help='number of episodes (default: 128)')
parser.add_argument('--updates_per_step',
type=int,
default=5,
metavar='N',
help='model updates per simulator step (default: 5)')
parser.add_argument('--num-stack',
type=int,
default=1,
help='number of frames to stack')
parser.add_argument('--model-suffix',
default="",
help='To resume training or not')
args = parser.parse_args()
env = NormalizedActions(gym.make(args.env_name))
writer = SummaryWriter()
env.seed(args.seed)
if torch.cuda.is_available():
device = torch.device("cuda:0")
torch.cuda.manual_seed(args.seed)
else:
device = torch.device("cpu")
torch.manual_seed(args.seed)
np.random.seed(args.seed)
obs_shape = env.observation_space.shape
obs_shape = (obs_shape[0] * args.num_stack, *obs_shape[1:])
parser.set_defaults(done_util=True)
parser.add_argument('--render', dest='render', action='store_true')
parser.add_argument('--no_render', dest='render', action='store_false')
parser.set_defaults(render=False)
parser.add_argument('--record', dest='record', action='store_true')
parser.add_argument('--no-record', dest='record', action='store_false')
parser.set_defaults(record=False)
args = parser.parse_args()
if __name__ == '__main__':
env_name = args.env
try:
env = NormalizedActions(envs.env_list[env_name](render=args.render))
except TypeError as err:
print('no argument render, assumping env.render will just work')
env = NormalizedActions(envs.env_list[env_name]())
assert np.any(np.abs(env.action_space.low) <= 1.) and np.any(
np.abs(env.action_space.high) <= 1.), 'Action space not normalizd'
if args.record:
env = gym.wrappers.Monitor(env,
'./data/vid/mpc/{}-{}'.format(
env_name, args.frame),
force=True)
env.reset()
env.seed(args.seed)
if args.gpu >= 0:
print("gpu ok")
ptu.set_gpu_mode(True, args.gpu)
# set env
if args.env_name == 'Humanoidrllab':
from rllab.envs.mujoco.humanoid_env import HumanoidEnv
from rllab.envs.normalized_env import normalize
env = normalize(HumanoidEnv())
max_episode_steps = float('inf')
if args.seed >= 0:
global seed_
seed_ = args.seed
else:
env = gym.make(args.env_name)
max_episode_steps=env._max_episode_steps
env=NormalizedActions(env)
if args.seed >= 0:
env.seed(args.seed)
if args.seed >= 0:
torch.manual_seed(args.seed)
np.random.seed(args.seed)
random.seed(args.seed)
torch.random.manual_seed(args.seed)
if torch.cuda.is_available():
torch.cuda.manual_seed(args.seed)
torch.backends.cudnn.deterministic = True
torch.backends.cudnn.benchmark = False
# set args
args.num_actions = env.action_space.shape[0]
args.max_action = env.action_space.high
if __name__ == '__main__':
env = sys.argv[1]
args = None
if env == 'mc':
args = args_mc
elif env == 'pd':
args = args_pd
elif env == 'll':
args = args_ll
else:
print('Environment not selected, Please choose from: mc, pd,ll')
exit(-1)
env = NormalizedActions(gym.make(args['env_name']))
env.seed(args['seed'])
torch.manual_seed(args['seed'])
np.random.seed(args['seed'])
agent = NAF(args['gamma'], args['tau'], args['hidden_size'],
env.observation_space.shape[0], env.action_space)
agent.load_model(f'models/naf_{args["env_name"]}')
replay_buffer = ReplayBuffer(args['replay_size'])
ounoise = OUNoise(env.action_space.shape[0]) if args['ou_noise'] else None
run()