diff --git a/mnist/main.py b/mnist/main.py
index 0a733cd48b..04dd254117 100644
--- a/mnist/main.py
+++ b/mnist/main.py
@@ -100,12 +100,14 @@ def main():
 
     device = torch.device("cuda" if use_cuda else "cpu")
 
-    kwargs = {'batch_size': args.batch_size}
+    train_kwargs = {'batch_size': args.batch_size}
+    test_kwargs = {'batch_size': args.test_batch_size}
     if use_cuda:
-        kwargs.update({'num_workers': 1,
+        cuda_kwargs = {'num_workers': 1,
                        'pin_memory': True,
-                       'shuffle': True},
-                     )
+                       'shuffle': True}
+        train_kwargs.update(cuda_kwargs)
+        test_kwargs.update(cuda_kwargs)
 
     transform=transforms.Compose([
         transforms.ToTensor(),
@@ -115,8 +117,8 @@ def main():
                        transform=transform)
     dataset2 = datasets.MNIST('../data', train=False,
                        transform=transform)
-    train_loader = torch.utils.data.DataLoader(dataset1,**kwargs)
-    test_loader = torch.utils.data.DataLoader(dataset2, **kwargs)
+    train_loader = torch.utils.data.DataLoader(dataset1,**train_kwargs)
+    test_loader = torch.utils.data.DataLoader(dataset2, **test_kwargs)
 
     model = Net().to(device)
     optimizer = optim.Adadelta(model.parameters(), lr=args.lr)