move to ext/sela

.gitignore

@@ -29,7 +29,7 @@ share/python-wheels/
 MANIFEST
 metagpt/tools/schemas/
 examples/data/search_kb/*.json
-sela/AutogluonModels
+metagpt/ext/sela/AutogluonModels
 
 # PyInstaller
 #  Usually these files are written by a python scripts from a template
@@ -189,4 +189,4 @@ cov.xml
 *-structure.json
 *.dot
 .python-version
-sela/results/*
+metagpt/ext/sela/results/*

metagpt/ext/sela/Greedy.py

@@ -1,6 +1,6 @@
 import random
 
-from sela.MCTS import MCTS
+from metagpt.ext.sela.MCTS import MCTS
 
 
 class Greedy(MCTS):

metagpt/ext/sela/MCTS.py

@@ -8,12 +8,12 @@ import shutil
 import numpy as np
 import pandas as pd
 
-from sela.data.custom_task import get_mle_bench_requirements, get_mle_task_id
-from sela.data.dataset import generate_task_requirement, get_split_dataset_path
-from sela.evaluation.evaluation import evaluate_score
-from sela.insights.instruction_generator import InstructionGenerator
-from sela.research_assistant import ResearchAssistant, TimeoutException
-from sela.utils import get_exp_pool_path, load_execute_notebook, mcts_logger
+from metagpt.ext.sela.data.custom_task import get_mle_bench_requirements, get_mle_task_id
+from metagpt.ext.sela.data.dataset import generate_task_requirement, get_split_dataset_path
+from metagpt.ext.sela.evaluation.evaluation import evaluate_score
+from metagpt.ext.sela.insights.instruction_generator import InstructionGenerator
+from metagpt.ext.sela.research_assistant import ResearchAssistant, TimeoutException
+from metagpt.ext.sela.utils import get_exp_pool_path, load_execute_notebook, mcts_logger
 from metagpt.tools.tool_recommend import ToolRecommender
 from metagpt.utils.common import read_json_file
 

metagpt/ext/sela/data.yaml

@@ -0,0 +1,3 @@
+datasets_dir: "path/to/datasets" # path to the datasets directory
+work_dir: ../workspace # path to the workspace directory
+role_dir: storage/SELA # path to the role directory

metagpt/ext/sela/data/custom_task.py

@@ -1,7 +1,7 @@
 import os
 
-from sela.data.dataset import SPECIAL_INSTRUCTIONS
-from sela.experimenter.mle_bench.instructions import (
+from metagpt.ext.sela.data.dataset import SPECIAL_INSTRUCTIONS
+from metagpt.ext.sela.experimenter.mle_bench.instructions import (
     ADDITIONAL_NOTES,
     INSTRUCTIONS,
     INSTRUCTIONS_OBFUSCATED,

metagpt/ext/sela/data/dataset.py

@@ -9,8 +9,8 @@ import pandas as pd
 import yaml
 from sklearn.model_selection import train_test_split
 
-from sela.insights.solution_designer import SolutionDesigner
-from sela.utils import DATA_CONFIG
+from metagpt.ext.sela.insights.solution_designer import SolutionDesigner
+from metagpt.ext.sela.utils import DATA_CONFIG
 
 BASE_USER_REQUIREMENT = """
 This is a {datasetname} dataset. Your goal is to predict the target column `{target_col}`.

metagpt/ext/sela/data/hf_data.py

@@ -7,14 +7,14 @@ import pandas as pd
 from datasets import load_dataset
 from PIL import Image
 
-from sela.data.dataset import (
+from metagpt.ext.sela.data.dataset import (
     ExpDataset,
     parse_args,
     process_dataset,
     save_datasets_dict_to_yaml,
 )
-from sela.insights.solution_designer import SolutionDesigner
-from sela.utils import DATA_CONFIG
+from metagpt.ext.sela.insights.solution_designer import SolutionDesigner
+from metagpt.ext.sela.utils import DATA_CONFIG
 
 HFDATSETS = [
     {"name": "sms_spam", "dataset_name": "ucirvine/sms_spam", "target_col": "label", "modality": "text"},

metagpt/ext/sela/evaluation/visualize_mcts.py

@@ -3,7 +3,7 @@ import textwrap
 import matplotlib.pyplot as plt
 import networkx as nx
 
-from sela.MCTS import Node
+from metagpt.ext.sela.MCTS import Node
 
 NODE_TEMPLATE = """\
 [Node {id}]

metagpt/ext/sela/experimenter/aug.py

@@ -1,7 +1,7 @@
-from sela.experimenter.experimenter import Experimenter
-from sela.insights.instruction_generator import InstructionGenerator
-from sela.research_assistant import ResearchAssistant
-from sela.utils import get_exp_pool_path
+from metagpt.ext.sela.experimenter.experimenter import Experimenter
+from metagpt.ext.sela.insights.instruction_generator import InstructionGenerator
+from metagpt.ext.sela.research_assistant import ResearchAssistant
+from metagpt.ext.sela.utils import get_exp_pool_path
 
 EXPS_PROMPT = """
 When doing the tasks, you can refer to the insights below:

metagpt/ext/sela/experimenter/autogluon.py

@@ -1,5 +1,5 @@
 from datetime import datetime
-from sela.experimenter.custom import CustomExperimenter
+from metagpt.ext.sela.experimenter.custom import CustomExperimenter
 import os
 import pandas as pd
 

metagpt/ext/sela/experimenter/autosklearn.py

@@ -1,7 +1,7 @@
 from datetime import datetime
 import pandas as pd
-from sela.experimenter.custom import CustomExperimenter
-from sela.evaluation.evaluation import evaluate_score
+from metagpt.ext.sela.experimenter.custom import CustomExperimenter
+from metagpt.ext.sela.evaluation.evaluation import evaluate_score
 from functools import partial
 
 

metagpt/ext/sela/experimenter/custom.py

@@ -2,9 +2,9 @@ import os
 
 import pandas as pd
 
-from sela.evaluation.evaluation import evaluate_score
-from sela.experimenter.experimenter import Experimenter
-from sela.MCTS import create_initial_state
+from metagpt.ext.sela.evaluation.evaluation import evaluate_score
+from metagpt.ext.sela.experimenter.experimenter import Experimenter
+from metagpt.ext.sela.MCTS import create_initial_state
 
 
 class CustomExperimenter(Experimenter):

metagpt/ext/sela/experimenter/experimenter.py

@@ -5,10 +5,10 @@ import os
 import numpy as np
 import pandas as pd
 
-from sela.evaluation.evaluation import evaluate_score
-from sela.MCTS import create_initial_state
-from sela.research_assistant import ResearchAssistant
-from sela.utils import DATA_CONFIG, save_notebook
+from metagpt.ext.sela.evaluation.evaluation import evaluate_score
+from metagpt.ext.sela.MCTS import create_initial_state
+from metagpt.ext.sela.research_assistant import ResearchAssistant
+from metagpt.ext.sela.utils import DATA_CONFIG, save_notebook
 
 
 class Experimenter:

metagpt/ext/sela/experimenter/mcts.py

@@ -1,13 +1,13 @@
 import shutil
 
-from sela.evaluation.evaluation import (
+from metagpt.ext.sela.evaluation.evaluation import (
     node_evaluate_score_mlebench,
     node_evaluate_score_sela,
 )
-from sela.evaluation.visualize_mcts import get_tree_text
-from sela.experimenter.experimenter import Experimenter
-from sela.Greedy import Greedy, Random
-from sela.MCTS import MCTS
+from metagpt.ext.sela.evaluation.visualize_mcts import get_tree_text
+from metagpt.ext.sela.experimenter.experimenter import Experimenter
+from metagpt.ext.sela.Greedy import Greedy, Random
+from metagpt.ext.sela.MCTS import MCTS
 
 
 class MCTSExperimenter(Experimenter):

metagpt/ext/sela/experimenter/mle_bench/instructions.py

@@ -8,7 +8,7 @@ You will be given a machine learning task. You must solve the task by training a
 
 BENCHMARK INSTRUCTIONS
 ------
-First, please read the instructions below which are specific to our benchmark setup. Note, these instructions can be found again in `sela.experimenter.mle_bench.instructions.py`.
+First, please read the instructions below which are specific to our benchmark setup. Note, these instructions can be found again in `experimenter/mle_bench/instructions.py`.
 
 - **Competition-specific instructions**: The instructions for the competition are available in `{dataset_dir}/description.md`. Read this file to understand the problem setup for the competition you are currently solving.
 - **Dataset**: The dataset files for the competition are available in `{dataset_dir}`.
@@ -28,7 +28,7 @@ You will be given a machine learning task. You must solve the task by training a
 
 BENCHMARK INSTRUCTIONS
 ------
-First, please read the instructions below which are specific to our benchmark setup. Note, these instructions can be found again in `sela.experimenter.mle_bench.instructions.py`.
+First, please read the instructions below which are specific to our benchmark setup. Note, these instructions can be found again in `experimenter/mle_bench/instructions.py`.
 
 - **Task-specific instructions**: The instructions for the task are available in `{dataset_dir}/description.md`. Read this file to understand the problem setup for the task you are currently solving.
 - **Dataset**: The dataset files for the task are available in `{dataset_dir}/`.

metagpt/ext/sela/insights/instruction_generator.py

@@ -3,8 +3,8 @@ import os
 import random
 from difflib import SequenceMatcher
 
-from sela.insights.solution_designer import SolutionDesigner
-from sela.utils import clean_json_from_rsp, load_data_config, mcts_logger
+from metagpt.ext.sela.insights.solution_designer import SolutionDesigner
+from metagpt.ext.sela.utils import clean_json_from_rsp, load_data_config, mcts_logger
 from metagpt.llm import LLM
 from metagpt.schema import Message
 

metagpt/ext/sela/insights/solution_designer.py

@@ -1,6 +1,6 @@
 import json
 
-from sela.utils import clean_json_from_rsp, load_data_config
+from metagpt.ext.sela.utils import clean_json_from_rsp, load_data_config
 from metagpt.llm import LLM
 
 DATA_CONFIG = load_data_config()

metagpt/ext/sela/research_assistant.py

@@ -6,7 +6,7 @@ import os
 
 from pydantic import model_validator
 
-from sela.utils import mcts_logger, save_notebook
+from metagpt.ext.sela.utils import mcts_logger, save_notebook
 from metagpt.actions.di.write_analysis_code import WriteAnalysisCode
 from metagpt.const import SERDESER_PATH
 from metagpt.roles.di.data_interpreter import DataInterpreter

metagpt/ext/sela/run_experiment.py

@@ -1,13 +1,13 @@
 import argparse
 import asyncio
 
-from sela.data.custom_task import get_mle_is_lower_better, get_mle_task_id
-from sela.experimenter.aug import AugExperimenter
-from sela.experimenter.autogluon import GluonExperimenter
-from sela.experimenter.autosklearn import AutoSklearnExperimenter
-from sela.experimenter.custom import CustomExperimenter
-from sela.experimenter.experimenter import Experimenter
-from sela.experimenter.mcts import MCTSExperimenter
+from metagpt.ext.sela.data.custom_task import get_mle_is_lower_better, get_mle_task_id
+from metagpt.ext.sela.experimenter.aug import AugExperimenter
+from metagpt.ext.sela.experimenter.autogluon import GluonExperimenter
+from metagpt.ext.sela.experimenter.autosklearn import AutoSklearnExperimenter
+from metagpt.ext.sela.experimenter.custom import CustomExperimenter
+from metagpt.ext.sela.experimenter.experimenter import Experimenter
+from metagpt.ext.sela.experimenter.mcts import MCTSExperimenter
 
 
 def get_args(cmd=True):

metagpt/ext/sela/scripts/visualize_experiment.py

@@ -1,9 +1,9 @@
 import networkx as nx
 
-from sela.evaluation.visualize_mcts import build_tree_recursive, visualize_tree
-from sela.MCTS import MCTS, create_initial_state, initialize_di_root_node
-from sela.run_experiment import get_args
-from sela.utils import DATA_CONFIG
+from metagpt.ext.sela.evaluation.visualize_mcts import build_tree_recursive, visualize_tree
+from metagpt.ext.sela.MCTS import MCTS, create_initial_state, initialize_di_root_node
+from metagpt.ext.sela.run_experiment import get_args
+from metagpt.ext.sela.utils import DATA_CONFIG
 
 if __name__ == "__main__":
     args = get_args()

sela/data.yaml

@@ -1,3 +0,0 @@
-datasets_dir: "D:/work/automl/datasets" # path to the datasets directory
-work_dir: ../workspace # path to the workspace directory
-role_dir: storage/SELA # path to the role directory