jayden118/rl-baselines

Reimplementing some basic RL algos

On the task list:

• Vanilla Policy Gradient
• PPO [Paper] (https://arxiv.org/abs/1707.06347) Code
• RDN (sort of done, but does not match efficiency of OpenAI's implementation Paper Code
• World models Convolutional Reservoir Computing Paper

Gym environements that can be solved:

CartPole-v0

python -m rl_baselines.reinforce --env CartPole-v0

LunarLander-v2

python -m rl_baselines.ppo --env LunarLander-v2

CarRacing-v0

python -m rl_baselines.rcrc --env CarRacing-v0

Trying out an agent

Whenever you launch a baseline, a run pops up in runs/.... You can check how an agent is currently performing by using.

python -m rl_baselines.test_agent --model=runs/...../checkpoint.pth --env=$(ENV_NAME)