PPO 跑CartPole-v1

gym-0.26.2
cartPole-v1

参考动手学强化学习书中的代码,并做了一些修改

代码

import gym
import torch
import torch.nn as nn
import torch.nn.functional as F
import numpy as np
import matplotlib.pyplot as plt
from tqdm import tqdmclass PolicyNet(nn.Module):def __init__(self, state_dim, hidden_dim, action_dim):super().__init__()self.fc1 = nn.Linear(state_dim, hidden_dim)self.fc2 = nn.Linear(hidden_dim, action_dim)def forward(self, x):x = F.relu(self.fc1(x))return F.softmax(self.fc2(x), dim=1)class ValueNet(nn.Module):def __init__(self, state_dim, hidden_dim):super().__init__()self.fc1 = nn.Linear(state_dim, hidden_dim)self.fc2 = nn.Linear(hidden_dim, 1)def forward(self, x):x = F.relu(self.fc1(x))return self.fc2(x)class PPO:"""PPO算法，采用截断的方式"""def __init__(self, state_dim, hidden_dim, action_dim, actor_lr, critic_lr, lmbda, epochs, eps, gamma, device):self.actor = PolicyNet(state_dim, hidden_dim, action_dim).to(device)self.critic = ValueNet(state_dim, hidden_dim).to(device)self.actor_optimizer = torch.optim.Adam(self.actor.parameters(), lr=actor_lr)self.critic_optimizer = torch.optim.Adam(self.critic.parameters(), lr=critic_lr)self.gamma = gammaself.lmbda = lmbdaself.epochs = epochs    # 一条序列的数据用来训练轮数self.eps = eps  # PPO 中阶段范围的参数self.device = devicedef take_action(self, state):state = torch.FloatTensor([state]).to(self.device)probs = self.actor(state)action_dist = torch.distributions.Categorical(probs)action = action_dist.sample()return action.item()def gae(self, td_delta):td_delta = td_delta.detach().numpy()advantages_list = []advantage = 0.0for delta in td_delta[::-1]:advantage = self.gamma * self.lmbda * advantage + deltaadvantages_list.append(advantage)advantages_list.reverse()return torch.FloatTensor(advantages_list)def update(self, transition_dist):states = torch.FloatTensor(transition_dist['states']).to(self.device)actions = torch.tensor(transition_dist['actions']).reshape((-1, 1)).to(self.device)rewards = torch.FloatTensor(transition_dist['rewards']).reshape((-1, 1)).to(self.device)next_states = torch.FloatTensor(transition_dist['next_states']).to(self.device)dones = torch.FloatTensor(transition_dist['dones']).reshape((-1, 1)).to(self.device)td_target = rewards + self.gamma * self.critic(next_states) * (1 - dones)td_delta = td_target - self.critic(states)# GAE 计算广义优势advantage = self.gae(td_delta.cpu()).to(self.device)old_log_probs = torch.log(self.actor(states).gather(1, actions)).detach()for _ in range(self.epochs):log_probs = torch.log(self.actor(states).gather(1, actions))ration = torch.exp(log_probs - old_log_probs)surr1 = ration * advantagesurr2 = torch.clamp(ration, 1-self.eps, 1+self.eps) * advantage # 截断actor_loss = torch.mean(-torch.min(surr1, surr2))   # PPO损失函数critic_loss = torch.mean(F.mse_loss(self.critic(states), td_target.detach()))self.actor_optimizer.zero_grad()self.critic_optimizer.zero_grad()actor_loss.backward()critic_loss.backward()self.actor_optimizer.step()self.critic_optimizer.step()def moving_average(a, window_size):cumulative_sum = np.cumsum(np.insert(a, 0, 0))middle = (cumulative_sum[window_size:] - cumulative_sum[:-window_size]) / window_sizer = np.arange(1, window_size-1, 2)begin = np.cumsum(a[:window_size-1])[::2] / rend = (np.cumsum(a[:-window_size:-1])[::2] / r)[::-1]return np.concatenate((begin, middle, end))def train():actor_lr = 1e-3critic_lr = 1e-2num_episodes = 500hidden_dim = 128gamma = 0.98lmbda = 0.95epochs = 10eps = 0.2device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")env_name = "CartPole-v1"env = gym.make(env_name)torch.manual_seed(0)state_dim = env.observation_space.shape[0]action_dim = env.action_space.nagent = PPO(state_dim, hidden_dim, action_dim, actor_lr, critic_lr, lmbda, epochs, eps, gamma, device)return_list = []for i in range(10):with tqdm(total=int(num_episodes / 10), desc='Iteration %d' % i) as pbar:for i_episode in range(int(num_episodes / 10)):episode_return = 0transition_dict = {'states': [], 'actions': [], 'next_states': [], 'rewards': [], 'dones': []}state, _ = env.reset()done, truncated = False, Falsewhile not done and not truncated:action = agent.take_action(state)next_state, reward, done, truncated, _ = env.step(action)done = done or truncated    # 这个地方要注意transition_dict['states'].append(state)transition_dict['actions'].append(action)transition_dict['next_states'].append(next_state)transition_dict['rewards'].append(reward)transition_dict['dones'].append(done)state = next_stateepisode_return += rewardreturn_list.append(episode_return)agent.update(transition_dict)if (i_episode + 1) % 10 == 0:pbar.set_postfix({'episode': '%d' % (num_episodes / 10 * i + i_episode + 1),'return': '%.3f' % np.mean(return_list[-10:])})pbar.update(1)episodes_list = list(range(len(return_list)))plt.plot(episodes_list, return_list)plt.xlabel('Episodes')plt.ylabel('Returns')plt.title(f'PPO on {env_name}')plt.show()mv_return = moving_average(return_list, 9)plt.plot(episodes_list, mv_return)plt.xlabel('Episodes')plt.ylabel('Returns')plt.title(f'PPO on {env_name}')plt.show()if __name__ == '__main__':train()