def watch(sess_):
with sess_.as_default():
env_monitor = GymEnvironment(env_name=env_name,
log_path=None,
render=True)
info_ = env_monitor.reset()[brain_name]
trainer_monitor = Trainer(ppo_model, sess_, info_, is_continuous,
use_observations, use_states, False)
steps_ = sess_.run([ppo_model.global_step])
print("Starting watcher.")
while True:
done = False
info_ = env_monitor.reset()[brain_name]
while not done:
info_ = trainer_monitor.take_action(info_,
env_monitor,
brain_name,
steps_,
normalize,
stochastic=False)
done = info_.local_done[0]
# summary-freq=<n> Frequency at which to save training statistics [default: 10000].
summary_freq = buffer_size * 5
# save-freq=<n> Frequency at which to save model [default: 50000].
save_freq = summary_freq
# train Whether to train model, or only run inference [default: False].
train_model = True
# render environment to display progress
render = True
# save recordings of episodes
record = False
os.environ[
"CUDA_VISIBLE_DEVICES"] = "-1" #"0"# 0 is effective GPU #"-1" # GPU is not efficient here
env_name = 'RocketLander-v0'
env = GymEnvironment(env_name=env_name, log_path="./PPO_log", skip_frames=5)
env_render = GymEnvironment(env_name=env_name,
log_path="./PPO_log_render",
render=True,
record=record)
fps = env_render.env.metadata.get('video.frames_per_second', 30)
print(str(env))
brain_name = env.external_brain_names[0]
tf.reset_default_graph()
ppo_model = create_agent_model(env,
lr=learning_rate,
h_size=hidden_units,
epsilon=epsilon,
hidden_units = 128
# lambd=<n> Lambda parameter for GAE [default: 0.95].
lambd = 0.95
# learning-rate=<rate> Model learning rate [default: 3e-4].
learning_rate = 3e-4
# normalize Whether to normalize the state input using running statistics [default: False].
normalize = False
# num-epoch=<n> Number of gradient descent steps per batch of experiences [default: 5].
num_epoch = 5
# num-layers=<n> Number of hidden layers between state/observation and outputs [default: 2].
num_layers = 1
# time-horizon=<n> How many steps to collect per agent before adding to buffer [default: 2048].
time_horizon = 1024
env_name = 'RocketLander-v0'
env = GymEnvironment(env_name=env_name, log_path="./PPO_log")
print(str(env))
brain_name = env.external_brain_names[0]
tf.reset_default_graph()
ppo_model = create_agent_model(env,
lr=learning_rate,
h_size=hidden_units,
epsilon=epsilon,
beta=beta,
max_step=max_steps,
normalize=normalize,
num_layers=num_layers)