End the game early in case of a timeout (#259)

End the game early in case of timeout or non yes/no answer.

Author: bovard

kaggle_environments/envs/llm_20_questions/llm_20_questions.py

@@ -24,6 +24,7 @@
 DONE = "DONE"
 INACTIVE = "INACTIVE"
 ACTIVE = "ACTIVE"
+TIMEOUT = "TIMEOUT"
 
 GUESS = "guess"
 ASK = "ask"
@@ -98,43 +99,40 @@ def guesser_action(active, inactive, step):
     if active.action and keyword_guessed(active.action):
         guessed = True
         score = 20 - int(step / 3)
-        active.reward = score
-        inactive.reward = score
-        active.status = DONE
-        inactive.status = DONE
-        active.observation.keyword = keyword
-        active.observation.category = category
+        end_game(active, inactive, score, DONE, DONE)
+    return guessed
+
+def end_game(active, inactive, reward, status, inactive_status):
+    active.observation.keyword = keyword
+    active.observation.category = category
     inactive.observation.keyword = keyword
     inactive.observation.category = category
-    return guessed
+    active.reward = reward
+    inactive.reward = reward
+    active.status = status
+    inactive.status = inactive_status
+
 
 def answerer_action(active, inactive):
     active.observation.keyword = keyword
     active.observation.category = category
     response = active.action
     if not response:
         response = "none"
-        active.status = ERROR
+        end_game(active, inactive, -1, ERROR, DONE)
     elif "yes" in response.lower():
         response = "yes"
     elif "no" in response.lower():
         response = "no"
     else:
         response = "maybe"
-        active.status = ERROR
+        end_game(active, inactive, -1, ERROR, DONE)
     active.observation.answers.append(response)
     inactive.observation.answers.append(response)
 
 def increment_turn(active, inactive, step, guessed):
     if step == 59 and not guessed:
-        active.observation.keyword = keyword
-        active.observation.category = category
-        inactive.observation.keyword = keyword
-        inactive.observation.category = category
-        active.reward = -1
-        inactive.reward = -1
-        active.status = DONE
-        inactive.status = DONE
+        end_game(active, inactive, -1, DONE, DONE)
     elif active.observation.turnType == "guess":
         active.observation.turnType = "ask"
     elif active.observation.turnType == "ask":
@@ -166,21 +164,33 @@ def interpreter(state, env):
 
     step = state[0].observation.step
 
+    end_early = (active1 and active1.status) in (TIMEOUT, ERROR) or (active2 and active2.status in (TIMEOUT, ERROR))
+
     if active1 is not None:
         guessed = False
         if active1.observation.role == GUESSER:
             guessed = guesser_action(active1, inactive1, step)
         else:
             answerer_action(active1, inactive1)
-        increment_turn(active1, inactive1, step, guessed)
+        if active1.status in (TIMEOUT, ERROR):
+            end_game(active1, inactive1, -1, active1.status, DONE)
+        elif end_early:
+            end_game(active1, inactive1, 0, DONE, DONE)
+        else:
+            increment_turn(active1, inactive1, step, guessed)
 
     if active2 is not None:
         guessed = False
         if active2.observation.role == GUESSER:
             guessed = guesser_action(active2, inactive2, step)
         else:
             answerer_action(active2, inactive2)
-        increment_turn(active2, inactive2, step, guessed)
+        if active2.status in (TIMEOUT, ERROR):
+            end_game(active2, inactive2, -1, active2.status, DONE)
+        elif end_early:
+            end_game(active2, inactive2, 0, DONE, DONE)
+        else:
+            increment_turn(active2, inactive2, step, guessed)
 
     return state