# save the configuration and other files
rsg_root = os.path.dirname(os.path.abspath(__file__)) + '/../'
log_dir = rsg_root + '/QuadrotorTrainingdata'
saver = ConfigurationSaver(
log_dir=log_dir + '/quadrotor_position_tracking',
save_items=[
rsg_root + 'raisim_gym/env/env/hummingbird/Environment.hpp',
cfg_abs_path
])
# create environment from the configuration file
if args.mode == "test": # for test mode, force # of env to 1
cfg['environment']['num_envs'] = 1
env = Environment(
RaisimGymEnv(__RSCDIR__, dump(cfg['environment'], Dumper=RoundTripDumper)))
if mode == 'train':
# Get algorithm
model = PPO2(
tensorboard_log=saver.data_dir,
policy=MlpPolicy,
policy_kwargs=dict(net_arch=[dict(pi=[96, 64], vf=[96, 64])]),
env=env,
gamma=0.998,
n_steps=math.floor(cfg['environment']['max_time'] /
cfg['environment']['control_dt']),
ent_coef=0,
learning_rate=1e-3,
vf_coef=0.5,
mode = args.mode
cfg_abs_path = parser.parse_args().cfg
cfg = YAML().load(open(cfg_abs_path, 'r'))
# save the configuration and other files
rsg_root = os.path.dirname(os.path.abspath(__file__)) + '/../cartpole'
log_dir = rsg_root + '/data'
saver = ConfigurationSaver(
log_dir=log_dir + '/Cartpole_tutorial',
save_items=[rsg_root + '/Environment.hpp', cfg_abs_path])
# create environment from the configuration file
if args.mode == "test": # for test mode, force # of env to 1
cfg['environment']['num_envs'] = 1
env = Environment(
RaisimGymEnv(current_dir + "/rsc",
dump(cfg['environment'], Dumper=RoundTripDumper)))
if mode == 'train':
# tensorboard, this will open your default browser.
TensorboardLauncher(saver.data_dir + '/PPO2_1')
# Get algorithm
model = PPO2(
tensorboard_log=saver.data_dir,
policy=MlpPolicy,
policy_kwargs=dict(net_arch=[dict(pi=[128, 128], vf=[128, 128])]),
env=env,
gamma=0.998,
n_steps=math.floor(cfg['environment']['max_time'] /
cfg['environment']['control_dt']),
ent_coef=0,
learning_rate=cfg['environment']['learning_rate'],