🌍 Join Neuralk-AI's tabular AI ecosystem!
💡 Bringing tabular models closer to industrial tasks
Neuralk-AI builds the first Tabular Foundational Model focused on industrial tasks, starting with Commerce.
We created TabBench, the first benchmark dedicated to evaluating and advancing tabular models on real-world use cases and ML workflows typical in industries like Commerce, such as product categorization, deduplication, and more.
What TabBench currently supports (frequently updated!):
- Real-world use cases: product categorization, deduplication
- Easily evaluate your model or dataset for each use case thanks to a steamlined Workflow logic (load → vectorize → predict → evaluate)
- Evaluation on both industrial datasets (private) & academic ones (OpenML)
- Classical ML & Tabular Foundation models: NICL, TabICL, TabPFNv2, XGBoost, CatBoost, LightGBM, MLP
- Built on Neuralk Foundry, an open-source, modular framework to customize your own industrial workflows
🚀 How to get started?
Install TabBench with pip:
pip install tabbench
or directly clone the repository:
git clone https://github.com/Neuralk-AI/TabBench
cd TabBench
Jump straight into our example notebooks to start exploring tabular models on industrial tasks:
| File | Description |
|---|---|
| 1 - Getting Started with TabBench | Discover how TabBench works and train your first tabular model on a Product Categorization task. |
| 2 - Adding a local or internet dataset | How to add your own datasets for evaluation (local, downloadable, or OpenML). |
| 3 - Use a custom model | How to integrate a new model in TabBench and use it on different use cases. |
For more information about TabBench, open-source code and tutorials, you can check our Github Page