Created by: TomorrowIsAnOtherDay
PARL aims to provide an easy way for users who want to transfer the algorithms to a new benchmark, and at the same time, it still provides the basic class for users to develop a customized algorithm.
That's why we decide to redesign the basic class in PARL. In the new architecture, you could easily transfer the algorithm(e.g. DQN) to a new benchmark, the only thing you need to do is define the forward function.
A simple example will be added to PARL after this PR is merged.