Rename NLU to CR

docker/docker-compose.yaml

@@ -47,7 +47,7 @@ services:
       <<: *healthcheck-defaults
     volumes: *model-volume
     entrypoint: python
-    command: src/emma_policy/commands/run_simbot_nlu.py
+    command: src/emma_policy/commands/run_simbot_cr.py
 
   instruction_predictor:
     container_name: instruction_predictor

src/emma_experience_hub/api/clients/simbot/__init__.py

@@ -5,7 +5,7 @@
     SimBotExtractedFeaturesClient,
     SimBotPydanticCacheClient,
 )
+from emma_experience_hub.api.clients.simbot.cr_intent import SimBotCRIntentClient
 from emma_experience_hub.api.clients.simbot.features import SimBotFeaturesClient
-from emma_experience_hub.api.clients.simbot.nlu_intent import SimBotNLUIntentClient
 from emma_experience_hub.api.clients.simbot.placeholder_vision import SimBotPlaceholderVisionClient
 from emma_experience_hub.api.clients.simbot.session_db import SimBotSessionDbClient

src/emma_experience_hub/api/clients/simbot/nlu_intent.py → src/emma_experience_hub/api/clients/simbot/cr_intent.py

@@ -4,8 +4,8 @@
 from emma_experience_hub.api.clients.emma_policy import EmmaPolicyClient
 
 
-class SimBotNLUIntentClient(EmmaPolicyClient):
-    """API Client for SimBot NLU."""
+class SimBotCRIntentClient(EmmaPolicyClient):
+    """API Client for SimBot CR."""
 
     def generate(
         self,

src/emma_experience_hub/api/controllers/simbot/clients.py

@@ -11,9 +11,9 @@
 from emma_experience_hub.api.clients.simbot import (
     SimbotActionPredictionClient,
     SimBotAuxiliaryMetadataClient,
+    SimBotCRIntentClient,
     SimBotExtractedFeaturesClient,
     SimBotFeaturesClient,
-    SimBotNLUIntentClient,
     SimBotPlaceholderVisionClient,
     SimBotSessionDbClient,
 )
@@ -26,7 +26,7 @@ class SimBotControllerClients(BaseModel, arbitrary_types_allowed=True):
     _exit = Event()
 
     features: SimBotFeaturesClient
-    nlu_intent: SimBotNLUIntentClient
+    cr_intent: SimBotCRIntentClient
     action_predictor: SimbotActionPredictionClient
     session_db: SimBotSessionDbClient
 
@@ -53,8 +53,8 @@ def from_simbot_settings(cls, simbot_settings: SimBotSettings) -> "SimBotControl
             session_db=SimBotSessionDbClient(
                 db_file=Path(simbot_settings.session_local_db_file),
             ),
-            nlu_intent=SimBotNLUIntentClient(
-                endpoint=simbot_settings.nlu_predictor_url,
+            cr_intent=SimBotCRIntentClient(
+                endpoint=simbot_settings.cr_predictor_url,
                 timeout=simbot_settings.client_timeout,
             ),
             action_predictor=SimbotActionPredictionClient(

src/emma_experience_hub/api/controllers/simbot/pipelines.py

@@ -4,7 +4,7 @@
 from emma_experience_hub.common.settings import SimBotSettings
 from emma_experience_hub.parsers.simbot import (
     SimBotActionPredictorOutputParser,
-    SimBotNLUOutputParser,
+    SimBotCROutputParser,
     SimBotPreviousActionParser,
     SimBotVisualGroundingOutputParser,
 )
@@ -54,9 +54,9 @@ def from_clients(
             environment_intent_extractor=SimBotEnvironmentIntentExtractionPipeline(),
             agent_intent_selector=SimBotAgentIntentSelectionPipeline(
                 features_client=clients.features,
-                nlu_intent_client=clients.nlu_intent,
-                nlu_intent_parser=SimBotNLUOutputParser(
-                    intent_type_delimiter=simbot_settings.nlu_predictor_intent_type_delimiter
+                cr_intent_client=clients.cr_intent,
+                cr_intent_parser=SimBotCROutputParser(
+                    intent_type_delimiter=simbot_settings.cr_predictor_intent_type_delimiter
                 ),
                 environment_error_pipeline=SimBotEnvironmentErrorCatchingPipeline(),
                 action_predictor_client=clients.action_predictor,

src/emma_experience_hub/common/settings/simbot.py

@@ -66,8 +66,8 @@ class SimBotSettings(BaseSettings):
 
     feature_extractor_url: AnyHttpUrl = AnyHttpUrl(url=f"{scheme}://0.0.0.0:5500", scheme=scheme)
 
-    nlu_predictor_url: AnyHttpUrl = AnyHttpUrl(url=f"{scheme}://0.0.0.0:5501", scheme=scheme)
-    nlu_predictor_intent_type_delimiter: str = " "
+    cr_predictor_url: AnyHttpUrl = AnyHttpUrl(url=f"{scheme}://0.0.0.0:5501", scheme=scheme)
+    cr_predictor_intent_type_delimiter: str = " "
 
     action_predictor_url: AnyHttpUrl = AnyHttpUrl(url=f"{scheme}://0.0.0.0:5502", scheme=scheme)
 

src/emma_experience_hub/datamodels/simbot/__init__.py

@@ -6,10 +6,10 @@
 from emma_experience_hub.datamodels.simbot.enums import (
     SimBotActionStatusType,
     SimBotActionType,
+    SimBotCRIntentType,
     SimBotDummyRawActions,
     SimBotEnvironmentIntentType,
     SimBotIntentType,
-    SimBotNLUIntentType,
     SimBotPhysicalInteractionIntentType,
     SimBotUserIntentType,
     SimBotVerbalInteractionIntentType,

src/emma_experience_hub/datamodels/simbot/enums/__init__.py

@@ -4,9 +4,9 @@
     SimBotDummyRawActions,
 )
 from emma_experience_hub.datamodels.simbot.enums.intents import (
+    SimBotCRIntentType,
     SimBotEnvironmentIntentType,
     SimBotIntentType,
-    SimBotNLUIntentType,
     SimBotPhysicalInteractionIntentType,
     SimBotUserIntentType,
     SimBotVerbalInteractionIntentType,

src/emma_experience_hub/datamodels/simbot/enums/intents.py

@@ -165,7 +165,7 @@ def is_verbal_interaction_intent_type(  # noqa: WPS602
     SimBotIntentType.generic_success,
 ]
 
-SimBotNLUIntentType = Literal[
+SimBotCRIntentType = Literal[
     SimBotIntentType.act_no_match,
     SimBotIntentType.act_missing_inventory,
     SimBotIntentType.act_too_many_matches,

src/emma_experience_hub/functions/simbot/agent_intent_selection/instruction_handler.py

@@ -5,14 +5,14 @@
 from emma_common.datamodels import EnvironmentStateTurn, SpeakerRole
 from emma_experience_hub.api.clients.simbot import (
     SimbotActionPredictionClient,
+    SimBotCRIntentClient,
     SimBotFeaturesClient,
-    SimBotNLUIntentClient,
 )
 from emma_experience_hub.datamodels.simbot import (
     SimBotAgentIntents,
+    SimBotCRIntentType,
     SimBotIntent,
     SimBotIntentType,
-    SimBotNLUIntentType,
     SimBotSession,
     SimBotUserSpeech,
     SimBotUtterance,
@@ -27,8 +27,8 @@ class SimBotActHandler:
     def __init__(
         self,
         features_client: SimBotFeaturesClient,
-        nlu_intent_client: SimBotNLUIntentClient,
-        nlu_intent_parser: NeuralParser[SimBotIntent[SimBotNLUIntentType]],
+        cr_intent_client: SimBotCRIntentClient,
+        cr_intent_parser: NeuralParser[SimBotIntent[SimBotCRIntentType]],
         action_predictor_client: SimbotActionPredictionClient,
         _enable_clarification_questions: bool = True,
         _enable_search_actions: bool = True,
@@ -38,8 +38,8 @@ def __init__(
     ) -> None:
         self._features_client = features_client
 
-        self._nlu_intent_client = nlu_intent_client
-        self._nlu_intent_parser = nlu_intent_parser
+        self._cr_intent_client = cr_intent_client
+        self._cr_intent_parser = cr_intent_parser
         self._action_predictor_client = action_predictor_client
 
         self._enable_clarification_questions = _enable_clarification_questions
@@ -50,24 +50,24 @@ def __init__(
 
     def run(self, session: SimBotSession) -> Optional[SimBotAgentIntents]:
         """Get the agent intent."""
-        # Check if the utterance has already been processed by the NLU
-        if self._utterance_has_been_processed_by_nlu(session):
+        # Check if the utterance has already been processed by the CR
+        if self._utterance_has_been_processed_by_cr(session):
             logger.debug("Executing utterance that triggered the search.")
             return SimBotAgentIntents(
                 physical_interaction=SimBotIntent(type=SimBotIntentType.act_one_match)
             )
 
-        # Otherwise, use the NLU to detect it
-        intents = self._process_utterance_with_nlu(session)
+        # Otherwise, use the CR to detect it
+        intents = self._process_utterance_with_cr(session)
         if self._should_search_target_object(session, intents):
             intents = self._handle_act_no_match_intent(session=session, intents=intents)
         elif self._should_search_missing_inventory(session, intents):
             intents = self._handle_act_missing_inventory_intent(session=session, intents=intents)
 
         return self._handle_search_holding_object(session=session, intents=intents)
 
-    def _utterance_has_been_processed_by_nlu(self, session: SimBotSession) -> bool:
-        """Determine if the utterance has already been processed by the NLU.
+    def _utterance_has_been_processed_by_cr(self, session: SimBotSession) -> bool:
+        """Determine if the utterance has already been processed by the CR.
 
         This happens when the intent was act no_match and search was completed.
         """
@@ -78,15 +78,15 @@ def _utterance_has_been_processed_by_nlu(self, session: SimBotSession) -> bool:
             and session.previous_turn.actions.is_successful
         )
 
-    def _process_utterance_with_nlu(self, session: SimBotSession) -> SimBotAgentIntents:
-        """Perform NLU on the utterance to determine what the agent should do next.
+    def _process_utterance_with_cr(self, session: SimBotSession) -> SimBotAgentIntents:
+        """Perform CR on the utterance to determine what the agent should do next.
 
         This is primarily used to determine whether the agent should act or ask for more
         information.
         """
         extracted_features = self._features_client.get_features(session.current_turn)
-        intent = self._nlu_intent_parser(
-            self._nlu_intent_client.generate(
+        intent = self._cr_intent_parser(
+            self._cr_intent_client.generate(
                 dialogue_history=session.current_turn.utterances,
                 environment_state_history=[EnvironmentStateTurn(features=extracted_features)],
                 inventory_entity=session.current_state.inventory.entity,
@@ -141,13 +141,13 @@ def _process_utterance_with_nlu(self, session: SimBotSession) -> SimBotAgentInte
             )
 
         raise NotImplementedError(
-            "All NLU intents are not accounted for. This means that NLU has returned an intent which does not map to either an interaction intent, or a response intent."
+            "All CR intents are not accounted for. This means that CR has returned an intent which does not map to either an interaction intent, or a response intent."
         )
 
     def _handle_act_no_match_intent(
         self, session: SimBotSession, intents: SimBotAgentIntents
     ) -> SimBotAgentIntents:
-        """Update the session based on the NLU output.
+        """Update the session based on the CR output.
 
         For `act_no_match`, push the current utterance in the utterance queue, and set the verbal
         interaction intent to `confirm_before_search`.
@@ -301,7 +301,7 @@ def _should_search_missing_inventory(
         return should_search_before_executing_instruction
 
     def _should_replace_missing_inventory_intent(
-        self, intent: SimBotIntent[SimBotNLUIntentType]
+        self, intent: SimBotIntent[SimBotCRIntentType]
     ) -> bool:
         """Is the missing inventory intent disabled?"""
         return (

src/emma_experience_hub/parsers/simbot/__init__.py

@@ -1,10 +1,10 @@
 from emma_experience_hub.parsers.simbot.action_predictor_output import (
     SimBotActionPredictorOutputParser,
 )
+from emma_experience_hub.parsers.simbot.cr_output import SimBotCROutputParser
 from emma_experience_hub.parsers.simbot.intent_from_action_status import (
     SimBotIntentFromActionStatusParser,
 )
-from emma_experience_hub.parsers.simbot.nlu_output import SimBotNLUOutputParser
 from emma_experience_hub.parsers.simbot.previous_action import SimBotPreviousActionParser
 from emma_experience_hub.parsers.simbot.visual_grounding_output import (
     SimBotVisualGroundingOutputParser,

src/emma_experience_hub/parsers/simbot/nlu_output.py → src/emma_experience_hub/parsers/simbot/cr_output.py

@@ -3,21 +3,21 @@
 from loguru import logger
 
 from emma_experience_hub.datamodels.simbot import (
+    SimBotCRIntentType,
     SimBotIntent,
     SimBotIntentType,
-    SimBotNLUIntentType,
 )
 from emma_experience_hub.parsers.parser import NeuralParser
 
 
-class SimBotNLUOutputParser(NeuralParser[SimBotIntent[SimBotNLUIntentType]]):
-    """Convert the output from the SimBot NLU module to a SimBot intent."""
+class SimBotCROutputParser(NeuralParser[SimBotIntent[SimBotCRIntentType]]):
+    """Convert the output from the SimBot CR module to a SimBot intent."""
 
     def __init__(self, intent_type_delimiter: str) -> None:
         self._intent_type_delimiter = intent_type_delimiter
 
-    def __call__(self, output_text: str) -> SimBotIntent[SimBotNLUIntentType]:
-        """Parses the intent generated by the NLU component.
+    def __call__(self, output_text: str) -> SimBotIntent[SimBotCRIntentType]:
+        """Parses the intent generated by the CR component.
 
         The model is trained with the following templates:
             - <act><one_match>
@@ -26,15 +26,15 @@ def __call__(self, output_text: str) -> SimBotIntent[SimBotNLUIntentType]:
             - <act><missing_inventory> object_name
             - <search>
         """
-        logger.debug(f"NLU output text: `{output_text}`")
+        logger.debug(f"CR output text: `{output_text}`")
 
         # Split the raw output text by the given delimiter. We assume it's a " " separating the
         # special tokens and the object_name.
         split_parts = output_text.split(self._intent_type_delimiter)
 
         # Get the intent type from the left-side of the template.
         intent_type = SimBotIntentType(split_parts[0])
-        intent_type = cast(SimBotNLUIntentType, intent_type)
+        intent_type = cast(SimBotCRIntentType, intent_type)
 
         # If it exists, get the object name from the right-side of the template
         object_name = " ".join(split_parts[1:]) if len(split_parts) > 1 else None

src/emma_experience_hub/pipelines/simbot/agent_intent_selection.py

@@ -4,14 +4,14 @@
 
 from emma_experience_hub.api.clients.simbot import (
     SimbotActionPredictionClient,
+    SimBotCRIntentClient,
     SimBotFeaturesClient,
-    SimBotNLUIntentClient,
 )
 from emma_experience_hub.datamodels.simbot import (
     SimBotAgentIntents,
+    SimBotCRIntentType,
     SimBotIntent,
     SimBotIntentType,
-    SimBotNLUIntentType,
     SimBotSession,
     SimBotUserIntentType,
 )
@@ -58,8 +58,8 @@ class SimBotAgentIntentSelectionPipeline:
     def __init__(
         self,
         features_client: SimBotFeaturesClient,
-        nlu_intent_client: SimBotNLUIntentClient,
-        nlu_intent_parser: NeuralParser[SimBotIntent[SimBotNLUIntentType]],
+        cr_intent_client: SimBotCRIntentClient,
+        cr_intent_parser: NeuralParser[SimBotIntent[SimBotCRIntentType]],
         action_predictor_client: SimbotActionPredictionClient,
         environment_error_pipeline: SimBotEnvironmentErrorCatchingPipeline,
         _enable_clarification_questions: bool = True,
@@ -72,8 +72,8 @@ def __init__(
         self._features_client = features_client
         self.act_handler = SimBotActHandler(
             features_client=features_client,
-            nlu_intent_client=nlu_intent_client,
-            nlu_intent_parser=nlu_intent_parser,
+            cr_intent_client=cr_intent_client,
+            cr_intent_parser=cr_intent_parser,
             action_predictor_client=action_predictor_client,
             _enable_clarification_questions=_enable_clarification_questions,
             _enable_search_actions=_enable_search_actions,

tests/fixtures/clients.py

@@ -4,8 +4,8 @@
 
 from emma_experience_hub.api.clients.simbot import (
     SimbotActionPredictionClient,
+    SimBotCRIntentClient,
     SimBotFeaturesClient,
-    SimBotNLUIntentClient,
 )
 from emma_experience_hub.datamodels import EmmaExtractedFeatures
 from tests.fixtures.simbot_arena_constants import create_placeholder_features_frames
@@ -26,13 +26,13 @@ def mock_features(*args: Any, **kwargs: Any) -> list[EmmaExtractedFeatures]:  #
 def mock_policy_response_goto_room(monkeypatch: MonkeyPatch) -> None:
     """Mock the responses of EMMA policy."""
 
-    def get_nlu(*args: Any, **kwargs: Any) -> str:  # noqa: WPS430
+    def get_cr(*args: Any, **kwargs: Any) -> str:  # noqa: WPS430
         return "<act><one_match>"
 
     def get_action(*args: Any, **kwargs: Any) -> str:  # noqa: WPS430
         return "goto breakroom<stop>."
 
-    monkeypatch.setattr(SimBotNLUIntentClient, "generate", get_nlu)
+    monkeypatch.setattr(SimBotCRIntentClient, "generate", get_cr)
     monkeypatch.setattr(SimbotActionPredictionClient, "generate", get_action)
 
 
@@ -41,13 +41,13 @@ def mock_policy_response_toggle_computer(monkeypatch: MonkeyPatch) -> None:
     """Mock the responses of EMMA policy when the input instruction is about turning on the
     computer."""
 
-    def get_nlu(*args: Any, **kwargs: Any) -> str:  # noqa: WPS430
+    def get_cr(*args: Any, **kwargs: Any) -> str:  # noqa: WPS430
         return "<act><one_match>"
 
     def get_action(*args: Any, **kwargs: Any) -> str:  # noqa: WPS430
         return "toggle computer <frame_token_1> <vis_token_1><stop>."
 
-    monkeypatch.setattr(SimBotNLUIntentClient, "generate", get_nlu)
+    monkeypatch.setattr(SimBotCRIntentClient, "generate", get_cr)
     monkeypatch.setattr(SimbotActionPredictionClient, "generate", get_action)
 
 
@@ -56,7 +56,7 @@ def mock_policy_response_search(monkeypatch: MonkeyPatch) -> None:
     """Mock the responses of EMMA policy when the input instruction is about searching an
     object."""
 
-    def get_nlu(*args: Any, **kwargs: Any) -> str:  # noqa: WPS430
+    def get_cr(*args: Any, **kwargs: Any) -> str:  # noqa: WPS430
         return "<search>"
 
     def get_object(*args: Any, **kwargs: Any) -> list[str]:  # noqa: WPS430
@@ -65,6 +65,6 @@ def get_object(*args: Any, **kwargs: Any) -> list[str]:  # noqa: WPS430
     def get_target(*args: Any, **kwargs: Any) -> str:  # noqa: WPS430
         return "goto object <frame_token_1> <vis_token_1> <stop>."
 
-    monkeypatch.setattr(SimBotNLUIntentClient, "generate", get_nlu)
+    monkeypatch.setattr(SimBotCRIntentClient, "generate", get_cr)
     monkeypatch.setattr(SimbotActionPredictionClient, "find_object_in_scene", get_object)
     monkeypatch.setattr(SimbotActionPredictionClient, "generate", get_target)