🌟 The Multi-Agent Framework: First AI Software Company, Towards Natural Language Programming
update text dataset
Yizhou Chi562af8c4e4b2
diff --git a/expo/data/dataset.py b/expo/data/dataset.pyindex 1494eb267..3b2017d1a 100644--- a/expo/data/dataset.py+++ b/expo/data/dataset.py@@ -268,8 +268,9 @@ class ExpDataset:print(f"Dataset info for {self.name} already exists")def save_datasetinfo(self, dataset_info):- with open(Path(self.dataset_dir, self.name, "dataset_info.json"), "w") as file:- json.dump(dataset_info, file, indent=4)+ with open(Path(self.dataset_dir, self.name, "dataset_info.json"), "w", encoding="utf-8") as file:+ # utf-8 encoding is required+ json.dump(dataset_info, file, indent=4, ensure_ascii=False)def save_split_datasets(self, df, split, target_col=None):path = Path(self.dataset_dir, self.name)diff --git a/expo/data/hf_data.py b/expo/data/hf_data.pyindex 9ed2b2c48..6f1f55f6d 100644--- a/expo/data/hf_data.py+++ b/expo/data/hf_data.py@@ -9,12 +9,22 @@ from expo.data.dataset import ExpDataset, process_dataset, save_datasets_dict_tofrom expo.insights.solution_designer import SolutionDesignerHFDATSETS = [- {"name": "sms_spam", "dataset_name": "ucirvine/sms_spam", "target_col": "label"},- {"name": "banking77", "dataset_name": "PolyAI/banking77", "target_col": "label"},- {"name": "gnad10", "dataset_name": "community-datasets/gnad10", "target_col": "label"},- {"name": "oxford-iiit-pet", "dataset_name": "timm/oxford-iiit-pet", "target_col": "label"},- {"name": "stanford_cars", "dataset_name": "tanganke/stanford_cars", "target_col": "label"},- {"name": "fashion_mnist", "dataset_name": "zalando-datasets/fashion_mnist", "target_col": "label"},+ # {"name": "sms_spam", "dataset_name": "ucirvine/sms_spam", "target_col": "label", "modality": "text"},+ # {"name": "banking77", "dataset_name": "PolyAI/banking77", "target_col": "label", "modality": "text"},+ # {"name": "gnad10", "dataset_name": "community-datasets/gnad10", "target_col": "label", "modality": "text"},+ {+ "name": "oxford-iiit-pet",+ "dataset_name": "timm/oxford-iiit-pet",+ "target_col": "label_cat_dog",+ "modality": "image",+ },+ {"name": "stanford_cars", "dataset_name": "tanganke/stanford_cars", "target_col": "label", "modality": "image"},+ {+ "name": "fashion_mnist",+ "dataset_name": "zalando-datasets/fashion_mnist",+ "target_col": "label",+ "modality": "image",+ },]@@ -27,41 +37,60 @@ class HFExpDataset(ExpDataset):self.name = nameself.dataset_dir = dataset_dirself.dataset_name = dataset_name+ self.modality = kwargs.get("modality", "")self.target_col = kwargs.get("target_col", "label")- self.dataset = load_dataset(dataset_name)+ self.dataset = load_dataset(self.dataset_name, trust_remote_code=True)super().__init__(self.name, dataset_dir, **kwargs)def get_raw_dataset(self):raw_dir = Path(self.dataset_dir, self.name, "raw")raw_dir.mkdir(parents=True, exist_ok=True)if os.path.exists(Path(raw_dir, "train.csv")):- df = pd.read_csv(Path(raw_dir, "train.csv"))+ df = pd.read_csv(Path(raw_dir, "train.csv"), encoding="utf-8")else:df = self.dataset["train"].to_pandas()- df.to_csv(Path(raw_dir, "train.csv"), index=False)+ df.to_csv(Path(raw_dir, "train.csv"), index=False, encoding="utf-8")if os.path.exists(Path(raw_dir, "test.csv")):- test_df = pd.read_csv(Path(raw_dir, "test.csv"), index=False)+ test_df = pd.read_csv(Path(raw_dir, "test.csv"), encoding="utf-8")else:- if "test" in self.dataset:+ if self.dataset and "test" in self.dataset:test_df = self.dataset["test"].to_pandas()- test_df.to_csv(Path(raw_dir, "test.csv"), index=False)+ test_df.to_csv(Path(raw_dir, "test.csv"), index=False, encoding="utf-8")else:test_df = Nonereturn df, test_df- # def get_df_head(self, raw_df):- # return raw_df.head()+ def get_df_head(self, raw_df):+ if self.modality == "text":+ examples = []+ for i in range(5):+ examples.append(raw_df.iloc[i].to_dict())+ return examples+ elif self.modality == "image":+ return ""++ def get_dataset_info(self):+ dataset_info = super().get_dataset_info()+ dataset = self.dataset+ dataset_info["description"] = dataset["train"].info.description+ return dataset_infoif __name__ == "__main__":dataset_dir = "D:/work/automl/datasets"- save_analysis_pool = True+ save_analysis_pool = False+ force_update = Falsedatasets_dict = {"datasets": {}}solution_designer = SolutionDesigner()for dataset_meta in HFDATSETS:hf_dataset = HFExpDataset(- dataset_meta["name"], dataset_dir, dataset_meta["dataset_name"], target_col=dataset_meta["target_col"]+ dataset_meta["name"],+ dataset_dir,+ dataset_meta["dataset_name"],+ target_col=dataset_meta["target_col"],+ force_update=force_update,+ modality=dataset_meta["modality"],)asyncio.run(process_dataset(hf_dataset, solution_designer, save_analysis_pool, datasets_dict))save_datasets_dict_to_yaml(datasets_dict, "hf_datasets.yaml")diff --git a/expo/datasets.yaml b/expo/datasets.yamlindex 512cbc292..051e8232d 100644--- a/expo/datasets.yaml+++ b/expo/datasets.yaml@@ -79,6 +79,14 @@ datasets:\ column `percent_pell_grant`.\nPerform data analysis, data preprocessing, feature\\ engineering, and modeling to predict the target. \nReport rmse on the eval\\ data. Do not plot or make any visualizations.\n"+ concrete-strength:+ dataset: concrete-strength+ metric: rmse+ target_col: Strength+ user_requirement: "This is a concrete-strength dataset. Your goal is to predict\+ \ the target column `Strength`.\nPerform data analysis, data preprocessing,\+ \ feature engineering, and modeling to predict the target. \nReport rmse on\+ \ the eval data. Do not plot or make any visualizations.\n"credit-g:dataset: credit-gmetric: f1@@ -135,6 +143,22 @@ datasets:\ column `class`.\nPerform data analysis, data preprocessing, feature engineering,\\ and modeling to predict the target. \nReport f1 weighted on the eval data.\\ Do not plot or make any visualizations.\n"+ smoker-status:+ dataset: smoker-status+ metric: f1+ target_col: smoking+ user_requirement: "This is a smoker-status dataset. Your goal is to predict the\+ \ target column `smoking`.\nPerform data analysis, data preprocessing, feature\+ \ engineering, and modeling to predict the target. \nReport f1 on the eval data.\+ \ Do not plot or make any visualizations.\n"+ software-defects:+ dataset: software-defects+ metric: f1+ target_col: defects+ user_requirement: "This is a software-defects dataset. Your goal is to predict\+ \ the target column `defects`.\nPerform data analysis, data preprocessing, feature\+ \ engineering, and modeling to predict the target. \nReport f1 on the eval data.\+ \ Do not plot or make any visualizations.\n"steel-plates-fault:dataset: steel-plates-faultmetric: f1 weighted@@ -151,28 +175,27 @@ datasets:\ the target column `Class`.\nPerform data analysis, data preprocessing, feature\\ engineering, and modeling to predict the target. \nReport f1 weighted on the\\ eval data. Do not plot or make any visualizations.\n"- concrete-strength:- dataset: concrete-strength- metric: rmse- target_col: Strength- user_requirement: "This is a concrete-strength dataset. Your goal is to predict\- \ the target column `Strength`.\nPerform data analysis, data preprocessing,\- \ feature engineering, and modeling to predict the target. \nReport rmse on\- \ the eval data. Do not plot or make any visualizations.\n"- smoker-status:- dataset: smoker-status- metric: f1- target_col: smoking- user_requirement: "This is a smoker-status dataset. Your goal is to predict the\- \ target column `smoking`.\nPerform data analysis, data preprocessing, feature\- \ engineering, and modeling to predict the target. \nReport f1 on the eval data.\+ banking77:+ dataset: banking77+ metric: f1 weighted+ target_col: label+ user_requirement: "This is a banking77 dataset. Your goal is to predict the target\+ \ column `label`.\nPerform data analysis, data preprocessing, feature engineering,\+ \ and modeling to predict the target. \nReport f1 weighted on the eval data.\\ Do not plot or make any visualizations.\n"- software-defects:- dataset: software-defects- metric: f1- target_col: defects- user_requirement: "This is a software-defects dataset. Your goal is to predict\- \ the target column `defects`.\nPerform data analysis, data preprocessing, feature\- \ engineering, and modeling to predict the target. \nReport f1 on the eval data.\+ gnad10:+ dataset: gnad10+ metric: f1 weighted+ target_col: label+ user_requirement: "This is a gnad10 dataset. Your goal is to predict the target\+ \ column `label`.\nPerform data analysis, data preprocessing, feature engineering,\+ \ and modeling to predict the target. \nReport f1 weighted on the eval data.\\ Do not plot or make any visualizations.\n"-+ sms_spam:+ dataset: sms_spam+ metric: f1+ target_col: label+ user_requirement: "This is a sms_spam dataset. Your goal is to predict the target\+ \ column `label`.\nPerform data analysis, data preprocessing, feature engineering,\+ \ and modeling to predict the target. \nReport f1 on the eval data. Do not plot\+ \ or make any visualizations.\n"