diff --git a/aepsych/models/base.py b/aepsych/models/base.py
index 7fbedf9b1..d5d564253 100644
--- a/aepsych/models/base.py
+++ b/aepsych/models/base.py
@@ -116,7 +116,7 @@ class AEPsychMixin(GPyTorchModel):
 
     extremum_solver = "Nelder-Mead"
     outcome_types: List[str] = []
-    train_inputs: Optional[Tuple[torch.Tensor]]
+    train_inputs: Optional[Tuple[torch.Tensor, ...]]
     train_targets: Optional[torch.Tensor]
 
     @property
@@ -393,7 +393,7 @@ def p_below_threshold(
 
 
 class AEPsychModelDeviceMixin(AEPsychMixin):
-    _train_inputs: Optional[Tuple[torch.Tensor]]
+    _train_inputs: Optional[Tuple[torch.Tensor, ...]]
     _train_targets: Optional[torch.Tensor]
 
     def set_train_data(self, inputs=None, targets=None, strict=False):
@@ -423,13 +423,17 @@ def device(self) -> torch.device:
             return torch.device("cpu")
 
     @property
-    def train_inputs(self) -> Optional[Tuple[torch.Tensor]]:
+    def train_inputs(self) -> Optional[Tuple[torch.Tensor, ...]]:
         if self._train_inputs is None:
             return None
 
         # makes sure the tensors are on the right device, move in place
+        _train_inputs = []
         for input in self._train_inputs:
-            input.to(self.device)
+            _train_inputs.append(input.to(self.device))
+
+        _tuple_inputs: Tuple[torch.Tensor, ...] = tuple(_train_inputs)
+        self._train_inputs = _tuple_inputs
 
         return self._train_inputs
 
diff --git a/tests_gpu/models/test_derivative_gp.py b/tests_gpu/models/test_derivative_gp.py
new file mode 100644
index 000000000..200ef62eb
--- /dev/null
+++ b/tests_gpu/models/test_derivative_gp.py
@@ -0,0 +1,39 @@
+#!/usr/bin/env python3
+# Copyright (c) Facebook, Inc. and its affiliates.
+# All rights reserved.
+
+# This source code is licensed under the license found in the
+# LICENSE file in the root directory of this source tree.
+
+import torch
+from aepsych import Config, SequentialStrategy
+from aepsych.models.derivative_gp import MixedDerivativeVariationalGP
+from botorch.fit import fit_gpytorch_mll
+from botorch.utils.testing import BotorchTestCase
+from gpytorch.likelihoods import BernoulliLikelihood
+from gpytorch.mlls.variational_elbo import VariationalELBO
+
+
+class TestDerivativeGP(BotorchTestCase):
+    def test_MixedDerivativeVariationalGP_gpu(self):
+        train_x = torch.cat(
+            (torch.tensor([1.0, 2.0, 3.0, 4.0]).unsqueeze(1), torch.zeros(4, 1)), dim=1
+        )
+        train_y = torch.tensor([1.0, 2.0, 3.0, 4.0])
+        m = MixedDerivativeVariationalGP(
+            train_x=train_x,
+            train_y=train_y,
+            inducing_points=train_x,
+            fixed_prior_mean=0.5,
+        ).cuda()
+
+        self.assertEqual(m.mean_module.constant.item(), 0.5)
+        self.assertEqual(
+            m.covar_module.base_kernel.raw_lengthscale.shape, torch.Size([1, 1])
+        )
+        mll = VariationalELBO(
+            likelihood=BernoulliLikelihood(), model=m, num_data=train_y.numel()
+        ).cuda()
+        mll = fit_gpytorch_mll(mll)
+        test_x = torch.tensor([[1.0, 0], [3.0, 1.0]]).cuda()
+        m(test_x)