Hi,
The current implementation of the Runner class does not allow the agent to know if an episode has completed. So we are not able to implement an algorithm that requires this information. Please modify the Runner class to pass done also as an argument to the learn function.