metadata
library_name: stable-baselines3
tags:
- LunarLander-v2
- deep-reinforcement-learning
- reinforcement-learning
- stable-baselines3
model-index:
- name: LunarLander-Kratuzen
results:
- task:
type: moon-landing-training
name: reinforcement-learning-training
dataset:
name: state-action-landing-data
type: reinforcement-learning-generated-data
metrics:
- type: mean_reward
value: 266.40 +/- 21.38
name: mean_reward
verified: false
🚀 PPO Agent: LunarLander-Kratuzen
This is a trained PPO (Proximal Policy Optimization) agent for the LunarLander-v2 environment, built with Stable-Baselines3.
Repo ID: KraTUZen/LunarLander
Model name: LunarLander-Kratuzen
📊 Performance
- Mean Reward: 266.40 ± 21.38
- Episodes Evaluated: 10
- ✅ Consistently lands successfully, showing stability and robustness.
🛠️ Usage
from huggingface_sb3 import load_from_hub
from stable_baselines3 import PPO
import gymnasium as gym
# Load model from Hugging Face Hub
model = load_from_hub(
repo_id="KraTUZen/LunarLander",
filename="LunarLander-Kratuzen.zip"
)
# Create environment
env = gym.make("LunarLander-v2")
# Run a quick evaluation loop
obs, info = env.reset()
for _ in range(20):
action = env.action_space.sample()
obs, reward, terminated, truncated, info = env.step(action)
if terminated or truncated:
obs, info = env.reset()
env.close()
📦 Training Setup
| Parameter | Value |
|---|---|
| Algorithm | PPO |
| Policy | MlpPolicy |
| Timesteps | 1,000,000 |
| n_steps | 1024 |
| batch_size | 64 |
| gamma | 0.999 |
| gae_lambda | 0.98 |
| ent_coef | 0.01 |
🎯 Key Takeaways
- Achieves high reward and stable landings.
- Ready-to-use with Hugging Face Hub.
- Reproducible training setup for reinforcement learning experiments.