diff --git a/eval/verification.py b/eval/verification.py
index 2b846c6..367062e 100644
--- a/eval/verification.py
+++ b/eval/verification.py
@@ -197,7 +197,8 @@ def evaluate(embeddings, actual_issame, nrof_folds=10, pca=0):
     return tpr, fpr, accuracy, val, val_std, far
 
 @torch.no_grad()
-def load_bin(path, image_size):
+def load_bin(path, image_size):\
+    device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
     try:
         with open(path, 'rb') as f:
             bins, issame_list = pickle.load(f)  # py2
@@ -217,7 +218,7 @@ def load_bin(path, image_size):
         for flip in [0, 1]:
             if flip == 1:
                 img = mx.ndarray.flip(data=img, axis=2)
-            data_list[flip][idx][:] = torch.from_numpy(img.asnumpy())
+            data_list[flip][idx][:] = torch.from_numpy(img.asnumpy()).float().to(device)
         if idx % 1000 == 0:
             print('loading bin', idx)
     print(data_list[0].shape)
@@ -225,6 +226,7 @@ def load_bin(path, image_size):
 
 @torch.no_grad()
 def test(data_set, backbone, batch_size, nfolds=10):
+    device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
     print('testing verification..')
     data_list = data_set[0]
     issame_list = data_set[1]
@@ -240,7 +242,7 @@ def test(data_set, backbone, batch_size, nfolds=10):
             _data = data[bb - batch_size: bb]
             time0 = datetime.datetime.now()
             img = ((_data / 255) - 0.5) / 0.5
-            net_out: torch.Tensor = backbone(img)
+            net_out: torch.cuda.Tensor = backbone(img.float().to(device))
             _embeddings = net_out.detach().cpu().numpy()
             time_now = datetime.datetime.now()
             diff = time_now - time0