best_score = -np.inf
scores = []
eps_history = []
state_dims = env.observation_space.shape[0]
num_actions = env.action_space.n
lr = 0.001
gamma = 0.99
agent = PolicyGradientAgent(lr=lr,
gamma=gamma,
state_dims=state_dims,
num_actions=num_actions,
env_name='lunar_lander',
checkpoint_dir='temp/')
if test_mode:
agent.load_model()
# env = gym.wrappers.Monitor(env, 'temp/lunar_lander',
# video_callable=lambda episode_id: True, force=True)
for count in range(num_games):
state = env.reset()
done = False
score = 0
while not done:
env.render()
action = agent.get_action(state)
new_state, reward, done, _ = env.step(action)
agent.reward_history.append(reward)
