Trained with Stable-Baselines3 PPO for 1M timesteps as part of the
Hugging Face Deep RL Course (Unit 8 — Proximal Policy Optimization).
Note: trained on LunarLander-v3 from current Gymnasium (the v2 environment
is no longer registered). The dynamics and reward structure are identical
to v2; tags reflect the certification benchmark name.
Evaluation: mean reward = 271.21 +/- 17.91 over 10 episodes.
