diff --git a/test/test_mps.py b/test/test_mps.py
index 0f4b2ea..05b42c7 100644
--- a/test/test_mps.py
+++ b/test/test_mps.py
@@ -4545,9 +4545,9 @@
             )
 
     def test_upsample_nearest2d(self):
-        def helper(N, C, H, W):
+        def helper(N, C, H, W, memory_format):
             inputCPU = torch.arange(N * C * H * W, device='cpu', dtype=torch.float,
-                                    requires_grad=True).reshape(N, C, H, W)
+                                    requires_grad=True).reshape(N, C, H, W).to(memory_format=memory_format)
             inputCPU.retain_grad()
             inputMPS = inputCPU.detach().to('mps').requires_grad_()
 
@@ -4573,8 +4573,9 @@
 
                     self.assertEqual(inputCPU.grad, inputMPS.grad)
 
-        helper(1, 1, 4, 4)
-        helper(7, 5, 3, 2)
+        for memory_format in [torch.channels_last, torch.contiguous_format]:
+            helper(1, 1, 4, 4, memory_format=memory_format)
+            helper(7, 5, 3, 2, memory_format=memory_format)
 
     def test_upsample_bilinear2d(self):
         def helper(N, C, H, W):