change tau value for boltzmann

algorithms/dqn.py

@@ -55,7 +55,7 @@ class DQN(nn.Module):
             q_values = self.predict(state)
             return torch.argmax(q_values).item()
 
-    def boltzmann_policy(self, state, n_action, tau=1, clip=(-500., 500.)):
+    def boltzmann_policy(self, state, n_action, tau=0.5, clip=(-500., 500.)):
         """
         Boltzmann policy builds a probability law on q values and returns an
         action selected randomly according to this law