Train transformer language models with reinforcement learning. - View it on GitHub