EricMingle69/ppo-SpaceInvadersNoFrameskip-v4

PPO Agent playing SpaceInvadersNoFrameskip-v4

This is a trained model of a PPO agent playing SpaceInvadersNoFrameskip-v4 using the stable-baselines3 library.

Trained on Apple Silicon (MPS) as part of the Hugging Face Deep RL Course, Unit 3.

Usage

import ale_py
from stable_baselines3 import PPO
from stable_baselines3.common.env_util import make_atari_env
from stable_baselines3.common.vec_env import VecFrameStack
from huggingface_sb3 import load_from_hub

checkpoint = load_from_hub("EricMingle69/ppo-SpaceInvadersNoFrameskip-v4", "ppo-SpaceInvadersNoFrameskip-v4.zip")
model = PPO.load(checkpoint)
env = VecFrameStack(make_atari_env("SpaceInvadersNoFrameskip-v4", n_envs=1), n_stack=4)

Evaluation

mean_reward = 452.80 +/- 101.73 over 50 episodes, deterministic=True, true game score (no reward clipping, no episodic-life).