Very much a WIP, APIs subject to changing.
Library for Deep Reinforcement Learning in Tensorflow and integrating with OpenAI Gym.
- Make setup.py, alternatively can just set
PYTHONPATHenvironment variable
Higher-level APIs, ex: A3CAgent class which takes of training, evaluating, monitoring, checkpointing, etc.
- CI, tests
- Parallel trajectory sampling. Either to do this with native Python threading or TFs threading.
- TRPO