Use standard tool for cleaning up metrics

parlai/crowdsourcing/tasks/model_chat/bot_agent.py

@@ -44,20 +44,15 @@ def act(self, timeout=None):
                 act_out = self.model_agent.act()
         else:
             act_out = self.model_agent.act()
-        act_out = Message(act_out)
-        # Wrap as a Message for compatibility with older ParlAI models
-
-        if 'metrics' in act_out:
-            # Metrics can't be serialized when saving results as a JSON
-            del act_out['metrics']
+        act_out = Message(act_out).json_safe_payload()
 
         if 'dict_lower' in self.opt and not self.opt['dict_lower']:
             # model is cased so we don't want to normalize the reply like below
             final_message_text = act_out['text']
         else:
             final_message_text = normalize_reply(act_out['text'])
 
-        act_out.force_set('text', final_message_text)
+        act_out['text'] = final_message_text
         assert ('episode_done' not in act_out) or (not act_out['episode_done'])
         self.turn_idx += 1
         return {**act_out, 'episode_done': False}

parlai/crowdsourcing/tasks/model_chat/worlds.py

@@ -12,8 +12,9 @@
 
 import numpy as np
 
-from parlai.core.worlds import validate
 from parlai.core.agents import create_agent_from_shared
+from parlai.core.message import Message
+from parlai.core.worlds import validate
 from parlai.crowdsourcing.utils.acceptability import AcceptabilityChecker
 from parlai.crowdsourcing.utils.worlds import CrowdOnboardWorld, CrowdTaskWorld
 from parlai.crowdsourcing.tasks.model_chat.bot_agent import TurkLikeAgent
@@ -224,10 +225,9 @@ def parley(self):
         for idx, agent in enumerate([self.agent, self.bot]):
             if not self.chat_done:
                 acts[idx] = agent.act(timeout=self.max_resp_time)
-                acts[idx] = Compatibility.maybe_fix_act(acts[idx])
-                if 'metrics' in acts[idx]:
-                    del acts[idx]['metrics']
-                    # Metrics can't be saved to JSON and are not needed here
+                acts[idx] = Message(
+                    Compatibility.maybe_fix_act(acts[idx])
+                ).json_safe_payload()
                 print(
                     f'Got act for agent idx {idx}, act was: {acts[idx]} and self.task_turn_idx: {self.task_turn_idx}.'
                 )

parlai/crowdsourcing/tasks/model_chat/worlds_image_chat.py

@@ -75,10 +75,9 @@ def _run_initial_turn(self) -> None:
 
         # Have the bot respond
         bot_first_act_raw = self.bot.act()
-        bot_first_act_raw = Compatibility.maybe_fix_act(bot_first_act_raw)
-        if 'metrics' in bot_first_act_raw:
-            del bot_first_act_raw['metrics']
-            # Metrics can't be saved to JSON and are not needed here
+        bot_first_act_raw = Message(
+            Compatibility.maybe_fix_act(bot_first_act_raw)
+        ).json_safe_payload()
         self.agent.observe(validate(bot_first_act_raw))
         bot_first_act = {
             'episode_done': False,
@@ -91,7 +90,7 @@ def _run_initial_turn(self) -> None:
         self.dialog.append(image_act)
         self.dialog.append(bot_first_act)
 
-    def _postprocess_acts(self, acts: List[Message], agent_idx: int):
+    def _postprocess_acts(self, acts: List[dict], agent_idx: int):
         """
         Show the bot the image again on every turn.
         """
@@ -100,7 +99,7 @@ def _postprocess_acts(self, acts: List[Message], agent_idx: int):
             # image-related fields needed by the model
             for key, value in self.image_act.items():
                 if key not in ['episode_done', 'id', 'text', 'agent_idx']:
-                    acts[agent_idx].force_set(key, value)
+                    acts[agent_idx][key] = value
 
     def get_final_chat_data(self) -> Dict[str, Any]:
         """