Finetuning Conversational Models for Auxiliary Tasks with Deep Reinforcement Learning A repository with my BS MIPT Diploma September first presentation February report