Skip to content

IgorSusmelj/barlowtwins

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

18 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

barlowtwins

PyTorch Implementation of Barlow Twins paper: Barlow Twins: Self-Supervised Learning via Redundancy Reduction

This is currently a work in progress. The code is a modified version of the SimSiam implementation here

  • Time per epoch is around 40 seconds on a V100 GPU
  • GPU usage is around 9 GBytes
  • The current version reaches around 84.7% test accuracy

Todo:

  • warmup learning rate from 0
  • report results on cifar-10
  • create PR to add to lightly

Installation

pip install -r requirements.txt

Dependencies

  • PyTorch
  • PyTorch Lightning
  • Torchvision
  • lightly

Benchmarks

We benchmark the BarlowTwins model on the CIFAR-10 dataset following the KNN evaluation protocol. Currently, the best effort achieved a test accuracy of 84.7%.

Accuracy Loss

Paper

Barlow Twins: Self-Supervised Learning via Redundancy Reduction