/deep_rl

WIP: Deep Reinforcement Learning with Tensorflow

Primary LanguagePythonMIT LicenseMIT

Deep RL

Very much a WIP, APIs subject to changing.

Library for Deep Reinforcement Learning in Tensorflow and integrating with OpenAI Gym.

TODO

  • Make setup.py, alternatively can just set PYTHONPATH environment 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