ins-pricing 0.4.4__py3-none-any.whl → 0.5.0__py3-none-any.whl

ins_pricing/docs/modelling/README.md

@@ -1,34 +0,0 @@
-# ins_pricing
-
-This directory contains reusable production-grade tooling and training frameworks, with a focus on the BayesOpt series.
-
-Key contents:
-- `core/bayesopt/`: core subpackage (data preprocessing, trainers, models, plotting, explainability)
-- `plotting/`: standalone plotting helpers (lift/roc/importance/geo)
-- `explain/`: explainability helpers (Permutation/Integrated Gradients/SHAP)
-- `core/BayesOpt.py`: compatibility entry point for legacy imports
-- `cli/BayesOpt_entry.py`: batch training CLI
-- `cli/BayesOpt_incremental.py`: incremental training CLI
-- `cli/utils/cli_common.py` / `cli/utils/notebook_utils.py`: shared CLI and notebook utilities
-- `examples/modelling/config_template.json` / `examples/modelling/config_incremental_template.json`: config templates
-- `cli/Explain_entry.py` / `cli/Explain_Run.py`: explainability entry points (load trained models)
-- `examples/modelling/config_explain_template.json` / `examples/modelling/Explain_Run.ipynb`: explainability demo
-
-Note: `examples/modelling/` is kept in the repo only and is not shipped in the PyPI package.
-Migration note: CLI entry points now live under `cli/` and demo assets are under `examples/modelling/`.
-
-Common usage:
-- CLI: `python ins_pricing/cli/BayesOpt_entry.py --config-json ...`
-- Notebook: `from ins_pricing.bayesopt import BayesOptModel`
-
-Explainability (load trained models under `Results/model` and explain a validation set):
-- CLI: `python ins_pricing/cli/Explain_entry.py --config-json ins_pricing/examples/modelling/config_explain_template.json`
-- Notebook: open `ins_pricing/examples/modelling/Explain_Run.ipynb` and run it
-
-Notes:
-- Models load from `output_dir/model` by default (override with `explain.model_dir`).
-- Validation data can be specified via `explain.validation_path`.
-
-Operational notes:
-- Training outputs are written to `plot/`, `Results/`, and `model/` by default.
-- Keep large data and secrets outside the repo and use environment variables or `.env`.

ins_pricing/frontend/QUICKSTART.md

@@ -1,152 +0,0 @@
-# Quick Start Guide
-
-Get started with the Insurance Pricing Model Training Frontend in 3 easy steps.
-
-## Prerequisites
-
-1. Install the `ins_pricing` package
-2. Install Gradio:
-   ```bash
-   pip install gradio>=4.0.0
-   ```
-
-## Step 1: Launch the Application
-
-### On Windows:
-Double-click `start_app.bat` or run:
-```bash
-python -m ins_pricing.frontend.app
-```
-
-### On Linux/Mac:
-Run the shell script:
-```bash
-./start_app.sh
-```
-
-Or use Python directly:
-```bash
-python -m ins_pricing.frontend.app
-```
-
-The web interface will automatically open at `http://localhost:7860`
-
-## Step 2: Configure Your Model
-
-### Option A: Upload Existing Config (Recommended)
-1. Go to the **Configuration** tab
-2. Click **"Upload JSON Config File"**
-3. Select a config file (e.g., `config_xgb_direct.json` from `examples/`)
-4. Click **"Load Config"**
-
-### Option B: Manual Configuration
-1. Go to the **Configuration** tab
-2. Scroll to **"Manual Configuration"**
-3. Fill in the required fields:
-   - **Data Directory**: Path to your data folder
-   - **Model List**: Model name(s)
-   - **Target Column**: Your target variable
-   - **Weight Column**: Your weight variable
-   - **Feature List**: Comma-separated features
-   - **Categorical Features**: Comma-separated categorical features
-4. Adjust other settings as needed
-5. Click **"Build Configuration"**
-
-## Step 3: Run Training
-
-1. Switch to the **Run Task** tab
-2. Click **"Run Task"**
-3. Watch real-time logs appear below
-
-Training will start automatically and logs will update in real-time!
-
-## New Features
-
-### FT Two-Step Workflow
-
-For advanced FT-Transformer → XGB/ResN training:
-
-1. **Prepare Base Config**: Create or load a base configuration
-2. **Go to FT Two-Step Workflow tab**
-3. **Step 1 - FT Embedding Generation**:
-   - Configure DDP settings
-   - Click "Prepare Step 1 Config"
-   - Copy the config to Configuration tab
-   - Run it in "Run Task" tab
-4. **Step 2 - Train XGB/ResN**:
-   - After Step 1 completes, click "Prepare Step 2 Configs"
-   - Choose which models to train (XGB, ResN, or both)
-   - Copy the generated configs and run them
-
-### Open Results Folder
-
-- In the **Run Task** tab, click **"📁 Open Results Folder"**
-- Automatically opens the output directory in your file explorer
-- Works on Windows, macOS, and Linux
-
-## Example Configuration
-
-Here's a minimal example to get started:
-
-```json
-{
-  "data_dir": "./Data",
-  "model_list": ["od"],
-  "model_categories": ["bc"],
-  "target": "response",
-  "weight": "weights",
-  "feature_list": ["age", "gender", "region"],
-  "categorical_features": ["gender", "region"],
-  "runner": {
-    "mode": "entry",
-    "model_keys": ["xgb"],
-    "max_evals": 50
-  }
-}
-```
-
-Save this as `my_first_config.json` and upload it!
-
-## Tips
-
-- **Save Your Config**: After building a configuration, save it using the "Save Configuration" button for reuse
-- **Check Logs**: Training logs update in real-time - watch for errors or progress indicators
-- **GPU Usage**: Toggle "Use GPU" checkbox in Training Settings to enable/disable GPU acceleration
-- **Model Selection**: Specify which models to train in "Model Keys" (xgb, resn, ft, gnn)
-- **Open Results**: Use the "📁 Open Results Folder" button to quickly access output files
-- **FT Workflow**: Use the dedicated FT tab for automated two-step FT → XGB/ResN training
-
-## Troubleshooting
-
-**Problem**: Interface doesn't load
-- **Solution**: Check that port 7860 is not in use, or specify a different port
-
-**Problem**: Configuration validation fails
-- **Solution**: Ensure all required fields are filled and feature lists are properly formatted
-
-**Problem**: Training doesn't start
-- **Solution**: Verify data paths exist and configuration is valid
-
-**Problem**: Results folder won't open
-- **Solution**: Make sure the task has run at least once to create the output directory
-
-**Problem**: Step 2 configs fail to generate
-- **Solution**: Ensure Step 1 completed successfully and embedding files exist
-
-## Next Steps
-
-- Explore advanced options in the Configuration tab
-- Try the FT Two-Step Workflow for better model performance
-- Experiment with different model combinations (xgb, resn, ft)
-- Try different split strategies
-- Use the Explain mode for model interpretability
-- Check the full [README.md](README.md) for detailed documentation
-
-## Support
-
-For issues or questions, refer to:
-- Full documentation: [README.md](README.md)
-- Example configs: `ins_pricing/examples/`
-- Package documentation: `ins_pricing/docs/`
-
-Happy modeling!

ins_pricing/modelling/core/BayesOpt.py

@@ -1,146 +0,0 @@
-from __future__ import annotations
-
-import torch
-
-try:
-    from .bayesopt.config_preprocess import (
-        BayesOptConfig,
-        DatasetPreprocessor,
-        OutputManager,
-        VersionManager,
-    )
-    from .bayesopt.core import BayesOptModel
-    from .bayesopt.models import (
-        FeatureTokenizer,
-        FTTransformerCore,
-        FTTransformerSklearn,
-        GraphNeuralNetSklearn,
-        MaskedTabularDataset,
-        ResBlock,
-        ResNetSequential,
-        ResNetSklearn,
-        ScaledTransformerEncoderLayer,
-        SimpleGraphLayer,
-        SimpleGNN,
-        TabularDataset,
-    )
-    from .bayesopt.trainers import (
-        FTTrainer,
-        GLMTrainer,
-        GNNTrainer,
-        ResNetTrainer,
-        TrainerBase,
-        XGBTrainer,
-        _xgb_cuda_available,
-    )
-    from .bayesopt.utils import (
-        EPS,
-        DistributedUtils,
-        IOUtils,
-        PlotUtils,
-        TorchTrainerMixin,
-        TrainingUtils,
-        compute_batch_size,
-        csv_to_dict,
-        ensure_parent_dir,
-        free_cuda,
-        infer_factor_and_cate_list,
-        plot_dlift_list,
-        plot_lift_list,
-        set_global_seed,
-        split_data,
-        tweedie_loss,
-    )
-except ImportError:  # pragma: no cover
-    from bayesopt.config_preprocess import (
-        BayesOptConfig,
-        DatasetPreprocessor,
-        OutputManager,
-        VersionManager,
-    )
-    from bayesopt.core import BayesOptModel
-    from bayesopt.models import (
-        FeatureTokenizer,
-        FTTransformerCore,
-        FTTransformerSklearn,
-        GraphNeuralNetSklearn,
-        MaskedTabularDataset,
-        ResBlock,
-        ResNetSequential,
-        ResNetSklearn,
-        ScaledTransformerEncoderLayer,
-        SimpleGraphLayer,
-        SimpleGNN,
-        TabularDataset,
-    )
-    from bayesopt.trainers import (
-        FTTrainer,
-        GLMTrainer,
-        GNNTrainer,
-        ResNetTrainer,
-        TrainerBase,
-        XGBTrainer,
-        _xgb_cuda_available,
-    )
-    from bayesopt.utils import (
-        EPS,
-        DistributedUtils,
-        IOUtils,
-        PlotUtils,
-        TorchTrainerMixin,
-        TrainingUtils,
-        compute_batch_size,
-        csv_to_dict,
-        ensure_parent_dir,
-        free_cuda,
-        infer_factor_and_cate_list,
-        plot_dlift_list,
-        plot_lift_list,
-        set_global_seed,
-        split_data,
-        tweedie_loss,
-    )
-
-__all__ = [
-    "BayesOptConfig",
-    "DatasetPreprocessor",
-    "OutputManager",
-    "VersionManager",
-    "BayesOptModel",
-    "FeatureTokenizer",
-    "FTTransformerCore",
-    "FTTransformerSklearn",
-    "GraphNeuralNetSklearn",
-    "MaskedTabularDataset",
-    "ResBlock",
-    "ResNetSequential",
-    "ResNetSklearn",
-    "ScaledTransformerEncoderLayer",
-    "SimpleGraphLayer",
-    "SimpleGNN",
-    "TabularDataset",
-    "FTTrainer",
-    "GLMTrainer",
-    "GNNTrainer",
-    "ResNetTrainer",
-    "TrainerBase",
-    "XGBTrainer",
-    "_xgb_cuda_available",
-    "EPS",
-    "DistributedUtils",
-    "IOUtils",
-    "PlotUtils",
-    "TorchTrainerMixin",
-    "TrainingUtils",
-    "compute_batch_size",
-    "csv_to_dict",
-    "ensure_parent_dir",
-    "free_cuda",
-    "infer_factor_and_cate_list",
-    "plot_dlift_list",
-    "plot_lift_list",
-    "set_global_seed",
-    "split_data",
-    "tweedie_loss",
-    "torch",
-]

ins_pricing/modelling/core/__init__.py

@@ -1 +0,0 @@
-"""Core modelling modules (bayesopt + evaluation)."""