/
model.py
174 lines (124 loc) · 7.83 KB
/
model.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
import tensorflow as tf
import numpy as np
import config
import time
import os
class SentenceCNN(object):
"""
A CNN for text classification.
Uses an embedding layer, followed by a convolutional, max-pooling and softmax layer.
"""
def __init__(self, sequence_length, num_classes, alphabet_size, sess):
self.sequence_length = sequence_length
self.num_classes = num_classes
self.alphabet_size = alphabet_size
self.sess = sess
# Placeholders for input, output and dropout
self.input_x = tf.placeholder(tf.int32, [None, alphabet_size, sequence_length], name="input_x")
self.input_y = tf.placeholder(tf.float32, [None, num_classes], name="input_y")
self.dropout_keep_prob = tf.placeholder(tf.float32, name="dropout_keep_prob")
# Keeping track of l2 regularization loss (optional)
self.l2_loss = tf.constant(0.0)
def inference(self):
# network weights
convolution_weights_1 = tf.Variable(tf.truncated_normal([self.alphabet_size, 7, 256], stddev=0.05))
convolution_bias_1 = tf.Variable(tf.constant(0.05, shape=[256]))
hidden_convolutional_layer_1 = tf.nn.relu(
tf.nn.conv2d(self.input_x, convolution_weights_1, strides=[1, 1, 1, 1], padding="SAME") + convolution_bias_1)
hidden_max_pooling_layer_1 = tf.nn.max_pool(hidden_convolutional_layer_1, ksize=[1, 256, 3, 1],
strides=[1, 3, 3, 1], padding="SAME")
convolution_weights_2 = tf.Variable(tf.truncated_normal([256, 7, 256], stddev=0.05))
convolution_bias_2 = tf.Variable(tf.constant(0.05, shape=[256]))
hidden_convolutional_layer_2 = tf.nn.relu(
tf.nn.conv2d(hidden_max_pooling_layer_1, convolution_weights_2, strides=[1, 1, 1, 1],
padding="SAME") + convolution_bias_2)
hidden_max_pooling_layer_2 = tf.nn.max_pool(hidden_convolutional_layer_2, ksize=[1, 256, 3, 1],
strides=[1, 3, 3, 1], padding="SAME")
convolution_weights_3 = tf.Variable(tf.truncated_normal([256, 3, 256], stddev=0.05))
convolution_bias_3 = tf.Variable(tf.constant(0.05, shape=[256]))
hidden_convolutional_layer_3 = tf.nn.relu(
tf.nn.conv2d(hidden_max_pooling_layer_2, convolution_weights_3,
strides=[1, 1, 1, 1], padding="SAME") + convolution_bias_3)
convolution_weights_4 = tf.Variable(tf.truncated_normal([256, 3, 256], stddev=0.05))
convolution_bias_4 = tf.Variable(tf.constant(0.05, shape=[256]))
hidden_convolutional_layer_4 = tf.nn.relu(
tf.nn.conv2d(hidden_convolutional_layer_3, convolution_weights_4,
strides=[1, 1, 1, 1], padding="SAME") + convolution_bias_4)
convolution_weights_5 = tf.Variable(tf.truncated_normal([256, 3, 256], stddev=0.05))
convolution_bias_5 = tf.Variable(tf.constant(0.05, shape=[256]))
hidden_convolutional_layer_5 = tf.nn.relu(
tf.nn.conv2d(hidden_convolutional_layer_4, convolution_weights_5,
strides=[1, 1, 1, 1], padding="SAME") + convolution_bias_5)
convolution_weights_6 = tf.Variable(tf.truncated_normal([256, 3, 256], stddev=0.05))
convolution_bias_6 = tf.Variable(tf.constant(0.05, shape=[256]))
hidden_convolutional_layer_6 = tf.nn.relu(
tf.nn.conv2d(hidden_convolutional_layer_5, convolution_weights_6,
strides=[1, 1, 1, 1], padding="SAME") + convolution_bias_6)
hidden_max_pooling_layer_6 = tf.nn.max_pool(hidden_convolutional_layer_6, ksize=[1, 256, 3, 1],
strides=[1, 3, 3, 1], padding="SAME")
hidden_max_pooling_layer_6_shape = hidden_max_pooling_layer_6.get_shape()[1] * \
hidden_max_pooling_layer_6.get_shape()[2] * \
hidden_max_pooling_layer_6.get_shape()[3]
hidden_max_pooling_layer_6_shape = hidden_max_pooling_layer_6_shape.value
hidden_convolutional_layer_6_flat = tf.reshape(hidden_max_pooling_layer_6, [-1, hidden_max_pooling_layer_6_shape])
feed_forward_weights_7 = tf.Variable(tf.truncated_normal([hidden_max_pooling_layer_6_shape, 1024], stddev=0.05))
feed_forward_bias_7 = tf.Variable(tf.constant(0.05, shape=[1024]))
feed_forward_layer_7 = tf.nn.relu(
tf.matmul(hidden_convolutional_layer_6_flat, feed_forward_weights_7) + feed_forward_bias_7)
# Add dropout
h_drop_1 = tf.nn.dropout(feed_forward_layer_7, self.dropout_keep_prob)
feed_forward_weights_8 = tf.Variable(tf.truncated_normal([1024, 1024], stddev=0.05))
feed_forward_bias_8 = tf.Variable(tf.constant(0.05, shape=[1024]))
feed_forward_layer_8 = tf.nn.relu(
tf.matmul(h_drop_1, feed_forward_weights_8) + feed_forward_bias_8)
# Add dropout
h_drop_2 = tf.nn.dropout(feed_forward_layer_8, self.dropout_keep_prob)
feed_forward_weights_9 = tf.Variable(tf.truncated_normal([1024, 2], stddev=0.05))
feed_forward_bias_9 = tf.Variable(tf.constant(0.05, shape=[2]))
self.output_layer = tf.matmul(h_drop_2, feed_forward_weights_9) + feed_forward_bias_9
def loss(self):
# self.l2_loss += tf.nn.l2_loss(W)
# self.l2_loss += tf.nn.l2_loss(b)
# self.scores = tf.nn.xw_plus_b(self.h_drop, W, b, name="scores")
self.predictions = tf.argmax(self.output_layer, 1, name="predictions")
# CalculateMean cross-entropy loss
with tf.name_scope("cross-entropy"):
losses = tf.nn.softmax_cross_entropy_with_logits(self.output_layer, self.input_y)
self.loss = tf.reduce_mean(losses)
# + config.l2_reg_lambda * self.l2_loss
self.val_loss_average = tf.train.ExponentialMovingAverage(0.9999, name='val_loss_mov_avg')
self.val_loss_average_op = self.val_loss_average.apply([self.loss])
# Accuracy
with tf.name_scope("accuracy"):
correct_predictions = tf.equal(self.predictions, tf.argmax(self.input_y, 1))
self.accuracy = tf.reduce_mean(tf.cast(correct_predictions, "float"), name="accuracy")
def train(self):
self.global_step = tf.Variable(0, name="global_step", trainable=False)
with tf.control_dependencies([self.val_loss_average_op]):
optimizer = tf.train.AdamOptimizer(config.learning_rate)
self.grads_and_vars = optimizer.compute_gradients(self.loss)
self.train_op = optimizer.apply_gradients(self.grads_and_vars, global_step=self.global_step)
def summary(self):
# Keep track of gradient values and sparsity (optional)
grad_summaries = []
for grad, var in self.grads_and_vars:
if grad is not None:
grad_hist_summary = tf.histogram_summary(var.op.name + '/gradients/hist', grad)
sparsity_summary = tf.scalar_summary(var.op.name + '/gradients/sparsity', tf.nn.zero_fraction(grad))
grad_summaries.append(grad_hist_summary)
grad_summaries.append(sparsity_summary)
grad_summaries_merged = tf.merge_summary(grad_summaries)
# Output directory for models and summaries
timestamp = str(int(time.time()))
print("Writing to %s\n" % config.out_dir)
# Summaries for loss and accuracy
loss_summary = tf.scalar_summary("loss", self.loss)
acc_summary = tf.scalar_summary("accuracy", self.accuracy)
# Train Summaries
self.train_summary_op = tf.merge_summary([loss_summary, acc_summary, grad_summaries_merged])
train_summary_dir = os.path.join(config.out_dir, "summaries", "train")
self.train_summary_writer = tf.train.SummaryWriter(train_summary_dir, self.sess.graph_def)
# Dev summaries
self.val_summary_op = tf.merge_summary([loss_summary, acc_summary])
val_summary_dir = os.path.join(config.out_dir, "summaries", "val")
self.val_summary_writer = tf.train.SummaryWriter(val_summary_dir, self.sess.graph_def)