SkipTrain

This repository contains the code for running the experiments in the paper "Energy-Aware Decentralized Learning with Intermittent Model Training" authored by Martijn de Vos, Akash Dhasade, Paolo Dini, Elia Guerra, Anne-Marie Kermarrec, Marco Miozzo, Rafael Pires, and Rishi Sharma.

Setting up

• Clone this repository

• Download its submodules

  git submodule update --init --recursive
  

• Create a new Conda environment

  conda create --name skiptrain python==3.8.16
  conda activate skiptrain
  

• Install requirements

  pip install -r requirements.txt
  

• Install SkipTrain

  pip3 install --editable .[dev]
  

• Download CIFAR-10 using download_dataset.py.

  python download_dataset.py
  

• (Optional) Download the FEMNIST from LEAF <https://github.com/TalwalkarLab/leaf> and place them in eval/data/.

Running the code

• The tutorial/ folder contains a working example on a 3-regular graph.

  • SkipTrain: tutorial/run_skiptrain.sh
  • SkipTrain Constrained: tutorial/run_skiptrain_constrained.sh
  • DPSGD: tutorial/run_dpsgd.sh
  • Greedy: tutorial/run_greedy.sh

• The topologies adopted for the simulation are available in the tutorial/ folder.

Citing

Cite us as :: ..

@inproceedings{skiptrain,

author = {}, title = {Energy-Aware Decentralized Learning with Intermittent Model Training}, year = {2023}, isbn = {}, publisher = {}, address = {}, url = {}, doi = {}, booktitle = {}, pages = {}, numpages = {}, keywords = {decentralized learning, machine learning, energy effiency, peer-to-peer}, location = {}, series = {} }