#!/usr/bin/env python3
# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
import json
import logging
import warnings
from typing import Any, Dict, List, Optional, Tuple, Union
import ax.service.utils.best_point as best_point_utils
import numpy as np
import pandas as pd
from ax.core.arm import Arm
from ax.core.data import Data
from ax.core.experiment import Experiment
from ax.core.generator_run import GeneratorRun
from ax.core.trial import Trial
from ax.core.types import (
TEvaluationOutcome,
TModelPredictArm,
TParameterization,
TParamValue,
)
from ax.modelbridge.generation_strategy import GenerationStrategy
from ax.modelbridge.modelbridge_utils import get_pending_observation_features
from ax.plot.base import AxPlotConfig
from ax.plot.contour import plot_contour
from ax.plot.exp_utils import exp_to_df
from ax.plot.helper import _format_dict, _get_in_sample_arms
from ax.plot.trace import optimization_trace_single_method
from ax.service.utils.dispatch import choose_generation_strategy
from ax.service.utils.instantiation import (
data_from_evaluations,
make_experiment,
raw_data_to_evaluation,
)
from ax.service.utils.storage import (
load_experiment_and_generation_strategy,
save_experiment_and_generation_strategy,
)
from ax.storage.json_store.decoder import (
generation_strategy_from_json,
object_from_json,
)
from ax.storage.json_store.encoder import object_to_json
from ax.utils.common.docutils import copy_doc
from ax.utils.common.logger import _round_floats_for_logging, get_logger
from ax.utils.common.typeutils import checked_cast, checked_cast_dict, not_none
from botorch.utils.sampling import manual_seed
logger = get_logger(__name__)
try: # We don't require SQLAlchemy by default.
from ax.storage.sqa_store.structs import DBSettings
except ModuleNotFoundError: # pragma: no cover
DBSettings = None
[docs]class AxClient:
"""
Convenience handler for management of experimentation cycle through a
service-like API. External system manages scheduling of the cycle and makes
calls to this client to get next suggestion in the experiment and log back
data from the evaluation of that suggestion.
Note: `AxClient` expects to only propose 1 arm (suggestion) per trial; support
for use cases that require use of batches is coming soon.
Two custom types used in this class for convenience are `TParamValue` and
`TParameterization`. Those are shortcuts for `Union[str, bool, float, int]`
and `Dict[str, Union[str, bool, float, int]]`, respectively.
Args:
generation_strategy: Optional generation strategy. If not set, one is
intelligently chosen based on properties of search space.
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.
enforce_sequential_optimization: Whether to enforce that when it is
reasonable to switch models during the optimization (as prescribed
by `num_arms` in generation strategy), Ax will wait for enough trials
to be completed with data to proceed. Defaults to True. If set to
False, Ax will keep generating new trials from the previous model
until enough data is gathered. Use this only if necessary;
otherwise, it is more resource-efficient to
optimize sequentially, by waiting until enough data is available to
use the next model.
random_seed: Optional integer random seed, set to fix the optimization
random seed for reproducibility. Works only for Sobol quasi-random
generator and for BoTorch-powered models. For the latter models, the
trials generated from the same optimization setup with the same seed,
will be mostly similar, but the exact parameter values may still vary
and trials latter in the optimizations will diverge more and more.
This is because a degree of randomness is essential for high performance
of the Bayesian optimization models and is not controlled by the seed.
Note: In multi-threaded environments, the random seed is thread-safe,
but does not actually guarantee reproducibility. Whether the outcomes
will be exactly the same for two same operations that use the random
seed, depends on whether the threads modify the random state in the
same order across the two operations.
verbose_logging: Whether Ax should log significant optimization events,
defaults to `True`.
"""
def __init__(
self,
generation_strategy: Optional[GenerationStrategy] = None,
db_settings: Any = None,
enforce_sequential_optimization: bool = True,
random_seed: Optional[int] = None,
verbose_logging: bool = True,
) -> None:
if not verbose_logging:
logger.setLevel(logging.WARNING)
else:
logger.info(
"Starting optimization with verbose logging. To disable logging, "
"set the `verbose_logging` argument to `False`. Note that float "
"values in the logs are rounded to 2 decimal points."
)
self._generation_strategy = generation_strategy
if db_settings and (not DBSettings or not isinstance(db_settings, DBSettings)):
raise ValueError(
"`db_settings` argument should be of type ax.storage.sqa_store."
"structs.DBSettings. To use `DBSettings`, you will need SQLAlchemy "
"installed in your environment (can be installed through pip)."
)
self.db_settings = db_settings
self._experiment: Optional[Experiment] = None
self._enforce_sequential_optimization = enforce_sequential_optimization
self._random_seed = random_seed
if random_seed is not None:
logger.warning(
f"Random seed set to {random_seed}. Note that this setting "
"only affects the Sobol quasi-random generator "
"and BoTorch-powered Bayesian optimization models. For the latter "
"models, setting random seed to the same number for two optimizations "
"will make the generated trials similar, but not exactly the same, "
"and over time the trials will diverge more."
)
# Trials, for which we received data since last `GenerationStrategy.gen`,
# used to make sure that generation strategy is updated with new data.
self._updated_trials: List[int] = []
# ------------------------ Public API methods. ------------------------
[docs] def create_experiment(
self,
parameters: List[Dict[str, Union[TParamValue, List[TParamValue]]]],
name: Optional[str] = None,
objective_name: Optional[str] = None,
minimize: bool = False,
parameter_constraints: Optional[List[str]] = None,
outcome_constraints: Optional[List[str]] = None,
status_quo: Optional[TParameterization] = None,
overwrite_existing_experiment: bool = False,
) -> None:
"""Create a new experiment and save it if DBSettings available.
Args:
parameters: List of dictionaries representing parameters in the
experiment search space. Required elements in the dictionaries
are: "name" (name of this parameter, string), "type" (type of the
parameter: "range", "fixed", or "choice", string), and "bounds"
for range parameters (list of two values, lower bound first),
"values" for choice parameters (list of values), and "value" for
fixed parameters (single value).
objective: Name of the metric used as objective in this experiment.
This metric must be present in `raw_data` argument to `complete_trial`.
name: Name of the experiment to be created.
minimize: Whether this experiment represents a minimization problem.
parameter_constraints: List of string representation of parameter
constraints, such as "x3 >= x4" or "x3 + x4 + x5 >= 2". For sum
constraints, any number of arguments is accepted, and acceptable
operators are "<=" and ">=".
outcome_constraints: List of string representation of outcome
constraints of form "metric_name >= bound", like "m1 <= 3."
status_quo: Parameterization of the current state of the system.
If set, this will be added to each trial to be evaluated alongside
test configurations.
overwrite_existing_experiment: If `DBSettings` were provided on
instantiation and the experiment being created has the same name
as some experiment already stored, whether to overwrite the
existing experiment. Defaults to False.
"""
if self.db_settings and not name:
raise ValueError( # pragma: no cover
"Must give the experiment a name if `db_settings` is not None."
)
if self.db_settings:
existing = None
try:
existing, _ = load_experiment_and_generation_strategy(
experiment_name=not_none(name), db_settings=self.db_settings
)
except ValueError: # Experiment does not exist, nothing to do.
pass
if existing and overwrite_existing_experiment:
logger.info(f"Overwriting existing experiment {name}.")
elif existing:
raise ValueError(
f"Experiment {name} exists; set the `overwrite_existing_"
"experiment` to `True` to overwrite with new experiment "
"or use `ax_client.load_experiment_from_database` to "
"continue an existing experiment."
)
self._experiment = make_experiment(
name=name,
parameters=parameters,
objective_name=objective_name,
minimize=minimize,
parameter_constraints=parameter_constraints,
outcome_constraints=outcome_constraints,
status_quo=status_quo,
)
if self._generation_strategy is None:
self._generation_strategy = choose_generation_strategy(
search_space=self._experiment.search_space,
enforce_sequential_optimization=self._enforce_sequential_optimization,
random_seed=self._random_seed,
)
self._save_experiment_and_generation_strategy_to_db_if_possible()
[docs] def get_next_trial(self) -> Tuple[TParameterization, int]:
"""
Generate trial with the next set of parameters to try in the iteration process.
Note: Service API currently supports only 1-arm trials.
Returns:
Tuple of trial parameterization, trial index
"""
trial = self.experiment.new_trial(generator_run=self._gen_new_generator_run())
logger.info(
f"Generated new trial {trial.index} with parameters "
f"{_round_floats_for_logging(item=not_none(trial.arm).parameters)}."
)
trial.mark_dispatched()
self._updated_trials = []
self._save_experiment_and_generation_strategy_to_db_if_possible()
return not_none(trial.arm).parameters, trial.index
[docs] def complete_trial(
self,
trial_index: int,
raw_data: TEvaluationOutcome,
metadata: Optional[Dict[str, str]] = None,
sample_size: Optional[int] = None,
) -> None:
"""
Completes the trial with given metric values and adds optional metadata
to it.
Args:
trial_index: Index of trial within the experiment.
raw_data: Evaluation data for the trial. Can be a mapping from
metric name to a tuple of mean and SEM, just a tuple of mean and
SEM if only one metric in optimization, or just the mean if there
is no SEM. Can also be a list of (fidelities, mapping from
metric name to a tuple of mean and SEM).
metadata: Additional metadata to track about this run.
"""
assert isinstance(
trial_index, int
), f"Trial index must be an int, got: {trial_index}." # pragma: no cover
trial = self.experiment.trials[trial_index]
if not isinstance(trial, Trial):
raise NotImplementedError(
"The Service API only supports `Trial`, not `BatchTrial`."
)
if metadata is not None:
trial._run_metadata = metadata
arm_name = not_none(trial.arm).name
evaluations = {
arm_name: raw_data_to_evaluation(
raw_data=raw_data, objective_name=self.objective_name
)
}
sample_sizes = {arm_name: sample_size} if sample_size else {}
data = data_from_evaluations(
evaluations=evaluations, trial_index=trial.index, sample_sizes=sample_sizes
)
# In service API, a trial may be completed multiple times (for multiple
# metrics, for example).
trial.mark_completed(allow_repeat_completion=True)
self.experiment.attach_data(data)
data_for_logging = _round_floats_for_logging(
item=evaluations[next(iter(evaluations.keys()))]
)
logger.info(
f"Completed trial {trial_index} with data: "
f"{_round_floats_for_logging(item=data_for_logging)}."
)
self._updated_trials.append(trial_index)
self._save_experiment_and_generation_strategy_to_db_if_possible()
[docs] def log_trial_failure(
self, trial_index: int, metadata: Optional[Dict[str, str]] = None
) -> None:
"""Mark that the given trial has failed while running.
Args:
trial_index: Index of trial within the experiment.
metadata: Additional metadata to track about this run.
"""
trial = self.experiment.trials[trial_index]
trial.mark_failed()
logger.info(f"Registered failure of trial {trial_index}.")
if metadata is not None:
trial._run_metadata = metadata
self._save_experiment_and_generation_strategy_to_db_if_possible()
[docs] def attach_trial(
self, parameters: TParameterization
) -> Tuple[TParameterization, int]:
"""Attach a new trial with the given parameterization to the experiment.
Args:
parameters: Parameterization of the new trial.
Returns:
Tuple of parameterization and trial index from newly created trial.
"""
trial = self.experiment.new_trial().add_arm(Arm(parameters=parameters))
trial.mark_dispatched()
logger.info(
"Attached custom parameterization "
f"{_round_floats_for_logging(item=parameters)} as trial {trial.index}."
)
self._save_experiment_and_generation_strategy_to_db_if_possible()
return not_none(trial.arm).parameters, trial.index
[docs] def get_trial_parameters(self, trial_index: int) -> TParameterization:
"""Retrieve the parameterization of the trial by the given index."""
if trial_index not in self.experiment.trials:
raise ValueError(f"Trial {trial_index} does not yet exist.")
trial = checked_cast(Trial, self.experiment.trials.get(trial_index))
return not_none(trial.arm).parameters
[docs] @copy_doc(best_point_utils.get_best_parameters)
def get_best_parameters(
self
) -> Optional[Tuple[TParameterization, Optional[TModelPredictArm]]]:
return best_point_utils.get_best_parameters(self.experiment)
[docs] def get_trials_data_frame(self) -> pd.DataFrame:
return exp_to_df(exp=self.experiment)
[docs] def get_recommended_max_parallelism(self) -> List[Tuple[int, int]]:
"""Recommends maximum number of trials that can be scheduled in parallel
at different stages of optimization.
Some optimization algorithms profit significantly from sequential
optimization (e.g. suggest a few points, get updated with data for them,
repeat). This setting indicates how many trials should be in flight
(generated, but not yet completed with data).
The output of this method is mapping of form
{num_trials -> max_parallelism_setting}, where the max_parallelism_setting
is used for num_trials trials. If max_parallelism_setting is -1, as
many of the trials can be ran in parallel, as necessary. If num_trials
in a tuple is -1, then the corresponding max_parallelism_setting
should be used for all subsequent trials.
For example, if the returned list is [(5, -1), (12, 6), (-1, 3)],
the schedule could be: run 5 trials in parallel, run 6 trials in
parallel twice, run 3 trials in parallel for as long as needed. Here,
'running' a trial means obtaining a next trial from `AxClient` through
get_next_trials and completing it with data when available.
Returns:
Mapping of form {num_trials -> max_parallelism_setting}.
"""
parallelism_settings = []
for step in self.generation_strategy._steps:
parallelism_settings.append(
(step.num_arms, step.recommended_max_parallelism or step.num_arms)
)
return parallelism_settings
[docs] def get_optimization_trace(
self, objective_optimum: Optional[float] = None
) -> AxPlotConfig:
"""Retrieves the plot configuration for optimization trace, which shows
the evolution of the objective mean over iterations.
Args:
objective_optimum: Optimal objective, if known, for display in the
visualization.
"""
if not self.experiment.trials:
raise ValueError("Cannot generate plot as there are no trials.")
objective_name = self.experiment.optimization_config.objective.metric.name
best_objectives = np.array(
[
[
checked_cast(Trial, trial).objective_mean
for trial in self.experiment.trials.values()
]
]
)
hover_labels = [
_format_dict(not_none(checked_cast(Trial, trial).arm).parameters)
for trial in self.experiment.trials.values()
]
return optimization_trace_single_method(
y=(
np.minimum.accumulate(best_objectives, axis=1)
if self.experiment.optimization_config.objective.minimize
else np.maximum.accumulate(best_objectives, axis=1)
),
optimum=objective_optimum,
title="Model performance vs. # of iterations",
ylabel=objective_name.capitalize(),
hover_labels=hover_labels,
)
[docs] def get_contour_plot(
self,
param_x: Optional[str] = None,
param_y: Optional[str] = None,
metric_name: Optional[str] = None,
) -> AxPlotConfig:
"""Retrieves a plot configuration for a contour plot of the response
surface. For response surfaces with more than two parameters,
selected two parameters will appear on the axes, and remaining parameters
will be affixed to the middle of their range. If contour params arguments
are not provided, the first two parameters in the search space will be
used. If contour metrics are not provided, objective will be used.
Args:
param_x: name of parameters to use on x-axis for
the contour response surface plots.
param_y: name of parameters to use on y-axis for
the contour response surface plots.
metric_name: Name of the metric, for which to plot the response
surface.
"""
if not self.experiment.trials:
raise ValueError("Cannot generate plot as there are no trials.")
if len(self.experiment.parameters) < 2:
raise ValueError(
"Cannot create a contour plot as experiment has less than 2 "
"parameters, but a contour-related argument was provided."
)
if (param_x or param_y) and not (param_x and param_y):
raise ValueError(
"If `param_x` is provided, `param_y` is "
"required as well, and vice-versa."
)
objective_name = self.objective_name
if not metric_name:
metric_name = objective_name
if not param_x or not param_y:
parameter_names = list(self.experiment.parameters.keys())
param_x = parameter_names[0]
param_y = parameter_names[1]
if param_x not in self.experiment.parameters:
raise ValueError(
f'Parameter "{param_x}" not found in the optimization search space.'
)
if param_y not in self.experiment.parameters:
raise ValueError(
f'Parameter "{param_y}" not found in the optimization search space.'
)
if metric_name not in self.experiment.metrics:
raise ValueError(
f'Metric "{metric_name}" is not associated with this optimization.'
)
if self.generation_strategy.model is not None:
try:
logger.info(
f"Retrieving contour plot with parameter '{param_x}' on X-axis "
f"and '{param_y}' on Y-axis, for metric '{metric_name}'. "
"Ramaining parameters are affixed to the middle of their range."
)
return plot_contour(
model=not_none(self.generation_strategy.model),
param_x=param_x,
param_y=param_y,
metric_name=metric_name,
)
except NotImplementedError:
# Some models don't implement '_predict', which is needed
# for the contour plots.
logger.info(
f"Model {self.generation_strategy.model} does not implement "
"`predict`, so it cannot be used to generate a response "
"surface plot."
)
raise ValueError(
f'Could not obtain contour plot of "{metric_name}" for parameters '
f'"{param_x}" and "{param_y}", as a model with predictive ability, '
"such as a Gaussian Process, has not yet been trained in the course "
"of this optimization."
)
[docs] def load_experiment_from_database(self, experiment_name: str) -> None:
"""Load an existing experiment from database using the `DBSettings`
passed to this `AxClient` on instantiation.
Args:
experiment_name: Name of the experiment.
Returns:
Experiment object.
"""
if not self.db_settings:
raise ValueError( # pragma: no cover
"Cannot load an experiment in the absence of the DB settings."
"Please initialize `AxClient` with DBSettings."
)
experiment, generation_strategy = load_experiment_and_generation_strategy(
experiment_name=experiment_name, db_settings=self.db_settings
)
self._experiment = experiment
logger.info(f"Loaded {experiment}.")
if generation_strategy is None: # pragma: no cover
self._generation_strategy = choose_generation_strategy(
search_space=self._experiment.search_space,
enforce_sequential_optimization=self._enforce_sequential_optimization,
random_seed=self._random_seed,
)
else:
self._generation_strategy = generation_strategy
logger.info(
f"Using generation strategy associated with the loaded experiment:"
f" {generation_strategy}."
)
[docs] def get_model_predictions(
self, metric_names: Optional[List[str]] = None
) -> Dict[int, Dict[str, Tuple[float, float]]]:
"""Retrieve model-estimated means and covariances for all metrics.
Note: this function retrieves the predictions for the 'in-sample' arms,
which means that the return mapping on this function will only contain
predictions for trials that have been completed with data.
Args:
metric_names: Names of the metrics, for which to retrieve predictions.
All metrics on experiment will be retrieved if this argument was
not specified.
Returns:
A mapping from trial index to a mapping of metric names to tuples
of predicted metric mean and SEM, of form:
{ trial_index -> { metric_name: ( mean, SEM ) } }.
"""
if self.generation_strategy.model is None: # pragma: no cover
raise ValueError("No model has been instantiated yet.")
if metric_names is None and self.experiment.metrics is None:
raise ValueError( # pragma: no cover
"No metrics to retrieve specified on the experiment or as "
"argument to `get_model_predictions`."
)
arm_info, _, _ = _get_in_sample_arms(
model=not_none(self.generation_strategy.model),
metric_names=set(metric_names)
if metric_names is not None
else set(not_none(self.experiment.metrics).keys()),
)
trials = checked_cast_dict(int, Trial, self.experiment.trials)
return {
trial_index: {
m: (
arm_info[not_none(trials[trial_index].arm).name].y_hat[m],
arm_info[not_none(trials[trial_index].arm).name].se_hat[m],
)
for m in arm_info[not_none(trials[trial_index].arm).name].y_hat
}
for trial_index in trials
if not_none(trials[trial_index].arm).name in arm_info
}
# ------------------ JSON serialization & storage methods. -----------------
[docs] def save_to_json_file(self, filepath: str = "ax_client_snapshot.json") -> None:
"""Save a JSON-serialized snapshot of this `AxClient`'s settings and state
to a .json file by the given path.
"""
with open(filepath, "w+") as file: # pragma: no cover
file.write(json.dumps(self.to_json_snapshot()))
logger.info(f"Saved JSON-serialized state of optimization to `{filepath}`.")
[docs] @staticmethod
def load_from_json_file(filepath: str = "ax_client_snapshot.json") -> "AxClient":
"""Restore an `AxClient` and its state from a JSON-serialized snapshot,
residing in a .json file by the given path.
"""
with open(filepath, "r") as file: # pragma: no cover
serialized = json.loads(file.read())
return AxClient.from_json_snapshot(serialized=serialized)
[docs] def to_json_snapshot(self) -> Dict[str, Any]:
"""Serialize this `AxClient` to JSON to be able to interrupt and restart
optimization and save it to file by the provided path.
Returns:
A JSON-safe dict representation of this `AxClient`.
"""
return {
"_type": self.__class__.__name__,
"experiment": object_to_json(self._experiment),
"generation_strategy": object_to_json(self._generation_strategy),
"_enforce_sequential_optimization": self._enforce_sequential_optimization,
"_updated_trials": object_to_json(self._updated_trials),
}
[docs] @staticmethod
def from_json_snapshot(serialized: Dict[str, Any]) -> "AxClient":
"""Recreate an `AxClient` from a JSON snapshot."""
experiment = object_from_json(serialized.pop("experiment"))
serialized_generation_strategy = serialized.pop("generation_strategy")
ax_client = AxClient(
generation_strategy=generation_strategy_from_json(
generation_strategy_json=serialized_generation_strategy
)
if serialized_generation_strategy is not None
else None,
enforce_sequential_optimization=serialized.pop(
"_enforce_sequential_optimization"
),
)
ax_client._experiment = experiment
ax_client._updated_trials = object_from_json(serialized.pop("_updated_trials"))
return ax_client
# ---------------------- Private helper methods. ---------------------
@property
def experiment(self) -> Experiment:
"""Returns the experiment set on this Ax client"""
if self._experiment is None:
raise ValueError(
"Experiment not set on Ax client. Must first "
"call load_experiment or create_experiment to use handler functions."
)
return not_none(self._experiment)
@property
def generation_strategy(self) -> GenerationStrategy:
"""Returns the generation strategy, set on this experiment."""
if self._generation_strategy is None:
raise ValueError(
"No generation strategy has been set on this optimization yet."
)
return not_none(self._generation_strategy)
@property
def objective_name(self) -> str:
"""Returns the name of the objective in this optimization."""
opt_config = not_none(self.experiment.optimization_config)
return opt_config.objective.metric.name
def _save_experiment_and_generation_strategy_to_db_if_possible(self) -> bool:
"""Saves attached experiment and generation strategy if DB settings are
set on this AxClient instance.
Returns:
bool: Whether the experiment was saved.
"""
if self.db_settings is not None:
save_experiment_and_generation_strategy(
experiment=self.experiment,
generation_strategy=self.generation_strategy,
db_settings=self.db_settings,
)
return False
def _get_new_data(self) -> Data:
"""
Returns new data since the last run of the generator.
Returns:
Latest data.
"""
return Data.from_multiple_data(
[self.experiment.lookup_data_for_trial(idx) for idx in self._updated_trials]
)
def _gen_new_generator_run(self, n: int = 1) -> GeneratorRun:
"""Generate new generator run for this experimebt.
Args:
n: Number of arms to generate.
"""
new_data = self._get_new_data()
# If random seed is not set for this optimization, context manager does
# nothing; otherwise, it sets the random seed for torch, but only for the
# scope of this call. This is important because torch seed is set globally,
# so if we just set the seed without the context manager, it can have
# serious negative impact on the performance of the models that employ
# stochasticity.
with manual_seed(seed=self._random_seed) and warnings.catch_warnings():
# Filter out GPYTorch warnings to avoid confusing users.
warnings.simplefilter("ignore")
return not_none(self.generation_strategy).gen(
experiment=self.experiment,
new_data=new_data,
n=n,
pending_observations=get_pending_observation_features(
experiment=self.experiment
),
)
# -------- Backward-compatibility with old save / load method names. -------
[docs] @staticmethod
def load_experiment(experiment_name: str) -> None:
raise NotImplementedError(
"Use `load_experiment_from_database` to load from SQL database or "
"`load_from_json_file` to load optimization state from .json file."
)
[docs] @staticmethod
def load(filepath: Optional[str] = None) -> None:
raise NotImplementedError(
"Use `load_experiment_from_database` to load from SQL database or "
"`load_from_json_file` to load optimization state from .json file."
)
[docs] @staticmethod
def save(filepath: Optional[str] = None) -> None:
raise NotImplementedError(
"Use `save_to_json_file` to save optimization state to .json file."
)
