"`orchestrator.run_all_trials` is called (but you can still use "
"`orchestrator.run_n_trials` to run a fixed number of trials)."
)
else:
self.logger.info(
msg + "number of trials ran by `orchestrator.run_all_trials` would be "
f"{total_trials - preexisting}."
)
def _validate_runner_and_implemented_metrics(self, experiment: Experiment) -> None:
"""Ensure that the experiment specifies runner and metrics; check that metrics
are not base ``Metric``-s, which do not implement fetching logic.
"""
# this will raise an exception if no runner is set on the experiment
self.runner
metrics_are_invalid = False
if not experiment.metrics:
msg = "`Orchestrator` requires that `experiment.metrics` not be None."
metrics_are_invalid = True
else:
msg = (
"`Orchestrator` requires that experiment specifies metrics "
"with implemented fetching logic."
)
base_metrics = {
m_name for m_name, m in experiment.metrics.items() if type(m) is Metric
}
if base_metrics:
msg += f" Metrics {base_metrics} do not implement fetching logic."
metrics_are_invalid = True
if metrics_are_invalid:
if self.options.validate_metrics:
raise UnsupportedError(msg)
else:
self.logger.warning(msg)
def _enforce_immutable_search_space_and_opt_config(self) -> None:
"""Experiments with immutable search space and optimization config don't
need to keep copies of those objects on each generator run in the experiment,
resulting in large performance gain in storage layer. In `Orchestrator`, we
force-set this immutability on `Experiment`, since Orchestrator experiments
are typically not human-in-the-loop.
"""
if self.experiment.immutable_search_space_and_opt_config:
return
self.logger.info(
f"`Orchestrator` requires experiment to have immutable search "
"space and optimization config. Setting property "
f"{Keys.IMMUTABLE_SEARCH_SPACE_AND_OPT_CONF.value} "
"to `True` on experiment."
)
self.experiment._properties[Keys.IMMUTABLE_SEARCH_SPACE_AND_OPT_CONF.value] = (
True
)
def _record_optimization_complete_message(self) -> None:
"""Adds a simple optimization completion message to this Orchestrator's markdown
messages.
"""
completion_msg = OPTIMIZATION_COMPLETION_MSG.format(
num_trials=len(self.experiment.trials),
experiment_name=(
self.experiment.name if self.experiment._name is not None else "unnamed"
),
)
if "Optimization complete" in self.markdown_messages:
self.markdown_messages["Optimization complete"].append(text=completion_msg)
else:
self.markdown_messages["Optimization complete"] = MessageOutput(
text=completion_msg,
priority=OutputPriority.DEBUG,
)
def _fetch_and_process_trials_data_results(
self,
trial_indices: Iterable[int],
) -> dict[int, dict[str, MetricFetchResult]]:
"""
Fetches results from experiment and modifies trial statuses depending on
success or failure.
"""
try:
kwargs = deepcopy(self.options.fetch_kwargs)
if self.trial_type is not None:
metrics = assert_is_instance(
self.experiment, MultiTypeExperiment
).metrics_for_trial_type(trial_type=none_throws(self.trial_type))
kwargs["metrics"] = metrics
results = self.experiment.fetch_trials_data_results(
trial_indices=trial_indices,
**kwargs,
)
except Exception as e:
self.logger.exception(
f"Failed to fetch data for trials {trial_indices} with error: {e}"
)
return {}
for trial_index, results_by_metric_name in results.items():
for metric_name, result in results_by_metric_name.items():
# If the fetch call succeeded, continue.
if result.is_ok():
continue
# Log the Err so the user is aware that something has failed, even if
# we do not do anything
metric_fetch_e = result.unwrap_err()
# If the metric is available while running just continue (we can try
# again later).
# NOTE: We don't need to report fetching errors in this case either
metric = self.experiment.metrics[metric_name]
status = self.experiment.trials[trial_index].status
if (
metric.is_available_while_running()
and status == TrialStatus.RUNNING
):
self.logger.info(
f"MetricFetchE INFO: Because {metric_name} is "
f"available_while_running and trial {trial_index} is still "
"RUNNING continuing the experiment and retrying on next "
"poll..."
)
continue
self.logger.error(
f"Failed to fetch {metric_name} for trial {trial_index} with "
f"status {status}, found {metric_fetch_e}."
)
self._num_metric_fetch_e_encountered += 1
self._report_metric_fetch_e(
trial=self.experiment.trials[trial_index],
metric_name=metric_name,
metric_fetch_e=metric_fetch_e,
)
# If the fetch failure was for a metric in the optimization config (an
# objective or constraint) mark the trial as failed
optimization_config = self.experiment.optimization_config
if (
optimization_config is not None
and metric_name in optimization_config.metrics.keys()
and not self.experiment.metrics[
metric_name
].is_reconverable_fetch_e(metric_fetch_e=metric_fetch_e)
):
status = self._mark_err_trial_status(
trial=self.experiment.trials[trial_index],
metric_name=metric_name,
metric_fetch_e=metric_fetch_e,
)
self.logger.warning(
f"MetricFetchE INFO: Because {metric_name} is an objective, "
f"marking trial {trial_index} as {status}."
)
self._num_trials_bad_due_to_err += 1
continue
self.logger.info(
"MetricFetchE INFO: Continuing optimization even though "
"MetricFetchE encountered."
)
continue
return results
def _report_metric_fetch_e(
self,
trial: BaseTrial,
metric_name: str,
metric_fetch_e: MetricFetchE,
) -> None:
pass
def _mark_err_trial_status(
self,
trial: BaseTrial,
metric_name: str | None = None,
metric_fetch_e: MetricFetchE | None = None,
) -> TrialStatus:
trial.mark_failed(unsafe=True)
return TrialStatus.FAILED
def _get_failure_rate_exceeded_error(
self,
num_bad_in_orchestrator: int,
num_ran_in_orchestrator: int,
) -> FailureRateExceededError:
return FailureRateExceededError(
FAILURE_EXCEEDED_MSG.format(
f_rate=self.options.tolerated_trial_failure_rate,
n_failed=num_bad_in_orchestrator,
n_ran=num_ran_in_orchestrator,
min_failed=self.options.min_failed_trials_for_failure_rate_check,
)
)
def _warn_if_non_terminal_trials(self) -> None:
"""Warns if there are any non-terminal trials on the experiment."""
non_terminal_trials = [
t.index for t in self.experiment.trials.values() if not t.status.is_terminal
]
if len(non_terminal_trials) > 0:
self.logger.warning(
f"Found {len(non_terminal_trials)} non-terminal trials on "
f"{self.experiment.name}: {non_terminal_trials}."
)
def _should_stop_due_to_global_stopping_strategy(self) -> tuple[bool, str]:
"""Check if optimization should stop due to global stopping strategy."""
if (
self.__ignore_global_stopping_strategy
or self.options.global_stopping_strategy is None
):
return False, ""
gss = none_throws(self.options.global_stopping_strategy)
num_trials = len(self.trials)
if num_trials > 1000:
self.logger.info(
f"There are {num_trials} trials; performing "
f"completion criterion check with {gss}..."
)
stop_optimization, global_stopping_msg = gss.should_stop_optimization(
experiment=self.experiment
)
return stop_optimization, global_stopping_msg
def _should_stop_due_to_total_trials(self) -> tuple[bool, str]:
"""Check if optimization should stop due to total number of trials."""
num_trials = len(self.trials)
should_stop = num_trials >= none_throws(self.options.total_trials)
return (
should_stop,
"Exceeding the total number of trials." if should_stop else "",
)
[docs]
def get_fitted_adapter(
orchestrator: Orchestrator, force_refit: bool = False
) -> Adapter:
"""Returns a fitted Adapter object. If the model is fit already, directly
returns the already fitted model. Otherwise, fits and returns a new one.
Args:
orchestrator:The Orchestrator object from which to get the fitted model.
force_refit: If True, will force a data lookup and a refit of the model.
Returns:
An ``Adapter`` object fitted to the observations of the Orchestrator's
experiment.
"""
gs = orchestrator.generation_strategy
adapter = gs.adapter # Optional[Adapter]
if adapter is None or force_refit: # Need to re-fit the model.
gs._curr._fit(experiment=orchestrator.experiment)
adapter = cast(Adapter, gs.adapter)
return adapter