You can deploy this model and start earning money today!

SarathL124/
ppo-LunarLander-v3

Name: ppo-LunarLander-v3
Availability: InStock

Views

No views yet

PPO LunarLander-v3 (Stable-Baselines3)

This repository contains a Proximal Policy Optimization (PPO) agent trained with Stable-Baselines3 on Gymnasium LunarLander-v3.
It includes the trained policy, evaluation results, and a short rollout video.

Demo Video

Evaluation Metrics

Evaluation setup

Environment: LunarLander-v3
Episodes: 10
Deterministic: True

Results (results.json)

Mean reward: 251.32
Std reward: 35.21

Training Snapshot

From the final training logs (near ~1M timesteps):

Total timesteps: ~1,000,000
Rollout ep_rew_mean: ~247–248
Rollout ep_len_mean: ~351
FPS: ~526
Learning rate: 3e-4
Clip range: 0.2
Approx KL: ~0.003–0.004
Explained variance: ~0.992–0.993

Note: ep_rew_mean is the rolling training reward; the evaluation reward above is measured separately over 10 episodes.

How to Use

Install dependencies

pip install stable-baselines3 swig gymnasium[box2d]

python

import gymnasium as gym
from stable_baselines3 import PPO

model = PPO.load("ppo_lunarlander")  # replace with your actual model filename in this repo

env = gym.make("LunarLander-v3", render_mode="human")  # use "rgb_array" for headless rendering
obs, info = env.reset()

done = False
while not done:
    action, _ = model.predict(obs, deterministic=True)
    obs, reward, terminated, truncated, info = env.step(action)
    done = terminated or truncated

env.close()

SarathL124/ppo-LunarLander-v3