This repository contains numerical experiments for "BEER: Fast O(1/T) Rate for Decentralized Nonconvex Optimization with Communication Compression" [PDF].

If you find this repo useful, please cite our paper

@article{zhao2022beer,
  title={BEER: Fast O (1/T) Rate for Decentralized Nonconvex Optimization with Communication Compression},
  author={Zhao, Haoyu and Li, Boyue and Li, Zhize and Richt{\'a}rik, Peter and Chi, Yuejie},
  journal={Advances in Neural Information Processing Systems},
  volume = {35},
  year={2022}
}

• beer/: framework for convolutional neural network experiments.

• experiments/experiments.ipynb: code for synthetic numerical experiments.

• experiments/mnist/: code and scripts for convolutional neural network experiments.

Please install [this package] first.

Then run pip install git+https://github.com/liboyue/beer.git to install this package.