diff --git a/implementations/gan/gan.py b/implementations/gan/gan.py
index d6f1d935..75286cb2 100644
--- a/implementations/gan/gan.py
+++ b/implementations/gan/gan.py
@@ -8,7 +8,6 @@
 
 from torch.utils.data import DataLoader
 from torchvision import datasets
-from torch.autograd import Variable
 
 import torch.nn as nn
 import torch.nn.functional as F
@@ -122,11 +121,11 @@ def forward(self, img):
     for i, (imgs, _) in enumerate(dataloader):
 
         # Adversarial ground truths
-        valid = Variable(Tensor(imgs.size(0), 1).fill_(1.0), requires_grad=False)
-        fake = Variable(Tensor(imgs.size(0), 1).fill_(0.0), requires_grad=False)
+        valid = torch.Tensor(imgs.size(0), 1).fill_(1.0)
+        fake = torch.Tensor(imgs.size(0), 1).fill_(0.0)
 
         # Configure input
-        real_imgs = Variable(imgs.type(Tensor))
+        real_imgs = imgs.type(torch.Tensor)
 
         # -----------------
         #  Train Generator
@@ -135,7 +134,7 @@ def forward(self, img):
         optimizer_G.zero_grad()
 
         # Sample noise as generator input
-        z = Variable(Tensor(np.random.normal(0, 1, (imgs.shape[0], opt.latent_dim))))
+        z = torch.Tensor(np.random.normal(0, 1, (imgs.shape[0], opt.latent_dim)))
 
         # Generate a batch of images
         gen_imgs = generator(z)