hot update Double DQN
This commit is contained in:
johnjim0816
@@ -5,7 +5,7 @@
|
||||
@Email: johnjim0816@gmail.com
|
||||
@Date: 2020-06-12 00:50:49
|
||||
@LastEditor: John
|
||||
LastEditTime: 2022-07-21 00:08:26
|
||||
LastEditTime: 2022-08-29 23:34:20
|
||||
@Discription:
|
||||
@Environment: python 3.7.7
|
||||
'''
|
||||
@@ -20,148 +20,87 @@ import torch.nn.functional as F
|
||||
import random
|
||||
import math
|
||||
import numpy as np
|
||||
|
||||
class ReplayBuffer:
|
||||
def __init__(self, capacity):
|
||||
self.capacity = capacity # 经验回放的容量
|
||||
self.buffer = [] # 缓冲区
|
||||
self.position = 0
|
||||
|
||||
def push(self, state, action, reward, next_state, done):
|
||||
''' 缓冲区是一个队列，容量超出时去掉开始存入的转移(transition)
|
||||
'''
|
||||
if len(self.buffer) < self.capacity:
|
||||
self.buffer.append(None)
|
||||
self.buffer[self.position] = (state, action, reward, next_state, done)
|
||||
self.position = (self.position + 1) % self.capacity
|
||||
|
||||
def sample(self, batch_size):
|
||||
batch = random.sample(self.buffer, batch_size) # 随机采出小批量转移
|
||||
state, action, reward, next_state, done = zip(*batch) # 解压成状态，动作等
|
||||
return state, action, reward, next_state, done
|
||||
|
||||
def __len__(self):
|
||||
''' 返回当前存储的量
|
||||
'''
|
||||
return len(self.buffer)
|
||||
|
||||
class MLP(nn.Module):
|
||||
def __init__(self, n_states,n_actions,hidden_dim=128):
|
||||
""" 初始化q网络，为全连接网络
|
||||
n_states: 输入的特征数即环境的状态维度
|
||||
n_actions: 输出的动作维度
|
||||
"""
|
||||
super(MLP, self).__init__()
|
||||
self.fc1 = nn.Linear(n_states, hidden_dim) # 输入层
|
||||
self.fc2 = nn.Linear(hidden_dim,hidden_dim) # 隐藏层
|
||||
self.fc3 = nn.Linear(hidden_dim, n_actions) # 输出层
|
||||
|
||||
def forward(self, x):
|
||||
# 各层对应的激活函数
|
||||
x = F.relu(self.fc1(x))
|
||||
x = F.relu(self.fc2(x))
|
||||
return self.fc3(x)
|
||||
|
||||
class DoubleDQN:
|
||||
def __init__(self, n_states, n_actions, model, memory, cfg):
|
||||
self.n_actions = n_actions # 总的动作个数
|
||||
self.device = torch.device(cfg.device) # 设备，cpu或gpu等
|
||||
self.gamma = cfg.gamma
|
||||
# e-greedy策略相关参数
|
||||
self.sample_count = 0
|
||||
self.epsilon_start = cfg.epsilon_start
|
||||
self.epsilon_end = cfg.epsilon_end
|
||||
self.epsilon_decay = cfg.epsilon_decay
|
||||
self.batch_size = cfg.batch_size
|
||||
self.policy_net = model.to(self.device)
|
||||
self.target_net = model.to(self.device)
|
||||
def __init__(self,models, memories, cfg):
|
||||
self.n_actions = cfg['n_actions']
|
||||
self.device = torch.device(cfg['device'])
|
||||
self.gamma = cfg['gamma']
|
||||
## e-greedy parameters
|
||||
self.sample_count = 0 # sample count for epsilon decay
|
||||
self.epsilon_start = cfg['epsilon_start']
|
||||
self.epsilon_end = cfg['epsilon_end']
|
||||
self.epsilon_decay = cfg['epsilon_decay']
|
||||
self.batch_size = cfg['batch_size']
|
||||
self.policy_net = models['Qnet'].to(self.device)
|
||||
self.target_net = models['Qnet'].to(self.device)
|
||||
# target_net copy from policy_net
|
||||
for target_param, param in zip(self.target_net.parameters(), self.policy_net.parameters()):
|
||||
target_param.data.copy_(param.data)
|
||||
# self.target_net.eval() # 不启用 BatchNormalization 和 Dropout
|
||||
# 可查parameters()与state_dict()的区别，前者require_grad=True
|
||||
self.optimizer = optim.Adam(self.policy_net.parameters(), lr=cfg.lr)
|
||||
self.loss = 0
|
||||
self.memory = memory
|
||||
# self.target_net.eval() # donnot use BatchNormalization or Dropout
|
||||
# the difference between parameters() and state_dict() is that parameters() require_grad=True
|
||||
self.optimizer = optim.Adam(self.policy_net.parameters(), lr=cfg['lr'])
|
||||
self.memory = memories['Memory']
|
||||
self.update_flag = False
|
||||
|
||||
def sample(self, state):
|
||||
'''选择动作
|
||||
def sample_action(self, state):
|
||||
''' sample action
|
||||
'''
|
||||
self.sample_count += 1
|
||||
self.epsilon = self.epsilon_end + (self.epsilon_start - self.epsilon_end) * math.exp(-1. * self.sample_count / self.epsilon_decay)
|
||||
if random.random() > self.epsilon:
|
||||
with torch.no_grad():
|
||||
# 先转为张量便于丢给神经网络,state元素数据原本为float64
|
||||
# 注意state=torch.tensor(state).unsqueeze(0)跟state=torch.tensor([state])等价
|
||||
state = torch.tensor(
|
||||
[state], device=self.device, dtype=torch.float32)
|
||||
# 如tensor([[-0.0798, -0.0079]], grad_fn=<AddmmBackward>)
|
||||
state = torch.tensor(state, device=self.device, dtype=torch.float32).unsqueeze(0)
|
||||
q_value = self.policy_net(state)
|
||||
# tensor.max(1)返回每行的最大值以及对应的下标，
|
||||
# 如torch.return_types.max(values=tensor([10.3587]),indices=tensor([0]))
|
||||
# 所以tensor.max(1)[1]返回最大值对应的下标，即action
|
||||
action = q_value.max(1)[1].item()
|
||||
else:
|
||||
action = random.randrange(self.n_actions)
|
||||
return action
|
||||
def predict(self, state):
|
||||
'''选择动作
|
||||
def predict_action(self, state):
|
||||
''' predict action
|
||||
'''
|
||||
with torch.no_grad():
|
||||
state = torch.tensor([state], device=self.device, dtype=torch.float32)
|
||||
state = torch.tensor(state, device=self.device, dtype=torch.float32).unsqueeze(0)
|
||||
q_value = self.policy_net(state)
|
||||
action = q_value.max(1)[1].item()
|
||||
return action
|
||||
def update(self):
|
||||
if len(self.memory) < self.batch_size: # 只有memory满了才会更新
|
||||
if len(self.memory) < self.batch_size: # when transitions in memory donot meet a batch, not update
|
||||
return
|
||||
# 从memory中随机采样transition
|
||||
state_batch, action_batch, reward_batch, next_state_batch, done_batch = self.memory.sample(
|
||||
self.batch_size)
|
||||
else:
|
||||
if not self.update_flag:
|
||||
print("Begin to update!")
|
||||
self.update_flag = True
|
||||
# sample a batch of transitions from replay buffer
|
||||
state_batch, action_batch, reward_batch, next_state_batch, done_batch = self.memory.sample(self.batch_size)
|
||||
# convert to tensor
|
||||
state_batch = torch.tensor(
|
||||
state_batch, device=self.device, dtype=torch.float)
|
||||
action_batch = torch.tensor(action_batch, device=self.device).unsqueeze(
|
||||
1) # 例如tensor([[1],...,[0]])
|
||||
reward_batch = torch.tensor(
|
||||
reward_batch, device=self.device, dtype=torch.float) # tensor([1., 1.,...,1])
|
||||
next_state_batch = torch.tensor(
|
||||
next_state_batch, device=self.device, dtype=torch.float)
|
||||
|
||||
done_batch = torch.tensor(np.float32(
|
||||
done_batch), device=self.device) # 将bool转为float然后转为张量
|
||||
# 计算当前(s_t,a)对应的Q(s_t, a)
|
||||
q_values = self.policy_net(state_batch)
|
||||
next_q_values = self.policy_net(next_state_batch)
|
||||
# 代入当前选择的action，得到Q(s_t|a=a_t)
|
||||
q_value = q_values.gather(dim=1, index=action_batch)
|
||||
'''以下是Nature DQN的q_target计算方式
|
||||
# 计算所有next states的Q'(s_{t+1})的最大值，Q'为目标网络的q函数
|
||||
next_q_state_value = self.target_net(
|
||||
next_state_batch).max(1)[0].detach() # 比如tensor([ 0.0060, -0.0171,...,])
|
||||
# 计算 q_target
|
||||
# 对于终止状态，此时done_batch[0]=1, 对应的expected_q_value等于reward
|
||||
q_target = reward_batch + self.gamma * next_q_state_value * (1-done_batch[0])
|
||||
'''
|
||||
'''以下是Double DQN q_target计算方式，与NatureDQN稍有不同'''
|
||||
next_target_values = self.target_net(
|
||||
next_state_batch)
|
||||
# 选出Q(s_t‘, a)对应的action，代入到next_target_values获得target net对应的next_q_value，即Q’(s_t|a=argmax Q(s_t‘, a))
|
||||
next_target_q_value = next_target_values.gather(1, torch.max(next_q_values, 1)[1].unsqueeze(1)).squeeze(1)
|
||||
q_target = reward_batch + self.gamma * next_target_q_value * (1-done_batch)
|
||||
self.loss = nn.MSELoss()(q_value, q_target.unsqueeze(1)) # 计算 均方误差loss
|
||||
# 优化模型
|
||||
self.optimizer.zero_grad() # zero_grad清除上一步所有旧的gradients from the last step
|
||||
# loss.backward()使用backpropagation计算loss相对于所有parameters(需要gradients)的微分
|
||||
self.loss.backward()
|
||||
for param in self.policy_net.parameters(): # clip防止梯度爆炸
|
||||
state_batch = torch.tensor(np.array(state_batch), device=self.device, dtype=torch.float)
|
||||
action_batch = torch.tensor(action_batch, device=self.device).unsqueeze(1) # shape(batchsize,1)
|
||||
reward_batch = torch.tensor(reward_batch, device=self.device, dtype=torch.float).unsqueeze(1) # shape(batchsize,1)
|
||||
next_state_batch = torch.tensor(np.array(next_state_batch), device=self.device, dtype=torch.float)
|
||||
done_batch = torch.tensor(np.float32(done_batch), device=self.device).unsqueeze(1) # shape(batchsize,1)
|
||||
# compute current Q(s_t|a=a_t)
|
||||
q_value_batch = self.policy_net(state_batch).gather(dim=1, index=action_batch) # shape(batchsize,1),requires_grad=True
|
||||
next_q_value_batch = self.policy_net(next_state_batch)
|
||||
'''the following is the way of computing Double DQN expected_q_value，a bit different from Nature DQN'''
|
||||
next_target_value_batch = self.target_net(next_state_batch)
|
||||
# choose action a from Q(s_t‘, a), next_target_values obtain next_q_value，which is Q’(s_t|a=argmax Q(s_t‘, a))
|
||||
next_target_q_value_batch = next_target_value_batch.gather(1, torch.max(next_q_value_batch, 1)[1].unsqueeze(1)) # shape(batchsize,1)
|
||||
expected_q_value_batch = reward_batch + self.gamma * next_target_q_value_batch * (1-done_batch)
|
||||
loss = nn.MSELoss()(q_value_batch , expected_q_value_batch)
|
||||
self.optimizer.zero_grad()
|
||||
loss.backward()
|
||||
# clip to avoid gradient explosion
|
||||
for param in self.policy_net.parameters():
|
||||
param.grad.data.clamp_(-1, 1)
|
||||
self.optimizer.step() # 更新模型
|
||||
self.optimizer.step()
|
||||
|
||||
def save(self,path):
|
||||
def save_model(self,path):
|
||||
from pathlib import Path
|
||||
# create path
|
||||
Path(path).mkdir(parents=True, exist_ok=True)
|
||||
torch.save(self.target_net.state_dict(), path+'checkpoint.pth')
|
||||
|
||||
def load(self,path):
|
||||
def load_model(self,path):
|
||||
self.target_net.load_state_dict(torch.load(path+'checkpoint.pth'))
|
||||
for target_param, param in zip(self.target_net.parameters(), self.policy_net.parameters()):
|
||||
param.data.copy_(target_param.data)
|
||||
|
||||
Reference in New Issue
Block a user