def main():
args = gym_ctrl_arg_parser().parse_args()
logger.configure(format_strs=['stdout', 'log', 'csv'],
log_suffix="OpenAI-ES-" + args.env + "_seed_" +
str(args.seed))
logger.log("Algorithm: OpenAI-ES-" + args.env + "_seed_" + str(args.seed))
train(args.env, num_timesteps=args.num_timesteps, seed=args.seed)
def main():
"""
Train on CartPole.
"""
args = gym_ctrl_arg_parser().parse_args()
logger.configure(format_strs=['stdout', 'log', 'csv'], log_suffix = "UberGA-"+args.env+"_seed_"+str(args.seed))
logger.log("Algorithm:UberGA-" + args.env + "_seed_" + str(args.seed))
env_id = args.env
seed = args.seed
generation = 0
with make_session() as sess:
env = make_gym_control_env(env_id, seed)
try:
model = simple_mlp(sess, env)
sess.run(tf.global_variables_initializer())
learn_sess = LearningSession(sess, model)
while True:
if generation >= 10000 or learn_sess.timesteps_so_far >= 5e6:
break
pop = learn_sess.generation(env, trials=1, population=POPULATION)
generation+=1
finally:
env.close()
def main():
args = gym_ctrl_arg_parser().parse_args()
logger.configure(format_strs=["stdout", 'log', 'csv'],
log_suffix="ACKTR-" + args.env)
import random
seed = 2
logger.log("Algorithm: ACKTR-" + args.env + "_seed_" + str(seed))
train(args.env, num_timesteps=args.num_timesteps, seed=seed)
def main():
args = gym_ctrl_arg_parser().parse_args()
logger.configure(format_strs=['stdout', 'log', 'csv'],
log_suffix="Dual_RAC-" + args.env)
logger.log("Algorithm: Dual_RAC-" + args.env)
import random
args.seed += random.randint(0, 2**32 - 1)
logger.log("Algorithm: SEED-" + str(args.seed))
train(args.env, num_timesteps=args.num_timesteps, seed=args.seed)
def main():
args = gym_ctrl_arg_parser().parse_args()
logger.configure(format_strs=['stdout', 'log', 'csv'],
log_suffix="PES-S2-" + args.env + "_seed_" +
str(args.seed))
import random
seed = args.seed + random.randint(0, 2**32 - 1)
logger.log("Algorithm: PES-S2-" + args.env + "_seed_" + str(seed))
train(args.env, num_timesteps=args.num_timesteps, seed=seed)
def main():
parser = gym_ctrl_arg_parser()
parser.add_argument('--policy',
help='Policy architecture',
choices=['cnn', 'lstm', 'lnlstm', 'mlp'],
default='mlp')
parser.add_argument('--lrschedule',
help='Learning rate schedule',
choices=['constant', 'linear'],
default='linear')
args = parser.parse_args()
logger.configure(format_strs=['stdout', 'log', 'csv'],
log_suffix="A2C-" + args.env)
train(args.env,
num_timesteps=args.num_timesteps,
seed=args.seed,
policy=args.policy,
lrschedule=args.lrschedule,
num_env=1,
env_name=args.env)
def main():
args = gym_ctrl_arg_parser().parse_args()
logger.configure(format_strs=['stdout', 'log', 'csv'],
log_suffix="PPO_NAC_Advantage-" + args.env)
logger.log("Algorithm: PPO_NAC_Advantage-" + args.env)
train(args.env, num_timesteps=args.num_timesteps, seed=args.seed)
def main():
args = gym_ctrl_arg_parser().parse_args()
logger.configure(format_strs=['stdout', 'log', 'csv'],
log_suffix="PPO_CMAES-" + args.env)
train(args.env, num_timesteps=args.num_timesteps, seed=args.seed)
