This is your work, valued
fairseq-rl. Modified version of fairseq, including new implementations for criterions using reinforcement learning methods.
