VICReg: Variance-Invariance-Covariance Regularization for Self-Supervised Learning

An easy-to-use implementation of VICReg in Pytorch.

VICReg: image from paper


To rerun experiments or try on your own dataset, first clone the repository and install requirements.txt.

$ pip install -r requirements.txt

Train VICReg

To train model use

$ python --path path/to/your/dataset/ 

VICReg implementation

  • 3 different loss function
  • VICReg model and training loop
  • Implementation for multiple GPUs
  • Compare results with paper

Any contribution is appreciated!