This is the repository code for "TabNSA: Native Sparse Attention for Efficient Tabular Data Learning".
📄 Paper: TabNSA: Native Sparse Attention for Efficient Tabular Data Learning
TabNSA introduces a native sparse attention mechanism tailored for tabular data. It achieves superior performance and accuracy on multiple benchmark datasets compared to existing models.
- Native sparse attention architecture for tabular inputs
- Significantly improved classification/regression performance
- Competitive results on standard tabular benchmarks
- Easily pluggable into PyTorch training pipelines
git clone https://github.com/yourusername/TabNSA.git
cd TabNSA
pip install -r requirements.txt