**Note:** The code is currently in an experimental version. Its sole purpose is to verify the referenced results. A user-friendly version of the code, accompanied by more detailed documentation and available as a pip package, will be released after the camera-ready submission.

#### Model
The "STab" folder contains all model-related files.

#### Examples
We provide Jupyter notebooks for training the model with each of the studied datasets.

#### Data
To download the same versions and splits for YE, HI, HE, AD, and JA, please follow the instructions in this repository:
[Download Link](https://www.dropbox.com/s/o53umyg6mn3zhxy/data.tar.gz?dl=1) -O revisiting_models_data.tar.gz

To download the same versions and splits for DI, HO, and OT, please follow the instructions in this repository:
[Download Link](https://huggingface.co/datasets/puhsu/tabular-benchmarks/resolve/main/data.tar) -O tabular-dl-tabr.tar.gz

To use the example notebooks, copy the dataset folder directly into the "Data" directory. We provide the Helena dataset in the exact format used in the paper, available in the specified directory.

#### Dependencies
We use torch for model implementation and the keras4torch library for easier training and evaluation. Additionally, our code makes use of elements from the following repositories:
- [FtTransformer PyTorch](https://github.com/lucidrains/tab-transformer-pytorch)
- [Stochastic Transformer Networks with Linear Competing Units](https://github.com/avoskou/Stochastic-Transformer-Networks-with-Linear-Competing-Units-Application-to-end-to-end-SL-Translation)
- [Keras 4 Torch](https://github.com/blueloveTH/keras4torch)

