用pytorch写添加了icm算法的ddpg代码并逐行添加中文注释

很好，下面是使用PyTorch编写的带有ICM算法的DDPG代码，并为每行代码添加了中文注释：

import numpy as np
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim

# 定义Actor网络
class Actor(nn.Module):
    def __init__(self, state_dim, action_dim, max_action):
        super(Actor, self).__init__()
        self.layer_1 = nn.Linear(state_dim, 400)
        self.layer_2 = nn.Linear(400, 300)
        self.layer_3 = nn.Linear(300, action_dim)
        self.max_action = max_action

    def forward(self, x):
        x = F.relu(self.layer_1(x))
        x = F.relu(self.layer_2(x))
        x = self.max_action * torch.tanh(self.layer_3(x))
        return x

# 定义Critic网络
class Critic(nn.Module):
    def __init__(self, state_dim, action_dim):
        super(Critic, self).__init__()
        # 定义输入层和第一层隐藏层，输入为状态维度+动作维度
        self.layer_1 = nn.Linear(state_dim + action_dim, 400)
        # 定义第二层隐藏层，输入为第一层的输出维度
        self.layer_2 = nn.Linear(400, 300)
        # 定义输出层
        self.layer_3 = nn.Linear(300, 1)

    def forward(self, x, u):
        # 将状态和动作拼接
        xu = torch.cat([x, u], 1)
        # 经过第一层和第二层隐藏层
        x = F.relu(self.layer_1(xu))
        x = F.relu(self.layer_2(x))
        # 输出Q值
        x = self.layer_3(x)
        return x

# 定义ICM网络
class ICM(nn.Module):
    def __init__(self, state_dim, action_dim, eta=0.01):
        super(ICM, self).__init__()
        # 定义反向模型（从下一个状态和当前状态预测动作）
        self.forward_model_layer_1 = nn.Linear(state_dim * 2, 256)
        self.forward_model_layer_2 = nn.Linear(256, action_dim)
        # 定义正向模型（从当前状态和动作预测下一个状态）
        self.inverse_model_layer_1 = nn.Linear(state_dim * 2, 256)
        self.inverse_model_layer_2 = nn.Linear(256, action_dim)
        # 定义特征提取器
        self.encoder_layer_1 = nn.Linear(state_dim, 256)
        self.encoder_layer_2 = nn.Linear(256, 256)
        # 定义激励器
        self.eta = eta

    def forward(self, state, next_state, action):
        # 计算特征提取器的输出
        phi1 = F.relu(self.encoder_layer_1(state))
        phi1 = F.relu(self.encoder_layer_2(phi1))
        phi2 = F.relu(self.encoder_layer_1(next_state))
        phi2 = F.relu(self.encoder_layer_2(phi2))

        # 计算奖励
        intrinsic_reward = self.eta * (phi2.detach() - phi1).pow(2).sum(1) / 2

        # 计算反向模型的输出
        inverse_input = torch.cat([phi1, phi2], 1)
        pred_action = F.relu(self.inverse_model_layer_1(inverse_input))
        pred_action = torch.tanh(self.inverse_model_layer_2(pred_action))

        # 计算正向模型的输出
        forward_input = torch.cat([phi1, action], 1)
        pred_next_state = F.relu(self.forward_model_layer_1(forward_input))
        pred_next_state = self.forward_model_layer_2(pred_next_state)

        return intrinsic_reward, pred_action, pred_next_state

# 定义DDPG类
class DDPG(object):
    def __init__(self, state_dim, action_dim, max_action):
        # 定义Actor和Critic网络
        self.actor = Actor(state_dim, action_dim, max_action).to(device)
        self.actor_target = Actor(state_dim, action_dim, max_action).to(device)
        self.actor_target.load_state_dict(self.actor.state_dict())
        self.actor_optimizer = optim.Adam(self.actor.parameters(), lr=1e-4)

        self.critic = Critic(state_dim, action_dim).to(device)
        self.critic_target = Critic(state_dim, action_dim).to(device)
        self.critic_target.load_state_dict(self.critic.state_dict())
        self.critic_optimizer = optim.Adam(self.critic.parameters(), lr=1e-3)

        # 定义ICM网络
        self.icm = ICM(state_dim, action_dim).to(device)
        self.icm_optimizer = optim.Adam(self.icm.parameters(), lr=1e-3)

        # 定义其他参数
        self.max_action = max_action

    def select_action(self, state):
        # 将状态转化为张量
        state = torch.FloatTensor(state.reshape(1, -1)).to(device)
        # 将状态输入actor网络，得到动作
        action = self.actor(state).cpu().data.numpy().flatten()
        return action

    def train(self, replay_buffer, batch_size=100, gamma=0.99, eta=0.01):
        # 从replay buffer随机采样一批数据
        state, action, next_state, reward, not_done = replay_buffer.sample(batch_size)
        state = torch.FloatTensor(state).to(device)
        action = torch.FloatTensor(action).to(device)
        next_state = torch.FloatTensor(next_state).to(device)
        reward = torch.FloatTensor(reward).to(device)
        not_done = torch.FloatTensor(1 - not_done).to(device)

        # 计算当前策略下的动作
        new_action = self.actor(state)
        # 计算Q值
        critic_loss = F.mse_loss(self.critic(state, action), reward + gamma * not_done * self.critic_target(next_state, self.actor_target(next_state)))
        self.critic_optimizer.zero_grad()
        critic_loss.backward()
        self.critic_optimizer.step()

        # 计算actor的损失
        actor_loss = -self.critic(state, new_action).mean()
        self.actor_optimizer.zero_grad()
        actor_loss.backward()
        self.actor_optimizer.step()

        # 计算ICM的损失
        intrinsic_reward, pred_action, pred_next_state = self.icm(state, next_state, action)
        icm_loss = F.mse_loss(pred_action, action) + F.mse_loss(pred_next_state, next_state)
        self.icm_optimizer.zero_grad()
        icm_loss.backward()
        self.icm_optimizer.step()

        # 更新目标网络
        for param, target_param in zip(self.critic.parameters(), self.critic_target.parameters()):
            target_param.data.copy_(0.995 * target_param.data + 0.005 * param.data)

        for param, target_param in zip(self.actor.parameters(), self.actor_target.parameters()):
            target_param.data.copy_(0.995 * target_param.data + 0.005 * param.data)

        # 返回ICM奖励和外部奖励之和
        return intrinsic_reward, critic_loss.item(), actor_loss.item(), icm_loss.item()

希望这个代码可以帮助你更好地理解DDPG和ICM算法。