evaluate.py

import gymnasium as gym
from stable_baselines3 import PPO
from stable_baselines3.common.evaluation import evaluate_policy
import numpy as np

# Load model
model = PPO.load("models/best_model")
env = gym.make("LunarLander-v2")

# Evaluate
mean_reward, std_reward = evaluate_policy(model, env, n_eval_episodes=20)
print(f"Mean Reward: {mean_reward:.2f} +/- {std_reward:.2f}")

# Run 5 episodes and print per-episode reward
print("\nSample episodes:")
for ep in range(5):
    obs, _ = env.reset()
    total_reward = 0
    done = False
    steps = 0
    while not done:
        action, _ = model.predict(obs, deterministic=True)
        obs, reward, terminated, truncated, _ = env.step(action)
        total_reward += reward
        done = terminated or truncated
        steps += 1
    print(f"  Episode {ep+1}: Reward = {total_reward:.2f}, Steps = {steps}")

env.close()