README.md

metadata

library_name: stable-baselines3
tags:
  - LunarLander-v2
  - deep-reinforcement-learning
  - reinforcement-learning
  - stable-baselines3
model-index:
  - name: PPO
    results:
      - task:
          type: reinforcement-learning
          name: reinforcement-learning
        dataset:
          name: LunarLander-v2
          type: LunarLander-v2
        metrics:
          - type: mean_reward
            value: 254.49 +/- 35.78
            name: mean_reward
            verified: false

PPO Agent playing LunarLander-v2

This is a trained model of a PPO agent playing LunarLander-v2 using the Stable-Baselines3 library.

Installation

Before running the model, follow these steps to set up a virtual environment, install dependencies, and run the model.

1️⃣ Create and activate a virtual environment

# Create a virtual environment named "py_env_lunar_lander"
python3 -m venv py_env_lunar_lander

# Activate the virtual environment
# On macOS/Linux:
source py_env_lunar_lander/bin/activate

2️⃣ Install dependencies

# Update package list and install system dependencies
sudo apt-get update
sudo apt-get install -y python3-opengl swig
apt install ffmpeg
apt install xvfb

# Install required Python packages
pip install --upgrade pip
pip install pyvirtualdisplay imageio huggingface_sb3 torch
pip install stable-baselines3==2.0.0a5 gymnasium[box2d]

Usage (with Stable-baselines3)

After installing dependencies, run the following script to load the trained model, run an episode, and generate a video.

import gymnasium as gym
import torch
import imageio
from stable_baselines3 import PPO
from huggingface_sb3 import load_from_hub
from IPython.display import Video

# Define the repo ID and model filename
repo_id = "aiamnoone/ppo-LunarLander-v2"  # Your Hugging Face repo ID
filename = "ppo-LunarLander-v2.zip"  # Model file name

# Download and load the model from Hugging Face Hub
model_path = load_from_hub(repo_id, filename)
model = PPO.load(model_path)

# Create the environment
env = gym.make("LunarLander-v2", render_mode="rgb_array")

obs = env.reset()[0]  # Get only the observation
done = False
frames = []

# Run one episode and collect frames
while not done:
    action, _ = model.predict(obs, deterministic=True)  # Use trained model
    obs, reward, done, truncated, info = env.step(action)
    frame = env.render()  # Get the frame (RGB image)
    frames.append(frame)  # Save the frame

env.close()

# Save the frames as a video
video_path = "lunar_lander.mp4"
imageio.mimsave(video_path, frames, fps=30)  # Adjust FPS if needed

# Display the video
print("Video saved as 'lunar_lander.mp4'. Open it to watch!")


...