I'm sorry the title might be misleading.
I'm currently working on a Q-Learning algorithm. I only can modify the Agent. And the runner doesn't implement episodes. Is there any way to make the agent store its learned q-values each time in a separate file, in order to reuse them the next time it runs.