-
Notifications
You must be signed in to change notification settings - Fork 0
/
ddqn.py
220 lines (183 loc) · 7.72 KB
/
ddqn.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
import sys
import gym
import csv
import time
import pylab
import random
import numpy as np
import matplotlib
from gym import wrappers
from collections import deque
from keras.layers import Dense, Reshape
from keras.optimizers import Adam
from keras.models import Sequential
EPISODES = 10000 # max number of episodes
# this is Double DQN Agent
# it uses Neural Network to approximate q function
# and replay memory & target q network
class DoubleDQNAgent:
def __init__(self, state_size, action_size):
# if you want to see learning, then change to True
self.render = False
self.load = False
self.evaluate = False
self.record = False
self.save_loc = './LunarLander_DoubleDQN'
# get size of state and action
self.state_size = state_size
self.action_size = action_size
# these is hyper parameters for the Double DQN
self.discount_factor = 0.99
self.learning_rate = 0.0001
self.epsilon = 1.0
self.epsilon_decay = 0.999995
self.epsilon_min = 0.01
self.batch_size = 64
self.train_start = 500
# create replay memory using deque
self.memory = deque(maxlen=2000)
# create main model and target model
self.model = self._build_model()
self.target_model = self._build_model()
# copy the model to target model
# --> initialize the target model so that the parameters of model & target model to be same
self.update_target_model()
if self.load:
self.load_model()
# approximate Q function using Neural Network
# state is input and Q Value of each action is output of network
def _build_model(self):
model = Sequential()
model.add(Dense(150, input_dim=self.state_size, activation='relu', kernel_initializer='glorot_uniform'))
model.add(Dense(150, activation='relu', kernel_initializer='glorot_uniform'))
model.add(Dense(self.action_size, activation='linear', kernel_initializer='glorot_uniform'))
model.summary()
model.compile(loss='mse', optimizer=Adam(lr=self.learning_rate))
return model
# after some time interval update the target model to be same with model
def update_target_model(self):
if not self.evaluate:
self.target_model.set_weights(self.model.get_weights())
# get action from model using epsilon-greedy policy
def get_action(self, state):
if np.random.rand() <= self.epsilon and not self.evaluate:
return np.random.randint(self.action_size)
else:
q_value = self.model.predict(state)
return np.argmax(q_value[0])
def get_qvals(self, state):
q_value = self.model.predict(state)
return q_value
# save sample <s,a,r,s'> to the replay memory
def replay_memory(self, state, action, reward, next_state, done):
self.memory.append((state, action, reward, next_state, done))
if self.epsilon > self.epsilon_min:
self.epsilon *= self.epsilon_decay
# pick samples randomly from replay memory (with batch_size)
def train_replay(self):
if len(self.memory) < self.train_start or self.evaluate:
return
batch_size = min(self.batch_size, len(self.memory))
mini_batch = random.sample(self.memory, batch_size)
update_input = np.zeros((batch_size, self.state_size))
update_target = np.zeros((batch_size, self.state_size))
action, reward, done = [], [], []
for i in range(batch_size):
update_input[i] = mini_batch[i][0]
action.append(mini_batch[i][1])
reward.append(mini_batch[i][2])
update_target[i] = mini_batch[i][3]
done.append(mini_batch[i][4])
target = self.model.predict(update_input)
target_next = self.model.predict(update_target)
target_val = self.target_model.predict(update_target)
for i in range(self.batch_size):
# like Q Learning, get maximum Q value at s'
# But from target model
if done[i]:
target[i][action[i]] = reward[i]
else:
# the key point of Double DQN
# selection of action is from model
# update is from target model
a = np.argmax(target_next[i])
target[i][action[i]] = reward[i] + self.discount_factor * (
target_val[i][a])
# make minibatch which includes target q value and predicted q value
# and do the model fit!
self.model.fit(update_input, target, batch_size=self.batch_size,
epochs=1, verbose=0)
# load the saved model
def load_model(self):
self.model.load_weights(self.save_loc + '.h5')
# save the model which is under training
def save_model(self):
if not self.evaluate:
self.model.save_weights(self.save_loc + '.h5')
if __name__ == "__main__":
env = gym.make('LunarLander-v2')
'''
state:
x position
y position
x velocity
y velocity
angle
angular velocity
left leg contacting ground
right leg contacting ground
'''
# get size of state and action from environment
state_size = env.observation_space.shape[0]
action_size = env.action_space.n
agent = DoubleDQNAgent(state_size, action_size)
if agent.record:
env = wrappers.Monitor(env, agent.save_loc + '/')
scores, episodes, filtered_scores, elapsed_times = [], [], [], []
start_time = time.time()
for e in range(EPISODES):
done = False
score = 0
state = env.reset()
state = np.reshape(state, [1, state_size])
while not done:
if agent.render:
env.render()
# get action for the current state and go one step in environment
action = agent.get_action(state)
next_state, reward, done, info = env.step(action)
next_state = np.reshape(next_state, [1, state_size])
# save the sample <s, a, r, s'> to the replay memory
agent.replay_memory(state, action, reward, next_state, done)
# every time step do the training
agent.train_replay()
score += reward
state = next_state
if done:
# every episode update the target model to be same with model
agent.update_target_model()
elapsed_times.append(time.time()-start_time)
scores.append(score)
episodes.append(e)
ave_score = np.mean(scores[-min(100, len(scores)):])
filtered_scores.append(ave_score)
if not agent.evaluate:
pylab.gcf().clear()
pylab.figure(figsize=(12, 8))
pylab.plot(episodes, scores, 'b', episodes, filtered_scores, 'orange')
pylab.savefig(agent.save_loc + ".png")
pylab.close()
print("episode: {:5} score: {:12.6} memory length: {:4} epsilon {:.3}"
.format(e, ave_score, len(agent.memory), agent.epsilon))
# if the mean of scores of last N episodes is bigger than X
# stop training
if ave_score >= 240 and not agent.evaluate:
np.savetxt(agent.save_loc + '.csv', filtered_scores, delimiter=",")
np.savetxt(agent.save_loc + '_time.csv', elapsed_times, delimiter=",")
agent.save_model()
time.sleep(5) # Delays for 5 seconds. You can also use a float value.
sys.exit()
# save the model every N episodes
if e % 100 == 0:
agent.save_model()
agent.save_model()