My First RL Project

PPO Agent playing LunarLander-v2

This is a trained model of a PPO agent playing LunarLander-v2 using the stable-baselines3 library.

Usage

code was done with gym env and stable-basline3 libraray

#Dependencies and stuff 
!apt install swig cmake
!pip install -r https://raw.githubusercontent.com/huggingface/deep-rl-class/main/notebooks/unit1/requirements-unit1.txt
!sudo apt-get update
!apt install python3-opengl
!apt install ffmpeg
!apt install xvfb
!pip3 install pyvirtualdisplay
# restart colab
import os
os.kill(os.getpid(), 9)
#display
from pyvirtualdisplay import Display

virtual_display = Display(visible=0, size=(1400, 900))
virtual_display.start()
# import libraries
import gymnasium as gym

from huggingface_sb3 import load_from_hub, package_to_hub
from huggingface_hub import (
    notebook_login,
) 

from stable_baselines3 import PPO
from stable_baselines3.common.env_util import make_vec_env
from stable_baselines3.common.evaluation import evaluate_policy
from stable_baselines3.common.monitor import Monitor

# Create environment
env = gym.make('LunarLander-v2')

#Define PPO 
model = PPO(
    policy="MlpPolicy",
    env=env,
    n_steps=1024,
    batch_size=64,
    n_epochs=4,
    gamma=0.999,
    gae_lambda=0.98,
    ent_coef=0.01,
    verbose=1,
)
# Train the agent
model.learn(total_timesteps=1000000)

# Save the model
model_name = "ppo-LunarLander-v2"
model.save(model_name)

#evaluate model
eval_env = Monitor(gym.make("LunarLander-v2"))
mean_reward, std_reward = evaluate_policy(model, eval_env, n_eval_episodes=10, deterministic=True)
print(f"mean_reward={mean_reward:.2f} +/- {std_reward}")

# create a video (for colab)

import gymnasium as gym
from stable_baselines3 import PPO
from IPython.display import Video, display
import os

env = gym.make('LunarLander-v2')

model_name = "ppo-LunarLander-v2"
model = PPO.load(model_name)

def record_video(env, model, video_length=500, prefix="ppo-lunarlander"):
    env = gym.wrappers.RecordVideo(env, video_folder=prefix, episode_trigger=lambda x: x == 0)
    obs = env.reset()
    for _ in range(video_length):
        action, _ = model.predict(obs)
        obs, _, done, _ = env.step(action)
        if done:
            obs = env.reset()
    env.close()

record_video(env, model, video_length=500, prefix="ppo-lunarlander")

video_path = "ppo-lunarlander/rl-video-episode-0.mp4"
display(Video(video_path))
...