ins-pricing 0.1.9__tar.gz → 0.2.0__tar.gz

{ins_pricing-0.1.9 → ins_pricing-0.2.0}/MANIFEST.in

@@ -1,3 +1,3 @@
 include README.md
 recursive-include ins_pricing *.md
-recursive-exclude ins_pricing/modelling/demo *
+recursive-exclude ins_pricing/examples *

{ins_pricing-0.1.9/Ins_Pricing.egg-info → ins_pricing-0.2.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: ins_pricing
-Version: 0.1.9
+Version: 0.2.0
 Summary: Reusable modelling, pricing, governance, and reporting utilities.
 Author: meishi125478
 License: Proprietary
@@ -64,7 +64,7 @@ This repository contains risk modeling and optimization notebooks, scripts, and
 - `ins_pricing/`: reusable training framework and CLI tools (BayesOpt subpackage)
 - `user_packages legacy/`: historical snapshot
 
-Note: `ins_pricing/modelling/demo/` is kept in the repo only and is not shipped in the PyPI package.
+Note: `ins_pricing/modelling/examples/` is kept in the repo only and is not shipped in the PyPI package.
 
 ## Quickstart
 
@@ -84,8 +84,8 @@ jupyter lab
 
 ## BayesOpt entry points
 
-- CLI batch training: `python ins_pricing/modelling/BayesOpt_entry.py --config-json <path>`
-- Incremental training: `python ins_pricing/modelling/BayesOpt_incremental.py --config-json <path>`
+- CLI batch training: `python ins_pricing/modelling/cli/BayesOpt_entry.py --config-json <path>`
+- Incremental training: `python ins_pricing/modelling/cli/BayesOpt_incremental.py --config-json <path>`
 - Python API: `from ins_pricing.modelling import BayesOptModel`
 
 ## Tests

{ins_pricing-0.1.9 → ins_pricing-0.2.0}/README.md

@@ -12,7 +12,7 @@ This repository contains risk modeling and optimization notebooks, scripts, and
 - `ins_pricing/`: reusable training framework and CLI tools (BayesOpt subpackage)
 - `user_packages legacy/`: historical snapshot
 
-Note: `ins_pricing/modelling/demo/` is kept in the repo only and is not shipped in the PyPI package.
+Note: `ins_pricing/modelling/examples/` is kept in the repo only and is not shipped in the PyPI package.
 
 ## Quickstart
 
@@ -32,8 +32,8 @@ jupyter lab
 
 ## BayesOpt entry points
 
-- CLI batch training: `python ins_pricing/modelling/BayesOpt_entry.py --config-json <path>`
-- Incremental training: `python ins_pricing/modelling/BayesOpt_incremental.py --config-json <path>`
+- CLI batch training: `python ins_pricing/modelling/cli/BayesOpt_entry.py --config-json <path>`
+- Incremental training: `python ins_pricing/modelling/cli/BayesOpt_incremental.py --config-json <path>`
 - Python API: `from ins_pricing.modelling import BayesOptModel`
 
 ## Tests

{ins_pricing-0.1.9 → ins_pricing-0.2.0}/ins_pricing/README.md

@@ -7,13 +7,13 @@ between modelling, production, governance, and reporting.
 
 ## Architecture
 
+- `cli/`: CLI entry points + shared utilities.
 - `modelling/`
-  - `bayesopt/`: BayesOpt training core (GLM / XGB / ResNet / FT / GNN).
+  - `core/`: BayesOpt training core (GLM / XGB / ResNet / FT / GNN).
   - `plotting/`: model-agnostic curves and geo visualizations.
   - `explain/`: permutation, gradients, and SHAP helpers.
-  - `BayesOpt_entry.py`: CLI runner for batch training.
-  - `BayesOpt_incremental.py`: CLI for incremental runs.
-  - `Pricing_Run.py`: lightweight pricing orchestration.
+- `docs/modelling/`: modelling documentation.
+- `examples/modelling/`: demo configs + notebooks (repo only; not packaged).
 - `pricing/`: factor tables, calibration, exposure, monitoring.
 - `production/`: scoring, metrics, drift/PSI.
 - `governance/`: registry, release, audit, approval workflow.
@@ -23,7 +23,7 @@ between modelling, production, governance, and reporting.
 
 1. Model training
    - Python API: `from ins_pricing.modelling import BayesOptModel`
-   - CLI: `python ins_pricing/modelling/BayesOpt_entry.py --config-json ...`
+   - CLI: `python ins_pricing/cli/BayesOpt_entry.py --config-json ...`
 2. Evaluation & visualization
    - Curves: `from ins_pricing.plotting import curves`
    - Importance: `from ins_pricing.plotting import importance`
@@ -42,7 +42,10 @@ between modelling, production, governance, and reporting.
 
 - `ins_pricing` exposes lightweight lazy imports so that `pricing/production/governance`
   can be used without installing heavy ML dependencies.
-- Demo notebooks/configs live in the repo under `ins_pricing/modelling/demo/` and are not shipped in the PyPI package.
+- Migration note: CLI entry points now live under `ins_pricing/cli/` and demo assets are under
+  `ins_pricing/examples/modelling/`. Update any scripts that referenced `ins_pricing/modelling/cli/*` or
+  `ins_pricing/modelling/examples/*`.
+- Demo notebooks/configs live in the repo under `ins_pricing/examples/modelling/` and are not shipped in the PyPI package.
 - Heavy dependencies are only required when you import or use the related modules:
   - BayesOpt: `torch`, `optuna`, `xgboost`, etc.
   - Explain: `torch` (gradients), `shap` (SHAP).

{ins_pricing-0.1.9 → ins_pricing-0.2.0}/ins_pricing/__init__.py

@@ -22,22 +22,14 @@ _MODELLING_EXPORTS = {
 }
 
 _LAZY_SUBMODULES = {
-    "bayesopt": "ins_pricing.modelling.bayesopt",
+    "bayesopt": "ins_pricing.modelling.core.bayesopt",
     "plotting": "ins_pricing.modelling.plotting",
     "explain": "ins_pricing.modelling.explain",
-    "BayesOpt": "ins_pricing.modelling.BayesOpt",
-    "BayesOpt_entry": "ins_pricing.modelling.BayesOpt_entry",
-    "BayesOpt_incremental": "ins_pricing.modelling.BayesOpt_incremental",
-    "Explain_entry": "ins_pricing.modelling.Explain_entry",
-    "Explain_Run": "ins_pricing.modelling.Explain_Run",
-    "Pricing_Run": "ins_pricing.modelling.Pricing_Run",
-    "cli_common": "ins_pricing.modelling.cli_common",
-    "notebook_utils": "ins_pricing.modelling.notebook_utils",
-    "watchdog_run": "ins_pricing.modelling.watchdog_run",
+    "BayesOpt": "ins_pricing.modelling.core.BayesOpt",
 }
 
 _PACKAGE_PATHS = {
-    "bayesopt": Path(__file__).resolve().parent / "modelling" / "bayesopt",
+    "bayesopt": Path(__file__).resolve().parent / "modelling" / "core" / "bayesopt",
     "plotting": Path(__file__).resolve().parent / "modelling" / "plotting",
     "explain": Path(__file__).resolve().parent / "modelling" / "explain",
 }

ins_pricing-0.2.0/ins_pricing/cli/BayesOpt_entry.py

@@ -0,0 +1,24 @@
+"""Thin wrapper for the BayesOpt CLI entry point.
+
+The main implementation lives in bayesopt_entry_runner.py.
+"""
+
+from __future__ import annotations
+
+from pathlib import Path
+import sys
+
+if __package__ in {None, ""}:
+    repo_root = Path(__file__).resolve().parents[2]
+    if str(repo_root) not in sys.path:
+        sys.path.insert(0, str(repo_root))
+
+try:
+    from .bayesopt_entry_runner import main
+except Exception:  # pragma: no cover
+    from ins_pricing.cli.bayesopt_entry_runner import main
+
+__all__ = ["main"]
+
+if __name__ == "__main__":
+    main()

{ins_pricing-0.1.9/ins_pricing/modelling → ins_pricing-0.2.0/ins_pricing/cli}/BayesOpt_incremental.py

@@ -8,8 +8,8 @@ of per-model incremental CSVs or a single incremental file when updating
 one dataset.
 
 Example:
-    python ins_pricing/modelling/BayesOpt_incremental.py \
-        --config-json ins_pricing/modelling/demo/config_incremental_template.json \
+    python ins_pricing/cli/BayesOpt_incremental.py \
+        --config-json ins_pricing/examples/modelling/config_incremental_template.json \
         --incremental-dir ./incremental_batches \
         --merge-keys policy_id vehicle_id \
         --model-keys glm xgb resn --plot-curves
@@ -17,80 +17,114 @@ Example:
 
 from __future__ import annotations
 
+from pathlib import Path
+import sys
+
+if __package__ in {None, ""}:
+    repo_root = Path(__file__).resolve().parents[2]
+    if str(repo_root) not in sys.path:
+        sys.path.insert(0, str(repo_root))
+
 import argparse
 import json
 from dataclasses import asdict
 from datetime import datetime
-from pathlib import Path
 from typing import Any, Dict, List, Optional, Sequence, Tuple
 
 import pandas as pd
-from sklearn.model_selection import train_test_split
 
 try:
-    from . import bayesopt as ropt  # type: ignore
-    from .cli_common import (  # type: ignore
+    from .. import bayesopt as ropt  # type: ignore
+    from .utils.cli_common import (  # type: ignore
         PLOT_MODEL_LABELS,
         PYTORCH_TRAINERS,
         build_model_names,
         dedupe_preserve_order,
-        load_config_json,
-        normalize_config_paths,
+        load_dataset,
         parse_model_pairs,
-        resolve_config_path,
+        resolve_data_path,
         resolve_path,
-        set_env,
+        split_train_test,
+    )
+    from .utils.cli_config import (  # type: ignore
+        add_config_json_arg,
+        resolve_and_load_config,
+        resolve_data_config,
+        resolve_split_config,
+        resolve_runtime_config,
+        resolve_output_dirs,
     )
 except Exception:  # pragma: no cover
     try:
         import bayesopt as ropt  # type: ignore
-        from cli_common import (  # type: ignore
+        from utils.cli_common import (  # type: ignore
             PLOT_MODEL_LABELS,
             PYTORCH_TRAINERS,
             build_model_names,
             dedupe_preserve_order,
-            load_config_json,
-            normalize_config_paths,
+            load_dataset,
             parse_model_pairs,
-            resolve_config_path,
+            resolve_data_path,
             resolve_path,
-            set_env,
+            split_train_test,
+        )
+        from utils.cli_config import (  # type: ignore
+            add_config_json_arg,
+            resolve_and_load_config,
+            resolve_data_config,
+            resolve_split_config,
+            resolve_runtime_config,
+            resolve_output_dirs,
         )
     except Exception:
         try:
-            import ins_pricing.bayesopt as ropt  # type: ignore
-            from ins_pricing.cli_common import (  # type: ignore
+            import ins_pricing.modelling.core.bayesopt as ropt  # type: ignore
+            from ins_pricing.cli.utils.cli_common import (  # type: ignore
                 PLOT_MODEL_LABELS,
                 PYTORCH_TRAINERS,
                 build_model_names,
                 dedupe_preserve_order,
-                load_config_json,
-                normalize_config_paths,
+                load_dataset,
                 parse_model_pairs,
-                resolve_config_path,
+                resolve_data_path,
                 resolve_path,
-                set_env,
+                split_train_test,
+            )
+            from ins_pricing.cli.utils.cli_config import (  # type: ignore
+                add_config_json_arg,
+                resolve_and_load_config,
+                resolve_data_config,
+                resolve_split_config,
+                resolve_runtime_config,
+                resolve_output_dirs,
             )
         except Exception:
             import BayesOpt as ropt  # type: ignore
-            from cli_common import (  # type: ignore
+            from utils.cli_common import (  # type: ignore
                 PLOT_MODEL_LABELS,
                 PYTORCH_TRAINERS,
                 build_model_names,
                 dedupe_preserve_order,
-                load_config_json,
-                normalize_config_paths,
+                load_dataset,
                 parse_model_pairs,
-                resolve_config_path,
+                resolve_data_path,
                 resolve_path,
-                set_env,
+                split_train_test,
+            )
+            from utils.cli_config import (  # type: ignore
+                add_config_json_arg,
+                resolve_and_load_config,
+                resolve_data_config,
+                resolve_split_config,
+                resolve_runtime_config,
+                resolve_output_dirs,
             )
 
 try:
-    from .run_logging import configure_run_logging  # type: ignore
+    from .utils.run_logging import configure_run_logging  # type: ignore
 except Exception:  # pragma: no cover
     try:
-        from run_logging import configure_run_logging  # type: ignore
+        from utils.run_logging import configure_run_logging  # type: ignore
     except Exception:  # pragma: no cover
         configure_run_logging = None  # type: ignore
 
@@ -103,10 +137,9 @@ def _parse_args() -> argparse.Namespace:
     parser = argparse.ArgumentParser(
         description="Incrementally retrain BayesOpt models using new batches of data."
     )
-    parser.add_argument(
-        "--config-json",
-        required=True,
-        help="Path to the JSON config that BayesOpt_entry.py uses."
+    add_config_json_arg(
+        parser,
+        help_text="Path to the JSON config that cli/BayesOpt_entry.py uses.",
     )
     parser.add_argument(
         "--model-names",
@@ -358,6 +391,15 @@ def _coerce_scalar(value: Any) -> Any:
     return value
 
 
+def _infer_format_from_path(path: Path) -> str:
+    suffix = path.suffix.lower()
+    if suffix in {".parquet", ".pq"}:
+        return "parquet"
+    if suffix in {".feather", ".ft"}:
+        return "feather"
+    return "csv"
+
+
 def _load_best_params(model: ropt.BayesOptModel, trainer, silent: bool = False) -> Optional[Dict[str, Any]]:
     label = trainer.label.lower()
     result_dir = Path(model.output_manager.result_dir)
@@ -393,10 +435,10 @@ def _to_serializable(obj: Any) -> Any:
 class IncrementalUpdateRunner:
     def __init__(self, args: argparse.Namespace) -> None:
         self.args = args
-        script_dir = Path(__file__).resolve().parent
-        self.config_path = resolve_config_path(args.config_json, script_dir)
-        cfg = load_config_json(
-            self.config_path,
+        script_dir = Path(__file__).resolve().parents[1]
+        self.config_path, self.cfg = resolve_and_load_config(
+            args.config_json,
+            script_dir,
             required_keys=[
                 "data_dir",
                 "model_list",
@@ -407,26 +449,57 @@ class IncrementalUpdateRunner:
                 "categorical_features",
             ],
         )
-        self.cfg = normalize_config_paths(cfg, self.config_path)
-        set_env(self.cfg.get("env", {}))
-        self.data_dir = Path(self.cfg["data_dir"])
-        self.data_dir.mkdir(parents=True, exist_ok=True)
-        self.prop_test = args.prop_test if args.prop_test is not None else self.cfg.get("prop_test", 0.25)
+        data_dir, data_format, data_path_template, dtype_map = resolve_data_config(
+            self.cfg,
+            self.config_path,
+            create_data_dir=True,
+        )
+        self.data_dir = data_dir
+        self.data_format = data_format
+        self.data_path_template = data_path_template
+        self.dtype_map = dtype_map
+        split_cfg = resolve_split_config(self.cfg)
+        runtime_cfg = resolve_runtime_config(self.cfg)
+        output_cfg = resolve_output_dirs(
+            self.cfg,
+            self.config_path,
+            output_override=args.output_dir,
+        )
+        self.runtime_cfg = runtime_cfg
+        self.prop_test = args.prop_test if args.prop_test is not None else split_cfg["prop_test"]
         self.rand_seed = args.rand_seed if args.rand_seed is not None else self.cfg.get("rand_seed", 13)
         self.epochs = args.epochs if args.epochs is not None else self.cfg.get("epochs", 50)
+        self.split_strategy = split_cfg["split_strategy"]
+        self.split_group_col = split_cfg["split_group_col"]
+        self.split_time_col = split_cfg["split_time_col"]
+        self.split_time_ascending = split_cfg["split_time_ascending"]
+        self.cv_strategy = split_cfg["cv_strategy"]
+        self.cv_group_col = split_cfg["cv_group_col"]
+        self.cv_time_col = split_cfg["cv_time_col"]
+        self.cv_time_ascending = split_cfg["cv_time_ascending"]
+        self.cv_splits = split_cfg["cv_splits"]
+        self.ft_oof_folds = split_cfg["ft_oof_folds"]
+        self.ft_oof_strategy = split_cfg["ft_oof_strategy"]
+        self.ft_oof_shuffle = split_cfg["ft_oof_shuffle"]
+        self.save_preprocess = runtime_cfg["save_preprocess"]
+        self.preprocess_artifact_path = runtime_cfg["preprocess_artifact_path"]
+        self.bo_sample_limit = runtime_cfg["bo_sample_limit"]
+        self.cache_predictions = runtime_cfg["cache_predictions"]
+        self.prediction_cache_dir = runtime_cfg["prediction_cache_dir"]
+        self.prediction_cache_format = runtime_cfg["prediction_cache_format"]
+        self.plot_path_style = runtime_cfg["plot_path_style"]
+        self.xgb_max_depth_max = runtime_cfg["xgb_max_depth_max"]
+        self.xgb_n_estimators_max = runtime_cfg["xgb_n_estimators_max"]
+        self.optuna_storage = runtime_cfg["optuna_storage"]
+        self.optuna_study_prefix = runtime_cfg["optuna_study_prefix"]
+        self.best_params_files = runtime_cfg["best_params_files"]
+        self.reuse_best_params = runtime_cfg["reuse_best_params"]
         self.plot_requested = bool(args.plot_curves or self.cfg.get("plot_curves", False))
         self.model_names = self._resolve_model_names(args.model_names)
         self.merge_keys = list(args.merge_keys or [])
         self.timestamp_col = args.timestamp_col
         self.timestamp_ascending = not args.timestamp_descending
-        output_root = args.output_dir or self.cfg.get("output_dir")
-        if isinstance(output_root, Path) and not output_root.is_absolute():
-            output_root = (self.config_path.parent / output_root).resolve()
-        if isinstance(output_root, str) and output_root.strip():
-            resolved = resolve_path(output_root, self.config_path.parent)
-            if resolved is not None:
-                output_root = str(resolved)
-        self.output_root = output_root
+        self.output_root = output_cfg["output_dir"]
 
         self.incremental_dir = None
         if args.incremental_dir is not None:
@@ -465,10 +538,18 @@ class IncrementalUpdateRunner:
         if not path or not path.exists():
             return None, None
         try:
-            df = pd.read_csv(path, low_memory=False)
+            df = load_dataset(
+                path,
+                data_format="auto",
+                dtype_map=self.dtype_map,
+                low_memory=False,
+            )
         except pd.errors.EmptyDataError:
             _log(f"Incremental file {path} is empty; treating as no-op.")
             return None, path
+        except Exception as exc:
+            _log(f"Failed to load incremental file {path}: {exc}")
+            return None, path
         return df, path
 
     def _merge_frames(self, base_df: pd.DataFrame, inc_df: Optional[pd.DataFrame]) -> pd.DataFrame:
@@ -507,7 +588,15 @@ class IncrementalUpdateRunner:
 
     def _write_dataset(self, df: pd.DataFrame, dest: Path, reason: str) -> None:
         dest.parent.mkdir(parents=True, exist_ok=True)
-        df.to_csv(dest, index=False)
+        fmt = str(self.data_format or "csv").lower()
+        if fmt == "auto":
+            fmt = _infer_format_from_path(dest)
+        if fmt == "parquet":
+            df.to_parquet(dest, index=False)
+        elif fmt == "feather":
+            df.reset_index(drop=True).to_feather(dest)
+        else:
+            df.to_csv(dest, index=False)
         _log(f"Wrote {len(df)} rows to {dest} ({reason}).")
 
     def _prepare_splits(self, merged: pd.DataFrame) -> Tuple[pd.DataFrame, pd.DataFrame]:
@@ -515,12 +604,19 @@ class IncrementalUpdateRunner:
             raise ValueError(f"prop_test must fall in (0, 1); got {self.prop_test}.")
         if len(merged) < 2:
             raise ValueError("Need at least two rows to form a train/test split.")
-        train_df, test_df = train_test_split(
+        train_df, test_df = split_train_test(
             merged,
-            test_size=self.prop_test,
-            random_state=self.rand_seed,
+            holdout_ratio=self.prop_test,
+            strategy=self.split_strategy,
+            group_col=self.split_group_col,
+            time_col=self.split_time_col,
+            time_ascending=self.split_time_ascending,
+            rand_seed=self.rand_seed,
+            reset_index_mode="always",
+            ratio_label="prop_test",
+            validate_ratio=False,
         )
-        return train_df.reset_index(drop=True), test_df.reset_index(drop=True)
+        return train_df, test_df
 
     def _requested_model_keys(self, trainer_map: Dict[str, Any]) -> List[str]:
         requested = self.args.model_keys
@@ -555,6 +651,7 @@ class IncrementalUpdateRunner:
             prop_test=self.prop_test,
             rand_seed=self.rand_seed,
             epochs=self.epochs,
+            use_gpu=bool(self.cfg.get("use_gpu", True)),
             use_resn_data_parallel=self.cfg.get("use_resn_data_parallel", False),
             use_ft_data_parallel=self.cfg.get("use_ft_data_parallel", True),
             use_gnn_data_parallel=self.cfg.get("use_gnn_data_parallel", False),
@@ -562,27 +659,52 @@ class IncrementalUpdateRunner:
             use_ft_ddp=self.cfg.get("use_ft_ddp", False),
             use_gnn_ddp=self.cfg.get("use_gnn_ddp", False),
             output_dir=str(self.output_root) if self.output_root else None,
-            xgb_max_depth_max=int(self.cfg.get("xgb_max_depth_max", 25)),
-            xgb_n_estimators_max=int(self.cfg.get("xgb_n_estimators_max", 500)),
+            xgb_max_depth_max=self.xgb_max_depth_max,
+            xgb_n_estimators_max=self.xgb_n_estimators_max,
             resn_weight_decay=self.cfg.get("resn_weight_decay"),
             final_ensemble=bool(self.cfg.get("final_ensemble", False)),
             final_ensemble_k=int(self.cfg.get("final_ensemble_k", 3)),
             final_refit=bool(self.cfg.get("final_refit", True)),
-            optuna_storage=self.cfg.get("optuna_storage"),
-            optuna_study_prefix=self.cfg.get("optuna_study_prefix"),
-            best_params_files=self.cfg.get("best_params_files"),
-            reuse_best_params=bool(self.cfg.get("reuse_best_params", False)),
+            optuna_storage=self.optuna_storage,
+            optuna_study_prefix=self.optuna_study_prefix,
+            best_params_files=self.best_params_files,
+            reuse_best_params=self.reuse_best_params,
             gnn_use_approx_knn=self.cfg.get("gnn_use_approx_knn", True),
             gnn_approx_knn_threshold=self.cfg.get("gnn_approx_knn_threshold", 50000),
             gnn_graph_cache=self.cfg.get("gnn_graph_cache"),
             gnn_max_gpu_knn_nodes=self.cfg.get("gnn_max_gpu_knn_nodes", 200000),
             gnn_knn_gpu_mem_ratio=self.cfg.get("gnn_knn_gpu_mem_ratio", 0.9),
             gnn_knn_gpu_mem_overhead=self.cfg.get("gnn_knn_gpu_mem_overhead", 2.0),
+            region_province_col=self.cfg.get("region_province_col"),
+            region_city_col=self.cfg.get("region_city_col"),
+            region_effect_alpha=self.cfg.get("region_effect_alpha"),
+            geo_feature_nmes=self.cfg.get("geo_feature_nmes"),
+            geo_token_hidden_dim=self.cfg.get("geo_token_hidden_dim"),
+            geo_token_layers=self.cfg.get("geo_token_layers"),
+            geo_token_dropout=self.cfg.get("geo_token_dropout"),
+            geo_token_k_neighbors=self.cfg.get("geo_token_k_neighbors"),
+            geo_token_learning_rate=self.cfg.get("geo_token_learning_rate"),
+            geo_token_epochs=self.cfg.get("geo_token_epochs"),
             ft_role=str(self.cfg.get("ft_role", "model")),
             ft_feature_prefix=str(self.cfg.get("ft_feature_prefix", "ft_emb")),
             ft_num_numeric_tokens=self.cfg.get("ft_num_numeric_tokens"),
             infer_categorical_max_unique=int(self.cfg.get("infer_categorical_max_unique", 50)),
             infer_categorical_max_ratio=float(self.cfg.get("infer_categorical_max_ratio", 0.05)),
+            cv_strategy=self.cv_strategy or self.split_strategy,
+            cv_group_col=self.cv_group_col or self.split_group_col,
+            cv_time_col=self.cv_time_col or self.split_time_col,
+            cv_time_ascending=self.cv_time_ascending,
+            cv_splits=self.cv_splits,
+            ft_oof_folds=self.ft_oof_folds,
+            ft_oof_strategy=self.ft_oof_strategy,
+            ft_oof_shuffle=self.ft_oof_shuffle,
+            save_preprocess=self.save_preprocess,
+            preprocess_artifact_path=self.preprocess_artifact_path,
+            plot_path_style=self.plot_path_style,
+            bo_sample_limit=self.bo_sample_limit,
+            cache_predictions=self.cache_predictions,
+            prediction_cache_dir=self.prediction_cache_dir,
+            prediction_cache_format=self.prediction_cache_format,
         )
 
         if self.plot_requested and not self.args.dry_run:
@@ -689,7 +811,12 @@ class IncrementalUpdateRunner:
         _log(f"Finished incremental update for {total_trained} dataset(s).")
 
     def _process_single_model(self, model_name: str) -> int:
-        base_path = self.data_dir / f"{model_name}.csv"
+        base_path = resolve_data_path(
+            self.data_dir,
+            model_name,
+            data_format=self.data_format,
+            path_template=self.data_path_template,
+        )
         if not base_path.exists():
             _log(f"Base dataset {base_path} not found; skipping {model_name}.")
             self.summary_records.append({
@@ -698,7 +825,12 @@ class IncrementalUpdateRunner:
             })
             return 0
 
-        base_df = pd.read_csv(base_path, low_memory=False)
+        base_df = load_dataset(
+            base_path,
+            data_format=self.data_format,
+            dtype_map=self.dtype_map,
+            low_memory=False,
+        )
         inc_df, inc_path = self._load_incremental_df(model_name)
         if inc_df is None and self.incremental_dir and self.args.strict_incremental and not self.args.train_without_incremental:
             raise FileNotFoundError(f"Missing incremental file for {model_name} under {self.incremental_dir}.")
@@ -711,7 +843,8 @@ class IncrementalUpdateRunner:
         if self.args.update_base_data and not self.args.dry_run:
             self._write_dataset(merged_df, base_path, "update_base_data")
         if self.args.persist_merged_dir and not self.args.dry_run:
-            dest = Path(self.args.persist_merged_dir).resolve() / f"{model_name}.csv"
+            suffix = base_path.suffix or ".csv"
+            dest = Path(self.args.persist_merged_dir).resolve() / f"{model_name}{suffix}"
             self._write_dataset(merged_df, dest, "persist_merged_dir")
 
         if not self._should_train(new_rows):

ins_pricing-0.2.0/ins_pricing/cli/Explain_Run.py

@@ -0,0 +1,25 @@
+from __future__ import annotations
+
+from pathlib import Path
+from typing import Optional
+
+try:
+    from .utils.notebook_utils import run_from_config, run_from_config_cli  # type: ignore
+except Exception:  # pragma: no cover
+    from utils.notebook_utils import run_from_config, run_from_config_cli  # type: ignore
+
+
+def run(config_json: str | Path) -> None:
+    """Run explain by config.json (runner.mode=explain)."""
+    run_from_config(config_json)
+
+
+def main(argv: Optional[list[str]] = None) -> None:
+    run_from_config_cli(
+        "Explain_Run: run explain by config.json (runner.mode=explain).",
+        argv,
+    )
+
+
+if __name__ == "__main__":
+    main()