diff --git a/model.py b/model.py
index d6cf038..cc548e0 100644
--- a/model.py
+++ b/model.py
@@ -102,7 +102,8 @@ class GaussianPolicy(nn.Module):
         # Enforcing Action Bound
         log_prob -= torch.log(self.action_scale * (1 - y_t.pow(2)) + epsilon)
         log_prob = log_prob.sum(1, keepdim=True)
-        return action, log_prob, torch.tanh(mean)
+        mean = torch.tanh(mean) * self.action_scale + self.action_bias
+        return action, log_prob, mean
 
     def to(self, device):
         self.action_scale = self.action_scale.to(device)