HPCC2025/dqn.py
weixin_46229132 34725a8edf first commit
2025-03-06 20:44:30 +08:00

132 lines
4.5 KiB
Python

import torch
import torch.nn as nn
import torch.optim as optim
import numpy as np
from collections import deque
import random
class DQN(nn.Module):
def __init__(self, state_dim, action_dim):
super(DQN, self).__init__()
self.network = nn.Sequential(
nn.Linear(state_dim, 128),
nn.ReLU(),
nn.Linear(128, 128),
nn.ReLU(),
nn.Linear(128, action_dim)
)
def forward(self, x):
return self.network(x)
class Agent:
def __init__(self, state_dim, action_dim):
self.state_dim = state_dim
self.action_dim = action_dim
# DQN网络
self.eval_net = DQN(state_dim, action_dim)
self.target_net = DQN(state_dim, action_dim)
self.target_net.load_state_dict(self.eval_net.state_dict())
# 训练参数
self.learning_rate = 0.001
self.gamma = 0.99
self.epsilon = 1.0
self.epsilon_min = 0.01
self.epsilon_decay = 0.995
self.memory = deque(maxlen=10000)
self.batch_size = 64
self.optimizer = optim.Adam(self.eval_net.parameters(), lr=self.learning_rate)
# 离散化动作空间
self.v_cuts_actions = [1, 2, 3, 4, 5] # 垂直切割数选项
self.h_cuts_actions = [1, 2, 3, 4, 5] # 水平切割数选项
# self.rho_actions = [0.0, 0.2, 0.4, 0.6, 0.8, 1.0] # 卸载率选项
def discretize_action(self, q_values):
"""将Q值转换为离散动作"""
action = []
# 分别为三个维度选择动作
idx = 0
# 垂直切割数
v_cuts_q = q_values[idx:idx+len(self.v_cuts_actions)]
v_cuts_idx = torch.argmax(v_cuts_q).item()
action.append(self.v_cuts_actions[v_cuts_idx])
idx += len(self.v_cuts_actions)
# 水平切割数
h_cuts_q = q_values[idx:idx+len(self.h_cuts_actions)]
h_cuts_idx = torch.argmax(h_cuts_q).item()
action.append(self.h_cuts_actions[h_cuts_idx])
idx += len(self.h_cuts_actions)
# # 卸载率
# rho_q = q_values[idx:idx+len(self.rho_actions)]
# rho_idx = torch.argmax(rho_q).item()
# action.append(self.rho_actions[rho_idx])
return np.array(action)
def get_action_dim(self):
"""获取离散化后的动作空间维度"""
return (len(self.v_cuts_actions) +
len(self.h_cuts_actions))
# len(self.rho_actions))
def choose_action(self, state):
if random.random() < self.epsilon:
# 随机选择动作
v_cuts = random.choice(self.v_cuts_actions)
h_cuts = random.choice(self.h_cuts_actions)
# rho = random.choice(self.rho_actions)
return np.array([v_cuts, h_cuts])
else:
# 根据Q值选择动作
state = torch.FloatTensor(state).unsqueeze(0)
q_values = self.eval_net(state)
return self.discretize_action(q_values[0])
def store_transition(self, state, action, reward, next_state, done):
self.memory.append((state, action, reward, next_state, done))
def learn(self):
if len(self.memory) < self.batch_size:
return
# 随机采样batch
batch = random.sample(self.memory, self.batch_size)
states = torch.FloatTensor([x[0] for x in batch])
actions = torch.FloatTensor([x[1] for x in batch])
rewards = torch.FloatTensor([x[2] for x in batch])
next_states = torch.FloatTensor([x[3] for x in batch])
dones = torch.FloatTensor([x[4] for x in batch])
# 计算当前Q值
current_q_values = self.eval_net(states)
# 计算目标Q值
next_q_values = self.target_net(next_states).detach()
max_next_q = torch.max(next_q_values, dim=1)[0]
target_q_values = rewards + (1 - dones) * self.gamma * max_next_q
# 计算损失
loss = nn.MSELoss()(current_q_values.mean(), target_q_values.mean())
# 更新网络
self.optimizer.zero_grad()
loss.backward()
self.optimizer.step()
# 更新epsilon
self.epsilon = max(self.epsilon_min, self.epsilon * self.epsilon_decay)
# 定期更新目标网络
if self.learn.counter % 100 == 0:
self.target_net.load_state_dict(self.eval_net.state_dict())
# 添加计数器属性
learn.counter = 0