diff --git a/ReinforcementLearning/PolicyGradient/actor_critic/actor_critic_replay_torch.py b/ReinforcementLearning/PolicyGradient/actor_critic/actor_critic_replay_torch.py
index f6fac17..56ba73d 100644
--- a/ReinforcementLearning/PolicyGradient/actor_critic/actor_critic_replay_torch.py
+++ b/ReinforcementLearning/PolicyGradient/actor_critic/actor_critic_replay_torch.py
@@ -51,7 +51,7 @@ def __init__(self, lr, input_dims, fc1_dims, fc2_dims,
         self.v = nn.Linear(self.fc2_dims, 1)
         self.optimizer = optim.Adam(self.parameters(), lr=lr)
 
-        self.device = T.device('cuda:0' if T.cuda.is_available() else 'cuda:1')
+        self.device = T.device('cuda:0' if T.cuda.is_available() else 'cpu')
         self.to(self.device)
 
     def forward(self, state):