Function test_agent

rl/cartpole_gym0.19.py:69–80

(model, env, n_episodes=20)

def test_agent(model, env, n_episodes=20):
  reward_per_episode = np.zeros(n_episodes)
  for it in range(n_episodes):
    done = False
    episode_reward = 0
    s = env.reset()
    while not done:
      a = epsilon_greedy(model, s, eps=0)
      s, r, done, info = env.step(a)
      episode_reward += r
    reward_per_episode[it] = episode_reward
  return np.mean(reward_per_episode)


def watch_agent(model, env, eps):