Open Akella17 opened 3 years ago
NAF might be a little bit outdated (I'm not planning to maintain this repository).
Is there a specific reason to use NAF?
I personally find it easier to tune soft actor-critic on new tasks and there is a bunch of well tested SAC implementations available. For example:
https://github.com/denisyarats/pytorch_sac or https://github.com/ikostrikov/jax-rl
I am working on an extension to NAF, so was am looking for a stable NAF implementation.
The NAF algorithm does not work on Pendulum or any of the PyBullet environments. @ikostrikov Do you have any guesses why that might be the case? Which environments did you experiment with this code on? In case you used different hyperparameters than the default values, could you mention the changes that need to be made to get the NAF algorithm working.