factor extraction pipeline ready (#16)

* run the code

* update code

* remove some redundant code

---------

Co-authored-by: xuyang1 <xuyang1@microsoft.com>

rdagent/factor_implementation/evolving/evaluators.py

@@ -3,28 +3,28 @@ from __future__ import annotations
 import re
 from typing import List
 
-from pandas.core.api import DataFrame as DataFrame
-
-from core.evolving_framework import Evaluator as EvolvingEvaluator
-from core.evolving_framework import Feedback, QueriedKnowledge
-from core.log import FinCoLog
-from core.utils import multiprocessing_wrapper
-from factor_implementation.evolving.evolvable_subjects import (
-    FactorImplementationList,
-)
-from factor_implementation.share_modules.conf import FactorImplementSettings
-from factor_implementation.share_modules.evaluator import (
+from rdagent.core.evolving_framework import Evaluator as EvolvingEvaluator
+from rdagent.core.evolving_framework import Feedback, QueriedKnowledge
+from rdagent.core.log import FinCoLog
+from rdagent.core.utils import multiprocessing_wrapper
+from rdagent.factor_implementation.evolving.evolvable_subjects import FactorImplementationList
+from rdagent.factor_implementation.share_modules.evaluator import (
     Evaluator as FactorImplementationEvaluator,
 )
-from factor_implementation.share_modules.evaluator import (
+from rdagent.factor_implementation.share_modules.evaluator import (
     FactorImplementationCodeEvaluator,
     FactorImplementationFinalDecisionEvaluator,
     FactorImplementationValueEvaluator,
 )
-from factor_implementation.share_modules.factor import (
+from rdagent.factor_implementation.share_modules.factor import (
     FactorImplementation,
     FactorImplementationTask,
 )
+from pandas.core.api import DataFrame as DataFrame
+
+from rdagent.factor_implementation.share_modules.factor_implementation_config import (
+    FactorImplementSettings,
+)
 
 
 class FactorImplementationSingleFeedback:

rdagent/factor_implementation/evolving/evolvable_subjects.py

@@ -1,8 +1,8 @@
 from __future__ import annotations
 
-from core.evolving_framework import EvolvableSubjects
-from core.log import FinCoLog
-from factor_implementation.share_modules.factor import (
+from rdagent.core.evolving_framework import EvolvableSubjects
+from rdagent.core.log import FinCoLog
+from rdagent.factor_implementation.share_modules.factor import (
     FactorImplementation,
     FactorImplementationTask,
 )

rdagent/factor_implementation/evolving/evolving_strategy.py

@@ -1,26 +1,29 @@
 from __future__ import annotations
 
 import json
+from pathlib import Path
 import random
 from abc import abstractmethod
 from copy import deepcopy
 from typing import TYPE_CHECKING
 
-from jinja2 import Template
-
-from core.evolving_framework import EvolvingStrategy, QueriedKnowledge
-from core.utils import multiprocessing_wrapper
-from factor_implementation.share_modules.conf import FactorImplementSettings
-from factor_implementation.share_modules.factor import (
+from rdagent.core.evolving_framework import EvolvingStrategy, QueriedKnowledge
+from rdagent.core.utils import multiprocessing_wrapper
+from rdagent.factor_implementation.share_modules.factor import (
     FactorImplementation,
     FactorImplementationTask,
     FileBasedFactorImplementation,
 )
-from factor_implementation.share_modules.prompt import (
-    FactorImplementationPrompts,
+from rdagent.core.prompts import Prompts
+from jinja2 import Template
+from rdagent.oai.llm_utils import APIBackend
+
+from rdagent.factor_implementation.share_modules.factor_implementation_config import (
+    FactorImplementSettings,
+)
+from rdagent.factor_implementation.share_modules.factor_implementation_utils import (
+    get_data_folder_intro,
 )
-from factor_implementation.share_modules.utils import get_data_folder_intro
-from oai.llm_utils import APIBackend
 
 if TYPE_CHECKING:
     from factor_implementation.evolving.evolvable_subjects import (
@@ -115,7 +118,9 @@ class FactorEvolvingStrategy(MultiProcessEvolvingStrategy):
             queried_former_failed_knowledge_to_render = queried_former_failed_knowledge
 
             system_prompt = Template(
-                FactorImplementationPrompts()["evolving_strategy_factor_implementation_v1_system"],
+                Prompts(file_path=Path(__file__).parent.parent / "prompts.yaml")[
+                    "evolving_strategy_factor_implementation_v1_system"
+                ],
             ).render(
                 data_info=get_data_folder_intro(),
                 queried_former_failed_knowledge=queried_former_failed_knowledge_to_render,
@@ -128,7 +133,9 @@ class FactorEvolvingStrategy(MultiProcessEvolvingStrategy):
             while True:
                 user_prompt = (
                     Template(
-                        FactorImplementationPrompts()["evolving_strategy_factor_implementation_v1_user"],
+                        Prompts(file_path=Path(__file__).parent.parent / "prompts.yaml")[
+                            "evolving_strategy_factor_implementation_v1_user"
+                        ],
                     )
                     .render(
                         factor_information_str=factor_information_str,
@@ -202,7 +209,9 @@ class FactorEvolvingStrategyWithGraph(MultiProcessEvolvingStrategy):
             queried_former_failed_knowledge_to_render = queried_former_failed_knowledge
 
             system_prompt = Template(
-                FactorImplementationPrompts()["evolving_strategy_factor_implementation_v1_system"],
+                Prompts(file_path=Path(__file__).parent.parent / "prompts.yaml")[
+                    "evolving_strategy_factor_implementation_v1_system"
+                ],
             ).render(
                 data_info=get_data_folder_intro(),
                 queried_former_failed_knowledge=queried_former_failed_knowledge_to_render,
@@ -222,7 +231,11 @@ class FactorEvolvingStrategyWithGraph(MultiProcessEvolvingStrategy):
                     and len(queried_former_failed_knowledge_to_render) != 0
                 ):
                     error_summary_system_prompt = (
-                        Template(FactorImplementationPrompts()["evolving_strategy_error_summary_v2_system"])
+                        Template(
+                            Prompts(file_path=Path(__file__).parent.parent / "prompts.yaml")[
+                                "evolving_strategy_error_summary_v2_system"
+                            ]
+                        )
                         .render(
                             factor_information_str=target_factor_task_information,
                             code_and_feedback=queried_former_failed_knowledge_to_render[
@@ -236,7 +249,11 @@ class FactorEvolvingStrategyWithGraph(MultiProcessEvolvingStrategy):
                     )
                     while True:
                         error_summary_user_prompt = (
-                            Template(FactorImplementationPrompts()["evolving_strategy_error_summary_v2_user"])
+                            Template(
+                                Prompts(file_path=Path(__file__).parent.parent / "prompts.yaml")[
+                                    "evolving_strategy_error_summary_v2_user"
+                                ]
+                            )
                             .render(
                                 queried_similar_component_knowledge=queried_similar_component_knowledge_to_render,
                             )
@@ -256,7 +273,9 @@ class FactorEvolvingStrategyWithGraph(MultiProcessEvolvingStrategy):
 
                 user_prompt = (
                     Template(
-                        FactorImplementationPrompts()["evolving_strategy_factor_implementation_v2_user"],
+                        Prompts(file_path=Path(__file__).parent.parent / "prompts.yaml")[
+                            "evolving_strategy_factor_implementation_v2_user"
+                        ],
                     )
                     .render(
                         factor_information_str=target_factor_task_information,

rdagent/factor_implementation/evolving/factor_implementation_evolving_cli.py

@@ -4,32 +4,29 @@ import subprocess
 from pathlib import Path
 
 import pandas as pd
-from fire.core import Fire
-from tqdm import tqdm
-
-from core.evolving_framework import EvoAgent, KnowledgeBase
-from core.utils import multiprocessing_wrapper
-from factor_implementation.evolving.evaluators import (
+from rdagent.core.evolving_framework import EvoAgent, KnowledgeBase
+from rdagent.core.utils import multiprocessing_wrapper
+from rdagent.factor_implementation.evolving.evaluators import (
     FactorImplementationEvaluatorV1,
     FactorImplementationsMultiEvaluator,
 )
-from factor_implementation.evolving.evolvable_subjects import (
-    FactorImplementationList,
-)
-from factor_implementation.evolving.evolving_strategy import (
+from rdagent.factor_implementation.evolving.evolvable_subjects import FactorImplementationList
+from rdagent.factor_implementation.evolving.evolving_strategy import (
     FactorEvolvingStrategy,
     FactorEvolvingStrategyWithGraph,
 )
-from factor_implementation.evolving.knowledge_management import (
+from rdagent.factor_implementation.evolving.knowledge_management import (
     FactorImplementationGraphKnowledgeBase,
     FactorImplementationGraphRAGStrategy,
     FactorImplementationKnowledgeBaseV1,
     FactorImplementationRAGStrategyV1,
 )
-from factor_implementation.share_modules.factor import (
+from rdagent.factor_implementation.share_modules.factor import (
     FactorImplementationTask,
     FileBasedFactorImplementation,
 )
+from fire.core import Fire
+from tqdm import tqdm
 
 ALPHA101_INIT_COMPONENTS = [
     "1. abs(): absolute value to certain columns",
@@ -107,11 +104,17 @@ class FactorImplementationEvolvingCli:
     def load_or_init_knowledge_base(self, former_knowledge_base_path: Path = None, component_init_list: list = []):
         if former_knowledge_base_path is not None and former_knowledge_base_path.exists():
             factor_knowledge_base = pickle.load(open(former_knowledge_base_path, "rb"))
-            if self.evolving_version == 1 and not isinstance(
-                factor_knowledge_base, FactorImplementationKnowledgeBaseV1,
-            ) or self.evolving_version == 2 and not isinstance(
-                factor_knowledge_base,
-                FactorImplementationGraphKnowledgeBase,
+            if (
+                self.evolving_version == 1
+                and not isinstance(
+                    factor_knowledge_base,
+                    FactorImplementationKnowledgeBaseV1,
+                )
+                or self.evolving_version == 2
+                and not isinstance(
+                    factor_knowledge_base,
+                    FactorImplementationGraphKnowledgeBase,
+                )
             ):
                 raise ValueError("The former knowledge base is not compatible with the current version")
         else:
@@ -259,7 +262,11 @@ class FactorImplementationEvolvingCli:
         print([feedback.final_decision if feedback is not None else None for feedback in feedbacks].count(True))
 
     def implement_amc(
-        self, evo_sub_path_str, former_knowledge_base_path_str, implementation_dump_path_str, slice_index,
+        self,
+        evo_sub_path_str,
+        former_knowledge_base_path_str,
+        implementation_dump_path_str,
+        slice_index,
     ):
         factor_implementations: FactorImplementationList = pickle.load(open(evo_sub_path_str, "rb"))
         factor_implementations.target_factor_tasks = factor_implementations.target_factor_tasks[

rdagent/factor_implementation/evolving/knowledge_management.py

@@ -8,10 +8,7 @@ from itertools import combinations
 from pathlib import Path
 from typing import Union
 
-from finco.graph import UndirectedGraph, UndirectedNode
-from jinja2 import Template
-
-from core.evolving_framework import (
+from rdagent.core.evolving_framework import (
     EvolvableSubjects,
     EvoStep,
     Knowledge,
@@ -19,19 +16,20 @@ from core.evolving_framework import (
     QueriedKnowledge,
     RAGStrategy,
 )
-from core.log import FinCoLog
-from factor_implementation.evolving.evaluators import (
-    FactorImplementationSingleFeedback,
-)
-from factor_implementation.share_modules.conf import FactorImplementSettings
-from factor_implementation.share_modules.factor import (
+from rdagent.core.log import FinCoLog
+from rdagent.factor_implementation.evolving.evaluators import FactorImplementationSingleFeedback
+from rdagent.factor_implementation.share_modules.factor import (
     FactorImplementation,
     FactorImplementationTask,
 )
-from factor_implementation.share_modules.prompt import (
-    FactorImplementationPrompts,
+from rdagent.core.prompts import Prompts
+from rdagent.knowledge_management.graph import UndirectedGraph, UndirectedNode
+from jinja2 import Template
+from rdagent.oai.llm_utils import APIBackend, calculate_embedding_distance_between_str_list
+
+from rdagent.factor_implementation.share_modules.factor_implementation_config import (
+    FactorImplementSettings,
 )
-from oai.llm_utils import APIBackend, calculate_embedding_distance_between_str_list
 
 
 class FactorImplementationKnowledge(Knowledge):
@@ -187,9 +185,9 @@ class FactorImplementationRAGStrategyV1(RAGStrategy):
                     )[-1]
                     for index in similar_indexes
                 ]
-                queried_knowledge.working_task_to_similar_successful_knowledge_dict[
-                    target_factor_task_information
-                ] = similar_successful_knowledge
+                queried_knowledge.working_task_to_similar_successful_knowledge_dict[target_factor_task_information] = (
+                    similar_successful_knowledge
+                )
         return queried_knowledge
 
 
@@ -212,7 +210,7 @@ class FactorImplementationGraphRAGStrategy(RAGStrategy):
     def __init__(self, knowledgebase: FactorImplementationGraphKnowledgeBase) -> None:
         super().__init__(knowledgebase)
         self.current_generated_trace_count = 0
-        self.prompt = FactorImplementationPrompts()
+        self.prompt = Prompts(file_path=Path(__file__).parent.parent / "prompts.yaml")
 
     def generate_knowledge(
         self,

rdagent/factor_implementation/share_modules/evaluator.py

@@ -3,19 +3,19 @@ from abc import ABC, abstractmethod
 from typing import Tuple
 
 import pandas as pd
-from finco.log import FinCoLog
-from jinja2 import Template
-
-from factor_implementation.share_modules.conf import FactorImplementSettings
 from factor_implementation.share_modules.factor import (
     FactorImplementation,
     FactorImplementationTask,
 )
-from factor_implementation.share_modules.prompt import (
-    FactorImplementationPrompts,
-)
+from factor_implementation.share_modules.prompt import FactorImplementationPrompts
+from finco.log import FinCoLog
+from jinja2 import Template
 from oai.llm_utils import APIBackend
 
+from rdagent.factor_implementation.share_modules.factor_implementation_config import (
+    FactorImplementSettings,
+)
+
 
 class Evaluator(ABC):
     @abstractmethod

rdagent/factor_implementation/share_modules/factor.py

@@ -6,17 +6,19 @@ from pathlib import Path
 from typing import Tuple, Union
 
 import pandas as pd
-from filelock import FileLock
-from finco.log import FinCoLog
-
-from factor_implementation.share_modules.conf import FactorImplementSettings
 from factor_implementation.share_modules.exception import (
     CodeFormatException,
     NoOutputException,
     RuntimeErrorException,
 )
+from filelock import FileLock
+from finco.log import FinCoLog
 from oai.llm_utils import md5_hash
 
+from rdagent.factor_implementation.share_modules.factor_implementation_config import (
+    FactorImplementSettings,
+)
+
 
 class FactorImplementationTask:
     # TODO: remove the factor_ prefix may be better
@@ -122,7 +124,8 @@ class FileBasedFactorImplementation(FactorImplementation):
                 raise ValueError(self.FB_CODE_NOT_SET)
         with FileLock(self.workspace_path / "execution.lock"):
             (Path.cwd() / "git_ignore_folder" / "factor_implementation_execution_cache").mkdir(
-                exist_ok=True, parents=True,
+                exist_ok=True,
+                parents=True,
             )
             if FactorImplementSettings().enable_execution_cache:
                 # NOTE: cache the result for the same code

rdagent/factor_implementation/share_modules/factor_implementation_config.py

@@ -1,6 +1,6 @@
 from pathlib import Path
 
-from finco.conf import FincoSettings
+from core.conf import FincoSettings
 
 
 class FactorImplementSettings(FincoSettings):

rdagent/factor_implementation/share_modules/factor_implementation_utils.py

@@ -5,7 +5,7 @@ import pandas as pd
 # render it with jinja
 from jinja2 import Template
 
-from factor_implementation.share_modules.conf import FIS
+from rdagent.factor_implementation.share_modules.factor_implementation_config import FIS
 
 TPL = """
 {{file_name}}

rdagent/factor_implementation/share_modules/prompt.py

@@ -1,22 +0,0 @@
-from pathlib import Path
-from typing import Dict
-
-import yaml
-from finco.utils import SingletonBaseClass
-
-
-class FactorImplementationPrompts(Dict, SingletonBaseClass):
-    def __init__(self):
-        super().__init__()
-        prompt_yaml_path = Path(__file__).parent / "prompts.yaml"
-
-        prompt_yaml_dict = yaml.load(
-            open(
-                prompt_yaml_path,
-                encoding="utf8",
-            ),
-            Loader=yaml.FullLoader,
-        )
-
-        for key, value in prompt_yaml_dict.items():
-            self[key] = value