feat: kaggle refactor (#489)

* init trail

* Add spec info

* auto unzip mlebench prepared data for out scenario

* successfully run example

* successfully run main

* simplify load traing

* extract load_from_raw_data

* split the fies(still buggy)
It should stop on ~20 epoch and reach the end

* some changes

* Fix bug to run example

* (success) until feature

* refine model and ensemble

* add metrics in ens.py

* update README &  spec.md

* ens change

* fix ens bug

* Delete rdagent/scenarios/kaggle/tpl_ex/aerial-cactus-identification/train.py

* add template_path in KG_conf

* fix test kaggle

* CI

* make test_import not check kaggle template codes

---------

Co-authored-by: Bowen Xian <xianbowen@outlook.com>

rdagent/app/kaggle/conf.py

@@ -44,6 +44,9 @@ class KaggleBasePropSetting(BasePropSetting):
     competition: str = ""
     """Kaggle competition name, e.g., 'sf-crime'"""
 
+    template_path: str = "rdagent/scenarios/kaggle/experiment/templates"
+    """Kaggle competition base templates path"""
+
     local_data_path: str = ""
     """Folder storing Kaggle competition data"""
 

rdagent/scenarios/kaggle/experiment/kaggle_experiment.py

@@ -38,7 +38,9 @@ class KGModelExperiment(ModelExperiment[ModelTask, KGFBWorkspace, ModelFBWorkspa
     def __init__(self, *args, source_feature_size: int = None, **kwargs) -> None:
         super().__init__(*args, **kwargs)
         self.experiment_workspace = KGFBWorkspace(
-            template_folder_path=Path(__file__).parent / f"{KAGGLE_IMPLEMENT_SETTING.competition}_template"
+            template_folder_path=Path(__file__).resolve()
+            / Path(KAGGLE_IMPLEMENT_SETTING.template_path).resolve()
+            / KAGGLE_IMPLEMENT_SETTING.competition
         )
         if len(self.based_experiments) > 0:
             self.experiment_workspace.inject_code(**self.based_experiments[-1].experiment_workspace.code_dict)
@@ -62,7 +64,9 @@ class KGFactorExperiment(FeatureExperiment[FactorTask, KGFBWorkspace, FactorFBWo
     def __init__(self, *args, source_feature_size: int = None, **kwargs) -> None:
         super().__init__(*args, **kwargs)
         self.experiment_workspace = KGFBWorkspace(
-            template_folder_path=Path(__file__).parent / f"{KAGGLE_IMPLEMENT_SETTING.competition}_template"
+            template_folder_path=Path(__file__).resolve()
+            / Path(KAGGLE_IMPLEMENT_SETTING.template_path).resolve()
+            / KAGGLE_IMPLEMENT_SETTING.competition
         )
         if len(self.based_experiments) > 0:
             self.experiment_workspace.inject_code(**self.based_experiments[-1].experiment_workspace.code_dict)