Improved debug readability

Config.py

@@ -90,11 +90,13 @@ class Config :
     printedCols = ["ID","FORM","UPOS","HEAD","DEPREL"]
     left = 5
     right = 5
+    historySize = 8
+    historyPopSize = 6
     print("state :", self.state, file=output)
-    print("stack :",[self.getAsFeature(ind, "ID") for ind in self.stack], file=output)
+    print("stack :",[int(self.getAsFeature(ind, "ID")) for ind in self.stack], file=output)
     print("nbUndone :", self.nbUndone, file=output)
-    print("history :",[str(trans) for trans in self.history], file=output)
-    print("historyPop :",[(str(c[0]),"dat:"+str(c[1]),"mvt:"+str(c[2]),"reward:"+str(c[3]),"state:"+str(c[4])) for c in self.historyPop], file=output)
+    print("history :",[str(trans) for trans in self.history[-historySize:]], file=output)
+    print("historyPop :",[(str(c[0]),"reward:"+str(c[3])) for c in self.historyPop[-historyPopSize:]], file=output)
     toPrint = []
     for lineIndex in range(self.wordIndex-left, self.wordIndex+right) :
       if lineIndex not in range(len(self.lines)) :

Train.py

@@ -271,6 +271,7 @@ def trainModelRl(debug, networkName, modelDir, filename, nbIter, batchSize, devF
         probaOracle = list_probas[fromState][1]
         
         if debug :
+          print("-"*80, file=sys.stderr)
           sentence.printForDebug(sys.stderr)
         action = selectAction(policy_net, state, transitionSet, sentence, missingLinks, probaRandom, probaOracle, fromState)