ding.bonus.a2c

ding.bonus.a2c

A2CAgent

Overview

Class of agent for training, evaluation and deployment of Reinforcement learning algorithm Advantage Actor Critic(A2C). For more information about the system design of RL agent, please refer to https://di-engine-docs.readthedocs.io/en/latest/03_system/agent.html.

Interface: __init__, train, deploy, collect_data, batch_evaluate, best

supported_env_list = list(supported_env_cfg.keys()) class-attribute instance-attribute

Overview

List of supported envs.

Examples: >>> from ding.bonus.a2c import A2CAgent >>> print(A2CAgent.supported_env_list)

best property

Overview

Load the best model from the checkpoint directory, which is by default in folder exp_name/ckpt/eval.pth.tar. The return value is the agent with the best model.

Returns: - (:obj:A2CAgent): The agent with the best model. Examples: >>> agent = A2CAgent(env_id='LunarLanderContinuous-v2') >>> agent.train() >>> agent = agent.best

.. note:: The best model is the model with the highest evaluation return. If this method is called, the current model will be replaced by the best model.

__init__(env_id=None, env=None, seed=0, exp_name=None, model=None, cfg=None, policy_state_dict=None)

Overview

Initialize agent for A2C algorithm.

Arguments: - env_id (:obj:str): The environment id, which is a registered environment name in gym or gymnasium. If env_id is not specified, env_id in cfg.env must be specified. If env_id is specified, env_id in cfg.env will be ignored. env_id should be one of the supported envs, which can be found in supported_env_list. - env (:obj:BaseEnv): The environment instance for training and evaluation. If env is not specified, env_id`` or ``cfg.env.env_id`` must be specified. ``env_id`` or ``cfg.env.env_id`` will be used to create environment instance. If ``env`` is specified, ``env_id`` and ``cfg.env.env_id`` will be ignored. - seed (:obj:int): The random seed, which is set before running the program. Default to 0. - exp_name (:obj:str): The name of this experiment, which will be used to create the folder to save log data. Default to None. If not specified, the folder name will be ``env_id``-``algorithm``. - model (:obj:torch.nn.Module): The model of A2C algorithm, which should be an instance of class :class:ding.model.VAC. If not specified, a default model will be generated according to the configuration. - cfg (:obj:Union[EasyDict, dict]): The configuration of A2C algorithm, which is a dict. Default to None. If not specified, the default configuration will be used. The default configuration can be found in ``ding/config/example/A2C/gym_lunarlander_v2.py``. - policy_state_dict (:obj:str`): The path of policy state dict saved by PyTorch a in local file. If specified, the policy will be loaded from this file. Default to None.

.. note:: An RL Agent Instance can be initialized in two basic ways. For example, we have an environment with id LunarLanderContinuous-v2 registered in gym, and we want to train an agent with A2C algorithm with default configuration. Then we can initialize the agent in the following ways: >>> agent = A2CAgent(env_id='LunarLanderContinuous-v2') or, if we want can specify the env_id in the configuration: >>> cfg = {'env': {'env_id': 'LunarLanderContinuous-v2'}, 'policy': ...... } >>> agent = A2CAgent(cfg=cfg) There are also other arguments to specify the agent when initializing. For example, if we want to specify the environment instance: >>> env = CustomizedEnv('LunarLanderContinuous-v2') >>> agent = A2CAgent(cfg=cfg, env=env) or, if we want to specify the model: >>> model = VAC(**cfg.policy.model) >>> agent = A2CAgent(cfg=cfg, model=model) or, if we want to reload the policy from a saved policy state dict: >>> agent = A2CAgent(cfg=cfg, policy_state_dict='LunarLanderContinuous-v2.pth.tar') Make sure that the configuration is consistent with the saved policy state dict.

train(step=int(10000000.0), collector_env_num=4, evaluator_env_num=4, n_iter_log_show=500, n_iter_save_ckpt=1000, context=None, debug=False, wandb_sweep=False)

Overview

Train the agent with A2C algorithm for step iterations with collector_env_num collector environments and evaluator_env_num evaluator environments. Information during training will be recorded and saved by wandb.

Arguments: - step (:obj:int): The total training environment steps of all collector environments. Default to 1e7. - collector_env_num (:obj:int): The collector environment number. Default to None. If not specified, it will be set according to the configuration. - evaluator_env_num (:obj:int): The evaluator environment number. Default to None. If not specified, it will be set according to the configuration. - n_iter_save_ckpt (:obj:int): The frequency of saving checkpoint every training iteration. Default to 1000. - context (:obj:str): The multi-process context of the environment manager. Default to None. It can be specified as spawn, fork or forkserver. - debug (:obj:bool): Whether to use debug mode in the environment manager. Default to False. If set True, base environment manager will be used for easy debugging. Otherwise, subprocess environment manager will be used. - wandb_sweep (:obj:bool): Whether to use wandb sweep, which is a hyper-parameter optimization process for seeking the best configurations. Default to False. If True, the wandb sweep id will be used as the experiment name. Returns: - (:obj:TrainingReturn): The training result, of which the attributions are: - wandb_url (:obj:str): The weight & biases (wandb) project url of the trainning experiment.

deploy(enable_save_replay=False, concatenate_all_replay=False, replay_save_path=None, seed=None, debug=False)

Overview

Deploy the agent with A2C algorithm by interacting with the environment, during which the replay video can be saved if enable_save_replay is True. The evaluation result will be returned.

Arguments: - enable_save_replay (:obj:bool): Whether to save the replay video. Default to False. - concatenate_all_replay (:obj:bool): Whether to concatenate all replay videos into one video. Default to False. If enable_save_replay is False, this argument will be ignored. If enable_save_replay is True and concatenate_all_replay is False, the replay video of each episode will be saved separately. - replay_save_path (:obj:str): The path to save the replay video. Default to None. If not specified, the video will be saved in exp_name/videos. - seed (:obj:Union[int, List]): The random seed, which is set before running the program. Default to None. If not specified, self.seed will be used. If seed is an integer, the agent will be deployed once. If seed is a list of integers, the agent will be deployed once for each seed in the list. - debug (:obj:bool): Whether to use debug mode in the environment manager. Default to False. If set True, base environment manager will be used for easy debugging. Otherwise, subprocess environment manager will be used. Returns: - (:obj:EvalReturn): The evaluation result, of which the attributions are: - eval_value (:obj:np.float32): The mean of evaluation return. - eval_value_std (:obj:np.float32): The standard deviation of evaluation return.

collect_data(env_num=8, save_data_path=None, n_sample=None, n_episode=None, context=None, debug=False)

Overview

Collect data with A2C algorithm for n_episode episodes with env_num collector environments. The collected data will be saved in save_data_path if specified, otherwise it will be saved in exp_name/demo_data.

Arguments: - env_num (:obj:int): The number of collector environments. Default to 8. - save_data_path (:obj:str): The path to save the collected data. Default to None. If not specified, the data will be saved in exp_name/demo_data. - n_sample (:obj:int): The number of samples to collect. Default to None. If not specified, n_episode must be specified. - n_episode (:obj:int): The number of episodes to collect. Default to None. If not specified, n_sample must be specified. - context (:obj:str): The multi-process context of the environment manager. Default to None. It can be specified as spawn, fork or forkserver. - debug (:obj:bool): Whether to use debug mode in the environment manager. Default to False. If set True, base environment manager will be used for easy debugging. Otherwise, subprocess environment manager will be used.

batch_evaluate(env_num=4, n_evaluator_episode=4, context=None, debug=False)

Overview

Evaluate the agent with A2C algorithm for n_evaluator_episode episodes with env_num evaluator environments. The evaluation result will be returned. The difference between methods batch_evaluate and deploy is that batch_evaluate will create multiple evaluator environments to evaluate the agent to get an average performance, while deploy will only create one evaluator environment to evaluate the agent and save the replay video.

Arguments: - env_num (:obj:int): The number of evaluator environments. Default to 4. - n_evaluator_episode (:obj:int): The number of episodes to evaluate. Default to 4. - context (:obj:str): The multi-process context of the environment manager. Default to None. It can be specified as spawn, fork or forkserver. - debug (:obj:bool): Whether to use debug mode in the environment manager. Default to False. If set True, base environment manager will be used for easy debugging. Otherwise, subprocess environment manager will be used. Returns: - (:obj:EvalReturn): The evaluation result, of which the attributions are: - eval_value (:obj:np.float32): The mean of evaluation return. - eval_value_std (:obj:np.float32): The standard deviation of evaluation return.

Full Source Code

../ding/bonus/a2c.py

1from typing import Optional, Union, List 2from ditk import logging 3from easydict import EasyDict 4import os 5import numpy as np 6import torch 7import treetensor.torch as ttorch 8from ding.framework import task, OnlineRLContext 9from ding.framework.middleware import CkptSaver, trainer, \ 10 wandb_online_logger, offline_data_saver, termination_checker, interaction_evaluator, StepCollector, \ 11 gae_estimator, final_ctx_saver 12from ding.envs import BaseEnv 13from ding.envs import setup_ding_env_manager 14from ding.policy import A2CPolicy 15from ding.utils import set_pkg_seed 16from ding.utils import get_env_fps, render 17from ding.config import save_config_py, compile_config 18from ding.model import VAC 19from ding.model import model_wrap 20from ding.bonus.common import TrainingReturn, EvalReturn 21from ding.config.example.A2C import supported_env_cfg 22from ding.config.example.A2C import supported_env 23 24 25class A2CAgent: 26 """ 27 Overview: 28 Class of agent for training, evaluation and deployment of Reinforcement learning algorithm \ 29 Advantage Actor Critic(A2C). 30 For more information about the system design of RL agent, please refer to \ 31 <https://di-engine-docs.readthedocs.io/en/latest/03_system/agent.html>. 32 Interface: 33 ``__init__``, ``train``, ``deploy``, ``collect_data``, ``batch_evaluate``, ``best`` 34 """ 35 supported_env_list = list(supported_env_cfg.keys()) 36 """ 37 Overview: 38 List of supported envs. 39 Examples: 40 >>> from ding.bonus.a2c import A2CAgent 41 >>> print(A2CAgent.supported_env_list) 42 """ 43 44 def __init__( 45 self, 46 env_id: str = None, 47 env: BaseEnv = None, 48 seed: int = 0, 49 exp_name: str = None, 50 model: Optional[torch.nn.Module] = None, 51 cfg: Optional[Union[EasyDict, dict]] = None, 52 policy_state_dict: str = None, 53 ) -> None: 54 """ 55 Overview: 56 Initialize agent for A2C algorithm. 57 Arguments: 58 - env_id (:obj:`str`): The environment id, which is a registered environment name in gym or gymnasium. \ 59 If ``env_id`` is not specified, ``env_id`` in ``cfg.env`` must be specified. \ 60 If ``env_id`` is specified, ``env_id`` in ``cfg.env`` will be ignored. \ 61 ``env_id`` should be one of the supported envs, which can be found in ``supported_env_list``. 62 - env (:obj:`BaseEnv`): The environment instance for training and evaluation. \ 63 If ``env`` is not specified, `env_id`` or ``cfg.env.env_id`` must be specified. \ 64 ``env_id`` or ``cfg.env.env_id`` will be used to create environment instance. \ 65 If ``env`` is specified, ``env_id`` and ``cfg.env.env_id`` will be ignored. 66 - seed (:obj:`int`): The random seed, which is set before running the program. \ 67 Default to 0. 68 - exp_name (:obj:`str`): The name of this experiment, which will be used to create the folder to save \ 69 log data. Default to None. If not specified, the folder name will be ``env_id``-``algorithm``. 70 - model (:obj:`torch.nn.Module`): The model of A2C algorithm, which should be an instance of class \ 71 :class:`ding.model.VAC`. \ 72 If not specified, a default model will be generated according to the configuration. 73 - cfg (:obj:`Union[EasyDict, dict]`): The configuration of A2C algorithm, which is a dict. \ 74 Default to None. If not specified, the default configuration will be used. \ 75 The default configuration can be found in ``ding/config/example/A2C/gym_lunarlander_v2.py``. 76 - policy_state_dict (:obj:`str`): The path of policy state dict saved by PyTorch a in local file. \ 77 If specified, the policy will be loaded from this file. Default to None. 78 79 .. note:: 80 An RL Agent Instance can be initialized in two basic ways. \ 81 For example, we have an environment with id ``LunarLanderContinuous-v2`` registered in gym, \ 82 and we want to train an agent with A2C algorithm with default configuration. \ 83 Then we can initialize the agent in the following ways: 84 >>> agent = A2CAgent(env_id='LunarLanderContinuous-v2') 85 or, if we want can specify the env_id in the configuration: 86 >>> cfg = {'env': {'env_id': 'LunarLanderContinuous-v2'}, 'policy': ...... } 87 >>> agent = A2CAgent(cfg=cfg) 88 There are also other arguments to specify the agent when initializing. 89 For example, if we want to specify the environment instance: 90 >>> env = CustomizedEnv('LunarLanderContinuous-v2') 91 >>> agent = A2CAgent(cfg=cfg, env=env) 92 or, if we want to specify the model: 93 >>> model = VAC(**cfg.policy.model) 94 >>> agent = A2CAgent(cfg=cfg, model=model) 95 or, if we want to reload the policy from a saved policy state dict: 96 >>> agent = A2CAgent(cfg=cfg, policy_state_dict='LunarLanderContinuous-v2.pth.tar') 97 Make sure that the configuration is consistent with the saved policy state dict. 98 """ 99 100 assert env_id is not None or cfg is not None, "Please specify env_id or cfg." 101 102 if cfg is not None and not isinstance(cfg, EasyDict): 103 cfg = EasyDict(cfg) 104 105 if env_id is not None: 106 assert env_id in A2CAgent.supported_env_list, "Please use supported envs: {}".format( 107 A2CAgent.supported_env_list 108 ) 109 if cfg is None: 110 cfg = supported_env_cfg[env_id] 111 else: 112 assert cfg.env.env_id == env_id, "env_id in cfg should be the same as env_id in args." 113 else: 114 assert hasattr(cfg.env, "env_id"), "Please specify env_id in cfg." 115 assert cfg.env.env_id in A2CAgent.supported_env_list, "Please use supported envs: {}".format( 116 A2CAgent.supported_env_list 117 ) 118 default_policy_config = EasyDict({"policy": A2CPolicy.default_config()}) 119 default_policy_config.update(cfg) 120 cfg = default_policy_config 121 122 if exp_name is not None: 123 cfg.exp_name = exp_name 124 self.cfg = compile_config(cfg, policy=A2CPolicy) 125 self.exp_name = self.cfg.exp_name 126 if env is None: 127 self.env = supported_env[cfg.env.env_id](cfg=cfg.env) 128 else: 129 assert isinstance(env, BaseEnv), "Please use BaseEnv as env data type." 130 self.env = env 131 132 logging.getLogger().setLevel(logging.INFO) 133 self.seed = seed 134 set_pkg_seed(self.seed, use_cuda=self.cfg.policy.cuda) 135 if not os.path.exists(self.exp_name): 136 os.makedirs(self.exp_name) 137 save_config_py(self.cfg, os.path.join(self.exp_name, 'policy_config.py')) 138 if model is None: 139 model = VAC(**self.cfg.policy.model) 140 self.policy = A2CPolicy(self.cfg.policy, model=model) 141 if policy_state_dict is not None: 142 self.policy.learn_mode.load_state_dict(policy_state_dict) 143 self.checkpoint_save_dir = os.path.join(self.exp_name, "ckpt") 144 145 def train( 146 self, 147 step: int = int(1e7), 148 collector_env_num: int = 4, 149 evaluator_env_num: int = 4, 150 n_iter_log_show: int = 500, 151 n_iter_save_ckpt: int = 1000, 152 context: Optional[str] = None, 153 debug: bool = False, 154 wandb_sweep: bool = False, 155 ) -> TrainingReturn: 156 """ 157 Overview: 158 Train the agent with A2C algorithm for ``step`` iterations with ``collector_env_num`` collector \ 159 environments and ``evaluator_env_num`` evaluator environments. Information during training will be \ 160 recorded and saved by wandb. 161 Arguments: 162 - step (:obj:`int`): The total training environment steps of all collector environments. Default to 1e7. 163 - collector_env_num (:obj:`int`): The collector environment number. Default to None. \ 164 If not specified, it will be set according to the configuration. 165 - evaluator_env_num (:obj:`int`): The evaluator environment number. Default to None. \ 166 If not specified, it will be set according to the configuration. 167 - n_iter_save_ckpt (:obj:`int`): The frequency of saving checkpoint every training iteration. \ 168 Default to 1000. 169 - context (:obj:`str`): The multi-process context of the environment manager. Default to None. \ 170 It can be specified as ``spawn``, ``fork`` or ``forkserver``. 171 - debug (:obj:`bool`): Whether to use debug mode in the environment manager. Default to False. \ 172 If set True, base environment manager will be used for easy debugging. Otherwise, \ 173 subprocess environment manager will be used. 174 - wandb_sweep (:obj:`bool`): Whether to use wandb sweep, \ 175 which is a hyper-parameter optimization process for seeking the best configurations. \ 176 Default to False. If True, the wandb sweep id will be used as the experiment name. 177 Returns: 178 - (:obj:`TrainingReturn`): The training result, of which the attributions are: 179 - wandb_url (:obj:`str`): The weight & biases (wandb) project url of the trainning experiment. 180 """ 181 182 if debug: 183 logging.getLogger().setLevel(logging.DEBUG) 184 logging.debug(self.policy._model) 185 # define env and policy 186 collector_env = self._setup_env_manager(collector_env_num, context, debug, 'collector') 187 evaluator_env = self._setup_env_manager(evaluator_env_num, context, debug, 'evaluator') 188 189 with task.start(ctx=OnlineRLContext()): 190 task.use( 191 interaction_evaluator( 192 self.cfg, 193 self.policy.eval_mode, 194 evaluator_env, 195 render=self.cfg.policy.eval.render if hasattr(self.cfg.policy.eval, "render") else False 196 ) 197 ) 198 task.use(CkptSaver(policy=self.policy, save_dir=self.checkpoint_save_dir, train_freq=n_iter_save_ckpt)) 199 task.use( 200 StepCollector( 201 self.cfg, 202 self.policy.collect_mode, 203 collector_env, 204 random_collect_size=self.cfg.policy.random_collect_size 205 if hasattr(self.cfg.policy, 'random_collect_size') else 0, 206 ) 207 ) 208 task.use(gae_estimator(self.cfg, self.policy.collect_mode)) 209 task.use(trainer(self.cfg, self.policy.learn_mode)) 210 task.use( 211 wandb_online_logger( 212 metric_list=self.policy._monitor_vars_learn(), 213 model=self.policy._model, 214 anonymous=True, 215 project_name=self.exp_name, 216 wandb_sweep=wandb_sweep, 217 ) 218 ) 219 task.use(termination_checker(max_env_step=step)) 220 task.use(final_ctx_saver(name=self.exp_name)) 221 task.run() 222 223 return TrainingReturn(wandb_url=task.ctx.wandb_url) 224 225 def deploy( 226 self, 227 enable_save_replay: bool = False, 228 concatenate_all_replay: bool = False, 229 replay_save_path: str = None, 230 seed: Optional[Union[int, List]] = None, 231 debug: bool = False 232 ) -> EvalReturn: 233 """ 234 Overview: 235 Deploy the agent with A2C algorithm by interacting with the environment, during which the replay video \ 236 can be saved if ``enable_save_replay`` is True. The evaluation result will be returned. 237 Arguments: 238 - enable_save_replay (:obj:`bool`): Whether to save the replay video. Default to False. 239 - concatenate_all_replay (:obj:`bool`): Whether to concatenate all replay videos into one video. \ 240 Default to False. If ``enable_save_replay`` is False, this argument will be ignored. \ 241 If ``enable_save_replay`` is True and ``concatenate_all_replay`` is False, \ 242 the replay video of each episode will be saved separately. 243 - replay_save_path (:obj:`str`): The path to save the replay video. Default to None. \ 244 If not specified, the video will be saved in ``exp_name/videos``. 245 - seed (:obj:`Union[int, List]`): The random seed, which is set before running the program. \ 246 Default to None. If not specified, ``self.seed`` will be used. \ 247 If ``seed`` is an integer, the agent will be deployed once. \ 248 If ``seed`` is a list of integers, the agent will be deployed once for each seed in the list. 249 - debug (:obj:`bool`): Whether to use debug mode in the environment manager. Default to False. \ 250 If set True, base environment manager will be used for easy debugging. Otherwise, \ 251 subprocess environment manager will be used. 252 Returns: 253 - (:obj:`EvalReturn`): The evaluation result, of which the attributions are: 254 - eval_value (:obj:`np.float32`): The mean of evaluation return. 255 - eval_value_std (:obj:`np.float32`): The standard deviation of evaluation return. 256 """ 257 258 if debug: 259 logging.getLogger().setLevel(logging.DEBUG) 260 # define env and policy 261 env = self.env.clone(caller='evaluator') 262 263 if seed is not None and isinstance(seed, int): 264 seeds = [seed] 265 elif seed is not None and isinstance(seed, list): 266 seeds = seed 267 else: 268 seeds = [self.seed] 269 270 returns = [] 271 images = [] 272 if enable_save_replay: 273 replay_save_path = os.path.join(self.exp_name, 'videos') if replay_save_path is None else replay_save_path 274 env.enable_save_replay(replay_path=replay_save_path) 275 else: 276 logging.warning('No video would be generated during the deploy.') 277 if concatenate_all_replay: 278 logging.warning('concatenate_all_replay is set to False because enable_save_replay is False.') 279 concatenate_all_replay = False 280 281 def single_env_forward_wrapper(forward_fn, cuda=True): 282 283 if self.cfg.policy.action_space == 'continuous': 284 forward_fn = model_wrap(forward_fn, wrapper_name='deterministic_sample').forward 285 elif self.cfg.policy.action_space == 'discrete': 286 forward_fn = model_wrap(forward_fn, wrapper_name='argmax_sample').forward 287 else: 288 raise NotImplementedError 289 290 def _forward(obs): 291 # unsqueeze means add batch dim, i.e. (O, ) -> (1, O) 292 obs = ttorch.as_tensor(obs).unsqueeze(0) 293 if cuda and torch.cuda.is_available(): 294 obs = obs.cuda() 295 action = forward_fn(obs, mode='compute_actor')["action"] 296 # squeeze means delete batch dim, i.e. (1, A) -> (A, ) 297 action = action.squeeze(0).detach().cpu().numpy() 298 return action 299 300 return _forward 301 302 forward_fn = single_env_forward_wrapper(self.policy._model, self.cfg.policy.cuda) 303 304 # reset first to make sure the env is in the initial state 305 # env will be reset again in the main loop 306 env.reset() 307 308 for seed in seeds: 309 env.seed(seed, dynamic_seed=False) 310 return_ = 0. 311 step = 0 312 obs = env.reset() 313 images.append(render(env)[None]) if concatenate_all_replay else None 314 while True: 315 action = forward_fn(obs) 316 obs, rew, done, info = env.step(action) 317 images.append(render(env)[None]) if concatenate_all_replay else None 318 return_ += rew 319 step += 1 320 if done: 321 break 322 logging.info(f'DQN deploy is finished, final episode return with {step} steps is: {return_}') 323 returns.append(return_) 324 325 env.close() 326 327 if concatenate_all_replay: 328 images = np.concatenate(images, axis=0) 329 import imageio 330 imageio.mimwrite(os.path.join(replay_save_path, 'deploy.mp4'), images, fps=get_env_fps(env)) 331 332 return EvalReturn(eval_value=np.mean(returns), eval_value_std=np.std(returns)) 333 334 def collect_data( 335 self, 336 env_num: int = 8, 337 save_data_path: Optional[str] = None, 338 n_sample: Optional[int] = None, 339 n_episode: Optional[int] = None, 340 context: Optional[str] = None, 341 debug: bool = False 342 ) -> None: 343 """ 344 Overview: 345 Collect data with A2C algorithm for ``n_episode`` episodes with ``env_num`` collector environments. \ 346 The collected data will be saved in ``save_data_path`` if specified, otherwise it will be saved in \ 347 ``exp_name/demo_data``. 348 Arguments: 349 - env_num (:obj:`int`): The number of collector environments. Default to 8. 350 - save_data_path (:obj:`str`): The path to save the collected data. Default to None. \ 351 If not specified, the data will be saved in ``exp_name/demo_data``. 352 - n_sample (:obj:`int`): The number of samples to collect. Default to None. \ 353 If not specified, ``n_episode`` must be specified. 354 - n_episode (:obj:`int`): The number of episodes to collect. Default to None. \ 355 If not specified, ``n_sample`` must be specified. 356 - context (:obj:`str`): The multi-process context of the environment manager. Default to None. \ 357 It can be specified as ``spawn``, ``fork`` or ``forkserver``. 358 - debug (:obj:`bool`): Whether to use debug mode in the environment manager. Default to False. \ 359 If set True, base environment manager will be used for easy debugging. Otherwise, \ 360 subprocess environment manager will be used. 361 """ 362 363 if debug: 364 logging.getLogger().setLevel(logging.DEBUG) 365 if n_episode is not None: 366 raise NotImplementedError 367 # define env and policy 368 env_num = env_num if env_num else self.cfg.env.collector_env_num 369 env = setup_ding_env_manager(self.env, env_num, context, debug, 'collector') 370 371 if save_data_path is None: 372 save_data_path = os.path.join(self.exp_name, 'demo_data') 373 374 # main execution task 375 with task.start(ctx=OnlineRLContext()): 376 task.use( 377 StepCollector( 378 self.cfg, self.policy.collect_mode, env, random_collect_size=self.cfg.policy.random_collect_size 379 ) 380 ) 381 task.use(offline_data_saver(save_data_path, data_type='hdf5')) 382 task.run(max_step=1) 383 logging.info( 384 f'A2C collecting is finished, more than {n_sample} samples are collected and saved in `{save_data_path}`' 385 ) 386 387 def batch_evaluate( 388 self, 389 env_num: int = 4, 390 n_evaluator_episode: int = 4, 391 context: Optional[str] = None, 392 debug: bool = False 393 ) -> EvalReturn: 394 """ 395 Overview: 396 Evaluate the agent with A2C algorithm for ``n_evaluator_episode`` episodes with ``env_num`` evaluator \ 397 environments. The evaluation result will be returned. 398 The difference between methods ``batch_evaluate`` and ``deploy`` is that ``batch_evaluate`` will create \ 399 multiple evaluator environments to evaluate the agent to get an average performance, while ``deploy`` \ 400 will only create one evaluator environment to evaluate the agent and save the replay video. 401 Arguments: 402 - env_num (:obj:`int`): The number of evaluator environments. Default to 4. 403 - n_evaluator_episode (:obj:`int`): The number of episodes to evaluate. Default to 4. 404 - context (:obj:`str`): The multi-process context of the environment manager. Default to None. \ 405 It can be specified as ``spawn``, ``fork`` or ``forkserver``. 406 - debug (:obj:`bool`): Whether to use debug mode in the environment manager. Default to False. \ 407 If set True, base environment manager will be used for easy debugging. Otherwise, \ 408 subprocess environment manager will be used. 409 Returns: 410 - (:obj:`EvalReturn`): The evaluation result, of which the attributions are: 411 - eval_value (:obj:`np.float32`): The mean of evaluation return. 412 - eval_value_std (:obj:`np.float32`): The standard deviation of evaluation return. 413 """ 414 415 if debug: 416 logging.getLogger().setLevel(logging.DEBUG) 417 # define env and policy 418 env_num = env_num if env_num else self.cfg.env.evaluator_env_num 419 env = setup_ding_env_manager(self.env, env_num, context, debug, 'evaluator') 420 421 # reset first to make sure the env is in the initial state 422 # env will be reset again in the main loop 423 env.launch() 424 env.reset() 425 426 evaluate_cfg = self.cfg 427 evaluate_cfg.env.n_evaluator_episode = n_evaluator_episode 428 429 # main execution task 430 with task.start(ctx=OnlineRLContext()): 431 task.use(interaction_evaluator(self.cfg, self.policy.eval_mode, env)) 432 task.run(max_step=1) 433 434 return EvalReturn(eval_value=task.ctx.eval_value, eval_value_std=task.ctx.eval_value_std) 435 436 @property 437 def best(self) -> 'A2CAgent': 438 """ 439 Overview: 440 Load the best model from the checkpoint directory, \ 441 which is by default in folder ``exp_name/ckpt/eval.pth.tar``. \ 442 The return value is the agent with the best model. 443 Returns: 444 - (:obj:`A2CAgent`): The agent with the best model. 445 Examples: 446 >>> agent = A2CAgent(env_id='LunarLanderContinuous-v2') 447 >>> agent.train() 448 >>> agent = agent.best 449 450 .. note:: 451 The best model is the model with the highest evaluation return. If this method is called, the current \ 452 model will be replaced by the best model. 453 """ 454 455 best_model_file_path = os.path.join(self.checkpoint_save_dir, "eval.pth.tar") 456 # Load best model if it exists 457 if os.path.exists(best_model_file_path): 458 policy_state_dict = torch.load(best_model_file_path, map_location=torch.device("cpu")) 459 self.policy.learn_mode.load_state_dict(policy_state_dict) 460 return self