Stable-Baselines 3 部分源代码解读 ./common/on_policy_algorithm.py

前言

阅读PPO相关的源码，了解一下标准库是如何建立PPO算法以及各种tricks的，以便于自己的复现。

在Pycharm里面一直跳转，可以看到PPO类是最终继承于基类，也就是这个py文件的内容。

所以阅读源码就先从这里开始。: )

import 包

import sys
import time
from typing import Any, Dict, List, Optional, Tuple, Type, TypeVar, Unionimport numpy as np
import torch as th
from gym import spacesfrom stable_baselines3.common.base_class import BaseAlgorithm
from stable_baselines3.common.buffers import DictRolloutBuffer, RolloutBuffer
from stable_baselines3.common.callbacks import BaseCallback
from stable_baselines3.common.policies import ActorCriticPolicy
from stable_baselines3.common.type_aliases import GymEnv, MaybeCallback, Schedule
from stable_baselines3.common.utils import obs_as_tensor, safe_mean
from stable_baselines3.common.vec_env import VecEnv

OnPolicyAlgorithm 类

这个类是PPO算法类的中间曾，夹在底层基类和上层PPO类的之间。

主要是同策略算法，例如：A2C和PPO算法。

policy、env和learning_rate三者与基类base-class.py的一致

n_steps表示每次更新前需要经过的时间步，作者在这里给出了n_steps * n_envs的例子，可能的意思是，如果环境是重复的多个，打算做并行训练的话，那么就是每个子环境的时间步乘以环境的数量

batch_size经验回放的最小批次信息

gamma、gae_lambda、clip_range、clip_range_vf均是具有默认值的参数，分别代表“折扣因子”、“GAE奖励中平衡偏置和方差的参数”、“为网络参数而限制幅度的范围”、“为值函数网络参数而限制幅度的范围”

normalize_advantage标志是否需要归一化优势（advantage）

ent_coef、vf_coef损失计算的熵系数

max_grad_norm最大的梯度长度，梯度下降的限幅

use_sde、sde_sample_freq是状态独立性探索，只适用于连续环境，与基类base-class.py的一致

target_kl限制每次更新时KL散度不能太大，因为clipping限幅不能防止大量更新

monitor_wrapper标志是否需要Gym库提供的监视器包装器

_init_setup_model是否建立模型，也就是是否在创建这个实例过程中创建初始化模型

class OnPolicyAlgorithm(BaseAlgorithm):"""The base for On-Policy algorithms (ex: A2C/PPO).:param policy: The policy model to use (MlpPolicy, CnnPolicy, ...):param env: The environment to learn from (if registered in Gym, can be str):param learning_rate: The learning rate, it can be a functionof the current progress remaining (from 1 to 0):param n_steps: The number of steps to run for each environment per update(i.e. batch size is n_steps * n_env where n_env is number of environment copies running in parallel):param gamma: Discount factor:param gae_lambda: Factor for trade-off of bias vs variance for Generalized Advantage Estimator.Equivalent to classic advantage when set to 1.:param ent_coef: Entropy coefficient for the loss calculation:param vf_coef: Value function coefficient for the loss calculation:param max_grad_norm: The maximum value for the gradient clipping:param use_sde: Whether to use generalized State Dependent Exploration (gSDE)instead of action noise exploration (default: False):param sde_sample_freq: Sample a new noise matrix every n steps when using gSDEDefault: -1 (only sample at the beginning of the rollout):param tensorboard_log: the log location for tensorboard (if None, no logging):param monitor_wrapper: When creating an environment, whether to wrap itor not in a Monitor wrapper.:param policy_kwargs: additional arguments to be passed to the policy on creation:param verbose: Verbosity level: 0 for no output, 1 for info messages (such as device or wrappers used), 2 fordebug messages:param seed: Seed for the pseudo random generators:param device: Device (cpu, cuda, ...) on which the code should be run.Setting it to auto, the code will be run on the GPU if possible.:param _init_setup_model: Whether or not to build the network at the creation of the instance:param supported_action_spaces: The action spaces supported by the algorithm."""def __init__(self,policy: Union[str, Type[ActorCriticPolicy]],env: Union[GymEnv, str],learning_rate: Union[float, Schedule],n_steps: int,gamma: float,gae_lambda: float,ent_coef: float,vf_coef: float,max_grad_norm: float,use_sde: bool,sde_sample_freq: int,tensorboard_log: Optional[str] = None,monitor_wrapper: bool = True,policy_kwargs: Optional[Dict[str, Any]] = None,verbose: int = 0,seed: Optional[int] = None,device: Union[th.device, str] = "auto",_init_setup_model: bool = True,supported_action_spaces: Optional[Tuple[spaces.Space, ...]] = None,):super().__init__(policy=policy,env=env,learning_rate=learning_rate,policy_kwargs=policy_kwargs,verbose=verbose,device=device,use_sde=use_sde,sde_sample_freq=sde_sample_freq,support_multi_env=True,seed=seed,tensorboard_log=tensorboard_log,supported_action_spaces=supported_action_spaces,)self.n_steps = n_stepsself.gamma = gammaself.gae_lambda = gae_lambdaself.ent_coef = ent_coefself.vf_coef = vf_coefself.max_grad_norm = max_grad_normself.rollout_buffer = None# 调用基类的_setup_model()模型if _init_setup_model:self._setup_model()def _setup_model(self) -> None:# 初始化学习率，让他可以调用self._setup_lr_schedule()# 设置随机数种子self.set_random_seed(self.seed)# 设置经验池子的类，如果观测空间是spaces.Dict类那么就赋值DictRolloutBuffer# 如果观测空间不是spaces.Dict类那么就赋值RolloutBufferbuffer_cls = DictRolloutBuffer if isinstance(self.observation_space, spaces.Dict) else RolloutBuffer# 根据类初始化实例经验池子# 初始化经验池子的是时候将设备信息、折扣率、GAE超参数和环境的数量也传进去了self.rollout_buffer = buffer_cls(self.n_steps,self.observation_space,self.action_space,device=self.device,gamma=self.gamma,gae_lambda=self.gae_lambda,n_envs=self.n_envs,)# 初始化策略，直接输入状态空间、动作空间、可调用的学习率、是否使用状态独立性探索，以及自己制定策略# 的时候自己家的模型的参数和激活函数self.policy = self.policy_class(  # pytype:disable=not-instantiableself.observation_space,self.action_space,self.lr_schedule,use_sde=self.use_sde,**self.policy_kwargs  # pytype:disable=not-instantiable)# 将策略放到GPU/CPU中self.policy = self.policy.to(self.device)def collect_rollouts(self,env: VecEnv,callback: BaseCallback,rollout_buffer: RolloutBuffer,n_rollout_steps: int,) -> bool:# 收集环境交互数据# 这个方法使用当前的策略并将交互历史填充到RolloutBuffer经验池子中# rollout的意思是无模型的概念，而不是有模型的RL或规划里面的rollout的概念# env 用于训练的环境# callback 在每个时间步都会调用的回调函数# rollout_buffer 将收集的经验放置到rollout_buffer中# 在每个环境中需要收集的条数# 返回值是True：如果rollout_buffer收集了这么多的经验；返回值是False：如果回调函数提前终止了# 这个rollouts。"""Collect experiences using the current policy and fill a ``RolloutBuffer``.The term rollout here refers to the model-free notion and should notbe used with the concept of rollout used in model-based RL or planning.:param env: The training environment:param callback: Callback that will be called at each step(and at the beginning and end of the rollout):param rollout_buffer: Buffer to fill with rollouts:param n_rollout_steps: Number of experiences to collect per environment:return: True if function returned with at least `n_rollout_steps`collected, False if callback terminated rollout prematurely."""assert self._last_obs is not None, "No previous observation was provided"# 将策略转变到评估模式# Switch to eval mode (this affects batch norm / dropout)self.policy.set_training_mode(False)# 重置经验池子，如果使用状态独立性探索，那么就重置策略的噪声n_steps = 0rollout_buffer.reset()# Sample new weights for the state dependent explorationif self.use_sde:self.policy.reset_noise(env.num_envs)# 回调函数执行on_rollout_start()命令，跳转定义时候没有看到具体定义callback.on_rollout_start()while n_steps < n_rollout_steps:# 如果使用了状态独立性探索，并且达到了探索频率的节点，那么就重置策略的噪声if self.use_sde and self.sde_sample_freq > 0 and n_steps % self.sde_sample_freq == 0:# Sample a new noise matrixself.policy.reset_noise(env.num_envs)# 在断开梯度的情况下，转换观测数据到tensor张量内，然后输入到策略中输出动作、价值和对数概率# 最后再将动作数据转移到numpy中with th.no_grad():# Convert to pytorch tensor or to TensorDictobs_tensor = obs_as_tensor(self._last_obs, self.device)actions, values, log_probs = self.policy(obs_tensor)actions = actions.cpu().numpy()# Rescale and perform action# 归一化动作信息，限制在动作空间的上下界clipped_actions = actions# Clip the actions to avoid out of bound errorif isinstance(self.action_space, spaces.Box):clipped_actions = np.clip(actions, self.action_space.low, self.action_space.high)# 将动作信息输入到环境中，输出新的观测、奖励数值、是否完成以及其他信息。new_obs, rewards, dones, infos = env.step(clipped_actions)# 处理回调函数和更新经验池子self.num_timesteps += env.num_envs# Give access to local variablescallback.update_locals(locals())if callback.on_step() is False:return Falseself._update_info_buffer(infos)n_steps += 1# 如果动作空间是离散空间的话，那么就转变成一个列向量if isinstance(self.action_space, spaces.Discrete):# Reshape in case of discrete actionactions = actions.reshape(-1, 1)# 判断数据是否是终止的# 终止之后计算累计奖励# Handle timeout by bootstraping with value function# see GitHub issue #633for idx, done in enumerate(dones):if (doneand infos[idx].get("terminal_observation") is not Noneand infos[idx].get("TimeLimit.truncated", False)):terminal_obs = self.policy.obs_to_tensor(infos[idx]["terminal_observation"])[0]with th.no_grad():terminal_value = self.policy.predict_values(terminal_obs)[0]rewards[idx] += self.gamma * terminal_value# 经验池子输入的是上一个状态、动作、奖励、上一个回合的开始状态、价值列表以及对数概率rollout_buffer.add(self._last_obs, actions, rewards, self._last_episode_starts, values, log_probs)self._last_obs = new_obsself._last_episode_starts = dones# 计算下一个状态的价值with th.no_grad():# Compute value for the last timestepvalues = self.policy.predict_values(obs_as_tensor(new_obs, self.device))# 计算回报和优势rollout_buffer.compute_returns_and_advantage(last_values=values, dones=dones)callback.on_rollout_end()return Truedef train(self) -> None:# 这个是父类的方法# 在子类的实际PPO类中做了重写"""Consume current rollout data and update policy parameters.Implemented by individual algorithms."""raise NotImplementedErrordef learn(self: SelfOnPolicyAlgorithm,total_timesteps: int,callback: MaybeCallback = None,log_interval: int = 1,tb_log_name: str = "OnPolicyAlgorithm",reset_num_timesteps: bool = True,progress_bar: bool = False,) -> SelfOnPolicyAlgorithm:iteration = 0# 初始化模型total_timesteps, callback = self._setup_learn(total_timesteps,callback,reset_num_timesteps,tb_log_name,progress_bar,)callback.on_training_start(locals(), globals())while self.num_timesteps < total_timesteps:# 这里开始执行上面的函数，在环境中收集数据，收集完了就继续训练# 如果出了故障了，就在接下来跳出循环continue_training = self.collect_rollouts(self.env, callback, self.rollout_buffer, n_rollout_steps=self.n_steps)if continue_training is False:break# 跌带次数+1,并根据当前的训练次数更新学习率iteration += 1self._update_current_progress_remaining(self.num_timesteps, total_timesteps)# 在控制台按照预先定义的频率输出相关信息# Display training infosif log_interval is not None and iteration % log_interval == 0:time_elapsed = max((time.time_ns() - self.start_time) / 1e9, sys.float_info.epsilon)fps = int((self.num_timesteps - self._num_timesteps_at_start) / time_elapsed)self.logger.record("time/iterations", iteration, exclude="tensorboard")if len(self.ep_info_buffer) > 0 and len(self.ep_info_buffer[0]) > 0:self.logger.record("rollout/ep_rew_mean", safe_mean([ep_info["r"] for ep_info in self.ep_info_buffer]))self.logger.record("rollout/ep_len_mean", safe_mean([ep_info["l"] for ep_info in self.ep_info_buffer]))self.logger.record("time/fps", fps)self.logger.record("time/time_elapsed", int(time_elapsed), exclude="tensorboard")self.logger.record("time/total_timesteps", self.num_timesteps, exclude="tensorboard")self.logger.dump(step=self.num_timesteps)self.train()callback.on_training_end()return selfdef _get_torch_save_params(self) -> Tuple[List[str], List[str]]:state_dicts = ["policy", "policy.optimizer"]return state_dicts, []