本文整理汇总了Python中baselines.trpo_mpi.nosharing_cnn_policy.CnnPolicy方法的典型用法代码示例。如果您正苦于以下问题：Python nosharing_cnn_policy.CnnPolicy方法的具体用法？Python nosharing_cnn_policy.CnnPolicy怎么用？Python nosharing_cnn_policy.CnnPolicy使用的例子？那么, 这里精选的方法代码示例或许可以为您提供帮助。您也可以进一步了解该方法所在类baselines.trpo_mpi.nosharing_cnn_policy的用法示例。
在下文中一共展示了nosharing_cnn_policy.CnnPolicy方法的3个代码示例，这些例子默认根据受欢迎程度排序。您可以为喜欢或者感觉有用的代码点赞，您的评价将有助于系统推荐出更棒的Python代码示例。
示例1: train
# 需要导入模块: from baselines.trpo_mpi import nosharing_cnn_policy [as 别名]
# 或者: from baselines.trpo_mpi.nosharing_cnn_policy import CnnPolicy [as 别名]
def train(env_id, num_timesteps, seed):
from baselines.trpo_mpi.nosharing_cnn_policy import CnnPolicy
from baselines.trpo_mpi import trpo_mpi
import baselines.common.tf_util as U
rank = MPI.COMM_WORLD.Get_rank()
sess = U.single_threaded_session()
sess.__enter__()
if rank == 0:
logger.configure()
else:
logger.configure(format_strs=[])
workerseed = seed + 10000 * MPI.COMM_WORLD.Get_rank()
set_global_seeds(workerseed)
env = make_atari(env_id)
def policy_fn(name, ob_space, ac_space): #pylint: disable=W0613
return CnnPolicy(name=name, ob_space=env.observation_space, ac_space=env.action_space)
env = bench.Monitor(env, logger.get_dir() and osp.join(logger.get_dir(), str(rank)))
env.seed(workerseed)
env = wrap_deepmind(env)
env.seed(workerseed)
trpo_mpi.learn(env, policy_fn, timesteps_per_batch=512, max_kl=0.001, cg_iters=10, cg_damping=1e-3,
max_timesteps=int(num_timesteps * 1.1), gamma=0.98, lam=1.0, vf_iters=3, vf_stepsize=1e-4, entcoeff=0.00)
env.close() 示例2: train
# 需要导入模块: from baselines.trpo_mpi import nosharing_cnn_policy [as 别名]
# 或者: from baselines.trpo_mpi.nosharing_cnn_policy import CnnPolicy [as 别名]
def train(env_id, num_timesteps, seed, num_cpu):
from baselines.trpo_mpi.nosharing_cnn_policy import CnnPolicy
from baselines.trpo_mpi import trpo_mpi
import baselines.common.tf_util as U
whoami = mpi_fork(num_cpu)
if whoami == "parent":
return
rank = MPI.COMM_WORLD.Get_rank()
sess = U.single_threaded_session()
sess.__enter__()
logger.session().__enter__()
if rank != 0:
logger.set_level(logger.DISABLED)
workerseed = seed + 10000 * MPI.COMM_WORLD.Get_rank()
set_global_seeds(workerseed)
env = gym.make(env_id)
def policy_fn(name, ob_space, ac_space): #pylint: disable=W0613
return CnnPolicy(name=name, ob_space=env.observation_space, ac_space=env.action_space)
env = bench.Monitor(env, osp.join(logger.get_dir(), "%i.monitor.json"%rank))
env.seed(workerseed)
gym.logger.setLevel(logging.WARN)
env = wrap_train(env)
num_timesteps /= 4 # because we're wrapping the envs to do frame skip
env.seed(workerseed)
trpo_mpi.learn(env, policy_fn, timesteps_per_batch=512, max_kl=0.001, cg_iters=10, cg_damping=1e-3,
max_timesteps=num_timesteps, gamma=0.98, lam=1.0, vf_iters=3, vf_stepsize=1e-4, entcoeff=0.00)
env.close() 示例3: train
# 需要导入模块: from baselines.trpo_mpi import nosharing_cnn_policy [as 别名]
# 或者: from baselines.trpo_mpi.nosharing_cnn_policy import CnnPolicy [as 别名]
def train(env_id, num_timesteps, seed):
from baselines.trpo_mpi.nosharing_cnn_policy import CnnPolicy
from baselines.trpo_mpi import trpo_mpi
import baselines.common.tf_util as U
rank = MPI.COMM_WORLD.Get_rank()
sess = U.single_threaded_session()
sess.__enter__()
if rank == 0:
logger.configure()
else:
logger.configure(format_strs=[])
workerseed = seed + 10000 * MPI.COMM_WORLD.Get_rank()
set_global_seeds(workerseed)
env = make_atari(env_id)
def policy_fn(name, ob_space, ac_space): #pylint: disable=W0613
return CnnPolicy(name=name, ob_space=env.observation_space, ac_space=env.action_space)
env = bench.Monitor(env, logger.get_dir() and osp.join(logger.get_dir(), str(rank)))
env.seed(workerseed)
gym.logger.setLevel(logging.WARN)
env = wrap_deepmind(env)
env.seed(workerseed)
trpo_mpi.learn(env, policy_fn, timesteps_per_batch=512, max_kl=0.001, cg_iters=10, cg_damping=1e-3,
max_timesteps=int(num_timesteps * 1.1), gamma=0.98, lam=1.0, vf_iters=3, vf_stepsize=1e-4, entcoeff=0.00)
env.close()