update tests

test/models/test_empirical_mean_var.py

@@ -88,7 +88,7 @@ def test_predict(self):
 
         # NOTE: Actual coordinates of test points are not used, only the indices are used.
 
-        with self.assertRaises(AssertionError):
+        with self.assertRaises(ValueError):
             self.model.predict(np.array([[-1, -1]]))
 
         X = np.array([[-1, -1, 0], [-1, -1, 1]])

test/models/test_gpytorch.py

@@ -234,6 +234,12 @@ def test_sample_from_single_posterior(self):
             samples.shape, (2, len(self.X)), "Sample shape mismatch for single posterior."
         )
 
+    def test_get_lengthscale_and_var(self):
+        """Test get_lengthscale_and_var method."""
+        lengthscales, variances = self.model.get_lengthscale_and_var()
+        self.assertGreaterEqual(np.min(variances), 0, "Negative variance.")
+        self.assertGreaterEqual(np.min(lengthscales), 0, "Negative lengthscale.")
+
 
 class TestGetGPyTorchModelListWithKnownHyperparams(unittest.TestCase):
     """Test the `get_gpytorch_modellist_w_known_hyperparams` function."""

test/test_confidence_region.py

@@ -224,7 +224,7 @@ def setUp(self):
 
     def test_update(self):
         """Test the update method."""
-        with self.assertRaises(AssertionError):
+        with self.assertRaises(ValueError):
             self.confidence_region.update(np.array([1, 1]), np.ones([2, 3]))
 
         self.confidence_region_new = copy.deepcopy(self.confidence_region)

vectoptal/confidence_region.py

@@ -89,7 +89,7 @@ def update(
         :type scale: np.ndarray
         """
         if covariance.shape[-1] != covariance.shape[-2]:
-            raise AssertionError("Covariance matrix must be square.")
+            raise ValueError("Covariance matrix must be square.")
         std = np.sqrt(np.diag(covariance.squeeze()))
 
         L = mean - std * scale

vectoptal/models/gpytorch.py

@@ -676,11 +676,11 @@ def get_lengthscale_and_var(self) -> tuple[np.ndarray, np.ndarray]:
         :return: Lengthscales and variances for each objective.
         :rtype: tuple[np.ndarray, np.ndarray]
         """
-        lengthscales = np.zeros(self.input_dim)
+        lengthscales = np.zeros((len(self.model.models), self.input_dim))
         variances = np.zeros(self.input_dim)
         for model_i, model in enumerate(self.model.models):
             cov_module = model.covar_module
-            lengthscale = cov_module.base_kernel.lengthscale.squeeze().numpy(force=True).item()
+            lengthscale = cov_module.base_kernel.lengthscale.squeeze().numpy(force=True)
             variance = cov_module.outputscale.squeeze().numpy(force=True).item()
 
             lengthscales[model_i] = lengthscale