ding.config

Config

class ding.config.Config(cfg_dict: Optional[dict] = None, cfg_text: Optional[str] = None, filename: Optional[str] = None)

Overview:

Base class for config.

Interface:

__init__, file_to_dict

Property:

cfg_dict

__init__(cfg_dict: Optional[dict] = None, cfg_text: Optional[str] = None, filename: Optional[str] = None) → None

Overview:

Init method. Create config including dict type config and text type config.

Arguments:

• cfg_dict (Optional[dict]): dict type config

• cfg_text (Optional[str]): text type config

• filename (Optional[str]): config file name

static _file_to_dict(filename: str) → Tuple[dict, str]

Overview:

Read config file and convert the config file to dict type config and text type config.

Arguments:

• filename (Optional[str]): config file name.

Returns:

• cfg_dict (Optional[dict]): dict type config

• cfg_text (Optional[str]): text type config

static file_to_dict(filename: str) → ding.config.config.Config

Overview:

Read config file and create config.

Arguments:

• filename (Optional[str]): config file name.

Returns:

• cfg_dict (Config): config class

read_config

ding.config.read_config(path: str) → Tuple[dict, dict]

Overview:

Read configuration from a file path(now only suport python file). And select some proper parts.

Arguments:

• path (str): Path of configuration file

Returns:

• cfg (Tuple[dict, dict]): A collection(tuple) of configuration dict, divided into main_config and create_cfg two parts.

save_config

ding.config.save_config(config_: dict, path: str, type_: str = 'py', save_formatted: bool = False) → None

Overview:

save configuration to python file or yaml file

Arguments:

• config (dict): Config dict

• path (str): Path of target yaml or target python file

• type (str): If type is yaml , save configuration to yaml file. If type is py , save configuration to python file.

• save_formatted (bool): If save_formatted is true, save formatted config to path. Formatted config can be read by serial_pipeline directly.

compile_config

ding.config.compile_config(cfg: easydict.EasyDict, env_manager: Optional[type] = None, policy: Optional[type] = None, learner: type = <class 'ding.worker.learner.base_learner.BaseLearner'>, collector: Optional[type] = None, evaluator: type = <class 'ding.worker.collector.interaction_serial_evaluator.InteractionSerialEvaluator'>, buffer: Optional[type] = None, env: Optional[type] = None, reward_model: Optional[type] = None, world_model: Optional[type] = None, seed: int = 0, auto: bool = False, create_cfg: Optional[dict] = None, save_cfg: bool = True, save_path: str = 'total_config.py', renew_dir: bool = True) → easydict.EasyDict

Overview:

Combine the input config information with other input information. Compile config to make it easy to be called by other programs

Arguments:

• cfg (EasyDict): Input config dict which is to be used in the following pipeline

• env_manager (type): Env_manager class which is to be used in the following pipeline

• policy (type): Policy class which is to be used in the following pipeline

• learner (type): Input learner class, defaults to BaseLearner

• collector (type): Input collector class, defaults to BaseSerialCollector

• evaluator (type): Input evaluator class, defaults to InteractionSerialEvaluator

• buffer (type): Input buffer class, defaults to IBuffer

• env (type): Environment class which is to be used in the following pipeline

• reward_model (type): Reward model class which aims to offer various and valuable reward

• seed (int): Random number seed

• auto (bool): Compile create_config dict or not

• create_cfg (dict): Input create config dict

• save_cfg (bool): Save config or not

• save_path (str): Path of saving file

• renew_dir (bool): Whether to new a directory for saving config.

Returns:

• cfg (EasyDict): Config after compiling