-
Notifications
You must be signed in to change notification settings - Fork 2
/
DQN+.py
340 lines (301 loc) · 12.7 KB
/
DQN+.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
import random
import gym
import numpy as np
import torch
import torch.nn.functional as F
import matplotlib.pyplot as plt
import rl_utils
from tqdm import tqdm
class Qnet(torch.nn.Module):
''' 只有一层隐藏层的Q网络 '''
def __init__(self, state_dim, hidden_dim, action_dim):
super(Qnet, self).__init__()
self.fc1 = torch.nn.Linear(state_dim, hidden_dim)
self.fc2 = torch.nn.Linear(hidden_dim, action_dim)
def forward(self, x):
x = F.relu(self.fc1(x))
return self.fc2(x)
class DQN:
''' DQN算法,包括Double DQN '''
def __init__(self,
state_dim,
hidden_dim,
action_dim,
learning_rate,
gamma,
epsilon,
target_update,
device,
dqn_type='VanillaDQN'):
self.action_dim = action_dim
self.q_net = Qnet(state_dim, hidden_dim, self.action_dim).to(device)
self.target_q_net = Qnet(state_dim, hidden_dim,
self.action_dim).to(device)
self.optimizer = torch.optim.Adam(self.q_net.parameters(),
lr=learning_rate)
self.gamma = gamma
self.epsilon = epsilon
self.target_update = target_update
self.count = 0
self.dqn_type = dqn_type
self.device = device
def take_action(self, state):
if np.random.random() < self.epsilon:
action = np.random.randint(self.action_dim)
else:
state = torch.tensor([state], dtype=torch.float).to(self.device)
action = self.q_net(state).argmax().item()
return action
def max_q_value(self, state):
state = torch.tensor([state], dtype=torch.float).to(self.device)
return self.q_net(state).max().item()
def update(self, transition_dict):
states = torch.tensor(transition_dict['states'],
dtype=torch.float).to(self.device)
actions = torch.tensor(transition_dict['actions']).view(-1, 1).to(
self.device)
rewards = torch.tensor(transition_dict['rewards'],
dtype=torch.float).view(-1, 1).to(self.device)
next_states = torch.tensor(transition_dict['next_states'],
dtype=torch.float).to(self.device)
dones = torch.tensor(transition_dict['dones'],
dtype=torch.float).view(-1, 1).to(self.device)
q_values = self.q_net(states).gather(1, actions) # Q值
# 下个状态的最大Q值
if self.dqn_type == 'DoubleDQN': # DQN与Double DQN的区别
max_action = self.q_net(next_states).max(1)[1].view(-1, 1)
max_next_q_values = self.target_q_net(next_states).gather(1, max_action)
else: # DQN的情况
max_next_q_values = self.target_q_net(next_states).max(1)[0].view(-1, 1)
q_targets = rewards + self.gamma * max_next_q_values * (1 - dones) # TD误差目标
dqn_loss = torch.mean(F.mse_loss(q_values, q_targets)) # 均方误差损失函数
self.optimizer.zero_grad() # PyTorch中默认梯度会累积,这里需要显式将梯度置为0
dqn_loss.backward() # 反向传播更新参数
self.optimizer.step()
if self.count % self.target_update == 0:
self.target_q_net.load_state_dict(
self.q_net.state_dict()) # 更新目标网络
self.count += 1
lr = 1e-2
lr = 1e-2
num_episodes = 200
hidden_dim = 128
gamma = 0.98
epsilon = 0.01
target_update = 50
buffer_size = 5000
minimal_size = 1000
batch_size = 64
device = torch.device("cuda") if torch.cuda.is_available() else torch.device(
"cpu")
env_name = 'Pendulum-v0'
env = gym.make(env_name)
state_dim = env.observation_space.shape[0]
action_dim = 11 # 将连续动作分成11个离散动作
def dis_to_con(discrete_action, env, action_dim): # 离散动作转回连续的函数
action_lowbound = env.action_space.low[0] # 连续动作的最小值
action_upbound = env.action_space.high[0] # 连续动作的最大值
return action_lowbound + (discrete_action /
(action_dim - 1)) * (action_upbound -
action_lowbound)
def train_DQN(agent, env, num_episodes, replay_buffer, minimal_size,
batch_size):
return_list = []
max_q_value_list = []
max_q_value = 0
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 = 0
state = env.reset()
done = False
while not done:
action = agent.take_action(state)
max_q_value = agent.max_q_value(
state) * 0.005 + max_q_value * 0.995 # 平滑处理
max_q_value_list.append(max_q_value) # 保存每个状态的最大Q值
action_continuous = dis_to_con(action, env,
agent.action_dim)
next_state, reward, done, _ = env.step([action_continuous])
replay_buffer.add(state, action, reward, next_state, done)
state = next_state
episode_return += reward
if replay_buffer.size() > minimal_size:
b_s, b_a, b_r, b_ns, b_d = replay_buffer.sample(
batch_size)
transition_dict = {
'states': b_s,
'actions': b_a,
'next_states': b_ns,
'rewards': b_r,
'dones': b_d
}
agent.update(transition_dict)
return_list.append(episode_return)
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)
return return_list, max_q_value_list
random.seed(0)
np.random.seed(0)
env.seed(0)
torch.manual_seed(0)
replay_buffer = rl_utils.ReplayBuffer(buffer_size)
agent = DQN(state_dim, hidden_dim, action_dim, lr, gamma, epsilon,
target_update, device)
return_list, max_q_value_list = train_DQN(agent, env, num_episodes,
replay_buffer, minimal_size,
batch_size)
episodes_list = list(range(len(return_list)))
mv_return = rl_utils.moving_average(return_list, 5)
plt.plot(episodes_list, mv_return)
plt.xlabel('Episodes')
plt.ylabel('Returns')
plt.title('DQN on {}'.format(env_name))
plt.show()
frames_list = list(range(len(max_q_value_list)))
plt.plot(frames_list, max_q_value_list)
plt.axhline(0, c='orange', ls='--')
plt.axhline(10, c='red', ls='--')
plt.xlabel('Frames')
plt.ylabel('Q value')
plt.title('DQN on {}'.format(env_name))
plt.show()
#### Double DQN
random.seed(0)
np.random.seed(0)
env.seed(0)
torch.manual_seed(0)
replay_buffer = rl_utils.ReplayBuffer(buffer_size)
agent = DQN(state_dim, hidden_dim, action_dim, lr, gamma, epsilon,
target_update, device, 'DoubleDQN')
return_list, max_q_value_list = train_DQN(agent, env, num_episodes,
replay_buffer, minimal_size,
batch_size)
episodes_list = list(range(len(return_list)))
mv_return = rl_utils.moving_average(return_list, 5)
plt.plot(episodes_list, mv_return)
plt.xlabel('Episodes')
plt.ylabel('Returns')
plt.title('Double DQN on {}'.format(env_name))
plt.show()
frames_list = list(range(len(max_q_value_list)))
plt.plot(frames_list, max_q_value_list)
plt.axhline(0, c='orange', ls='--')
plt.axhline(10, c='red', ls='--')
plt.xlabel('Frames')
plt.ylabel('Q value')
plt.title('Double DQN on {}'.format(env_name))
plt.show()
###### Dueiling DQN
class VAnet(torch.nn.Module):
''' 只有一层隐藏层的A网络和V网络 '''
def __init__(self, state_dim, hidden_dim, action_dim):
super(VAnet, self).__init__()
self.fc1 = torch.nn.Linear(state_dim, hidden_dim) # 共享网络部分
self.fc_A = torch.nn.Linear(hidden_dim, action_dim)
self.fc_V = torch.nn.Linear(hidden_dim, 1)
def forward(self, x):
A = self.fc_A(F.relu(self.fc1(x)))
V = self.fc_V(F.relu(self.fc1(x)))
Q = V + A - A.mean(1).view(-1, 1) # Q值由V值和A值计算得到
return Q
class DQN:
''' DQN算法,包括Double DQN和Dueling DQN '''
def __init__(self,
state_dim,
hidden_dim,
action_dim,
learning_rate,
gamma,
epsilon,
target_update,
device,
dqn_type='VanillaDQN'):
self.action_dim = action_dim
if dqn_type == 'DuelingDQN': # Dueling DQN采取不一样的网络框架
self.q_net = VAnet(state_dim, hidden_dim,
self.action_dim).to(device)
self.target_q_net = VAnet(state_dim, hidden_dim,
self.action_dim).to(device)
else:
self.q_net = Qnet(state_dim, hidden_dim,
self.action_dim).to(device)
self.target_q_net = Qnet(state_dim, hidden_dim,
self.action_dim).to(device)
self.optimizer = torch.optim.Adam(self.q_net.parameters(),
lr=learning_rate)
self.gamma = gamma
self.epsilon = epsilon
self.target_update = target_update
self.count = 0
self.dqn_type = dqn_type
self.device = device
def take_action(self, state):
if np.random.random() < self.epsilon:
action = np.random.randint(self.action_dim)
else:
state = torch.tensor([state], dtype=torch.float).to(self.device)
action = self.q_net(state).argmax().item()
return action
def max_q_value(self, state):
state = torch.tensor([state], dtype=torch.float).to(self.device)
return self.q_net(state).max().item()
def update(self, transition_dict):
states = torch.tensor(transition_dict['states'],
dtype=torch.float).to(self.device)
actions = torch.tensor(transition_dict['actions']).view(-1, 1).to(
self.device)
rewards = torch.tensor(transition_dict['rewards'],
dtype=torch.float).view(-1, 1).to(self.device)
next_states = torch.tensor(transition_dict['next_states'],
dtype=torch.float).to(self.device)
dones = torch.tensor(transition_dict['dones'],
dtype=torch.float).view(-1, 1).to(self.device)
q_values = self.q_net(states).gather(1, actions)
if self.dqn_type == 'DoubleDQN':
max_action = self.q_net(next_states).max(1)[1].view(-1, 1)
max_next_q_values = self.target_q_net(next_states).gather(
1, max_action)
else:
max_next_q_values = self.target_q_net(next_states).max(1)[0].view(
-1, 1)
q_targets = rewards + self.gamma * max_next_q_values * (1 - dones)
dqn_loss = torch.mean(F.mse_loss(q_values, q_targets))
self.optimizer.zero_grad()
dqn_loss.backward()
self.optimizer.step()
if self.count % self.target_update == 0:
self.target_q_net.load_state_dict(self.q_net.state_dict())
self.count += 1
random.seed(0)
np.random.seed(0)
env.seed(0)
torch.manual_seed(0)
replay_buffer = rl_utils.ReplayBuffer(buffer_size)
agent = DQN(state_dim, hidden_dim, action_dim, lr, gamma, epsilon,
target_update, device, 'DuelingDQN')
return_list, max_q_value_list = train_DQN(agent, env, num_episodes,
replay_buffer, minimal_size,
batch_size)
episodes_list = list(range(len(return_list)))
mv_return = rl_utils.moving_average(return_list, 5)
plt.plot(episodes_list, mv_return)
plt.xlabel('Episodes')
plt.ylabel('Returns')
plt.title('Dueling DQN on {}'.format(env_name))
plt.show()
frames_list = list(range(len(max_q_value_list)))
plt.plot(frames_list, max_q_value_list)
plt.axhline(0, c='orange', ls='--')
plt.axhline(10, c='red', ls='--')
plt.xlabel('Frames')
plt.ylabel('Q value')
plt.title('Dueling DQN on {}'.format(env_name))
plt.show()