-
Notifications
You must be signed in to change notification settings - Fork 0
/
lstm_model.py
217 lines (132 loc) · 8.63 KB
/
lstm_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
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
import tensorflow as tf
import numpy as np
from tensorflow.contrib import rnn
from six.moves import xrange
class SentimentModel(object):
def __init__(self, vocabulary_size, hidden_size, num_layers, dropout, gradient_clip,
learning_rate, max_sequence_length, batch_size, lr_decay_rate, forward_only= False):
self.vocabulary_size = vocabulary_size
self.num_layers= num_layers
self.dropout= dropout
self.gradient_clip= gradient_clip
self.learning_rate= tf.Variable(float(learning_rate), trainable= False, dtype= tf.float32)
self.max_sequence_length= max_sequence_length
self.batch_size= batch_size
self.global_step= tf.Variable(0, trainable= False)
self.learning_rate_decay_op= self.learning_rate.assign(self.learning_rate * lr_decay_rate)
self.num_classes= 2
self.input= tf.placeholder(tf.int32, shape=[None, max_sequence_length], name= 'input')
self.output= tf.placeholder(tf.int32, shape=[None, self.num_classes], name='output')
self.input_length= tf.placeholder(tf.int32, shape=[None], name='end_of_sequence')
# def loss(labels, logits):
# local_inputs= tf.cast(logits, tf.float32)
#
# losses=tf.cast(tf.nn.softmax_cross_entropy_with_logits(labels= labels, logits= local_inputs), tf.float32)
# sum_loss= tf.reduce_sum(losses)
# mean_loss= tf.reduce_mean(losses)
# return losses, sum_loss, mean_loss
self.dropout_keep_embedd_prob= tf.placeholder(tf.float32, name= "dropout_embedding_probability")
self.dropout_keep_input_prob= tf.placeholder(tf.float32, name= 'dropout_input_probability')
self.dropout_keep_output_prob= tf.placeholder(tf.float32, name="dropout_output_probability")
# def embedding(hidden_size):
#
# w= tf.get_variable("W_embed", shape= [self.vocabulary_size, hidden_size],
# initializer= tf.random_uniform_initializer(-1.0, 1.0))
# embedded_tokens= tf.nn.embedding_lookup(w, self.input)
# embedded_tokens_drop= tf.nn.dropout(embedded_tokens, keep_prob=self.dropout_keep_embedd_prob)
# return embedded_tokens_drop
# with tf.variable_scope("rnn_input", reuse= True):
# rnn_input= [embedding(hidden_size)[:, i, :] for i in range(self.max_sequence_length)]
with tf.variable_scope("embedding"), tf.device("/cpu:0"):
W = tf.get_variable("W",
[self.vocabulary_size, hidden_size],
initializer=tf.random_uniform_initializer(-1.0, 1.0))
embedded_tokens = tf.nn.embedding_lookup(W, self.input)
embedded_tokens_drop = tf.nn.dropout(embedded_tokens, self.dropout_keep_embedd_prob)
rnn_input = [embedded_tokens_drop[:, i, :] for i in range(self.max_sequence_length)]
def single_cell():
return rnn.DropoutWrapper(rnn.LSTMCell(hidden_size, initializer= tf.random_uniform_initializer(-1.0, 1.0),
state_is_tuple= True), input_keep_prob=self.dropout_keep_input_prob ,
output_keep_prob=self.dropout_keep_output_prob )
cell= single_cell()
if num_layers>1:
cell = rnn.MultiRNNCell([single_cell() for _ in range(num_layers)], state_is_tuple= True)
initial_state = cell.zero_state(self.batch_size, tf.float32)
rnn_output, rnn_state= rnn.static_rnn(cell= cell, inputs=rnn_input ,
initial_state= initial_state,
sequence_length= self.input_length)
w_t= tf.get_variable("w_layer", [hidden_size, self.num_classes],
initializer= tf.truncated_normal_initializer(stddev=0.1))
b_t= tf.get_variable("b_layer", [self.num_classes], initializer= tf.constant_initializer(0.1))
self.scores= tf.nn.xw_plus_b(rnn_state[-1][0], w_t, b_t)
self.y= tf.nn.softmax(self.scores)
self.prediction= tf.argmax(self.scores, 1)
self.losses = tf.nn.softmax_cross_entropy_with_logits(logits=self.scores,
labels=self.output,
name="ce_losses")
self.sum_loss = tf.reduce_sum(self.losses)
self.mean_loss = tf.reduce_mean(self.losses)
#self.losses, sum_loss, mean_loss= loss(self.scores, self.output)
self.correct_predictions = tf.equal(self.prediction, tf.argmax(self.output, 1))
self.accuracy = tf.reduce_mean(tf.cast(self.correct_predictions, "float"), name="accuracy")
params= tf.trainable_variables()
if not forward_only:
self.gradient_norms = []
self.updates = []
opt= tf.train.AdamOptimizer(self.learning_rate)
gradients= tf.gradients(self.losses, params)
clipped_gradients, norm = tf.clip_by_global_norm(gradients, self.gradient_clip)
self.gradient_norms.append(norm)
self.updates.append(opt.apply_gradients(zip(clipped_gradients, params),
global_step=self.global_step))
self.saver= tf.train.Saver(tf.global_variables())
def step(self, session, inputs, outputs, input_length, forward_only= False):
input_feed={}
input_feed[self.input.name]= inputs
input_feed[self.output.name]= outputs
input_feed[self.input_length.name]= input_length
input_feed[self.dropout_keep_embedd_prob.name]= self.dropout
input_feed[self.dropout_keep_input_prob.name]=self.dropout
input_feed[self.dropout_keep_output_prob.name]=self.dropout
if not forward_only:
output_feed= [self.updates, # Update Op that does SGD.
self.gradient_norms, # Gradient norm.
self.mean_loss]
else:
output_feed= [self.mean_loss, self.y, self.accuracy]
outputs= session.run(output_feed, input_feed)
if not forward_only:
return outputs[1], outputs[2], None
else:
return outputs[0], outputs[1], outputs[2]
def current_step_record(self, train=False, test=False):
if train:
self.train_batch_no=0
if test:
self.test_batch_no=0
def get_batch(self, train_data, test_data= None):
num_classes=2
if not test_data:
train_targets = (train_data.transpose()[-1]).transpose()
train_ohe= np.eye(num_classes)[train_targets]
train_input_size= (train_data.transpose()[-2]).transpose()
train_input= (train_data.transpose()[0:-2]).transpose()
train_num_batches= len(train_input)/self.batch_size
input_batch= train_input[self.train_batch_no:self.train_batch_no+200]
output_batch= train_ohe[self.train_batch_no: self.train_batch_no+200]
input_size_batch= train_input_size[self.train_batch_no:self.train_batch_no+200]
self.train_batch_no+=200
self.train_batch_no = self.train_batch_no % len(train_data)
return input_batch, output_batch, input_size_batch
else:
test_targets= (test_data.transpose()[-1]).transpose()
test_ohe= np.eye(num_classes)[test_targets]
test_input_size= (test_data.transpose()[-2]).transpose()
test_input= (train_data.transpose()[0:-2]).transpose()
test_num_batches= len(test_input)/self.batch_size
test_input= test_input[:len(test_input)-(len(test_input)%self.batch_size)]
test_input_batch= test_input[self.test_batch_no:self.test_batch_no+200]
test_output_batch= test_ohe[self.test_batch_no: self.test_batch_no+200]
test_input_size_batch= test_input_size[self.test_batch_no:self.test_batch_no+200]
self.test_batch_no+=200
return test_input_batch, test_output_batch, test_input_size_batch