refactor: move selector codes to an independent folder (#1059)

* refactor: move exp_gen selection files into select directory

* fix selector value in data_science conf

* fix selector template name

* fix CI

---------

Co-authored-by: Young <afe.young@gmail.com>

rdagent/app/data_science/conf.py

@@ -84,7 +84,7 @@ class DataScienceBasePropSetting(KaggleBasePropSetting):
     """The maximum number of traces to grow before merging"""
 
     #### multi-trace:checkpoint selector
-    selector_name: str = "rdagent.scenarios.data_science.proposal.exp_gen.ckp_select.LatestCKPSelector"
+    selector_name: str = "rdagent.scenarios.data_science.proposal.exp_gen.select.expand.LatestCKPSelector"
     """The name of the selector to use"""
     sota_count_window: int = 5
     """The number of trials to consider for SOTA count"""
@@ -92,7 +92,7 @@ class DataScienceBasePropSetting(KaggleBasePropSetting):
     """The threshold for SOTA count"""
 
     #### multi-trace: SOTA experiment selector
-    sota_exp_selector_name: str = "rdagent.scenarios.data_science.proposal.exp_gen.sota_exp_select.GlobalSOTASelector"
+    sota_exp_selector_name: str = "rdagent.scenarios.data_science.proposal.exp_gen.select.submit.GlobalSOTASelector"
     """The name of the SOTA experiment selector to use"""
 
     ### multi-trace:inject optimals for multi-trace

rdagent/oai/backend/litellm.py

@@ -98,7 +98,7 @@ class LiteLLMAPIBackend(APIBackend):
         if response_format and not supports_response_schema(model=LITELLM_SETTINGS.chat_model):
             # Deepseek will enter this branch
             logger.warning(
-                f"{LogColors.RED}Model {LITELLM_SETTINGS.chat_model} does not support response schema, ignoring response_format argument.{LogColors.END}",
+                f"{LogColors.YELLOW}Model {LITELLM_SETTINGS.chat_model} does not support response schema, ignoring response_format argument.{LogColors.END}",
                 tag="llm_messages",
             )
             response_format = None

rdagent/scenarios/data_science/proposal/exp_gen/select/submit.py

@@ -115,11 +115,9 @@ class AutoSOTAexpSelector(SOTAexpSelector):
                         Description: {desc}
                         Final score: {current_final_score}\n\n"""
 
-            system_prompt = T(".prompts_selector:auto_sota_selector.system").r(
-                scenario=trace.scen.get_scenario_all_desc()
-            )
+            system_prompt = T(".prompts:auto_sota_selector.system").r(scenario=trace.scen.get_scenario_all_desc())
 
-            user_prompt = T(".prompts_selector:auto_sota_selector.user").r(
+            user_prompt = T(".prompts:auto_sota_selector.user").r(
                 historical_sota_exp_with_desc_and_scores=SOAT_exp_with_desc_and_scores,
             )