Merge 7ce1389 into 63dd0cd

Author: sdaulton

botorch/acquisition/fixed_feature.py

@@ -16,11 +16,11 @@
 
 import torch
 from botorch.acquisition.acquisition import AcquisitionFunction
+from botorch.acquisition.wrapper import AbstractAcquisitionFunctionWrapper
 from torch import Tensor
-from torch.nn import Module
 
 
-class FixedFeatureAcquisitionFunction(AcquisitionFunction):
+class FixedFeatureAcquisitionFunction(AbstractAcquisitionFunctionWrapper):
     """A wrapper around AquisitionFunctions to fix a subset of features.
 
     Example:
@@ -56,8 +56,7 @@ def __init__(
                 combination of `Tensor`s and numbers which can be broadcasted
                 to form a tensor with trailing dimension size of `d_f`.
         """
-        Module.__init__(self)
-        self.acq_func = acq_function
+        AbstractAcquisitionFunctionWrapper.__init__(self, acq_function=acq_function)
         dtype = torch.float
         device = torch.device("cpu")
         self.d = d
@@ -126,24 +125,13 @@ def forward(self, X: Tensor):
         X_full = self._construct_X_full(X)
         return self.acq_func(X_full)
 
-    @property
-    def X_pending(self):
-        r"""Return the `X_pending` of the base acquisition function."""
-        try:
-            return self.acq_func.X_pending
-        except (ValueError, AttributeError):
-            raise ValueError(
-                f"Base acquisition function {type(self.acq_func).__name__} "
-                "does not have an `X_pending` attribute."
-            )
-
-    @X_pending.setter
-    def X_pending(self, X_pending: Optional[Tensor]):
+    def set_X_pending(self, X_pending: Optional[Tensor]):
         r"""Sets the `X_pending` of the base acquisition function."""
         if X_pending is not None:
-            self.acq_func.X_pending = self._construct_X_full(X_pending)
+            full_X_pending = self._construct_X_full(X_pending)
         else:
-            self.acq_func.X_pending = X_pending
+            full_X_pending = None
+        self.acq_func.set_X_pending(full_X_pending)
 
     def _construct_X_full(self, X: Tensor) -> Tensor:
         r"""Constructs the full input for the base acquisition function.

botorch/acquisition/penalized.py

@@ -15,9 +15,8 @@
 
 import torch
 from botorch.acquisition.acquisition import AcquisitionFunction
-from botorch.acquisition.analytic import AnalyticAcquisitionFunction
 from botorch.acquisition.objective import GenericMCObjective
-from botorch.exceptions import UnsupportedError
+from botorch.acquisition.wrapper import AbstractAcquisitionFunctionWrapper
 from torch import Tensor
 
 
@@ -139,7 +138,7 @@ def forward(self, X: Tensor) -> Tensor:
         return regularization_term
 
 
-class PenalizedAcquisitionFunction(AcquisitionFunction):
+class PenalizedAcquisitionFunction(AbstractAcquisitionFunctionWrapper):
     r"""Single-outcome acquisition function regularized by the given penalty.
 
     The usage is similar to:
@@ -161,29 +160,16 @@ def __init__(
             penalty_func: The regularization function.
             regularization_parameter: Regularization parameter used in optimization.
         """
-        super().__init__(model=raw_acqf.model)
-        self.raw_acqf = raw_acqf
+        AcquisitionFunction.__init__(self, model=raw_acqf.model)
+        AbstractAcquisitionFunctionWrapper.__init__(self, acq_function=raw_acqf)
         self.penalty_func = penalty_func
         self.regularization_parameter = regularization_parameter
 
     def forward(self, X: Tensor) -> Tensor:
-        raw_value = self.raw_acqf(X=X)
+        raw_value = self.acq_func(X=X)
         penalty_term = self.penalty_func(X)
         return raw_value - self.regularization_parameter * penalty_term
 
-    @property
-    def X_pending(self) -> Optional[Tensor]:
-        return self.raw_acqf.X_pending
-
-    def set_X_pending(self, X_pending: Optional[Tensor] = None) -> None:
-        if not isinstance(self.raw_acqf, AnalyticAcquisitionFunction):
-            self.raw_acqf.set_X_pending(X_pending=X_pending)
-        else:
-            raise UnsupportedError(
-                "The raw acquisition function is Analytic and does not account "
-                "for X_pending yet."
-            )
-
 
 def group_lasso_regularizer(X: Tensor, groups: List[List[int]]) -> Tensor:
     r"""Computes the group lasso regularization function for the given point.