handeyildirim/ppo-CleanRL-LunarLander-v2

PPO Agent Playing LunarLander-v2

This is a trained model of a PPO agent playing LunarLander-v2, implemented from scratch (CleanRL-style) for Unit 8 Part 1 of the Hugging Face Deep RL Course.

• Training: 2,000,000 timesteps, 8 parallel envs, 512 steps per rollout
• Evaluation: 10 episodes, greedy policy — mean reward -54.03 +/- 12.59

See ppo.py for the implementation and hyperparameters.json for the full configuration.