-
Notifications
You must be signed in to change notification settings - Fork 1
/
Advantage_Actor_Critic.py
175 lines (152 loc) · 7.25 KB
/
Advantage_Actor_Critic.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
import gym
import tensorflow as tf
import tensorflow.contrib.layers as layers
import numpy as np
import random
from collections import deque
import baselines.common.tf_util as U
import time
GAMMA = 1
INITIAL_EPSILON = 1
FINAL_EPSILON = 0.01
REPLAY_SIZE = 1000
BATCH_SIZE = 10
UPDATE_SEQ = 1000
TRAIN_SEQ = 1
class Advantage_Actor_Critic():
def __init__(self, env):
self.replay_buffer = deque()
self.time_step = 0
self.action_dim = env.action_space.n
self.critic_func_input, self.critic_func = self.create_network(1, False, "critic_func")
self.actor_func_input, self.actor_func = self.create_network(self.action_dim, True, "actor_func")
self.critic_y_input, self.critic_optimizer = self.create_training_critic_method()
# self.I, self.delta, self.actor_optimizer = self.create_training_actor_method()
self.I, self.delta, self.actor_optimizer = self.create_actor_train_method()
self.saver = tf.train.Saver()
self.session = tf.InteractiveSession()
self.session.run(tf.initialize_all_variables())
checkpoint = tf.train.get_checkpoint_state("saved_networks")
if checkpoint and checkpoint.model_checkpoint_path:
self.saver.restore(self.session, checkpoint.model_checkpoint_path)
print("Successfully loaded:", checkpoint.model_checkpoint_path)
else:
print("Could not find old network weights")
def create_network(self, output_num, is_actor, scope):
with tf.variable_scope(scope, reuse=False):
state_input = tf.placeholder('float', [None, 4])
out = layers.fully_connected(state_input, num_outputs=16, activation_fn=tf.nn.relu)
out = layers.fully_connected(out, num_outputs=16, activation_fn=tf.nn.relu)
out = layers.fully_connected(out, num_outputs=16, activation_fn=tf.nn.relu)
out = layers.fully_connected(out, num_outputs=output_num, activation_fn=None)
if is_actor:
out =layers.softmax(out)
return state_input, out
def create_training_critic_method(self):
y_input = tf.placeholder("float", [None, 1])
cost = tf.reduce_mean(tf.square(y_input - self.critic_func))
optimizer = tf.train.AdamOptimizer(1e-03).minimize(cost)
return y_input, optimizer
# def create_training_actor_method(self):
# I = tf.placeholder("float", [None, 1])
# delta = tf.placeholder("float", [None, 1])
# self.state_actor = tf.placeholder("float", [None, self.action_dim])
# self.action_input = tf.placeholder("float", [None, self.action_dim])
# actor_vars = tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES, scope="actor_func")
# actor = tf.reduce_sum(tf.multiply(self.state_actor, self.action_input), reduction_indices=1)
# coefficient = tf.multiply(-1.0, tf.multiply(I, delta))
# actor = tf.reshape(actor, tf.shape(coefficient))
# coefficient = tf.div(coefficient, actor)
# log_actor = tf.multiply(coefficient, self.actor_func)
# gredients = tf.gradients(log_actor, actor_vars)
# optimizer = tf.train.AdamOptimizer(1e-04).apply_gradients(zip(gredients, actor_vars))
# return I, delta, optimizer
def create_actor_train_method(self):
I = tf.placeholder("float", [None, 1])
delta = tf.placeholder("float", [None, 1])
self.action_input = tf.placeholder("float", [None, self.action_dim])
actor = tf.reduce_sum(tf.multiply(self.actor_func, self.action_input), reduction_indices=1)
cost = tf.reduce_mean(I * delta * tf.log(actor))
optimizer = tf.train.AdamOptimizer(1e-04).minimize(-cost)
return I, delta, optimizer
def perceive(self, state, action, reward, next_state, done, I, step_num):
one_hot_action = np.zeros(self.action_dim)
one_hot_action[action] = 1
self.replay_buffer.append((state, one_hot_action, reward, next_state, done, I))
if len(self.replay_buffer) > REPLAY_SIZE:
self.replay_buffer.popleft()
if len(self.replay_buffer) >= BATCH_SIZE and step_num % TRAIN_SEQ == 0:
self.train_network()
def train_network(self):
minibatch = random.sample(self.replay_buffer, BATCH_SIZE)
state_batch = []
action_batch = []
reward_batch = []
next_state_batch = []
I_batch = []
for state, action, reward, next_state, _, I in minibatch:
state_batch.append(state)
action_batch.append(action)
reward_batch.append(reward)
next_state_batch.append(next_state)
I_batch.append(np.array([I]))
y_batch = []
next_state_critic_batch = self.critic_func.eval(feed_dict={self.critic_func_input: next_state_batch})
state_critic_batch = self.critic_func.eval(feed_dict={self.critic_func_input: state_batch})
# print("v function", state_critic_batch[0])
# time.sleep(0.1)
state_actor_batch = self.actor_func.eval(feed_dict={self.actor_func_input: state_batch})
# print("pi function", state_actor_batch[0])
# time.sleep(0.1)
for i in range(0, BATCH_SIZE):
done = minibatch[i][4]
if done:
y_batch.append(np.array([reward_batch[i]]))
else:
y_batch.append(np.array([reward_batch[i]]) + GAMMA * next_state_critic_batch[i])
delta_batch = y_batch - state_critic_batch
self.critic_optimizer.run(feed_dict={
self.critic_y_input: y_batch,
self.critic_func_input: state_batch
})
# print()
# print("befor v function", state_critic_batch[0])
# after_critic_batch = self.critic_func.eval(feed_dict={self.critic_func_input: state_batch})
# print("after v function", after_critic_batch[0])
# self.actor_optimizer.run(feed_dict={
# self.actor_func_input: state_batch,
# self.I: I_batch,
# self.delta: delta_batch,
# self.action_input: action_batch,
# self.state_actor: state_actor_batch
# })
self.actor_optimizer.run(feed_dict={
self.action_input: action_batch,
self.actor_func_input: state_batch,
self.I: I_batch,
self.delta: delta_batch,
})
# print()
# print("befor pi function", state_actor_batch[0])
# after_actor_batch = self.actor_func.eval(feed_dict={self.actor_func_input: state_batch})
# print("after pi function", after_actor_batch[0])
def select_action(self, state):
action_dis = self.actor_func.eval(feed_dict={
self.actor_func_input: [state]
})[0]
# print(action_dis)
# time.sleep(0.1)
action = np.random.choice(self.action_dim, 1, p=action_dis)
return action[0]
def action(self, state):
return np.argmax(self.actor_func.eval(feed_dict={
self.actor_func_input: [state]
})[0])
def weight_variable(self, shape):
initial = tf.truncated_normal(shape)
return tf.Variable(initial)
def bias_variable(self, shape):
initial = tf.constant(0.01, shape=shape)
return tf.Variable(initial)
def conv2d(self, x, w):
return tf.nn.conv2d(x, w, strides=[1, 1, 1, 1], padding='SAME')