Support Ukraine 🇺🇦 Help Provide Humanitarian Aid to Ukraine.

A library for state-of-the-art self-supervised learning from images

Get Started
Tutorials
Docs
GitHub

Check it out in the intro video

Powered by PyTorch

Built on top of PyTorch which allows using all of its components.

SOTA Self-Supervision methods

Reproducible reference implementation of SOTA self-supervision approaches (like SimCLR, MoCo, PIRL, SwAV etc) and their components that can be reused. Also supports supervised trainings.

Benchmark tasks

Variety of benchmarks tasks (linear image classification, full finetuning, semi-supervised, low-shot, nearest neighbor, object detection) available to evaluate models.

Scalable

Easy to train model on 1-gpu, multi-gpu and multi-node. Seamless scaling to large scale data and model sizes with FP16, LARC etc

Get Started

  1. Install VISSL:

    via conda:
    conda create -n vissl python=3.8
conda activate vissl
conda install -c pytorch pytorch=1.7.1 torchvision cudatoolkit=10.2
conda install -c vissl -c iopath -c conda-forge -c pytorch -c defaults apex vissl

  2. Download SimCLR yaml config and builtin distributed launcher: 

    cd /tmp/ && mkdir -p /tmp/configs/config
wget -q -O configs/__init__.py https://dl.fbaipublicfiles.com/vissl/tutorials/configs/__init__.py
wget -q -O configs/config/quick_1gpu_resnet50_simclr.yaml https://dl.fbaipublicfiles.com/vissl/tutorials/configs/quick_1gpu_resnet50_simclr.yaml
wget -q  https://dl.fbaipublicfiles.com/vissl/tutorials/run_distributed_engines.py

  3. Try training SimCLR model on 1-gpu: 

    python3 run_distributed_engines.py config=quick_1gpu_resnet50_simclr config.DATA.TRAIN.DATA_SOURCES=[synthetic]