def main():
args = control_arg_parser().parse_args()
# Training
# ENV_path = get_dir(os.path.join(args.log_dir, args.env))
# log_dir = os.path.join(ENV_path, args.method + "-" +
# '{}'.format(args.seed)) + "-" + \
# datetime.datetime.now().strftime("%m-%d-%H-%M")
# logger.configure(dir=log_dir)
# save_args(args)
# train_copos(args.env, num_timesteps=args.num_timesteps * 1e6, seed=args.seed, trial=args.seed,
# hist_len=args.hist_len, block_high=float(args.block_high), nsteps=args.nsteps,
# method=args.method, hid_size=args.hid_size, give_state=bool(args.give_state), vf_iters=args.epoch)
#Render
load_path = '/Users/zhirong/Documents/ReinforcementLearning/tmp/LunarLanderContinuousPOMDP-v0/copos-guided-try-diffinput-0-07-05-13-34/checkpoints/00976.ckpt'
video_path = '/Users/zhirong/Documents/ReinforcementLearning/tmp/LunarLanderContinuousPOMDP-v0/copos-guided-try-diffinput-0-07-05-13-34/'
render(hid_size=args.hid_size,
load_path=load_path,
video_path=video_path,
env_id=args.env,
seed=0,
hist_len=args.hist_len,
block_high=args.block_high,
give_state=0)
def main():
args = control_arg_parser().parse_args()
if args.env == 'LunarLanderContinuousPOMDP-v0':
newenv(hist_len=args.hist_len)
if args.train is True:
ENV_path = get_dir(os.path.join(args.log_dir, args.env))
log_dir = os.path.join(ENV_path, args.method +"-"+
'{0}'.format(args.seed))+"-" +\
datetime.datetime.now().strftime("%m-%d-%H-%M")
# if MPI.COMM_WORLD.Get_rank() == 0:
logger.configure(dir=log_dir)
save_args(args)
train(args.env,
num_timesteps=args.num_timesteps,
seed=args.seed,
nsteps=args.nsteps,
batch_size=args.batch_size,
epoch=args.epoch,
method=args.method,
net_size=args.net_size,
ncpu=args.ncpu,
i_trial=args.seed,
load_path=args.load_path,
use_entr=int(args.use_entr))
if args.render is True:
video_path = osp.split(osp.split(args.load_path)[0])[0]
render(args.env,
nsteps=args.nsteps,
batch_size=args.batch_size,
net_size=args.net_size,
load_path=args.load_path,
video_path=video_path,
iters=args.iters)
def main():
args = control_arg_parser().parse_args()
# rank = MPI.COMM_WORLD.Get_rank()
# print("This is rank:", rank, "Seed:", args.seed)
if args.env == 'LunarLanderContinuousPOMDP-v0':
newenv(hist_len=args.hist_len,
block_high=float(args.block_high),
policy_name=args.policy_name)
# if args.train is True:
ENV_path = get_dir(os.path.join(args.log_dir, args.env))
log_dir = os.path.join(ENV_path, args.method +"-"+
'{}'.format(args.seed))+"-" +\
datetime.datetime.now().strftime("%m-%d-%H-%M")
logger.configure(dir=log_dir)
# logger.log("This is rank {}".format(rank))
save_args(args)
train(args.env,
num_timesteps=args.num_timesteps,
seed=args.seed,
hist_len=args.hist_len,
nsteps=args.nsteps,
batch_size=args.batch_size,
epoch=args.epoch,
env_name=args.env,
method=args.method,
net_size=tuple(args.net_size),
ncpu=args.ncpu,
policy_name=args.policy_name,
load_path=args.load_path,
use_entr=int(args.use_entr),
rank=args.seed,
checkpoint=args.checkpoint,
filter_size=args.filter_size)
def main():
# args = mujoco_arg_parser().parse_args()
args = control_arg_parser().parse_args()
args.seed = 0
log_path = get_dir("/Users/zhirong/Documents/Masterthesis-code/tmp")
# log_path = get_dir("/home/zhi/Documents/ReinforcementLearning/tmp")
ENV_path = get_dir(os.path.join(log_path, args.env))
log_dir = os.path.join(ENV_path, datetime.datetime.now().strftime("trpo-%m-%d-%H-%M-%S"))
logger.configure(dir=log_dir)
train(args.env, num_timesteps=args.num_timesteps, seed=args.seed)
def main():
args = control_arg_parser().parse_args()
# Training
ENV_path = get_dir(os.path.join(args.log_dir, args.env))
log_dir = os.path.join(ENV_path, args.method + "-" +
'{}'.format(args.seed)) + "-" + \
datetime.datetime.now().strftime("%m-%d-%H-%M")
logger.configure(dir=log_dir)
save_args(args)
train_copos(args.env, num_timesteps=args.num_timesteps * 1e6, seed=args.seed, trial=args.seed,
hist_len=args.hist_len, block_high=float(args.block_high), nsteps=args.nsteps,
method=args.method, hid_size=args.hid_size, give_state=bool(args.give_state), vf_iters=args.epoch)
def main():
args = control_arg_parser().parse_args()
ENV_path = get_dir(os.path.join(args.log_dir, args.env))
log_dir = os.path.join(ENV_path, args.method + "-" +
'{}'.format(args.seed)) + "-" + \
datetime.datetime.now().strftime("%m-%d-%H-%M")
logger.configure(dir=log_dir)
# logger.log("This is rank {}".format(rank))
save_args(args)
train(args.env,
num_timesteps=args.num_timesteps,
seed=args.seed,
trial=args.seed,
hist_len=args.hist_len)
def main():
args = control_arg_parser().parse_args()
ENV_path = get_dir(os.path.join(args.log_dir, args.env))
log_dir = os.path.join(ENV_path, args.method + "-" +
'{}'.format(args.seed)) + "-" + \
datetime.datetime.now().strftime("%m-%d-%H-%M")
logger.configure(dir=log_dir)
save_args(args)
# if args.env == 'LunarLanderContinuousPOMDP-v0':
# newenv(hist_len=args.hist_len, block_high=float(args.block_high), policy_name=args.policy_name)
train_copos(args.env,
num_timesteps=args.num_timesteps * 1e6,
seed=args.seed,
trial=args.seed,
hist_len=args.hist_len,
block_high=float(args.block_high),
nsteps=args.nsteps,
method=args.method,
hid_size=args.hid_size,
give_state=args.give_state,
vf_iters=args.epoch)
