Skip to content

Aleph-Alpha/NeurIPS-WANT-submission-efficient-parallelization-layouts

Repository files navigation

Excel sheets and scripts for NeurIPS WANT workshop publication:

Cite Us

To cite this project, you can use the following BibTeX citation.

@misc{hagemann2023efficient,
      title={Efficient Parallelization Layouts for Large-Scale Distributed Model Training}, 
      author={Johannes Hagemann and Samuel Weinbach and Konstantin Dobler and Maximilian Schall and Gerard de Melo},
      year={2023},
      eprint={2311.05610},
      archivePrefix={arXiv},
      primaryClass={cs.LG}
}

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 3

  •  
  •  
  •  

Languages