-
Notifications
You must be signed in to change notification settings - Fork 13
/
cnn3.py
171 lines (113 loc) · 5.06 KB
/
cnn3.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
import tensorflow as tf
from data_reader import load_data
import numpy as np
from uncompress import *
import os
slim = tf.contrib.slim
def lrelu(alpha):
def op(inputs):
return tf.maximum(alpha * inputs, inputs, name='leaky_relu')
return op
def conv_net(input):
with slim.arg_scope([slim.conv2d, slim.fully_connected], #using scope to avoid mentioning the paramters repeatdely
activation_fn=lrelu(0.005),
weights_initializer=tf.truncated_normal_initializer(0.0, 0.01),
weights_regularizer=slim.l2_regularizer(0.0005)):
#net = slim.max_pool2d(input,(1,4),(1,4), padding='VALID', scope='pool_0')
net = slim.conv2d(input, 512, (5,86796), 1, padding='SAME', scope='conv_1')
net = slim.max_pool2d(net, (4,1),4, padding='VALID', scope='pool_2')
net = slim.conv2d(net, 512, (5,1), 1, scope='conv_3')
net = slim.max_pool2d(net, (4,1),4, padding='VALID', scope='pool_4')
net = slim.conv2d(net, 1024, (5,1), 1, scope='conv_4')
net = slim.flatten(net, scope='flatten_5')
'''net = slim.fully_connected(net, 1024, scope='fc_6',activation_fn=tf.nn.softmax)
net = slim.fully_connected(net, 256, scope='fc_7',activation_fn=tf.nn.softmax)
net = slim.fully_connected(net, 2, scope='fc_8',activation_fn=tf.nn.softmax)'''
net = slim.fully_connected(net, 4096, scope='fc5')
net = slim.dropout(net, 0.5, scope='dropout6')
net = slim.fully_connected(net, 4096, scope='fc7')
net = slim.dropout(net, 0.5, scope='dropout8')
net = slim.fully_connected(net,2, activation_fn=None, scope='fc9')
return net
def one_hot(batch_size,Y):
B = np.zeros((batch_size,2))
B[np.arange(batch_size),Y] = 1
return B
if __name__=='__main__':
#os.environ['CUDA_VISIBLE_DEVICES'] = ''
# print one_hot(3,np.array((1,0,1)))
# exit(0)
# Training Parameters
learning_rate = 0.00001
num_epoch = 5
batch_size = 2
display_step = 1
input_size = 50
num_classes = 2
X = tf.placeholder(tf.float32, [None, input_size,86796,1])
Y = tf.placeholder(tf.float32, [None, num_classes])
#logits = conv_net(X)
#prediction = tf.nn.softmax(logits)
prediction = conv_net(X)
# Define loss and optimizer
'''loss_op = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(
logits=logits, labels=Y))'''
loss_op = slim.losses.softmax_cross_entropy(prediction, Y)
tf.summary.scalar('loss',loss_op)
optimizer = tf.train.AdagradOptimizer(learning_rate=learning_rate)
train_op = optimizer.minimize(loss_op)
# Evaluate model
correct_pred = tf.equal(tf.argmax(prediction, 1), tf.argmax(Y, 1))
accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))
tf.summary.scalar('accuracy',accuracy)
# Initialize the variables (i.e. assign their default value)
init = tf.global_variables_initializer()
data_X,data_Y = load_data()
indices = np.random.permutation(np.arange(data_X.shape[0]))
data_X = data_X[indices,:,:]
data_Y = data_Y[indices]
merged = tf.summary.merge_all()
saver = tf.train.Saver()
with tf.Session() as sess:
train_writer = tf.summary.FileWriter("cnn_logs_8/",
sess.graph)
# Run the initializer
sess.run(init)
'''
print 'restoring session'
saver.restore(sess, "logs3/epoch0i180.ckpt")
print 'done loading'
# exit(0) '''
i = 0
print 'started training'
for epoch in range(num_epoch):
for step in range(data_X.shape[0]/batch_size):
batch_x, batch_y = data_X[step*batch_size:(step+1)*batch_size],\
data_Y[step*batch_size:(step+1)*batch_size]
batch_x = uncompress(batch_x,86796)
# print batch_y
batch_y = one_hot(batch_size,batch_y)
batch_y = np.repeat(batch_y,50,axis=0)
# print batch_y
assert(batch_x.shape[0]==batch_y.shape[0])
# print batch_x.shape
# print batch_y.shape
# exit(0)
# Run optimization op (backprop)
_,summary = sess.run([train_op,merged], feed_dict={X: batch_x, Y: batch_y})
train_writer.add_summary(summary, i)
if step % display_step == 0:
# Calculate batch loss and accuracy
loss, acc,summary = sess.run([loss_op, accuracy,merged], feed_dict={X: batch_x,
Y: batch_y})
print("LR : "+str(learning_rate)+" Epoch : " + str(epoch) + " Step " + str(step) + ", Minibatch Loss= " + \
"{:.4f}".format(loss) + ", Training Accuracy= " + \
"{:.3f}".format(acc))
# train_writer.add_summary(summary, step)
if i%20 == 0:
print 'saving checkpoint'
save_path = saver.save(sess, os.path.join('cnn_logs_8','epoch'+str(epoch)+\
'i'+str(i)+'.ckpt'))
print("Model saved in path: %s" % save_path)
i+=1
# print("Optimization Finished!")