diff --git a/Train.py b/Train.py
index 666c925e834940cb460ff054048b90f9532a32e8..44306d0c3876b26ac27e6011fc394094a8e538db 100644
--- a/Train.py
+++ b/Train.py
@@ -274,6 +274,9 @@ def trainModelRl(debug, networkName, modelDir, filename, nbIter, batchSize, devF
         reward_ = rewarding(appliable, sentence, action, missingLinks, rewardFunc)
         reward = torch.FloatTensor([reward_]).to(getDevice())
 
+        if debug :
+          print("Reward :", reward_, file=sys.stderr)
+
         newState = None
         toState = strategy[fromState][action.name][1] if action.name in strategy[fromState] else -1
         if appliable :