LA Deep RL Meetup 1

Repo for first meetup of LA Deep RL group. You can find the meetup slides in slides.pdf in this repo.

Dev Setup

Developed on OS X 10.15.3 under Python 3.7.6. You can setup your dev environment as follows:

• Clone the repo: git clone https://github.com/nickjalbert/la_deep_rl_meet_1.git
• Enter directory: cd la_deep_rl_meet_1
• Create a virtualenv: virtualenv -p `which python3` la_deep_rl_meet_1
• Activate your virtualenv: source la_deep_rl_meet_1/bin/activate
• Install dependencies:
  • pip install -r requirements.txt or
  • pip install gym ray ray[rllib] ray[debug] pandas requests tensorflow

Files

• environment.py - skeleton of the Corridor environment
• solutions/environment.py - a full implementation fo the Corridor environment
• random_agent.py - skeleton of a random agent
• solutions/random_agent.py - a full implementation of a random agent
• ppo_agent.py - a full implementation of an agent that uses a state of the art PPO algorithm

Commandline Environment Demo

>>> import gym
>>> env = gym.make('CartPole-v0')
>>> env.reset()
>>> env.step(0)

Run environment unit tests

• Skeleton: python environment.py
• Full implementation: cd solutions/ && python environment.py

Run PPO agent

• cd solutions/ && python ppo_agent.py