def ad_connect(username, password, hosts):
valid, message = validate_connect(username, password, hosts)
if not valid:
return None, message
hosts = get_valid_hosts(hosts)
host = first(hosts)
address = 'ldap://{}'.format(host)
conn = ldap.initialize(address)
conn.protocol_version = 3
conn.set_option(ldap.OPT_REFERRALS, 0)
try:
conn.simple_bind_s(username, password)
return conn, 'Succesfully authenticated to {}'.format(address)
except ldap.INVALID_CREDENTIALS:
return None, 'Invalid credentials for {} on {}'.format(
username, address)
except ldap.SERVER_DOWN:
next_hosts = hosts[1:]
if next_hosts:
get_logger().warn('Server is down at %s. Trying next host...',
address)
return ad_connect(username, password, next_hosts)
return None, 'Server is down at {}'.format(address)
except ldap.LDAPError as e:
if type(e.message) == dict and e.message.has_key('desc'):
return None, 'LDAP Error: ' + e.message['desc']
return None, 'LDAP Error: ' + e
return conn
def __init__(self, conn, base, filter=None, scope=ldap.SCOPE_SUBTREE):
self.logger = get_logger(self)
self.conn = conn
self.base = base
self.filter = filter
self.scope = scope
self.entry_mapping = None
def __init__(self, env, config, logger=None):
# directory for training outputs
if not os.path.exists(config.output_path):
os.makedirs(config.output_path)
# store hyper-params
self.config = config
self.logger = logger
if logger is None:
self.logger = get_logger(config.log_path)
self.env = env
temp = self.env.observation_dim
self.observation_dim = [
temp[1], temp[2], temp[0] * self.config.history_mul
]
self.move_action_dim = int((self.env.action_dim - 1) / 2)
self.attack_action_dim = int((self.env.action_dim - 1) / 2)
self.lr = self.config.learning_rate
self.scheduler = LinearSchedule(self.config.rand_begin, \
self.config.rand_end,\
self.config.rand_steps)
self._path_net = "/".join(_path.split('/')[:-1]) \
+ "/policy_gradient_net_" \
+ str(self.config.history_mul) + "/policy_gradient.ckpt"
# build model
self.build()
def __init__(self, env, config, r_seed, logger=None):
"""
Initialize Policy Gradient Class
Args:
env: an OpenAI Gym environment
config: class with hyperparameters
logger: logger instance from the logging module
You do not need to implement anything in this function. However,
you will need to use self.discrete, self.observation_dim,
self.action_dim, and self.lr in other methods.
"""
# directory for training outputs
if not os.path.exists(config.output_path):
os.makedirs(config.output_path)
# store hyperparameters
self.config = config
self.r_seed = r_seed
self.logger = logger
if logger is None:
self.logger = get_logger(config.log_path)
self.env = env
self.env.seed(self.r_seed)
# discrete vs continuous action space
self.discrete = isinstance(env.action_space, gym.spaces.Discrete)
self.observation_dim = self.env.observation_space.shape[0]
self.action_dim = self.env.action_space.n if self.discrete else self.env.action_space.shape[
0]
self.lr = self.config.learning_rate
# build model
self.build()