ins-pricing 0.2.9__py3-none-any.whl → 0.3.1__py3-none-any.whl

ins_pricing/modelling/core/bayesopt/config_components.py

@@ -0,0 +1,351 @@
+"""Nested configuration components for BayesOptConfig.
+
+This module provides focused configuration dataclasses that group related settings
+together, improving maintainability and reducing the cognitive load of the main
+BayesOptConfig class.
+
+Usage:
+    config = BayesOptConfig(
+        model_nme="pricing_model",
+        resp_nme="claim",
+        weight_nme="exposure",
+        factor_nmes=["age", "gender"],
+        distributed=DistributedConfig(use_ft_ddp=True),
+        gnn=GNNConfig(use_approx_knn=False),
+    )
+"""
+
+from __future__ import annotations
+
+from dataclasses import dataclass, field
+from typing import Any, Dict, List, Optional
+
+
+@dataclass
+class DistributedConfig:
+    """Configuration for distributed training (DDP/DataParallel).
+
+    Attributes:
+        use_resn_data_parallel: Use DataParallel for ResNet
+        use_ft_data_parallel: Use DataParallel for FT-Transformer
+        use_gnn_data_parallel: Use DataParallel for GNN
+        use_resn_ddp: Use DistributedDataParallel for ResNet
+        use_ft_ddp: Use DistributedDataParallel for FT-Transformer
+        use_gnn_ddp: Use DistributedDataParallel for GNN
+    """
+
+    use_resn_data_parallel: bool = False
+    use_ft_data_parallel: bool = False
+    use_gnn_data_parallel: bool = False
+    use_resn_ddp: bool = False
+    use_ft_ddp: bool = False
+    use_gnn_ddp: bool = False
+
+    @classmethod
+    def from_flat_dict(cls, d: Dict[str, Any]) -> "DistributedConfig":
+        """Create from a flat dictionary with prefixed keys."""
+        return cls(
+            use_resn_data_parallel=bool(d.get("use_resn_data_parallel", False)),
+            use_ft_data_parallel=bool(d.get("use_ft_data_parallel", False)),
+            use_gnn_data_parallel=bool(d.get("use_gnn_data_parallel", False)),
+            use_resn_ddp=bool(d.get("use_resn_ddp", False)),
+            use_ft_ddp=bool(d.get("use_ft_ddp", False)),
+            use_gnn_ddp=bool(d.get("use_gnn_ddp", False)),
+        )
+
+
+@dataclass
+class GNNConfig:
+    """Configuration for Graph Neural Network training.
+
+    Attributes:
+        use_approx_knn: Use approximate k-NN for graph construction
+        approx_knn_threshold: Row count threshold for approximate k-NN
+        graph_cache: Path to cache/load adjacency matrix
+        max_gpu_knn_nodes: Max nodes for GPU k-NN construction
+        knn_gpu_mem_ratio: Fraction of GPU memory for k-NN
+        knn_gpu_mem_overhead: Temporary memory overhead multiplier
+    """
+
+    use_approx_knn: bool = True
+    approx_knn_threshold: int = 50000
+    graph_cache: Optional[str] = None
+    max_gpu_knn_nodes: int = 200000
+    knn_gpu_mem_ratio: float = 0.9
+    knn_gpu_mem_overhead: float = 2.0
+
+    @classmethod
+    def from_flat_dict(cls, d: Dict[str, Any]) -> "GNNConfig":
+        """Create from a flat dictionary with prefixed keys."""
+        return cls(
+            use_approx_knn=bool(d.get("gnn_use_approx_knn", True)),
+            approx_knn_threshold=int(d.get("gnn_approx_knn_threshold", 50000)),
+            graph_cache=d.get("gnn_graph_cache"),
+            max_gpu_knn_nodes=int(d.get("gnn_max_gpu_knn_nodes", 200000)),
+            knn_gpu_mem_ratio=float(d.get("gnn_knn_gpu_mem_ratio", 0.9)),
+            knn_gpu_mem_overhead=float(d.get("gnn_knn_gpu_mem_overhead", 2.0)),
+        )
+
+
+@dataclass
+class GeoTokenConfig:
+    """Configuration for geographic token embeddings.
+
+    Attributes:
+        feature_nmes: Feature column names for geo tokens
+        hidden_dim: Hidden dimension for geo token network
+        layers: Number of layers in geo token network
+        dropout: Dropout rate
+        k_neighbors: Number of neighbors for geo tokens
+        learning_rate: Learning rate for geo token training
+        epochs: Training epochs for geo tokens
+    """
+
+    feature_nmes: Optional[List[str]] = None
+    hidden_dim: int = 32
+    layers: int = 2
+    dropout: float = 0.1
+    k_neighbors: int = 10
+    learning_rate: float = 1e-3
+    epochs: int = 50
+
+    @classmethod
+    def from_flat_dict(cls, d: Dict[str, Any]) -> "GeoTokenConfig":
+        """Create from a flat dictionary with prefixed keys."""
+        return cls(
+            feature_nmes=d.get("geo_feature_nmes"),
+            hidden_dim=int(d.get("geo_token_hidden_dim", 32)),
+            layers=int(d.get("geo_token_layers", 2)),
+            dropout=float(d.get("geo_token_dropout", 0.1)),
+            k_neighbors=int(d.get("geo_token_k_neighbors", 10)),
+            learning_rate=float(d.get("geo_token_learning_rate", 1e-3)),
+            epochs=int(d.get("geo_token_epochs", 50)),
+        )
+
+
+@dataclass
+class RegionConfig:
+    """Configuration for region/geographic effects.
+
+    Attributes:
+        province_col: Column name for province/state
+        city_col: Column name for city
+        effect_alpha: Regularization alpha for region effects
+    """
+
+    province_col: Optional[str] = None
+    city_col: Optional[str] = None
+    effect_alpha: float = 50.0
+
+    @classmethod
+    def from_flat_dict(cls, d: Dict[str, Any]) -> "RegionConfig":
+        """Create from a flat dictionary with prefixed keys."""
+        return cls(
+            province_col=d.get("region_province_col"),
+            city_col=d.get("region_city_col"),
+            effect_alpha=float(d.get("region_effect_alpha", 50.0)),
+        )
+
+
+@dataclass
+class FTTransformerConfig:
+    """Configuration for FT-Transformer model.
+
+    Attributes:
+        role: Model role ('model', 'embedding', 'unsupervised_embedding')
+        feature_prefix: Prefix for generated embedding features
+        num_numeric_tokens: Number of numeric tokens
+    """
+
+    role: str = "model"
+    feature_prefix: str = "ft_emb"
+    num_numeric_tokens: Optional[int] = None
+
+    @classmethod
+    def from_flat_dict(cls, d: Dict[str, Any]) -> "FTTransformerConfig":
+        """Create from a flat dictionary with prefixed keys."""
+        return cls(
+            role=str(d.get("ft_role", "model")),
+            feature_prefix=str(d.get("ft_feature_prefix", "ft_emb")),
+            num_numeric_tokens=d.get("ft_num_numeric_tokens"),
+        )
+
+
+@dataclass
+class XGBoostConfig:
+    """Configuration for XGBoost model.
+
+    Attributes:
+        max_depth_max: Maximum tree depth for hyperparameter tuning
+        n_estimators_max: Maximum number of estimators for tuning
+    """
+
+    max_depth_max: int = 25
+    n_estimators_max: int = 500
+
+    @classmethod
+    def from_flat_dict(cls, d: Dict[str, Any]) -> "XGBoostConfig":
+        """Create from a flat dictionary with prefixed keys."""
+        return cls(
+            max_depth_max=int(d.get("xgb_max_depth_max", 25)),
+            n_estimators_max=int(d.get("xgb_n_estimators_max", 500)),
+        )
+
+
+@dataclass
+class CVConfig:
+    """Configuration for cross-validation.
+
+    Attributes:
+        strategy: CV strategy ('random', 'group', 'time', 'stratified')
+        splits: Number of CV splits
+        group_col: Column for group-based CV
+        time_col: Column for time-based CV
+        time_ascending: Whether to sort time ascending
+    """
+
+    strategy: str = "random"
+    splits: Optional[int] = None
+    group_col: Optional[str] = None
+    time_col: Optional[str] = None
+    time_ascending: bool = True
+
+    @classmethod
+    def from_flat_dict(cls, d: Dict[str, Any]) -> "CVConfig":
+        """Create from a flat dictionary with prefixed keys."""
+        return cls(
+            strategy=str(d.get("cv_strategy", "random")),
+            splits=d.get("cv_splits"),
+            group_col=d.get("cv_group_col"),
+            time_col=d.get("cv_time_col"),
+            time_ascending=bool(d.get("cv_time_ascending", True)),
+        )
+
+
+@dataclass
+class FTOOFConfig:
+    """Configuration for FT-Transformer out-of-fold predictions.
+
+    Attributes:
+        folds: Number of OOF folds
+        strategy: OOF strategy
+        shuffle: Whether to shuffle data
+    """
+
+    folds: Optional[int] = None
+    strategy: Optional[str] = None
+    shuffle: bool = True
+
+    @classmethod
+    def from_flat_dict(cls, d: Dict[str, Any]) -> "FTOOFConfig":
+        """Create from a flat dictionary with prefixed keys."""
+        return cls(
+            folds=d.get("ft_oof_folds"),
+            strategy=d.get("ft_oof_strategy"),
+            shuffle=bool(d.get("ft_oof_shuffle", True)),
+        )
+
+
+@dataclass
+class OutputConfig:
+    """Configuration for output and caching.
+
+    Attributes:
+        output_dir: Base output directory
+        optuna_storage: Optuna study storage path
+        optuna_study_prefix: Prefix for Optuna study names
+        best_params_files: Mapping of trainer keys to param files
+        save_preprocess: Whether to save preprocessing artifacts
+        preprocess_artifact_path: Path for preprocessing artifacts
+        plot_path_style: Plot path style ('nested' or 'flat')
+        cache_predictions: Whether to cache predictions
+        prediction_cache_dir: Directory for prediction cache
+        prediction_cache_format: Format for prediction cache ('parquet' or 'csv')
+    """
+
+    output_dir: Optional[str] = None
+    optuna_storage: Optional[str] = None
+    optuna_study_prefix: Optional[str] = None
+    best_params_files: Optional[Dict[str, str]] = None
+    save_preprocess: bool = False
+    preprocess_artifact_path: Optional[str] = None
+    plot_path_style: str = "nested"
+    cache_predictions: bool = False
+    prediction_cache_dir: Optional[str] = None
+    prediction_cache_format: str = "parquet"
+
+    @classmethod
+    def from_flat_dict(cls, d: Dict[str, Any]) -> "OutputConfig":
+        """Create from a flat dictionary with prefixed keys."""
+        return cls(
+            output_dir=d.get("output_dir"),
+            optuna_storage=d.get("optuna_storage"),
+            optuna_study_prefix=d.get("optuna_study_prefix"),
+            best_params_files=d.get("best_params_files"),
+            save_preprocess=bool(d.get("save_preprocess", False)),
+            preprocess_artifact_path=d.get("preprocess_artifact_path"),
+            plot_path_style=str(d.get("plot_path_style", "nested")),
+            cache_predictions=bool(d.get("cache_predictions", False)),
+            prediction_cache_dir=d.get("prediction_cache_dir"),
+            prediction_cache_format=str(d.get("prediction_cache_format", "parquet")),
+        )
+
+
+@dataclass
+class EnsembleConfig:
+    """Configuration for ensemble training.
+
+    Attributes:
+        final_ensemble: Whether to use final ensemble
+        final_ensemble_k: Number of models in ensemble
+        final_refit: Whether to refit after ensemble
+    """
+
+    final_ensemble: bool = False
+    final_ensemble_k: int = 3
+    final_refit: bool = True
+
+    @classmethod
+    def from_flat_dict(cls, d: Dict[str, Any]) -> "EnsembleConfig":
+        """Create from a flat dictionary with prefixed keys."""
+        return cls(
+            final_ensemble=bool(d.get("final_ensemble", False)),
+            final_ensemble_k=int(d.get("final_ensemble_k", 3)),
+            final_refit=bool(d.get("final_refit", True)),
+        )
+
+
+@dataclass
+class TrainingConfig:
+    """Core training configuration.
+
+    Attributes:
+        prop_test: Proportion of data for validation
+        rand_seed: Random seed for reproducibility
+        epochs: Number of training epochs
+        use_gpu: Whether to use GPU
+        reuse_best_params: Whether to reuse best params
+        resn_weight_decay: Weight decay for ResNet
+        bo_sample_limit: Sample limit for Bayesian optimization
+    """
+
+    prop_test: float = 0.25
+    rand_seed: Optional[int] = None
+    epochs: int = 100
+    use_gpu: bool = True
+    reuse_best_params: bool = False
+    resn_weight_decay: float = 1e-4
+    bo_sample_limit: Optional[int] = None
+
+    @classmethod
+    def from_flat_dict(cls, d: Dict[str, Any]) -> "TrainingConfig":
+        """Create from a flat dictionary with prefixed keys."""
+        return cls(
+            prop_test=float(d.get("prop_test", 0.25)),
+            rand_seed=d.get("rand_seed"),
+            epochs=int(d.get("epochs", 100)),
+            use_gpu=bool(d.get("use_gpu", True)),
+            reuse_best_params=bool(d.get("reuse_best_params", False)),
+            resn_weight_decay=float(d.get("resn_weight_decay", 1e-4)),
+            bo_sample_limit=d.get("bo_sample_limit"),
+        )

ins_pricing/modelling/core/bayesopt/config_preprocess.py

@@ -366,10 +366,9 @@ class DatasetPreprocessor:
     def __init__(self, train_df: pd.DataFrame, test_df: pd.DataFrame,
                  config: BayesOptConfig) -> None:
         self.config = config
-        # Use shallow copy to avoid unnecessary memory overhead
-        # Deep copies only made when actually modifying data
-        self.train_data = train_df.copy(deep=False)
-        self.test_data = test_df.copy(deep=False)
+        # Copy inputs to avoid mutating caller-provided DataFrames.
+        self.train_data = train_df.copy()
+        self.test_data = test_df.copy()
         self.num_features: List[str] = []
         self.train_oht_data: Optional[pd.DataFrame] = None
         self.test_oht_data: Optional[pd.DataFrame] = None

ins_pricing/modelling/core/bayesopt/core.py

@@ -48,7 +48,10 @@ class _CVSplitter:
 # =============================================================================
 class BayesOptModel(BayesOptPlottingMixin, BayesOptExplainMixin):
     def __init__(self, train_data, test_data,
-                 model_nme, resp_nme, weight_nme, factor_nmes: Optional[List[str]] = None, task_type='regression',
+                 config: Optional[BayesOptConfig] = None,
+                 # Backward compatibility: individual parameters (DEPRECATED)
+                 model_nme=None, resp_nme=None, weight_nme=None,
+                 factor_nmes: Optional[List[str]] = None, task_type='regression',
                  binary_resp_nme=None,
                  cate_list=None, prop_test=0.25, rand_seed=None,
                  epochs=100, use_gpu=True,
@@ -108,6 +111,10 @@ class BayesOptModel(BayesOptPlottingMixin, BayesOptExplainMixin):
         Args:
             train_data: Training DataFrame.
             test_data: Test DataFrame.
+            config: BayesOptConfig instance with all configuration (RECOMMENDED).
+                If provided, all other parameters are ignored.
+
+            # DEPRECATED: Individual parameters (use config instead)
             model_nme: Model name prefix used in outputs.
             resp_nme: Target column name.
             weight_nme: Sample weight column name.
@@ -136,101 +143,153 @@ class BayesOptModel(BayesOptPlottingMixin, BayesOptExplainMixin):
             final_ensemble: Enable k-fold model averaging at the final stage.
             final_ensemble_k: Number of folds for averaging.
             final_refit: Refit on full data using best stopping point.
+
+        Examples:
+            # New style (recommended):
+            config = BayesOptConfig(
+                model_nme="my_model",
+                resp_nme="target",
+                weight_nme="weight",
+                factor_nmes=["feat1", "feat2"]
+            )
+            model = BayesOptModel(train_df, test_df, config=config)
+
+            # Old style (deprecated, for backward compatibility):
+            model = BayesOptModel(
+                train_df, test_df,
+                model_nme="my_model",
+                resp_nme="target",
+                weight_nme="weight",
+                factor_nmes=["feat1", "feat2"]
+            )
         """
-        inferred_factors, inferred_cats = infer_factor_and_cate_list(
-            train_df=train_data,
-            test_df=test_data,
-            resp_nme=resp_nme,
-            weight_nme=weight_nme,
-            binary_resp_nme=binary_resp_nme,
-            factor_nmes=factor_nmes,
-            cate_list=cate_list,
-            infer_categorical_max_unique=int(infer_categorical_max_unique),
-            infer_categorical_max_ratio=float(infer_categorical_max_ratio),
-        )
+        # Detect which API is being used
+        if config is not None:
+            # New API: config object provided
+            if isinstance(config, BayesOptConfig):
+                cfg = config
+            else:
+                raise TypeError(
+                    f"config must be a BayesOptConfig instance, got {type(config).__name__}"
+                )
+        else:
+            # Old API: individual parameters (backward compatibility)
+            # Show deprecation warning
+            import warnings
+            warnings.warn(
+                "Passing individual parameters to BayesOptModel.__init__ is deprecated. "
+                "Use the 'config' parameter with a BayesOptConfig instance instead:\n"
+                "  config = BayesOptConfig(model_nme=..., resp_nme=..., ...)\n"
+                "  model = BayesOptModel(train_data, test_data, config=config)\n"
+                "Individual parameters will be removed in v0.4.0.",
+                DeprecationWarning,
+                stacklevel=2
+            )
 
-        cfg = BayesOptConfig(
-            model_nme=model_nme,
-            task_type=task_type,
-            resp_nme=resp_nme,
-            weight_nme=weight_nme,
-            factor_nmes=list(inferred_factors),
-            binary_resp_nme=binary_resp_nme,
-            cate_list=list(inferred_cats) if inferred_cats else None,
-            prop_test=prop_test,
-            rand_seed=rand_seed,
-            epochs=epochs,
-            use_gpu=use_gpu,
-            xgb_max_depth_max=int(xgb_max_depth_max),
-            xgb_n_estimators_max=int(xgb_n_estimators_max),
-            use_resn_data_parallel=use_resn_data_parallel,
-            use_ft_data_parallel=use_ft_data_parallel,
-            use_resn_ddp=use_resn_ddp,
-            use_gnn_data_parallel=use_gnn_data_parallel,
-            use_ft_ddp=use_ft_ddp,
-            use_gnn_ddp=use_gnn_ddp,
-            gnn_use_approx_knn=gnn_use_approx_knn,
-            gnn_approx_knn_threshold=gnn_approx_knn_threshold,
-            gnn_graph_cache=gnn_graph_cache,
-            gnn_max_gpu_knn_nodes=gnn_max_gpu_knn_nodes,
-            gnn_knn_gpu_mem_ratio=gnn_knn_gpu_mem_ratio,
-            gnn_knn_gpu_mem_overhead=gnn_knn_gpu_mem_overhead,
-            output_dir=output_dir,
-            optuna_storage=optuna_storage,
-            optuna_study_prefix=optuna_study_prefix,
-            best_params_files=best_params_files,
-            ft_role=str(ft_role or "model"),
-            ft_feature_prefix=str(ft_feature_prefix or "ft_emb"),
-            ft_num_numeric_tokens=ft_num_numeric_tokens,
-            reuse_best_params=bool(reuse_best_params),
-            resn_weight_decay=float(resn_weight_decay)
-            if resn_weight_decay is not None
-            else 1e-4,
-            final_ensemble=bool(final_ensemble),
-            final_ensemble_k=int(final_ensemble_k),
-            final_refit=bool(final_refit),
-            cv_strategy=str(cv_strategy or "random"),
-            cv_splits=cv_splits,
-            cv_group_col=cv_group_col,
-            cv_time_col=cv_time_col,
-            cv_time_ascending=bool(cv_time_ascending),
-            ft_oof_folds=ft_oof_folds,
-            ft_oof_strategy=ft_oof_strategy,
-            ft_oof_shuffle=bool(ft_oof_shuffle),
-            save_preprocess=bool(save_preprocess),
-            preprocess_artifact_path=preprocess_artifact_path,
-            plot_path_style=str(plot_path_style or "nested"),
-            bo_sample_limit=bo_sample_limit,
-            cache_predictions=bool(cache_predictions),
-            prediction_cache_dir=prediction_cache_dir,
-            prediction_cache_format=str(prediction_cache_format or "parquet"),
-            region_province_col=region_province_col,
-            region_city_col=region_city_col,
-            region_effect_alpha=float(region_effect_alpha)
-            if region_effect_alpha is not None
-            else 50.0,
-            geo_feature_nmes=list(geo_feature_nmes)
-            if geo_feature_nmes is not None
-            else None,
-            geo_token_hidden_dim=int(geo_token_hidden_dim)
-            if geo_token_hidden_dim is not None
-            else 32,
-            geo_token_layers=int(geo_token_layers)
-            if geo_token_layers is not None
-            else 2,
-            geo_token_dropout=float(geo_token_dropout)
-            if geo_token_dropout is not None
-            else 0.1,
-            geo_token_k_neighbors=int(geo_token_k_neighbors)
-            if geo_token_k_neighbors is not None
-            else 10,
-            geo_token_learning_rate=float(geo_token_learning_rate)
-            if geo_token_learning_rate is not None
-            else 1e-3,
-            geo_token_epochs=int(geo_token_epochs)
-            if geo_token_epochs is not None
-            else 50,
-        )
+            # Validate required parameters
+            if model_nme is None:
+                raise ValueError("model_nme is required when not using config parameter")
+            if resp_nme is None:
+                raise ValueError("resp_nme is required when not using config parameter")
+            if weight_nme is None:
+                raise ValueError("weight_nme is required when not using config parameter")
+
+            # Infer categorical features if needed
+            inferred_factors, inferred_cats = infer_factor_and_cate_list(
+                train_df=train_data,
+                test_df=test_data,
+                resp_nme=resp_nme,
+                weight_nme=weight_nme,
+                binary_resp_nme=binary_resp_nme,
+                factor_nmes=factor_nmes,
+                cate_list=cate_list,
+                infer_categorical_max_unique=int(infer_categorical_max_unique),
+                infer_categorical_max_ratio=float(infer_categorical_max_ratio),
+            )
+
+            # Construct config from individual parameters
+            cfg = BayesOptConfig(
+                model_nme=model_nme,
+                task_type=task_type,
+                resp_nme=resp_nme,
+                weight_nme=weight_nme,
+                factor_nmes=list(inferred_factors),
+                binary_resp_nme=binary_resp_nme,
+                cate_list=list(inferred_cats) if inferred_cats else None,
+                prop_test=prop_test,
+                rand_seed=rand_seed,
+                epochs=epochs,
+                use_gpu=use_gpu,
+                xgb_max_depth_max=int(xgb_max_depth_max),
+                xgb_n_estimators_max=int(xgb_n_estimators_max),
+                use_resn_data_parallel=use_resn_data_parallel,
+                use_ft_data_parallel=use_ft_data_parallel,
+                use_resn_ddp=use_resn_ddp,
+                use_gnn_data_parallel=use_gnn_data_parallel,
+                use_ft_ddp=use_ft_ddp,
+                use_gnn_ddp=use_gnn_ddp,
+                gnn_use_approx_knn=gnn_use_approx_knn,
+                gnn_approx_knn_threshold=gnn_approx_knn_threshold,
+                gnn_graph_cache=gnn_graph_cache,
+                gnn_max_gpu_knn_nodes=gnn_max_gpu_knn_nodes,
+                gnn_knn_gpu_mem_ratio=gnn_knn_gpu_mem_ratio,
+                gnn_knn_gpu_mem_overhead=gnn_knn_gpu_mem_overhead,
+                output_dir=output_dir,
+                optuna_storage=optuna_storage,
+                optuna_study_prefix=optuna_study_prefix,
+                best_params_files=best_params_files,
+                ft_role=str(ft_role or "model"),
+                ft_feature_prefix=str(ft_feature_prefix or "ft_emb"),
+                ft_num_numeric_tokens=ft_num_numeric_tokens,
+                reuse_best_params=bool(reuse_best_params),
+                resn_weight_decay=float(resn_weight_decay)
+                if resn_weight_decay is not None
+                else 1e-4,
+                final_ensemble=bool(final_ensemble),
+                final_ensemble_k=int(final_ensemble_k),
+                final_refit=bool(final_refit),
+                cv_strategy=str(cv_strategy or "random"),
+                cv_splits=cv_splits,
+                cv_group_col=cv_group_col,
+                cv_time_col=cv_time_col,
+                cv_time_ascending=bool(cv_time_ascending),
+                ft_oof_folds=ft_oof_folds,
+                ft_oof_strategy=ft_oof_strategy,
+                ft_oof_shuffle=bool(ft_oof_shuffle),
+                save_preprocess=bool(save_preprocess),
+                preprocess_artifact_path=preprocess_artifact_path,
+                plot_path_style=str(plot_path_style or "nested"),
+                bo_sample_limit=bo_sample_limit,
+                cache_predictions=bool(cache_predictions),
+                prediction_cache_dir=prediction_cache_dir,
+                prediction_cache_format=str(prediction_cache_format or "parquet"),
+                region_province_col=region_province_col,
+                region_city_col=region_city_col,
+                region_effect_alpha=float(region_effect_alpha)
+                if region_effect_alpha is not None
+                else 50.0,
+                geo_feature_nmes=list(geo_feature_nmes)
+                if geo_feature_nmes is not None
+                else None,
+                geo_token_hidden_dim=int(geo_token_hidden_dim)
+                if geo_token_hidden_dim is not None
+                else 32,
+                geo_token_layers=int(geo_token_layers)
+                if geo_token_layers is not None
+                else 2,
+                geo_token_dropout=float(geo_token_dropout)
+                if geo_token_dropout is not None
+                else 0.1,
+                geo_token_k_neighbors=int(geo_token_k_neighbors)
+                if geo_token_k_neighbors is not None
+                else 10,
+                geo_token_learning_rate=float(geo_token_learning_rate)
+                if geo_token_learning_rate is not None
+                else 1e-3,
+                geo_token_epochs=int(geo_token_epochs)
+                if geo_token_epochs is not None
+                else 50,
+            )
         self.config = cfg
         self.model_nme = cfg.model_nme
         self.task_type = cfg.task_type