Thinformer

Thinformer provides a fast, high-quality approximation to the scaled dot-product attention mechanism in Transformers.

For a detailed description of the Thinformer algorithm and its strong approximation guarantees, see Low-Rank Thinning.

@article{carrell2025low,
  title={Low-Rank Thinning},
  author={Carrell, Annabelle Michael and Gong, Albert and Shetty, Abhishek and Dwivedi, Raaz and Mackey, Lester},
  journal={arXiv preprint arXiv:2502.12063},
  year={2025}
}

Getting Started

To install the thinformer package, use the following pip command:

pip install git+https://github.com/microsoft/thinformer.git

Then, simply use ThinformerAttention as a drop-in replacement for a standard attention layer:

from thinformer import ThinformerAttention

attention_layer = ThinformerAttention()

# Assumes:
# - query has shape (B, T, H, E)
# - key has shape (B, S, H, E)
# - value has shape (B, S, H, D)

attn_output, attn_output_weights = attention_layer(query, key, value)

For an example usage, see our T2T-ViT ImageNet classification experiments.

This package has been tested with the following operating system, Python, and PyTorch combinations:

• Ubuntu 20.04, Python 3.12.9, Torch 2.4.0
• Ubuntu 20.04, Python 3.12.9, Torch 2.6.0
• Ubuntu 22.04.5, Python 3.12.9, Torch 2.8.0

Contributing

This project welcomes contributions and suggestions. Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your contribution. For details, visit https://cla.opensource.microsoft.com.

When you submit a pull request, a CLA bot will automatically determine whether you need to provide a CLA and decorate the PR appropriately (e.g., status check, comment). Simply follow the instructions provided by the bot. You will only need to do this once across all repos using our CLA.

This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact [email protected] with any additional questions or comments.

Trademarks

This project may contain trademarks or logos for projects, products, or services. Authorized use of Microsoft trademarks or logos is subject to and must follow Microsoft's Trademark & Brand Guidelines. Use of Microsoft trademarks or logos in modified versions of this project must not cause confusion or imply Microsoft sponsorship. Any use of third-party trademarks or logos are subject to those third-party's policies.