def _thunk():
env = make_env.make_env(env_id)
env.seed(seed + rank)
env = bench.Monitor(env, logger.get_dir() and os.path.join(logger.get_dir(), str(rank)),
allow_early_resets=True)
gym.logger.setLevel(logging.WARN)
return env
def _thunk():
env = make_env.make_env(env_id, max_episode_len=max_episode_len)
env.discrete_action_input = True
env.seed(seed + rank)
env = bench.Monitor(env,
logger.get_dir()
and os.path.join(logger.get_dir(), str(rank)),
allow_early_resets=True)
gym.logger.setLevel(logging.WARN)
return env
def train(logdir,
env_id,
lr,
num_timesteps,
seed,
timesteps_per_batch,
cont=False):
from sandbox.ppo_sgd import mlp_policy
from sandbox.ppo_sgd import pposgd_simple
from rl import logger
from rl.common import set_global_seeds, tf_util as U
from rl import bench
from gym.envs.registration import register
import multiagent
import make_env
logger.configure(logdir, format_strs=['log', 'json', 'tensorboard'])
U.make_session(num_cpu=1).__enter__()
set_global_seeds(seed)
env = make_env.make_env(env_id)
def policy_fn(name, ob_space, ac_space, id):
pi = mlp_policy.MlpPolicy(name=name,
ob_space=ob_space,
ac_space=ac_space,
hid_size=64,
num_hid_layers=2,
id=id)
return pi
env = bench.Monitor(
env,
logger.get_dir() and osp.join(logger.get_dir(), "monitor.json"))
env.seed(seed)
gym.logger.setLevel(logging.WARN)
pposgd_simple.learn(env,
policy_fn,
max_timesteps=num_timesteps,
timesteps_per_batch=timesteps_per_batch,
clip_param=0.2,
entcoeff=0.0,
optim_epochs=10,
optim_stepsize=lr,
optim_batchsize=64,
gamma=0.99,
lam=0.95,
schedule='linear',
cont=cont)
env.close()
return None
def _make_env():
env = gym.make(env_id)
env = MAWrapper(env)
env = bench.Monitor(env, logger.get_dir())
return env
def _make_env():
env = make_env(env_id) # gym.make(env_id)
env = bench.Monitor(env, logger.get_dir())
return env
def create_env():
env = make_env.make_env('simple_spread')
env.seed(3)
env = bench.Monitor(env, '/tmp/', allow_early_resets=True)
set_global_seeds(3)
return env
def _make_env(rank):
env = gym.make('RoboSumo-Ant-vs-Ant-v0')
env = bench.Monitor(env, logger.get_dir() and os.path.join(logger.get_dir(), str(rank)))
return env
