Make "device" attribute of contextual models a property instead (#1611)

Summary:
## Motivation

As of cornellius-gp/gpytorch#2234, the parent class of BoTorch kernels now has a property "device." This means that if a subclass tries to set `self.device`, it will error. This is why the BoTorch CI is currently breaking: https://github.com/pytorch/botorch/actions/runs/3841992968/jobs/6542850176

Pull Request resolved: #1611

Test Plan: Tests should pass

Reviewed By: saitcakmak, Balandat

Differential Revision: D42354199

Pulled By: esantorella

fbshipit-source-id: c53e5b508dd75f4116870cd30ab90d11cd3eb573

botorch/models/kernels/contextual_lcea.py

@@ -61,7 +61,7 @@ def __init__(
         self.decomposition = decomposition
         self.batch_shape = batch_shape
         self.train_embedding = train_embedding
-        self.device = device
+        self._device = device
 
         num_param = len(next(iter(decomposition.values())))
         self.context_list = list(decomposition.keys())
@@ -128,6 +128,10 @@ def __init__(
         )
         self.register_constraint("raw_outputscale_list", Positive())
 
+    @property
+    def device(self) -> Optional[torch.device]:
+        return self._device
+
     @property
     def outputscale_list(self) -> Tensor:
         return self.raw_outputscale_list_constraint.transform(self.raw_outputscale_list)

botorch/models/kernels/contextual_sac.py

@@ -57,7 +57,7 @@ def __init__(
 
         super().__init__(batch_shape=batch_shape)
         self.decomposition = decomposition
-        self.device = device
+        self._device = device
 
         num_param = len(next(iter(decomposition.values())))
         for active_parameters in decomposition.values():
@@ -86,6 +86,10 @@ def __init__(
             )
         self.kernel_dict = ModuleDict(self.kernel_dict)
 
+    @property
+    def device(self) -> Optional[torch.device]:
+        return self._device
+
     def forward(
         self,
         x1: Tensor,