Bump up unit test coverage (#3162)

Summary:
let's keep it up

Pull Request resolved: #3162

Reviewed By: jandylin

Differential Revision: D92409159

Pulled By: Balandat

fbshipit-source-id: e1f5aafdef62696807434ab01f9f0a3fbea7d335

Author: Balandat

test_community/acquisition/test_bayesian_active_learning.py

@@ -21,6 +21,25 @@
 
 
 class TestQStatisticalDistanceActiveLearning(BotorchTestCase):
+    def test_invalid_distance_metric(self):
+        """Test that invalid distance_metric raises ValueError."""
+        torch.manual_seed(1)
+        tkwargs = {"device": self.device, "dtype": torch.double}
+        input_dim = 2
+
+        model = get_fully_bayesian_model(
+            train_X=torch.rand(4, input_dim, **tkwargs),
+            train_Y=torch.rand(4, 1, **tkwargs),
+            num_models=3,
+            **tkwargs,
+        )
+
+        with self.assertRaises(ValueError):
+            qStatisticalDistanceActiveLearning(
+                model=model,
+                distance_metric="invalid_metric",
+            )
+
     def test_q_statistical_distance_active_learning(self):
         torch.manual_seed(1)
         tkwargs = {"device": self.device}

test_community/models/test_np_regression.py

@@ -37,6 +37,32 @@ def initialize(self):
             self.n_context,
         )
 
+    def test_default_hidden_dims(self):
+        """Test that default hidden dimensions are used when not provided."""
+        x_dim = 2
+        y_dim = 1
+        r_dim = 8
+        z_dim = 8
+        n_context = 20
+
+        # Create model without specifying hidden dimensions (use defaults)
+        model = NeuralProcessModel(
+            train_X=torch.rand(100, x_dim),
+            train_Y=torch.rand(100, y_dim),
+            r_hidden_dims=None,
+            z_hidden_dims=None,
+            decoder_hidden_dims=None,
+            x_dim=x_dim,
+            y_dim=y_dim,
+            r_dim=r_dim,
+            z_dim=z_dim,
+            n_context=n_context,
+        )
+
+        # Test that the model works with default dimensions
+        output = model(model.train_X, model.train_Y)
+        self.assertEqual(output.loc.shape, (80, y_dim))
+
     def test_r_encoder(self):
         self.initialize()
         input = torch.rand(100, self.x_dim + self.y_dim)