try:
with ServerProxy(addr, allow_none=True) as proxy:
config = proxy.get_actor_config()
num_episodes = config["num_episodes"]
episodes_per_update = config["episodes_per_update"]
eps = config["max_eps"]
min_eps = config["min_eps"]
eps_step = config["eps_step"]
frame_skip_min = config["frame_skip_min"]
frame_skip_max = config["frame_skip_max"]
actor = Actor(**config)
actor.update_policy_net(proxy.get_policy_net().data)
for i in range(1, num_episodes + 1):
reward = actor.act_one_episode(eps, frame_skip_min, frame_skip_max)
if args.verbose:
print("Episode {0}: Total reward = {1} | eps = {2}".format(
i, reward, eps))
if i % episodes_per_update == 0:
start_time = time.time()
policy_net = proxy.add_memories_and_get_policy_net(
actor.memory)
actor.update_policy_net(policy_net.data)
actor.memory.clear()
end_time = time.time()
if args.time:
print("Wall time taken to update: {0} sec".format(
round(end_time - start_time, 2)))
