Added TabDDPM model implementation under Models/TabDDPMSiddharth-main

[Siddharth7269]

Description:
This PR brings in my custom Tabular Denoising Diffusion Probabilistic Model (TabDDPM) implementation, organized in Models/TabDDPMSiddharth. It includes:

Core model code

tabddpm.py – the main diffusion class defining forward/backward processes

modules.py – neural network building blocks (MLP layers, embedding layers, etc.)

utils.py – data loaders, preprocessing, and helper functions

Training & evaluation

train_tabddpm.py – end-to-end training script with CLI arguments

evaluate_tabddpm.py – script to compute statistical and ML-based metrics on synthetic vs. real data

Example notebooks

demo_tabddpm_adult.ipynb – demo on the UCI Adult dataset

demo_tabddpm_car.ipynb – demo on the Car Evaluation dataset

Datasets & outputs

data/Adult.csv, data/car.csv, etc. – raw UCI CSVs

synthetic/ – generated synthetic datasets (e.g. Adult_synth.csv)

checkpoints/ – saved model weights for reproducibility

Configuration

config.yaml – default hyperparameters (learning rate, batch size, timesteps)

How to test / run:

Install dependencies (e.g. requirements.txt lists torch, pandas, numpy).

Train on a small dataset:

bash
Copy
Edit
python Models/TabDDPMSiddharth/train_tabddpm.py
--config Models/TabDDPMSiddharth/config.yaml
--data Models/TabDDPMSiddharth/data/Adult.csv
--epochs 50
Generate synthetic samples:

bash
Copy
Edit
python Models/TabDDPMSiddharth/evaluate_tabddpm.py
--checkpoint Models/TabDDPMSiddharth/checkpoints/last.pt
--output Models/TabDDPMSiddharth/synthetic/Adult_synth.csv
Review notebooks by opening demo_tabddpm_*.ipynb in Jupyter Lab.

Notes & Next Steps:

The raw CSVs are ~100 MB in total; consider moving to Git LFS or external storage.

We may refactor shared utilities (e.g. data loading) into a common Models/common directory.

Future enhancements: support conditional sampling, integration with the main benchmarking suite.

Please let me know if you’d like any restructuring or additional documentation!