def gen(
self,
n: int,
bounds: List[Tuple[float, float]],
linear_constraints: Optional[Tuple[np.ndarray, np.ndarray]] = None,
fixed_features: Optional[Dict[int, float]] = None,
model_gen_options: Optional[TConfig] = None,
rounding_func: Optional[Callable[[np.ndarray], np.ndarray]] = None,
) -> Tuple[np.ndarray, np.ndarray]:
"""Generate new candidates.
Args:
n: Number of candidates to generate.
bounds: A list of (lower, upper) tuples for each column of X.
linear_constraints: A tuple of (A, b). For k linear constraints on
d-dimensional x, A is (k x d) and b is (k x 1) such that
A x <= b.
fixed_features: A map {feature_index: value} for features that
should be fixed to a particular value during generation.
rounding_func: A function that rounds an optimization result
appropriately (e.g., according to `round-trip` transformations)
but *unused here*.
Returns:
2-element tuple containing
- (n x d) array of generated points.
- Uniform weights, an n-array of ones for each point.
"""
tf_indices = tunable_feature_indices(
bounds=bounds, fixed_features=fixed_features
)
if len(tf_indices) > 0:
self.init_engine(len(tf_indices))
points, weights = super().gen(
n=n,
bounds=bounds,
linear_constraints=linear_constraints,
fixed_features=fixed_features,
model_gen_options=model_gen_options,
rounding_func=rounding_func,
)
if self.engine:
self.init_position = not_none(self.engine).num_generated
return (points, weights)
def gen(
self,
n: int,
bounds: List[Tuple[float, float]],
linear_constraints: Optional[Tuple[np.ndarray, np.ndarray]] = None,
fixed_features: Optional[Dict[int, float]] = None,
model_gen_options: Optional[TConfig] = None,
rounding_func: Optional[Callable[[np.ndarray], np.ndarray]] = None,
) -> Tuple[np.ndarray, np.ndarray]:
"""Generate new candidates.
Args:
n: Number of candidates to generate.
bounds: A list of (lower, upper) tuples for each column of X.
Defined on [0, 1]^d.
linear_constraints: A tuple of (A, b). For k linear constraints on
d-dimensional x, A is (k x d) and b is (k x 1) such that
A x <= b.
fixed_features: A map {feature_index: value} for features that
should be fixed to a particular value during generation.
model_gen_options: A config dictionary that is passed along to the
model.
rounding_func: A function that rounds an optimization result
appropriately (e.g., according to `round-trip` transformations).
Returns:
2-element tuple containing
- (n x d) array of generated points.
- Uniform weights, an n-array of ones for each point.
"""
tf_indices = tunable_feature_indices(bounds=bounds,
fixed_features=fixed_features)
if fixed_features:
fixed_feature_indices = np.array(list(fixed_features.keys()))
else:
fixed_feature_indices = np.array([])
validate_bounds(bounds=bounds,
fixed_feature_indices=fixed_feature_indices)
attempted_draws = 0
max_draws = None
if model_gen_options:
max_draws = model_gen_options.get("max_rs_draws")
if max_draws is not None:
max_draws: int = int(max_draws)
# Always rejection sample, but this only rejects if there are
# constraints or actual duplicates and deduplicate is specified.
points, attempted_draws = rejection_sample(
gen_unconstrained=self._gen_unconstrained,
n=n,
d=len(bounds),
tunable_feature_indices=tf_indices,
linear_constraints=linear_constraints,
deduplicate=self.deduplicate,
max_draws=max_draws,
fixed_features=fixed_features,
rounding_func=rounding_func,
existing_points=self.generated_points,
)
# pyre-fixme[16]: `RandomModel` has no attribute `attempted_draws`.
self.attempted_draws = attempted_draws
if self.deduplicate:
if self.generated_points is None:
self.generated_points = points
else:
# pyre-fixme[6]: Expected `Collection[ndarray]` for 1st param but
# got `List[Optional[ndarray]]`.
self.generated_points = np.vstack(
[self.generated_points, points])
return (points, np.ones(len(points)))
def gen(
self,
n: int,
bounds: List[Tuple[float, float]],
linear_constraints: Optional[Tuple[np.ndarray, np.ndarray]] = None,
fixed_features: Optional[Dict[int, float]] = None,
model_gen_options: Optional[TConfig] = None,
rounding_func: Optional[Callable[[np.ndarray], np.ndarray]] = None,
) -> Tuple[np.ndarray, np.ndarray]:
"""Generate new candidates.
Args:
n: Number of candidates to generate.
bounds: A list of (lower, upper) tuples for each column of X.
Defined on [0, 1]^d.
linear_constraints: A tuple of (A, b). For k linear constraints on
d-dimensional x, A is (k x d) and b is (k x 1) such that
A x <= b.
fixed_features: A map {feature_index: value} for features that
should be fixed to a particular value during generation.
model_gen_options: A config dictionary that is passed along to the
model.
rounding_func: A function that rounds an optimization result
appropriately (e.g., according to `round-trip` transformations).
Returns:
2-element tuple containing
- (n x d) array of generated points.
- Uniform weights, an n-array of ones for each point.
"""
tf_indices = tunable_feature_indices(
bounds=bounds, fixed_features=fixed_features
)
if fixed_features:
fixed_feature_indices = np.array(list(fixed_features.keys()))
else:
fixed_feature_indices = np.array([])
validate_bounds(bounds=bounds, fixed_feature_indices=fixed_feature_indices)
attempted_draws = 0
max_draws = None
if model_gen_options:
max_draws = model_gen_options.get("max_rs_draws")
if max_draws is not None:
# pyre-fixme[6]: Expected `Union[bytes, str, typing.SupportsInt]`
# for 1st param but got
# `Union[botorch.acquisition.acquisition.AcquisitionFunction, float,
# int, str]`.
# pyre-fixme[35]: Target cannot be annotated.
max_draws: int = int(max_draws)
try:
# Always rejection sample, but this only rejects if there are
# constraints or actual duplicates and deduplicate is specified.
# If rejection sampling fails, fall back to polytope sampling
points, attempted_draws = rejection_sample(
gen_unconstrained=self._gen_unconstrained,
n=n,
d=len(bounds),
tunable_feature_indices=tf_indices,
linear_constraints=linear_constraints,
deduplicate=self.deduplicate,
max_draws=max_draws,
fixed_features=fixed_features,
rounding_func=rounding_func,
existing_points=self.generated_points,
)
except SearchSpaceExhausted as e:
if self.fallback_to_sample_polytope:
logger.info(
"Rejection sampling exceeded specified maximum draws."
"Falling back on polytope sampler"
)
# If rejection sampling fails, try polytope sampler.
polytope_sampler = HitAndRunPolytopeSampler(
inequality_constraints=self._convert_inequality_constraints(
linear_constraints
),
equality_constraints=self._convert_equality_constraints(
len(bounds), fixed_features
),
interior_point=self._get_last_point(),
bounds=self._convert_bounds(bounds),
)
points = polytope_sampler.draw(n).numpy()
else:
raise e
# pyre-fixme[16]: `RandomModel` has no attribute `attempted_draws`.
self.attempted_draws = attempted_draws
if self.deduplicate:
if self.generated_points is None:
self.generated_points = points
else:
self.generated_points = np.vstack([self.generated_points, points])
return points, np.ones(len(points))
