This repo contains an example of contrastive representation learning. An embedding can be trained in such a way that similar samples (sharing the same label) are pulled together, while dissimilar ones (with different labels) are pushed apart.
pip install -e .
python scripts/main.py fit --config config/mnist.yaml