DQN is no longer a single-file script #518
Description
I just realized that it removed the SB3 dependence but introduces the dependence on from cleanrl_utils.buffers import ReplayBuffer.
Is this by design? It could be more helpful if I could pip install cleanrl to install this package. For example, if people hope run CleanRL on Google Colab for a free GPU, they do not need to clone this repo and install every dependencies again.