rlpyt
rlpyt copied to clipboard
Dqn nonframe state
- Add dqn for non-frame state spaces
- Add loading and evaluation code for the trained model
Example_9 file contains using dqn for cartpole with a custom model and evaluation function.
Thank you! That explains why DQN did not work for my tasks. Your pull request is, however, quite complicated. Wouldn't just changing
self.replay_buffer = ReplayCls(**replay_kwargs)
into
self.replay_buffer = ReplayCls(**replay_kwargs) if self.ReplayBufferCls is None else self.ReplayBufferCls(**replay_kwargs)
in the DQN algorithm solve that issue?
@petrikvladimir It does if you set the class as input. I just added the same settings as the atari env for non-atari envs to do it automatically.
@astooke, is there any issue with this merge? This would really be a useful functionality to have officially