Fix model chat speakers and personas (#4273)

* Layer changes * Fixes * Work on CI checks * Fixes * Fixes
facebookresearch · Jan 4, 2022 · 6df9361 · 6df9361
1 parent e686a02
commit 6df9361
Show file tree

Hide file tree

Showing 11 changed files with 205 additions and 145 deletions.
diff --git a/parlai/crowdsourcing/tasks/model_chat/frontend/main.js b/parlai/crowdsourcing/tasks/model_chat/frontend/main.js
@@ -31,7 +31,13 @@ function MainApp() {
       renderSidePane={({ mephistoContext: { taskConfig }, appContext: { taskContext } }) => (
         <DefaultTaskDescription
           chatTitle={taskConfig.task_title}
-          taskDescriptionHtml={taskConfig.left_pane_text}
+          taskDescriptionHtml={
+              taskConfig.left_pane_text.replace(
+                  "[persona_string_1]", taskContext.human_persona_string_1,
+              ).replace(
+                  "[persona_string_2]", taskContext.human_persona_string_2,
+              )
+          }
         >
           {(taskContext.hasOwnProperty('image_src') && taskContext['image_src']) ? (
             <div>
@@ -45,24 +51,23 @@ function MainApp() {
         </DefaultTaskDescription>
       )}
       renderTextResponse={
-        ({ 
-          mephistoContext: { taskConfig }, 
+        ({
+          mephistoContext: { taskConfig },
           appContext: { appSettings },
           onMessageSend,
           active,
 
         }) => (
-          <ResponseComponent 
+          <ResponseComponent
             appSettings={appSettings}
             taskConfig={taskConfig}
             active={active}
             onMessageSend={onMessageSend}
           />
-        )  
+        )
       }
     />
   );
 }
 
 ReactDOM.render(<MainApp />, document.getElementById("app"));
-
diff --git a/parlai/crowdsourcing/tasks/model_chat/run.py b/parlai/crowdsourcing/tasks/model_chat/run.py
@@ -14,7 +14,6 @@
 
 from parlai.crowdsourcing.tasks.model_chat.impl import run_task
 from parlai.crowdsourcing.utils.mturk import MTurkRunScriptConfig
-import parlai.crowdsourcing.tasks.model_chat.worlds as world_module
 
 
 TASK_DIRECTORY = os.path.dirname(os.path.abspath(__file__))
@@ -40,7 +39,7 @@ class ScriptConfig(MTurkRunScriptConfig):
 
 @hydra.main(config_path="hydra_configs", config_name="scriptconfig")
 def main(cfg: DictConfig) -> None:
-    run_task(cfg=cfg, task_directory=TASK_DIRECTORY, world_module=world_module)
+    run_task(cfg=cfg, task_directory=TASK_DIRECTORY)
 
 
 if __name__ == "__main__":

diff --git a/parlai/crowdsourcing/tasks/model_chat/utils.py b/parlai/crowdsourcing/tasks/model_chat/utils.py
@@ -381,13 +381,16 @@ def _check_final_chat_data(
                 for actual_message, expected_message in zip(
                     actual_value[key_inner], expected_value_inner
                 ):
-                    self.assertEqual(
-                        {k: v for k, v in actual_message.items() if k != 'message_id'},
-                        {
-                            k: v
-                            for k, v in expected_message.items()
-                            if k != 'message_id'
-                        },
+                    clean_actual_message = {
+                        k: v for k, v in actual_message.items() if k != 'message_id'
+                    }
+                    clean_expected_message = {
+                        k: v for k, v in expected_message.items() if k != 'message_id'
+                    }
+                    self.assertDictEqual(
+                        clean_actual_message,
+                        clean_expected_message,
+                        f'The following dictionaries are different: {clean_actual_message} and {clean_expected_message}',
                     )
             elif key_inner == 'task_description':
                 for (key_inner2, expected_value_inner2) in expected_value_inner.items():

diff --git a/parlai/crowdsourcing/tasks/model_chat/worlds.py b/parlai/crowdsourcing/tasks/model_chat/worlds.py
@@ -139,6 +139,9 @@ def __init__(self, opt, agent, bot):
         self.task_turn_idx = 0
         self.num_turns = num_turns
 
+        self.agent.agent_id = 'Speaker 1'
+        self.bot.agent_id = 'Speaker 2'
+
         self.dialog = []
         self.tag = f'conversation_id {agent.mephisto_agent.db_id}'
         self.task_type = 'sandbox' if opt['is_sandbox'] else 'live'
@@ -447,7 +450,7 @@ def _run_initial_turn(self) -> None:
             # Display the previous two utterances
             human_first_msg = {
                 'episode_done': False,
-                'id': self.agent.id,
+                'id': self.agent.agent_id,
                 'text': self.context_info['person1_seed_utterance'],
                 'fake_start': True,
                 'agent_idx': 0,
@@ -456,7 +459,7 @@ def _run_initial_turn(self) -> None:
                 human_first_msg[k] = v
             bot_first_msg = {
                 'episode_done': False,
-                'id': self.bot.id,
+                'id': self.bot.agent_id,
                 'text': self.context_info['person2_seed_utterance'],
                 'fake_start': True,
                 'agent_idx': 1,
@@ -472,12 +475,20 @@ def _run_initial_turn(self) -> None:
 
         elif self.opt['conversation_start_mode'] == 'hi':
             print('[Displaying "Hi!" only as per Meena task.]')
+            if self.personas is not None:
+                human_persona_strings = [s.strip() for s in self.personas[0]]
+            else:
+                human_persona_strings = ['', '']
             human_first_msg = {
                 'episode_done': False,
-                'id': self.agent.id,
+                'id': self.agent.agent_id,
                 'text': 'Hi!',
                 'fake_start': True,
                 'agent_idx': 0,
+                'task_data': {
+                    'human_persona_string_1': human_persona_strings[0],
+                    'human_persona_string_2': human_persona_strings[1],
+                },
             }
             for k, v in control_msg.items():
                 human_first_msg[k] = v
@@ -487,7 +498,9 @@ def _run_initial_turn(self) -> None:
             self.bot.observe(validate(human_first_msg))
 
             first_bot_act = self.bot.act()
-            first_bot_act = Compatibility.maybe_fix_act(first_bot_act)
+            first_bot_act = Compatibility.backward_compatible_force_set(
+                first_bot_act, 'id', self.bot.agent_id
+            )
 
             self.agent.observe(validate(first_bot_act))
 
@@ -630,8 +643,6 @@ def make_world(opt, agents):
     statistics_condition = opt['statistics_condition']
     context_generator = opt['context_generator']
 
-    agents[0].agent_id = "Worker"
-
     # Get context: personas, previous utterances, etc.
     if context_generator is not None:
         context_info = context_generator.get_context()

diff --git a/parlai/crowdsourcing/tasks/model_chat/worlds_image_chat.py b/parlai/crowdsourcing/tasks/model_chat/worlds_image_chat.py
@@ -78,6 +78,7 @@ def _run_initial_turn(self) -> None:
         bot_first_act_raw = Message(
             Compatibility.maybe_fix_act(bot_first_act_raw)
         ).json_safe_payload()
+        bot_first_act_raw['id'] = self.bot.agent_id
         self.agent.observe(validate(bot_first_act_raw))
         bot_first_act = {
             'episode_done': False,
@@ -134,8 +135,6 @@ def shutdown(self):
 
 def make_world(opt, agents):
 
-    agents[0].agent_id = "Worker"
-
     # We are showing an image to the worker and bot, so grab the image path and other
     # context info
     image_idx, model_name, no_more_work = opt['image_stack'].get_next_image(

diff --git a/tests/crowdsourcing/tasks/model_chat/expected_states/final_chat_data.json b/tests/crowdsourcing/tasks/model_chat/expected_states/final_chat_data.json
@@ -7,16 +7,20 @@
     "dialog": [
         {
             "episode_done": false,
-            "id": "Worker",
+            "id": "Speaker 1",
             "text": "Hi!",
             "fake_start": true,
             "agent_idx": 0,
-            "message_id": "4f72800c-e1ff-4411-8ee5-286336c27859"
+            "message_id": "4f72800c-e1ff-4411-8ee5-286336c27859",
+            "task_data": {
+                "human_persona_string_1": "",
+                "human_persona_string_2": ""
+            }
         },
         {
             "agent_idx": 1,
             "text": "I don't know.",
-            "id": "FixedResponseAgent",
+            "id": "Speaker 2",
             "problem_data": {
                 "bucket_0": false,
                 "bucket_1": false,
@@ -29,12 +33,12 @@
         {
             "agent_idx": 0,
             "text": "What are you nervous about?",
-            "id": "Worker"
+            "id": "Speaker 1"
         },
         {
             "agent_idx": 1,
             "text": "I don't know.",
-            "id": "FixedResponseAgent",
+            "id": "Speaker 2",
             "problem_data": {
                 "bucket_0": false,
                 "bucket_1": false,
@@ -47,12 +51,12 @@
         {
             "agent_idx": 0,
             "text": "Do you have any plans for the weekend?",
-            "id": "Worker"
+            "id": "Speaker 1"
         },
         {
             "agent_idx": 1,
             "text": "I don't know.",
-            "id": "FixedResponseAgent",
+            "id": "Speaker 2",
             "problem_data": {
                 "bucket_0": false,
                 "bucket_1": false,
@@ -65,12 +69,12 @@
         {
             "agent_idx": 0,
             "text": "Yeah that sounds great! I like to bike and try new restaurants.",
-            "id": "Worker"
+            "id": "Speaker 1"
         },
         {
             "agent_idx": 1,
             "text": "I don't know.",
-            "id": "FixedResponseAgent",
+            "id": "Speaker 2",
             "problem_data": {
                 "bucket_0": false,
                 "bucket_1": false,
@@ -83,12 +87,12 @@
         {
             "agent_idx": 0,
             "text": "Oh, Italian food is great. I also love Thai and Indian.",
-            "id": "Worker"
+            "id": "Speaker 1"
         },
         {
             "agent_idx": 1,
             "text": "I don't know.",
-            "id": "FixedResponseAgent",
+            "id": "Speaker 2",
             "problem_data": {
                 "bucket_0": false,
                 "bucket_1": false,
@@ -101,12 +105,12 @@
         {
             "agent_idx": 0,
             "text": "Hmmm - anything with peanuts? Or I like when they have spicy licorice-like herbs.",
-            "id": "Worker"
+            "id": "Speaker 1"
         },
         {
             "agent_idx": 1,
             "text": "I don't know.",
-            "id": "FixedResponseAgent",
+            "id": "Speaker 2",
             "problem_data": {
                 "bucket_0": false,
                 "bucket_1": false,

diff --git a/tests/crowdsourcing/tasks/model_chat/expected_states/final_chat_data__image_chat.json b/tests/crowdsourcing/tasks/model_chat/expected_states/final_chat_data__image_chat.json
@@ -16,69 +16,69 @@
         },
         {
             "episode_done": false,
-            "id": "TransresnetAgent",
+            "id": "Speaker 2",
             "text": "I must learn that bird's name!",
             "agent_idx": 1
         },
         {
             "agent_idx": 0,
             "text": "Response 1",
-            "id": "Worker"
+            "id": "Speaker 1"
         },
         {
             "agent_idx": 1,
             "text": "My, aren't you a pretty bird?",
-            "id": "TransresnetAgent"
+            "id": "Speaker 2"
         },
         {
             "agent_idx": 0,
             "text": "Response 2",
-            "id": "Worker"
+            "id": "Speaker 1"
         },
         {
             "agent_idx": 1,
             "text": "My, aren't you a pretty bird?",
-            "id": "TransresnetAgent"
+            "id": "Speaker 2"
         },
         {
             "agent_idx": 0,
             "text": "Response 3",
-            "id": "Worker"
+            "id": "Speaker 1"
         },
         {
             "agent_idx": 1,
             "text": "My, aren't you a pretty bird?",
-            "id": "TransresnetAgent"
+            "id": "Speaker 2"
         },
         {
             "agent_idx": 0,
             "text": "Response 4",
-            "id": "Worker"
+            "id": "Speaker 1"
         },
         {
             "agent_idx": 1,
             "text": "My, aren't you a pretty bird?",
-            "id": "TransresnetAgent"
+            "id": "Speaker 2"
         },
         {
             "agent_idx": 0,
             "text": "Response 5",
-            "id": "Worker"
+            "id": "Speaker 1"
         },
         {
             "agent_idx": 1,
             "text": "My, aren't you a pretty bird?",
-            "id": "TransresnetAgent"
+            "id": "Speaker 2"
         },
         {
             "agent_idx": 0,
             "text": "Response 6",
-            "id": "Worker"
+            "id": "Speaker 1"
         },
         {
             "agent_idx": 1,
             "text": "My, aren't you a pretty bird?",
-            "id": "TransresnetAgent",
+            "id": "Speaker 2",
             "final_rating": 0
         }
     ],