本文整理汇总了Python中baselines.her.ddpg.DDPG属性的典型用法代码示例。如果您正苦于以下问题：Python ddpg.DDPG属性的具体用法？Python ddpg.DDPG怎么用？Python ddpg.DDPG使用的例子？那么恭喜您, 这里精选的属性代码示例或许可以为您提供帮助。您也可以进一步了解该属性所在类baselines.her.ddpg的用法示例。
在下文中一共展示了ddpg.DDPG属性的5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: prepare_params
# 需要导入模块: from baselines.her import ddpg [as 别名]
# 或者: from baselines.her.ddpg import DDPG [as 别名]
def prepare_params(kwargs):
# DDPG params
ddpg_params = dict()
env_name = kwargs['env_name']
def make_env():
return gym.make(env_name)
kwargs['make_env'] = make_env
tmp_env = cached_make_env(kwargs['make_env'])
assert hasattr(tmp_env, '_max_episode_steps')
kwargs['T'] = tmp_env._max_episode_steps
tmp_env.reset()
kwargs['max_u'] = np.array(kwargs['max_u']) if type(kwargs['max_u']) == list else kwargs['max_u']
kwargs['gamma'] = 1. - 1. / kwargs['T']
if 'lr' in kwargs:
kwargs['pi_lr'] = kwargs['lr']
kwargs['Q_lr'] = kwargs['lr']
del kwargs['lr']
for name in ['buffer_size', 'hidden', 'layers',
'network_class',
'polyak',
'batch_size', 'Q_lr', 'pi_lr',
'norm_eps', 'norm_clip', 'max_u',
'action_l2', 'clip_obs', 'scope', 'relative_goals']:
ddpg_params[name] = kwargs[name]
kwargs['_' + name] = kwargs[name]
del kwargs[name]
kwargs['ddpg_params'] = ddpg_params
return kwargs 示例2: configure_ddpg
# 需要导入模块: from baselines.her import ddpg [as 别名]
# 或者: from baselines.her.ddpg import DDPG [as 别名]
def configure_ddpg(dims, params, reuse=False, use_mpi=True, clip_return=True):
sample_her_transitions = configure_her(params)
# Extract relevant parameters.
gamma = params['gamma']
rollout_batch_size = params['rollout_batch_size']
ddpg_params = params['ddpg_params']
input_dims = dims.copy()
# DDPG agent
env = cached_make_env(params['make_env'])
env.reset()
ddpg_params.update({'input_dims': input_dims, # agent takes an input observations
'T': params['T'],
'clip_pos_returns': True, # clip positive returns
'clip_return': (1. / (1. - gamma)) if clip_return else np.inf, # max abs of return
'rollout_batch_size': rollout_batch_size,
'subtract_goals': simple_goal_subtract,
'sample_transitions': sample_her_transitions,
'gamma': gamma,
})
ddpg_params['info'] = {
'env_name': params['env_name'],
}
policy = DDPG(reuse=reuse, **ddpg_params, use_mpi=use_mpi)
return policy 示例3: prepare_params
# 需要导入模块: from baselines.her import ddpg [as 别名]
# 或者: from baselines.her.ddpg import DDPG [as 别名]
def prepare_params(kwargs):
# DDPG params
ddpg_params = dict()
env_name = kwargs['env_name']
def make_env():
return gym.make(env_name)
kwargs['make_env'] = make_env
tmp_env = cached_make_env(kwargs['make_env'])
assert hasattr(tmp_env, '_max_episode_steps')
kwargs['T'] = tmp_env._max_episode_steps
tmp_env.reset()
kwargs['max_u'] = np.array(kwargs['max_u']) if isinstance(kwargs['max_u'], list) else kwargs['max_u']
kwargs['gamma'] = 1. - 1. / kwargs['T']
if 'lr' in kwargs:
kwargs['pi_lr'] = kwargs['lr']
kwargs['Q_lr'] = kwargs['lr']
del kwargs['lr']
for name in ['buffer_size', 'hidden', 'layers',
'network_class',
'polyak',
'batch_size', 'Q_lr', 'pi_lr',
'norm_eps', 'norm_clip', 'max_u',
'action_l2', 'clip_obs', 'scope', 'relative_goals']:
ddpg_params[name] = kwargs[name]
kwargs['_' + name] = kwargs[name]
del kwargs[name]
kwargs['ddpg_params'] = ddpg_params
return kwargs 示例4: configure_ddpg
# 需要导入模块: from baselines.her import ddpg [as 别名]
# 或者: from baselines.her.ddpg import DDPG [as 别名]
def configure_ddpg(dims, params, reuse=False, use_mpi=True, clip_return=True):
sample_her_transitions = configure_her(params)
# Extract relevant parameters.
gamma = params['gamma']
rollout_batch_size = params['rollout_batch_size']
ddpg_params = params['ddpg_params']
input_dims = dims.copy()
# DDPG agent
env = cached_make_env(params['make_env'])
env.reset()
ddpg_params.update({'input_dims': input_dims, # agent takes an input observations
'T': params['T'],
'clip_pos_returns': True, # clip positive returns
'clip_return': (1. / (1. - gamma)) if clip_return else np.inf, # max abs of return
'rollout_batch_size': rollout_batch_size,
'subtract_goals': simple_goal_subtract,
'sample_transitions': sample_her_transitions,
'gamma': gamma,
'bc_loss': params['bc_loss'],
'q_filter': params['q_filter'],
'num_demo': params['num_demo'],
'demo_batch_size': params['demo_batch_size'],
'prm_loss_weight': params['prm_loss_weight'],
'aux_loss_weight': params['aux_loss_weight'],
})
ddpg_params['info'] = {
'env_name': params['env_name'],
}
policy = DDPG(reuse=reuse, **ddpg_params, use_mpi=use_mpi)
return policy 示例5: prepare_params
# 需要导入模块: from baselines.her import ddpg [as 别名]
# 或者: from baselines.her.ddpg import DDPG [as 别名]
def prepare_params(kwargs):
# DDPG params
ddpg_params = dict()
env_name = kwargs['env_name']
def make_env(subrank=None):
env = gym.make(env_name)
if subrank is not None and logger.get_dir() is not None:
try:
from mpi4py import MPI
mpi_rank = MPI.COMM_WORLD.Get_rank()
except ImportError:
MPI = None
mpi_rank = 0
logger.warn('Running with a single MPI process. This should work, but the results may differ from the ones publshed in Plappert et al.')
max_episode_steps = env._max_episode_steps
env = Monitor(env,
os.path.join(logger.get_dir(), str(mpi_rank) + '.' + str(subrank)),
allow_early_resets=True)
# hack to re-expose _max_episode_steps (ideally should replace reliance on it downstream)
env = gym.wrappers.TimeLimit(env, max_episode_steps=max_episode_steps)
return env
kwargs['make_env'] = make_env
tmp_env = cached_make_env(kwargs['make_env'])
assert hasattr(tmp_env, '_max_episode_steps')
kwargs['T'] = tmp_env._max_episode_steps
kwargs['max_u'] = np.array(kwargs['max_u']) if isinstance(kwargs['max_u'], list) else kwargs['max_u']
kwargs['gamma'] = 1. - 1. / kwargs['T']
if 'lr' in kwargs:
kwargs['pi_lr'] = kwargs['lr']
kwargs['Q_lr'] = kwargs['lr']
del kwargs['lr']
for name in ['buffer_size', 'hidden', 'layers',
'network_class',
'polyak',
'batch_size', 'Q_lr', 'pi_lr',
'norm_eps', 'norm_clip', 'max_u',
'action_l2', 'clip_obs', 'scope', 'relative_goals']:
ddpg_params[name] = kwargs[name]
kwargs['_' + name] = kwargs[name]
del kwargs[name]
kwargs['ddpg_params'] = ddpg_params
return kwargs