def gen(
self,
n: int,
search_space: Optional[SearchSpace] = None,
optimization_config: Optional[OptimizationConfig] = None,
pending_observations: Optional[Dict[str,
List[ObservationFeatures]]] = None,
fixed_features: Optional[ObservationFeatures] = None,
model_gen_options: Optional[TConfig] = None,
) -> GeneratorRun:
"""
Args:
n: Number of points to generate
search_space: Search space
optimization_config: Optimization config
pending_observations: A map from metric name to pending
observations for that metric.
fixed_features: An ObservationFeatures object containing any
features that should be fixed at specified values during
generation.
model_gen_options: A config dictionary that is passed along to the
model.
"""
t_gen_start = time.time()
if pending_observations is None:
pending_observations = {}
if fixed_features is None:
fixed_features = ObservationFeatures({})
# Get modifiable versions
if search_space is None:
search_space = self._model_space
search_space = search_space.clone()
if optimization_config is None:
optimization_config = (self._optimization_config.clone()
if self._optimization_config is not None
else None)
else:
optimization_config = optimization_config.clone()
# TODO(T34225037): replace deepcopy with native clone() in Lazarus
pending_observations = deepcopy(pending_observations)
fixed_features = deepcopy(fixed_features)
# Transform
for t in self.transforms.values():
search_space = t.transform_search_space(search_space)
if optimization_config is not None:
optimization_config = t.transform_optimization_config(
optimization_config=optimization_config,
modelbridge=self,
fixed_features=fixed_features,
)
for metric, po in pending_observations.items():
pending_observations[
metric] = t.transform_observation_features(po)
fixed_features = t.transform_observation_features([fixed_features
])[0]
# Apply terminal transform and gen
observation_features, weights, best_obsf = self._gen(
n=n,
search_space=search_space,
optimization_config=optimization_config,
pending_observations=pending_observations,
fixed_features=fixed_features,
model_gen_options=model_gen_options,
)
# Apply reverse transforms
# pyre-fixme[6]: Expected `Sequence[_T]` for 1st param but got `ValuesView[Tr...
for t in reversed(self.transforms.values()): # noqa T484
observation_features = t.untransform_observation_features(
observation_features)
if best_obsf is not None:
best_obsf = t.untransform_observation_features([best_obsf])[0]
best_point_predictions = None
try:
model_predictions = self.predict(observation_features)
if best_obsf is not None:
best_point_predictions = extract_arm_predictions(
model_predictions=self.predict([best_obsf]), arm_idx=0)
except NotImplementedError: # pragma: no cover
model_predictions = None
best_arm = (None if best_obsf is None else gen_arms(
observation_features=[best_obsf],
arms_by_signature=self._arms_by_signature,
)[0])
gr = GeneratorRun(
arms=gen_arms(
observation_features=observation_features,
arms_by_signature=self._arms_by_signature,
),
weights=weights,
optimization_config=optimization_config,
search_space=search_space,
model_predictions=model_predictions,
best_arm_predictions=None if best_arm is None else
(best_arm, best_point_predictions),
fit_time=self.fit_time_since_gen,
gen_time=time.time() - t_gen_start,
)
self.fit_time_since_gen = 0.0
return gr
def gen(
self,
n: int,
search_space: Optional[SearchSpace] = None,
optimization_config: Optional[OptimizationConfig] = None,
pending_observations: Optional[Dict[str, List[ObservationFeatures]]] = None,
fixed_features: Optional[ObservationFeatures] = None,
model_gen_options: Optional[TConfig] = None,
) -> GeneratorRun:
"""
Args:
n: Number of points to generate
search_space: Search space
optimization_config: Optimization config
pending_observations: A map from metric name to pending
observations for that metric.
fixed_features: An ObservationFeatures object containing any
features that should be fixed at specified values during
generation.
model_gen_options: A config dictionary that is passed along to the
model.
"""
t_gen_start = time.time()
# Get modifiable versions
if search_space is None:
search_space = self._model_space
orig_search_space = search_space
search_space = search_space.clone()
base_gen_args = self._get_transformed_gen_args(
search_space=search_space,
optimization_config=optimization_config,
pending_observations=pending_observations,
fixed_features=fixed_features,
)
# Apply terminal transform and gen
observation_features, weights, best_obsf, gen_metadata = self._gen(
n=n,
search_space=base_gen_args.search_space,
optimization_config=base_gen_args.optimization_config,
pending_observations=base_gen_args.pending_observations,
fixed_features=base_gen_args.fixed_features,
model_gen_options=model_gen_options,
)
# Apply reverse transforms
for t in reversed(self.transforms.values()): # noqa T484
observation_features = t.untransform_observation_features(
observation_features
)
if best_obsf is not None:
best_obsf = t.untransform_observation_features([best_obsf])[0]
# Clamp the untransformed data to the original search space if
# we don't fit/gen OOD points
if not self._fit_out_of_design:
observation_features = clamp_observation_features(
observation_features, orig_search_space
)
if best_obsf is not None:
best_obsf = clamp_observation_features([best_obsf], orig_search_space)[
0
]
best_point_predictions = None
try:
model_predictions = self.predict(observation_features)
if best_obsf is not None:
best_point_predictions = extract_arm_predictions(
model_predictions=self.predict([best_obsf]), arm_idx=0
)
except NotImplementedError: # pragma: no cover
model_predictions = None
if best_obsf is None:
best_arm = None
else:
best_arms, _ = gen_arms(
observation_features=[best_obsf],
arms_by_signature=self._arms_by_signature,
)
best_arm = best_arms[0]
arms, candidate_metadata = gen_arms(
observation_features=observation_features,
arms_by_signature=self._arms_by_signature,
)
# If experiment has immutable search space and metrics, no need to
# save them on generator runs.
immutable = getattr(
self, "_experiment_has_immutable_search_space_and_opt_config", False
)
optimization_config = None if immutable else base_gen_args.optimization_config
gr = GeneratorRun(
arms=arms,
weights=weights,
optimization_config=optimization_config,
search_space=None if immutable else base_gen_args.search_space,
model_predictions=model_predictions,
best_arm_predictions=None
if best_arm is None
else (best_arm, best_point_predictions),
fit_time=self.fit_time_since_gen,
gen_time=time.time() - t_gen_start,
model_key=self._model_key,
model_kwargs=self._model_kwargs,
bridge_kwargs=self._bridge_kwargs,
gen_metadata=gen_metadata,
model_state_after_gen=self._get_serialized_model_state(),
candidate_metadata_by_arm_signature=candidate_metadata,
)
self.fit_time_since_gen = 0.0
return gr
def gen(
self,
n: int,
search_space: Optional[SearchSpace] = None,
optimization_config: Optional[OptimizationConfig] = None,
pending_observations: Optional[Dict[str,
List[ObservationFeatures]]] = None,
fixed_features: Optional[ObservationFeatures] = None,
model_gen_options: Optional[TConfig] = None,
) -> GeneratorRun:
"""
Args:
n: Number of points to generate
search_space: Search space
optimization_config: Optimization config
pending_observations: A map from metric name to pending
observations for that metric.
fixed_features: An ObservationFeatures object containing any
features that should be fixed at specified values during
generation.
model_gen_options: A config dictionary that is passed along to the
model.
"""
t_gen_start = time.time()
if pending_observations is None:
pending_observations = {}
if fixed_features is None:
fixed_features = ObservationFeatures({})
# Get modifiable versions
if search_space is None:
search_space = self._model_space
search_space = search_space.clone()
if optimization_config is None:
optimization_config = (
# pyre-fixme[16]: `Optional` has no attribute `clone`.
self._optimization_config.clone()
if self._optimization_config is not None else None)
else:
optimization_config = optimization_config.clone()
# TODO(T34225037): replace deepcopy with native clone() in Ax
pending_observations = deepcopy(pending_observations)
fixed_features = deepcopy(fixed_features)
# Transform
for t in self.transforms.values():
search_space = t.transform_search_space(search_space)
if optimization_config is not None:
optimization_config = t.transform_optimization_config(
optimization_config=optimization_config,
modelbridge=self,
fixed_features=fixed_features,
)
for metric, po in pending_observations.items():
pending_observations[
metric] = t.transform_observation_features(po)
fixed_features = t.transform_observation_features([fixed_features
])[0]
# Apply terminal transform and gen
observation_features, weights, best_obsf, gen_metadata = self._gen(
n=n,
search_space=search_space,
optimization_config=optimization_config,
pending_observations=pending_observations,
fixed_features=fixed_features,
model_gen_options=model_gen_options,
)
# Apply reverse transforms
for t in reversed(self.transforms.values()): # noqa T484
observation_features = t.untransform_observation_features(
observation_features)
if best_obsf is not None:
best_obsf = t.untransform_observation_features([best_obsf])[0]
best_point_predictions = None
try:
model_predictions = self.predict(observation_features)
if best_obsf is not None:
best_point_predictions = extract_arm_predictions(
model_predictions=self.predict([best_obsf]), arm_idx=0)
except NotImplementedError: # pragma: no cover
model_predictions = None
if best_obsf is None:
best_arm = None
else:
best_arms, _ = gen_arms(
observation_features=[best_obsf],
arms_by_signature=self._arms_by_signature,
)
best_arm = best_arms[0]
arms, candidate_metadata = gen_arms(
observation_features=observation_features,
arms_by_signature=self._arms_by_signature,
)
# If experiment has immutable search space and metrics, no need to
# save them on generator runs.
immutable = getattr(
self, "_experiment_has_immutable_search_space_and_opt_config",
False)
gr = GeneratorRun(
arms=arms,
weights=weights,
optimization_config=None if immutable else optimization_config,
search_space=None if immutable else search_space,
model_predictions=model_predictions,
best_arm_predictions=None if best_arm is None else
(best_arm, best_point_predictions),
fit_time=self.fit_time_since_gen,
gen_time=time.time() - t_gen_start,
model_key=self._model_key,
model_kwargs=self._model_kwargs,
bridge_kwargs=self._bridge_kwargs,
gen_metadata=gen_metadata,
model_state_after_gen=self._get_serialized_model_state(),
candidate_metadata_by_arm_signature=candidate_metadata,
)
self.fit_time_since_gen = 0.0
return gr
def model_best_point(
self,
search_space: Optional[SearchSpace] = None,
optimization_config: Optional[OptimizationConfig] = None,
pending_observations: Optional[Dict[str, List[ObservationFeatures]]] = None,
fixed_features: Optional[ObservationFeatures] = None,
model_gen_options: Optional[TConfig] = None,
) -> Optional[Tuple[Arm, Optional[TModelPredictArm]]]:
# Get modifiable versions
if search_space is None:
search_space = self._model_space
search_space = search_space.clone()
base_gen_args = self._get_transformed_gen_args(
search_space=search_space,
optimization_config=optimization_config,
pending_observations=pending_observations,
fixed_features=fixed_features,
)
array_model_gen_args = self._get_transformed_model_gen_args(
search_space=base_gen_args.search_space,
pending_observations=base_gen_args.pending_observations,
fixed_features=base_gen_args.fixed_features,
model_gen_options=None,
optimization_config=base_gen_args.optimization_config,
)
search_space_digest = array_model_gen_args.search_space_digest
xbest = self._model_best_point(
bounds=search_space_digest.bounds,
objective_weights=array_model_gen_args.objective_weights,
outcome_constraints=array_model_gen_args.outcome_constraints,
linear_constraints=array_model_gen_args.linear_constraints,
fixed_features=array_model_gen_args.fixed_features,
model_gen_options=model_gen_options,
target_fidelities=search_space_digest.target_fidelities,
)
if xbest is None:
return None
best_obsf = ObservationFeatures(
parameters={p: float(xbest[i]) for i, p in enumerate(self.parameters)}
)
for t in reversed(self.transforms.values()): # noqa T484
best_obsf = t.untransform_observation_features([best_obsf])[0]
best_point_predictions = extract_arm_predictions(
model_predictions=self.predict([best_obsf]), arm_idx=0
)
best_arms, _ = gen_arms(
observation_features=[best_obsf],
arms_by_signature=self._arms_by_signature,
)
best_arm = best_arms[0]
return best_arm, best_point_predictions
