diff --git a/carn/sample.py b/carn/sample.py
index a6a94ac..748177c 100644
--- a/carn/sample.py
+++ b/carn/sample.py
@@ -16,7 +16,6 @@ def parse_args():
     parser.add_argument("--model", type=str)
     parser.add_argument("--ckpt_path", type=str)
     parser.add_argument("--group", type=int, default=1)
-    parser.add_argument("--reduce_upsample", action="store_true", default=False)
     parser.add_argument("--sample_dir", type=str)
     parser.add_argument("--test_data_dir", type=str, default="dataset/Urban100")
     parser.add_argument("--cuda", action="store_true")
@@ -33,7 +32,7 @@ def save_image(tensor, filename):
     im.save(filename)
 
 
-def sample(net, dataset, cfg):
+def sample(net, device, dataset, cfg):
     scale = cfg.scale
     for step, (hr, lr, name) in enumerate(dataset):
         if "DIV2K" in dataset.name:
@@ -42,19 +41,19 @@ def sample(net, dataset, cfg):
             h_half, w_half = int(h/2), int(w/2)
             h_chop, w_chop = h_half + cfg.shave, w_half + cfg.shave
 
-            lr_patch = torch.FloatTensor(4, 3, h_chop, w_chop)
+            lr_patch = torch.tensor((4, 3, h_chop, w_chop), dtype=torch.float)
             lr_patch[0].copy_(lr[:, 0:h_chop, 0:w_chop])
             lr_patch[1].copy_(lr[:, 0:h_chop, w-w_chop:w])
             lr_patch[2].copy_(lr[:, h-h_chop:h, 0:w_chop])
             lr_patch[3].copy_(lr[:, h-h_chop:h, w-w_chop:w])
-            lr_patch = Variable(lr_patch, volatile=True).cuda()
+            lr_patch = lr_patch.to(device)
             
-            sr = net(lr_patch, cfg.scale).data
+            sr = net(lr_patch, cfg.scale).detach()
             
             h, h_half, h_chop = h*scale, h_half*scale, h_chop*scale
             w, w_half, w_chop = w*scale, w_half*scale, w_chop*scale
 
-            result = torch.FloatTensor(3, h, w).cuda()
+            result = torch.tensor((3, h, w), dtype=torch.float).to(device)
             result[:, 0:h_half, 0:w_half].copy_(sr[0, :, 0:h_half, 0:w_half])
             result[:, 0:h_half, w_half:w].copy_(sr[1, :, 0:h_half, w_chop-w+w_half:w_chop])
             result[:, h_half:h, 0:w_half].copy_(sr[2, :, h_chop-h+h_half:h_chop, 0:w_half])
@@ -63,9 +62,9 @@ def sample(net, dataset, cfg):
             t2 = time.time()
         else:
             t1 = time.time()
-            lr = Variable(lr.unsqueeze(0), volatile=True).cuda()
-            sr = net(lr, cfg.scale).data[0]
-            lr = lr.data[0]
+            lr = lr.unsqueeze(0).to(device)
+            sr = net(lr, cfg.scale).detach().squeeze(0)
+            lr = lr.squeeze(0)
             t2 = time.time()
         
         model_name = cfg.ckpt_path.split(".")[0].split("/")[-1]
@@ -80,17 +79,14 @@ def sample(net, dataset, cfg):
                               "x{}".format(cfg.scale),
                               "HR")
         
-        if not os.path.exists(sr_dir):
-            os.makedirs(sr_dir)
-            
-        if not os.path.exists(hr_dir):
-            os.makedirs(hr_dir)
+        os.makedirs(sr_dir, exist_ok=True)
+        os.makedirs(hr_dir, exist_ok=True)
 
         sr_im_path = os.path.join(sr_dir, "{}".format(name.replace("HR", "SR")))
         hr_im_path = os.path.join(hr_dir, "{}".format(name))
 
         save_image(sr, sr_im_path)
-        # save_image(hr, hr_im_path)
+        save_image(hr, hr_im_path)
         print("Saved {} ({}x{} -> {}x{}, {:.3f}s)"
             .format(sr_im_path, lr.shape[1], lr.shape[2], sr.shape[1], sr.shape[2], t2-t1))
 
@@ -98,23 +94,23 @@ def sample(net, dataset, cfg):
 def main(cfg):
     module = importlib.import_module("model.{}".format(cfg.model))
     net = module.Net(multi_scale=True, 
-                     group=cfg.group, 
-                     reduce_upsample=cfg.reduce_upsample)
+                     group=cfg.group)
     print(json.dumps(vars(cfg), indent=4, sort_keys=True))
 
     state_dict = torch.load(cfg.ckpt_path)
     new_state_dict = OrderedDict()
     for k, v in state_dict.items():
-        print(k)
         name = k
         # name = k[7:] # remove "module."
         new_state_dict[name] = v
 
     net.load_state_dict(new_state_dict)
-    net.cuda()
+    
+    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+    net = net.to(device)
     
     dataset = TestDataset(cfg.test_data_dir, cfg.scale)
-    sample(net, dataset, cfg)
+    sample(net, device, dataset, cfg)
  
 
 if __name__ == "__main__":
diff --git a/checkpoint/carn.pth b/checkpoint/carn.pth
index 2ca3dc9..2b9fb1e 100644
Binary files a/checkpoint/carn.pth and b/checkpoint/carn.pth differ
diff --git a/checkpoint/carn_m.pth b/checkpoint/carn_m.pth
index 2c699d2..8d53d10 100644
Binary files a/checkpoint/carn_m.pth and b/checkpoint/carn_m.pth differ