#!/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.
import re
from typing import Any, Dict, Type
from ax.benchmark.problems.hpo.torchvision import PyTorchCNNTorchvisionBenchmarkProblem
from ax.core import ObservationFeatures
from ax.core.arm import Arm
from ax.core.batch_trial import BatchTrial
from ax.core.data import Data
from ax.core.experiment import Experiment
from ax.core.generator_run import GeneratorRun
from ax.core.map_data import MapData, MapKeyInfo
from ax.core.metric import Metric
from ax.core.multi_type_experiment import MultiTypeExperiment
from ax.core.objective import MultiObjective, Objective, ScalarizedObjective
from ax.core.optimization_config import (
MultiObjectiveOptimizationConfig,
OptimizationConfig,
)
from ax.core.outcome_constraint import OutcomeConstraint
from ax.core.parameter import ChoiceParameter, FixedParameter, RangeParameter
from ax.core.parameter_constraint import (
OrderConstraint,
ParameterConstraint,
SumConstraint,
)
from ax.core.parameter_distribution import ParameterDistribution
from ax.core.risk_measures import RiskMeasure
from ax.core.runner import Runner
from ax.core.search_space import RobustSearchSpace, SearchSpace
from ax.core.trial import Trial
from ax.early_stopping.strategies import (
LogicalEarlyStoppingStrategy,
PercentileEarlyStoppingStrategy,
ThresholdEarlyStoppingStrategy,
)
from ax.modelbridge.completion_criterion import CompletionCriterion
from ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy
from ax.modelbridge.registry import _encode_callables_as_references
from ax.modelbridge.transforms.base import Transform
from ax.models.torch.botorch_modular.model import BoTorchModel
from ax.models.torch.botorch_modular.surrogate import Surrogate
from ax.models.winsorization_config import WinsorizationConfig
from ax.storage.botorch_modular_registry import CLASS_TO_REGISTRY
from ax.storage.transform_registry import TRANSFORM_REGISTRY
from ax.utils.common.serialization import serialize_init_args
from ax.utils.common.typeutils import not_none
[docs]def experiment_to_dict(experiment: Experiment) -> Dict[str, Any]:
"""Convert Ax experiment to a dictionary."""
return {
"__type": experiment.__class__.__name__,
"name": experiment._name,
"description": experiment.description,
"experiment_type": experiment.experiment_type,
"search_space": experiment.search_space,
"optimization_config": experiment.optimization_config,
"tracking_metrics": list(experiment._tracking_metrics.values()),
"runner": experiment.runner,
"status_quo": experiment.status_quo,
"time_created": experiment.time_created,
"trials": experiment.trials,
"is_test": experiment.is_test,
"data_by_trial": experiment.data_by_trial,
"properties": experiment._properties,
"default_data_type": experiment._default_data_type,
}
[docs]def multi_type_experiment_to_dict(experiment: MultiTypeExperiment) -> Dict[str, Any]:
"""Convert AE multitype experiment to a dictionary."""
multi_type_dict = {
"default_trial_type": experiment._default_trial_type,
"_metric_to_canonical_name": experiment._metric_to_canonical_name,
"_metric_to_trial_type": experiment._metric_to_trial_type,
"_trial_type_to_runner": experiment._trial_type_to_runner,
}
multi_type_dict.update(experiment_to_dict(experiment))
return multi_type_dict
[docs]def batch_to_dict(batch: BatchTrial) -> Dict[str, Any]:
"""Convert Ax batch to a dictionary."""
return {
"__type": batch.__class__.__name__,
"index": batch.index,
"trial_type": batch.trial_type,
"ttl_seconds": batch.ttl_seconds,
"status": batch.status,
"status_quo": batch.status_quo,
"status_quo_weight_override": batch._status_quo_weight_override,
"time_created": batch.time_created,
"time_completed": batch.time_completed,
"time_staged": batch.time_staged,
"time_run_started": batch.time_run_started,
"abandoned_reason": batch.abandoned_reason,
"run_metadata": batch.run_metadata,
"stop_metadata": batch.stop_metadata,
"generator_run_structs": batch.generator_run_structs,
"runner": batch.runner,
"abandoned_arms_metadata": batch._abandoned_arms_metadata,
"num_arms_created": batch._num_arms_created,
"optimize_for_power": batch.optimize_for_power,
"generation_step_index": batch._generation_step_index,
"properties": batch._properties,
}
[docs]def trial_to_dict(trial: Trial) -> Dict[str, Any]:
"""Convert Ax trial to a dictionary."""
return {
"__type": trial.__class__.__name__,
"index": trial.index,
"trial_type": trial.trial_type,
"ttl_seconds": trial.ttl_seconds,
"status": trial.status,
"time_created": trial.time_created,
"time_completed": trial.time_completed,
"time_staged": trial.time_staged,
"time_run_started": trial.time_run_started,
"abandoned_reason": trial.abandoned_reason,
"run_metadata": trial.run_metadata,
"stop_metadata": trial.stop_metadata,
"generator_run": trial.generator_run,
"runner": trial.runner,
"num_arms_created": trial._num_arms_created,
"generation_step_index": trial._generation_step_index,
"properties": trial._properties,
}
[docs]def range_parameter_to_dict(parameter: RangeParameter) -> Dict[str, Any]:
"""Convert Ax range parameter to a dictionary."""
return {
"__type": parameter.__class__.__name__,
"name": parameter.name,
"parameter_type": parameter.parameter_type,
"lower": parameter.lower,
"upper": parameter.upper,
"log_scale": parameter.log_scale,
"logit_scale": parameter.logit_scale,
"digits": parameter.digits,
"is_fidelity": parameter.is_fidelity,
"target_value": parameter.target_value,
}
[docs]def choice_parameter_to_dict(parameter: ChoiceParameter) -> Dict[str, Any]:
"""Convert Ax choice parameter to a dictionary."""
return {
"__type": parameter.__class__.__name__,
"is_ordered": parameter.is_ordered,
"is_task": parameter.is_task,
"name": parameter.name,
"parameter_type": parameter.parameter_type,
"values": parameter.values,
"is_fidelity": parameter.is_fidelity,
"target_value": parameter.target_value,
"dependents": parameter.dependents if parameter.is_hierarchical else None,
}
[docs]def fixed_parameter_to_dict(parameter: FixedParameter) -> Dict[str, Any]:
"""Convert Ax fixed parameter to a dictionary."""
return {
"__type": parameter.__class__.__name__,
"name": parameter.name,
"parameter_type": parameter.parameter_type,
"value": parameter.value,
"is_fidelity": parameter.is_fidelity,
"target_value": parameter.target_value,
"dependents": parameter.dependents if parameter.is_hierarchical else None,
}
[docs]def order_parameter_constraint_to_dict(
parameter_constraint: OrderConstraint,
) -> Dict[str, Any]:
"""Convert Ax order parameter constraint to a dictionary."""
return {
"__type": parameter_constraint.__class__.__name__,
"lower_name": parameter_constraint.lower_parameter.name,
"upper_name": parameter_constraint.upper_parameter.name,
}
[docs]def sum_parameter_constraint_to_dict(
parameter_constraint: SumConstraint,
) -> Dict[str, Any]:
"""Convert Ax sum parameter constraint to a dictionary."""
return {
"__type": parameter_constraint.__class__.__name__,
"parameter_names": parameter_constraint._parameter_names,
"is_upper_bound": parameter_constraint._is_upper_bound,
# SumParameterConstraint constructor takes in absolute value of
# the bound and transforms it based on the is_upper_bound value
"bound": abs(parameter_constraint._bound),
}
[docs]def parameter_constraint_to_dict(
parameter_constraint: ParameterConstraint,
) -> Dict[str, Any]:
"""Convert Ax sum parameter constraint to a dictionary."""
return {
"__type": parameter_constraint.__class__.__name__,
"constraint_dict": parameter_constraint.constraint_dict,
"bound": parameter_constraint.bound,
}
[docs]def arm_to_dict(arm: Arm) -> Dict[str, Any]:
"""Convert Ax arm to a dictionary."""
return {
"__type": arm.__class__.__name__,
"parameters": arm.parameters,
"name": arm._name,
}
[docs]def search_space_to_dict(search_space: SearchSpace) -> Dict[str, Any]:
"""Convert Ax search space to a dictionary."""
return {
"__type": search_space.__class__.__name__,
"parameters": list(search_space.parameters.values()),
"parameter_constraints": search_space.parameter_constraints,
}
[docs]def robust_search_space_to_dict(rss: RobustSearchSpace) -> Dict[str, Any]:
"""Convert robust search space to a dictionary."""
return {
"__type": rss.__class__.__name__,
"parameters": list(rss._parameters.values()),
"parameter_distributions": rss.parameter_distributions,
"num_samples": rss.num_samples,
"environmental_variables": list(rss._environmental_variables.values()),
"parameter_constraints": rss.parameter_constraints,
}
[docs]def parameter_distribution_to_dict(dist: ParameterDistribution) -> Dict[str, Any]:
"""Convert a parameter distribution to a dictionary."""
return {
"__type": dist.__class__.__name__,
"parameters": dist.parameters,
"distribution_class": dist.distribution_class,
"distribution_parameters": dist.distribution_parameters,
"multiplicative": dist.multiplicative,
}
[docs]def metric_to_dict(metric: Metric) -> Dict[str, Any]:
"""Convert Ax metric to a dictionary."""
properties = metric.serialize_init_args(obj=metric)
properties["__type"] = metric.__class__.__name__
return properties
[docs]def objective_to_dict(objective: Objective) -> Dict[str, Any]:
"""Convert Ax objective to a dictionary."""
return {
"__type": objective.__class__.__name__,
"metric": objective.metric,
"minimize": objective.minimize,
}
[docs]def multi_objective_to_dict(objective: MultiObjective) -> Dict[str, Any]:
"""Convert Ax objective to a dictionary."""
return {
"__type": objective.__class__.__name__,
"objectives": objective.objectives,
"weights": objective.weights,
}
[docs]def scalarized_objective_to_dict(objective: ScalarizedObjective) -> Dict[str, Any]:
"""Convert Ax objective to a dictionary."""
return {
"__type": objective.__class__.__name__,
"metrics": objective.metrics,
"weights": objective.weights,
"minimize": objective.minimize,
}
[docs]def outcome_constraint_to_dict(outcome_constraint: OutcomeConstraint) -> Dict[str, Any]:
"""Convert Ax outcome constraint to a dictionary."""
return {
"__type": outcome_constraint.__class__.__name__,
"metric": outcome_constraint.metric,
"op": outcome_constraint.op,
"bound": outcome_constraint.bound,
"relative": outcome_constraint.relative,
}
[docs]def optimization_config_to_dict(
optimization_config: OptimizationConfig,
) -> Dict[str, Any]:
"""Convert Ax optimization config to a dictionary."""
return {
"__type": optimization_config.__class__.__name__,
"objective": optimization_config.objective,
"outcome_constraints": optimization_config.outcome_constraints,
"risk_measure": optimization_config.risk_measure,
}
[docs]def multi_objective_optimization_config_to_dict(
multi_objective_optimization_config: MultiObjectiveOptimizationConfig,
) -> Dict[str, Any]:
"""Convert Ax optimization config to a dictionary."""
return {
"__type": multi_objective_optimization_config.__class__.__name__,
"objective": multi_objective_optimization_config.objective,
"outcome_constraints": multi_objective_optimization_config.outcome_constraints,
"objective_thresholds": multi_objective_optimization_config.objective_thresholds, # noqa E501
"risk_measure": multi_objective_optimization_config.risk_measure,
}
[docs]def generator_run_to_dict(generator_run: GeneratorRun) -> Dict[str, Any]:
"""Convert Ax generator run to a dictionary."""
gr = generator_run
cand_metadata = gr.candidate_metadata_by_arm_signature
return {
"__type": gr.__class__.__name__,
"arms": gr.arms,
"weights": gr.weights,
"optimization_config": gr.optimization_config,
"search_space": gr.search_space,
"time_created": gr.time_created,
"model_predictions": gr.model_predictions,
"best_arm_predictions": gr.best_arm_predictions,
"generator_run_type": gr.generator_run_type,
"index": gr.index,
"fit_time": gr.fit_time,
"gen_time": gr.gen_time,
"model_key": gr._model_key,
"model_kwargs": gr._model_kwargs,
"bridge_kwargs": gr._bridge_kwargs,
"gen_metadata": gr._gen_metadata,
"model_state_after_gen": gr._model_state_after_gen,
"generation_step_index": gr._generation_step_index,
"candidate_metadata_by_arm_signature": cand_metadata,
}
[docs]def runner_to_dict(runner: Runner) -> Dict[str, Any]:
"""Convert Ax runner to a dictionary."""
properties = runner.serialize_init_args(obj=runner)
properties["__type"] = runner.__class__.__name__
return properties
[docs]def data_to_dict(data: Data) -> Dict[str, Any]:
"""Convert Ax data to a dictionary."""
properties = data.serialize_init_args(obj=data)
properties["__type"] = data.__class__.__name__
return properties
[docs]def map_data_to_dict(map_data: MapData) -> Dict[str, Any]:
"""Convert Ax map data to a dictionary."""
properties = map_data.serialize_init_args(obj=map_data)
properties["__type"] = map_data.__class__.__name__
return properties
# pyre-fixme[24]: Generic type `MapKeyInfo` expects 1 type parameter.
[docs]def map_key_info_to_dict(mki: MapKeyInfo) -> Dict[str, Any]:
"""Convert Ax map data metadata to a dictionary."""
properties = serialize_init_args(object=mki)
properties["__type"] = mki.__class__.__name__
return properties
[docs]def generation_step_to_dict(generation_step: GenerationStep) -> Dict[str, Any]:
"""Converts Ax generation step to a dictionary."""
return {
"__type": generation_step.__class__.__name__,
"model": generation_step.model,
"num_trials": generation_step.num_trials,
"min_trials_observed": generation_step.min_trials_observed,
"completion_criteria": generation_step.completion_criteria,
"max_parallelism": generation_step.max_parallelism,
"use_update": generation_step.use_update,
"enforce_num_trials": generation_step.enforce_num_trials,
"model_kwargs": _encode_callables_as_references(
generation_step.model_kwargs or {}
),
"model_gen_kwargs": _encode_callables_as_references(
generation_step.model_gen_kwargs or {}
),
"index": generation_step.index,
"should_deduplicate": generation_step.should_deduplicate,
}
[docs]def generation_strategy_to_dict(
generation_strategy: GenerationStrategy,
) -> Dict[str, Any]:
"""Converts Ax generation strategy to a dictionary."""
if generation_strategy.uses_non_registered_models:
raise ValueError( # pragma: no cover
"Generation strategies that use custom models provided through "
"callables cannot be serialized and stored."
)
return {
"__type": generation_strategy.__class__.__name__,
"db_id": generation_strategy._db_id,
"name": generation_strategy.name,
"steps": generation_strategy._steps,
"curr_index": generation_strategy._curr.index,
"generator_runs": generation_strategy._generator_runs,
"had_initialized_model": generation_strategy.model is not None,
"experiment": generation_strategy._experiment,
}
[docs]def completion_criterion_to_dict(criterion: CompletionCriterion) -> Dict[str, Any]:
"""Convert Ax CompletionCriterion to a dictionary."""
properties = criterion.serialize_init_args(obj=criterion)
properties["__type"] = criterion.__class__.__name__
return properties
[docs]def observation_features_to_dict(obs_features: ObservationFeatures) -> Dict[str, Any]:
"""Converts Ax observation features to a dictionary"""
return {
"__type": obs_features.__class__.__name__,
"parameters": obs_features.parameters,
"trial_index": obs_features.trial_index,
"start_time": obs_features.start_time,
"end_time": obs_features.end_time,
"random_split": obs_features.random_split,
"metadata": obs_features.metadata,
}
[docs]def botorch_model_to_dict(model: BoTorchModel) -> Dict[str, Any]:
"""Convert Ax model to a dictionary."""
return {
"__type": model.__class__.__name__,
"surrogate": model.surrogate,
"surrogate_options": model.surrogate_options,
"acquisition_class": model.acquisition_class,
"botorch_acqf_class": model._botorch_acqf_class,
"acquisition_options": model.acquisition_options or {},
"refit_on_update": model.refit_on_update,
"refit_on_cv": model.refit_on_cv,
"warm_start_refit": model.warm_start_refit,
}
[docs]def surrogate_to_dict(surrogate: Surrogate) -> Dict[str, Any]:
"""Convert Ax surrogate to a dictionary."""
dict_representation = {"__type": surrogate.__class__.__name__}
dict_representation.update(surrogate._serialize_attributes_as_kwargs())
return dict_representation
# pyre-fixme[2]: Parameter annotation cannot contain `Any`.
[docs]def botorch_modular_to_dict(class_type: Type[Any]) -> Dict[str, Any]:
"""Convert any class to a dictionary."""
for _class in CLASS_TO_REGISTRY:
if issubclass(class_type, _class):
registry = CLASS_TO_REGISTRY[_class]
if class_type not in registry:
raise ValueError( # pragma: no cover
f"Class `{class_type.__name__}` not in Type[{_class.__name__}] "
"registry, please add it. BoTorch object registries are "
"located in `ax/storage/botorch_modular_registry.py`."
)
return {
"__type": f"Type[{_class.__name__}]",
"index": registry[class_type],
"class": f"{_class}",
}
raise ValueError(
f"{class_type} does not have a corresponding parent class in "
"CLASS_TO_REGISTRY."
)
# pyre-fixme[2]: Parameter annotation cannot contain `Any`.
[docs]def botorch_component_to_dict(input_obj: Type[Any]) -> Dict[str, Any]:
class_type = input_obj.__class__
state_dict = input_obj.state_dict()
# Cast dict values to float to avoid errors with Tensors.
state_dict = {k: float(v) for k, v in state_dict.items()}
return {
"__type": f"{class_type.__name__}",
"index": class_type,
"class": f"{class_type}",
"state_dict": state_dict,
}
[docs]def percentile_early_stopping_strategy_to_dict(
strategy: PercentileEarlyStoppingStrategy,
) -> Dict[str, Any]:
"""Convert Ax percentile early stopping strategy to a dictionary."""
return {
"__type": strategy.__class__.__name__,
"metric_names": strategy.metric_names,
"percentile_threshold": strategy.percentile_threshold,
"min_progression": strategy.min_progression,
"min_curves": strategy.min_curves,
"trial_indices_to_ignore": strategy.trial_indices_to_ignore,
"true_objective_metric_name": strategy.true_objective_metric_name,
"seconds_between_polls": strategy.seconds_between_polls,
"normalize_progressions": strategy.normalize_progressions,
}
[docs]def threshold_early_stopping_strategy_to_dict(
strategy: ThresholdEarlyStoppingStrategy,
) -> Dict[str, Any]:
"""Convert Ax metric-threshold early stopping strategy to a dictionary."""
return {
"__type": strategy.__class__.__name__,
"metric_names": strategy.metric_names,
"metric_threshold": strategy.metric_threshold,
"min_progression": strategy.min_progression,
"trial_indices_to_ignore": strategy.trial_indices_to_ignore,
"true_objective_metric_name": strategy.true_objective_metric_name,
"normalize_progressions": strategy.normalize_progressions,
}
[docs]def logical_early_stopping_strategy_to_dict(
strategy: LogicalEarlyStoppingStrategy,
) -> Dict[str, Any]:
return {
"__type": strategy.__class__.__name__,
"left": strategy.left,
"right": strategy.right,
}
[docs]def winsorization_config_to_dict(config: WinsorizationConfig) -> Dict[str, Any]:
"""Convert Ax winsorization config to a dictionary."""
return {
"__type": config.__class__.__name__,
"lower_quantile_margin": config.lower_quantile_margin,
"upper_quantile_margin": config.upper_quantile_margin,
"lower_boundary": config.lower_boundary,
"upper_boundary": config.upper_boundary,
}
[docs]def pytorch_cnn_torchvision_benchmark_problem_to_dict(
problem: PyTorchCNNTorchvisionBenchmarkProblem,
) -> Dict[str, Any]:
# unit tests for this in benchmark suite
return { # pragma: no cover
"__type": problem.__class__.__name__,
"name": not_none(re.compile("(?<=::).*").search(problem.name)).group(),
"num_trials": problem.num_trials,
"infer_noise": problem.infer_noise,
}
[docs]def risk_measure_to_dict(
risk_measure: RiskMeasure,
) -> Dict[str, Any]:
"""Convert a RiskMeasure to a dictionary."""
return {
"__type": risk_measure.__class__.__name__,
"risk_measure": risk_measure.risk_measure,
"options": risk_measure.options,
}
