RL frameworks

To be fleshed out

My primary experience is with TensorFlow, therefore any methods that don't support it aren't considered.

A good source of information

• https://winderresearch.com/a-comparison-of-reinforcement-learning-frameworks-dopamine-rllib-keras-rl-coach-trfl-tensorforce-coach-and-more/

Keras-RL had last commit in 2019, basically dead now although still functional. (e.g. documentation is partially completed)

Tensorforce looks pretty good, RLLib looks even better but is quite different.

Google Dopamine and TF-Agents are both from Google researchers. Dopamine aims at fast prototyping vs TF-Agent for production-level rl algorithms.

Actually garage looks pretty good, but it's documentation is less exposed than other methods.