Core Module

baconian.core.core.Basic

class baconian.core.core.Basic(name: str, status=None)

Basic class within the whole framework

INIT_STATUS = None

STATUS_LIST = ('TRAIN', 'TEST')

__init__(name: str, status=None)

Init a new Basic instance.

Parameters:

• name (str) – name of the object, can be determined to generate log path, handle tensorflow name scope etc.
• status (Status) – A status instance Status to indicate the status of the object

allow_duplicate_name = False

get_status() → dict

Return the object’s status, a dictionary.

init(*args, **kwargs)

Initialize the object

load(*args, **kwargs)

Load the parameters from training checkpoints.

name

The name(id) of object, a string.

required_key_dict = ()

save(*args, **kwargs)

Save the parameters in training checkpoints.

set_status(val)

Set the object’s status.

status_list

Status list of the object, (‘TRAIN’, ‘TEST’).

baconian.core.core.Env

class baconian.core.core.Env(name: str = 'env', copy_from_env=None)

Abstract class for environment

INIT_STATUS = 'CREATED'

STATUS_LIST = ('JUST_RESET', 'INITED', 'TRAIN', 'TEST', 'CREATED')

__init__(name: str = 'env', copy_from_env=None)

get_state()

Get the status of the environment.

init()

Set the status to ‘INITED’.

key_list = ()

reset()

Set the status to ‘JUST_RESET’, and update new reset point

seed(seed=None)

Parameters:seed (int) – seed to generate random number Returns:seed of the unwrapped environment Return type:int

step(action)

Parameters:action (method) – agent’s action, the environment will react responding to action

baconian.core.core.EnvSpec

class baconian.core.core.EnvSpec(obs_space: baconian.common.spaces.base.Space, action_space: baconian.common.spaces.base.Space)

__init__(obs_space: baconian.common.spaces.base.Space, action_space: baconian.common.spaces.base.Space)

action_shape

action_space

Returns:Action space of environment Return type:Space

static flat(space: baconian.common.spaces.base.Space, obs_or_action: (<class 'numpy.ndarray'>, <class 'list'>))

flat the input obs or action :param space: space of environment :type space: Space :param obs_or_action: action or observation space :type obs_or_action: (np.ndarray, list) :return: flatten action or observation space :rtype: Space

flat_action(action: (<class 'numpy.ndarray'>, <class 'list'>))

Parameters:action ((np.ndarray, list)) – action taken by agent Returns:flatten action parameter Return type:np.ndarray

flat_action_dim

Returns:the dimension(length) of flatten action space Return type:int

flat_obs(obs: (<class 'numpy.ndarray'>, <class 'list'>))

Parameters:obs ((np.ndarray, list)) – observation of the agent Returns:flatten observation parameter Return type:np.ndarray

flat_obs_dim

Returns:the dimension(length) of flatten observation space Return type:int

obs_shape

obs_space

Returns:Observation space of environment Return type:Space