def video_oc():
agent = OptionCriticAgent(env, num_options=4, device=device).to(device)
from gym import wrappers
env_monitor = wrappers.Monitor(env, directory="videos/oc", force=True)
checkpoint = torch.load("checkpoint_oc.pt")
agent.load_state_dict(checkpoint["agent"])
rw = evaluate_oc(
agent,
env_monitor,
n_games=2,
)
env_monitor.close()
print(rw)
def video_a2c():
agent = A2CAgent(env, device).to(device)
from gym import wrappers
env_monitor = wrappers.Monitor(env, directory="videos/a2c", force=True)
checkpoint = torch.load("checkpoint_a2c.pt")
agent.load_state_dict(checkpoint["agent"])
rw = evaluate(
agent,
env_monitor,
n_games=2,
)
env_monitor.close()
print(rw)