The RL selector model is missing for the data distillation part of code.

[cgsdfc]

The paper I am reading recently, Data Distillation for Controlling Specificity in Dialogue Generation, feels like a very interesting and novel idea to me, so I am trying out some experiments on it. My experiment is based on the code in the Distill/ directory. Currently, I am nearly able to run the distillation pipline on the dataset provided in data/. But I am curious about the code for the Reinforcement learning model that is used to select the most appropriate generative model from the pool obtained in the distillation process. The RL model uses the adversarial evaluation model in fact, which is already provided so I am glad. But the RL used in Distill is not the same as that in adversarial evaluation and for the former, I can't find it in the codebase. Would you so kindly implement that for us? Looking forward to that.