This is a pytorch implementation of SimCLR.
- cudatoolkit=10.2
- pytorch=1.6.0
- torchvision=0.7.0
- opencv
- pandas
- pyyaml
To train a model simply run:
python cifar.py --config CONFIGFILE --gpu-id GPUID
where CONFIGFILE is the yaml configuration file. The yaml file for CIFAR is config-file.yaml.
To evaluate the model with a linear model run:
python cifar_eval.py --config CONFIGFILE --gpu-id GPUID