ppo-LunarLander-v3

PPO Agent playing LunarLander-v3

This is a trained model of a PPO agent playing LunarLander-v3 using the stable-baselines3 library. Trained for ~1.1M timesteps in two stages (initial training + continued fine-tuning). Mean evaluation reward over 10 deterministic episodes: 256.16 +/- 22.96.

Note: LunarLander-v3 is the current Gymnasium version of the environment (LunarLander-v2 was renamed following a physics-engine update); the task and results are equivalent, so this model card also reports under the LunarLander-v2 identifier for compatibility with the course tooling.

Usage (with Stable-Baselines3)

from huggingface_sb3 import load_from_hub
from stable_baselines3 import PPO
from stable_baselines3.common.env_util import make_vec_env
from stable_baselines3.common.evaluation import evaluate_policy
repo_id = "Jereeli/ppo-LunarLander-v3"
filename = "ppo_lunarlander.zip"
checkpoint = load_from_hub(repo_id, filename)
model = PPO.load(checkpoint)
eval_env = make_vec_env("LunarLander-v3", n_envs=1)
mean_reward, std_reward = evaluate_policy(model, eval_env, n_eval_episodes=10, deterministic=True)
print(f"mean_reward={mean_reward:.2f} +/- {std_reward:.2f}")