Gym by OpenAI is an open source interface to reinforcement learning (RL) tasks. It comes with a set of standard tasks useful for benchmarking RL algorithms (CartPole, Pendulum, Walking Humanoid, etc.), and an easy, standardized way to create a learning environment for a research problem. Stable Baselines 3 provides a set of reliable, rigorously tested PyTorch implementations of popular Deep RL algorithms to facilitate results replication. Together they can be used for fast and easy RL benchmarks and prototyping.
In this workshop we will:
- Discover the python gym library and explore standard RL tasks,
- Use Stable Baselines 3 to benchmark RL models in a Gym environment,
- Learn how to create a Gym environment for a specific learning task.