Skip to content

WELCOME TO 🧬 FUSIONBENCH

arXiv GitHub License PyPI - Version Downloads Static Badge Static Badge Static Badge

Install FusionBench

Prerequisites:

  • Python 3.10 or later (some features may not work as expected for earlier versions)

Install the latest version of fusion-bench from GitHub repository and install it in editable mode by passing the -e flag to pip.

git clone https://github.com/tanganke/fusion_bench.git
cd fusion_bench

# checkout to use a specific version. for example, v0.1.6
# git checkout v0.1.6

pip install -e . # install the package in editable mode

fusion-bench can also be installed from PyPI as a library and toolkit for deep model fusion.

pip install fusion-bench

# you can also install a specific version
# pip install fusion-bench==0.1.6

Installing fusion-bench will also install the latest stable PyTorch if you don't have it already.

Next Steps

  • Learn More about FusionBench

    Learn the basic concepts of FusionBench and the command line interface (CLI) as well as the programmatic usage of FusionBench.

    Read More

  • Learn More About Deep Model Fusion

    Read an introduction to deep model fusion and learn about its key concepts, techniques, and applications.

    Read More

Contributing to FusionBench

  • Any questions or comments can be directed to the GitHub Issues page for this project.
  • Any contributions or pull requests are welcome. If you find any mistakes or have suggestions for improvements, please feel free to raise an issue or submit a pull request.

Introduction to Deep Model Fusion (The Learn From Model Paradigm)

Deep model fusion is a technique that merges, ensemble, or fuse multiple deep neural networks to obtain a unified model. It can be used to improve the performance and robustness of model or to combine the strengths of different models, such as fuse multiple task-specific models to create a multi-task model. For a more detailed introduction to deep model fusion, you can refer to W. Li, 2023, 'Deep Model Fusion: A Survey'. In this benchmark, we evaluate the performance of different fusion methods on a variety of datasets and tasks. ...

Read More

Citation

If you find this benchmark useful, please consider citing our work:

@article{tang2024fusionbench,
  title={Fusionbench: A comprehensive benchmark of deep model fusion},
  author={Tang, Anke and Shen, Li and Luo, Yong and Hu, Han and Du, Bo and Tao, Dacheng},
  journal={arXiv preprint arXiv:2406.03280},
  year={2024}
}