Merge branch 'master' into cifartut_improve

Author: holly1238

intermediate_source/reinforcement_q_learning.py

@@ -426,7 +426,8 @@ def optimize_model():
     expected_state_action_values = (next_state_values * GAMMA) + reward_batch
 
     # Compute Huber loss
-    loss = F.smooth_l1_loss(state_action_values, expected_state_action_values.unsqueeze(1))
+    criterion = nn.SmoothL1Loss()
+    loss = criterion(state_action_values, expected_state_action_values.unsqueeze(1))
 
     # Optimize the model
     optimizer.zero_grad()