-
Notifications
You must be signed in to change notification settings - Fork 41
/
pair.py
197 lines (163 loc) · 9.82 KB
/
pair.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
# encoding: utf-8
# @author: zxding
# email: d.z.x@qq.com
import numpy as np
import tensorflow as tf
from sklearn.model_selection import KFold
import sys, os, time, codecs, pdb
from utils.tf_funcs import *
from utils.prepare_data import *
FLAGS = tf.app.flags.FLAGS
# >>>>>>>>>>>>>>>>>>>> For Model <<<<<<<<<<<<<<<<<<<< #
## embedding parameters ##
tf.app.flags.DEFINE_string('w2v_file', '../data/w2v_200.txt', 'embedding file')
tf.app.flags.DEFINE_integer('embedding_dim', 200, 'dimension of word embedding')
tf.app.flags.DEFINE_integer('embedding_dim_pos', 50, 'dimension of position embedding')
## input struct ##
tf.app.flags.DEFINE_integer('max_sen_len', 30, 'max number of tokens per sentence')
## model struct ##
tf.app.flags.DEFINE_integer('n_hidden', 100, 'number of hidden unit')
tf.app.flags.DEFINE_integer('n_class', 2, 'number of distinct class')
# >>>>>>>>>>>>>>>>>>>> For Data <<<<<<<<<<<<<<<<<<<< #
tf.app.flags.DEFINE_string('log_file_name', '', 'name of log file')
# >>>>>>>>>>>>>>>>>>>> For Training <<<<<<<<<<<<<<<<<<<< #
tf.app.flags.DEFINE_integer('training_iter', 10, 'number of train iter')
tf.app.flags.DEFINE_string('scope', 'P_cause', 'RNN scope')
# not easy to tune , a good posture of using data to train model is very important
tf.app.flags.DEFINE_integer('batch_size', 32, 'number of example per batch')
tf.app.flags.DEFINE_float('learning_rate', 0.005, 'learning rate')
tf.app.flags.DEFINE_float('keep_prob1', 0.5, 'word embedding training dropout keep prob')
tf.app.flags.DEFINE_float('keep_prob2', 1.0, 'softmax layer dropout keep prob')
tf.app.flags.DEFINE_float('l2_reg', 0.00001, 'l2 regularization')
def build_model(word_embedding, pos_embedding, x, sen_len, keep_prob1, keep_prob2, distance, y, RNN = biLSTM):
x = tf.nn.embedding_lookup(word_embedding, x)
inputs = tf.reshape(x, [-1, FLAGS.max_sen_len, FLAGS.embedding_dim])
inputs = tf.nn.dropout(inputs, keep_prob=keep_prob1)
sen_len = tf.reshape(sen_len, [-1])
def get_s(inputs, name):
with tf.name_scope('word_encode'):
inputs = RNN(inputs, sen_len, n_hidden=FLAGS.n_hidden, scope=FLAGS.scope+'word_layer' + name)
with tf.name_scope('word_attention'):
sh2 = 2 * FLAGS.n_hidden
w1 = get_weight_varible('word_att_w1' + name, [sh2, sh2])
b1 = get_weight_varible('word_att_b1' + name, [sh2])
w2 = get_weight_varible('word_att_w2' + name, [sh2, 1])
s = att_var(inputs,sen_len,w1,b1,w2)
s = tf.reshape(s, [-1, 2 * 2 * FLAGS.n_hidden])
return s
s = get_s(inputs, name='cause_word_encode')
dis = tf.nn.embedding_lookup(pos_embedding, distance)
s = tf.concat([s, dis], 1)
s1 = tf.nn.dropout(s, keep_prob=keep_prob2)
w_pair = get_weight_varible('softmax_w_pair', [4 * FLAGS.n_hidden + FLAGS.embedding_dim_pos, FLAGS.n_class])
b_pair = get_weight_varible('softmax_b_pair', [FLAGS.n_class])
pred_pair = tf.nn.softmax(tf.matmul(s1, w_pair) + b_pair)
reg = tf.nn.l2_loss(w_pair) + tf.nn.l2_loss(b_pair)
return pred_pair, reg
def print_training_info():
print('\n\n>>>>>>>>>>>>>>>>>>>>TRAINING INFO:\n')
print('batch-{}, lr-{}, kb1-{}, kb2-{}, l2_reg-{}'.format(
FLAGS.batch_size, FLAGS.learning_rate, FLAGS.keep_prob1, FLAGS.keep_prob2, FLAGS.l2_reg))
print('training_iter-{}, scope-{}\n'.format(FLAGS.training_iter, FLAGS.scope))
def get_batch_data(x, sen_len, keep_prob1, keep_prob2, distance, y, batch_size, test=False):
for index in batch_index(len(y), batch_size, test):
feed_list = [x[index], sen_len[index], keep_prob1, keep_prob2, distance[index], y[index]]
yield feed_list, len(index)
def run():
save_dir = 'pair_data/{}/'.format(FLAGS.scope)
if not os.path.exists(save_dir):
os.makedirs(save_dir)
if FLAGS.log_file_name:
sys.stdout = open(save_dir + FLAGS.log_file_name, 'w')
print_time()
tf.reset_default_graph()
# Model Code Block
word_idx_rev, word_id_mapping, word_embedding, pos_embedding = load_w2v(FLAGS.embedding_dim, FLAGS.embedding_dim_pos, 'data_combine/clause_keywords.csv', FLAGS.w2v_file)
word_embedding = tf.constant(word_embedding, dtype=tf.float32, name='word_embedding')
pos_embedding = tf.constant(pos_embedding, dtype=tf.float32, name='pos_embedding')
print('build model...')
x = tf.placeholder(tf.int32, [None, 2, FLAGS.max_sen_len])
sen_len = tf.placeholder(tf.int32, [None, 2])
keep_prob1 = tf.placeholder(tf.float32)
keep_prob2 = tf.placeholder(tf.float32)
distance = tf.placeholder(tf.int32, [None])
y = tf.placeholder(tf.float32, [None, FLAGS.n_class])
placeholders = [x, sen_len, keep_prob1, keep_prob2, distance, y]
pred_pair, reg = build_model(word_embedding, pos_embedding, x, sen_len, keep_prob1, keep_prob2, distance, y)
loss_op = - tf.reduce_mean(y * tf.log(pred_pair)) + reg * FLAGS.l2_reg
optimizer = tf.train.AdamOptimizer(learning_rate=FLAGS.learning_rate).minimize(loss_op)
true_y_op = tf.argmax(y, 1)
pred_y_op = tf.argmax(pred_pair, 1)
acc_op = tf.reduce_mean(tf.cast(tf.equal(true_y_op, pred_y_op), tf.float32))
print('build model done!\n')
# Training Code Block
print_training_info()
tf_config = tf.ConfigProto()
tf_config.gpu_options.allow_growth = True
with tf.Session(config=tf_config) as sess:
keep_rate_list, acc_subtask_list, p_pair_list, r_pair_list, f1_pair_list = [], [], [], [], []
o_p_pair_list, o_r_pair_list, o_f1_pair_list = [], [], []
for fold in range(1,11):
sess.run(tf.global_variables_initializer())
# train for one fold
print('############# fold {} begin ###############'.format(fold))
# Data Code Block
train_file_name = 'fold{}_train.txt'.format(fold, FLAGS)
test_file_name = 'fold{}_test.txt'.format(fold)
tr_pair_id_all, tr_pair_id, tr_y, tr_x, tr_sen_len, tr_distance = load_data_2nd_step(save_dir + train_file_name, word_id_mapping, max_sen_len = FLAGS.max_sen_len)
te_pair_id_all, te_pair_id, te_y, te_x, te_sen_len, te_distance = load_data_2nd_step(save_dir + test_file_name, word_id_mapping, max_sen_len = FLAGS.max_sen_len)
max_acc_subtask, max_f1 = [-1.]*2
print('train docs: {} test docs: {}'.format(len(tr_x), len(te_x)))
for i in xrange(FLAGS.training_iter):
start_time, step = time.time(), 1
# train
for train, _ in get_batch_data(tr_x, tr_sen_len, FLAGS.keep_prob1, FLAGS.keep_prob2, tr_distance, tr_y, FLAGS.batch_size):
_, loss, pred_y, true_y, acc = sess.run(
[optimizer, loss_op, pred_y_op, true_y_op, acc_op], feed_dict=dict(zip(placeholders, train)))
print('step {}: train loss {:.4f} acc {:.4f}'.format(step, loss, acc))
step = step + 1
# test
test = [te_x, te_sen_len, 1., 1., te_distance, te_y]
loss, pred_y, true_y, acc = sess.run([loss_op, pred_y_op, true_y_op, acc_op], feed_dict=dict(zip(placeholders, test)))
print('\nepoch {}: test loss {:.4f}, acc {:.4f}, cost time: {:.1f}s\n'.format(i, loss, acc, time.time()-start_time))
if acc > max_acc_subtask:
max_acc_subtask = acc
print('max_acc_subtask: {:.4f} \n'.format(max_acc_subtask))
# p, r, f1, o_p, o_r, o_f1, keep_rate = prf_2nd_step(te_pair_id_all, te_pair_id, pred_y, fold, save_dir)
p, r, f1, o_p, o_r, o_f1, keep_rate = prf_2nd_step(te_pair_id_all, te_pair_id, pred_y)
if f1 > max_f1:
max_keep_rate, max_p, max_r, max_f1 = keep_rate, p, r, f1
print('original o_p {:.4f} o_r {:.4f} o_f1 {:.4f}'.format(o_p, o_r, o_f1))
print 'pair filter keep rate: {}'.format(keep_rate)
print('test p {:.4f} r {:.4f} f1 {:.4f}'.format(p, r, f1))
print('max_p {:.4f} max_r {:.4f} max_f1 {:.4f}\n'.format(max_p, max_r, max_f1))
print 'Optimization Finished!\n'
print('############# fold {} end ###############'.format(fold))
# fold += 1
acc_subtask_list.append(max_acc_subtask)
keep_rate_list.append(max_keep_rate)
p_pair_list.append(max_p)
r_pair_list.append(max_r)
f1_pair_list.append(max_f1)
o_p_pair_list.append(o_p)
o_r_pair_list.append(o_r)
o_f1_pair_list.append(o_f1)
print_training_info()
all_results = [acc_subtask_list, keep_rate_list, p_pair_list, r_pair_list, f1_pair_list, o_p_pair_list, o_r_pair_list, o_f1_pair_list]
acc_subtask, keep_rate, p_pair, r_pair, f1_pair, o_p_pair, o_r_pair, o_f1_pair = map(lambda x: np.array(x).mean(), all_results)
print('\nOriginal pair_predict: test f1 in 10 fold: {}'.format(np.array(o_f1_pair_list).reshape(-1,1)))
print('average : p {:.4f} r {:.4f} f1 {:.4f}\n'.format(o_p_pair, o_r_pair, o_f1_pair))
print('\nAverage keep_rate: {:.4f}\n'.format(keep_rate))
print('\nFiltered pair_predict: test f1 in 10 fold: {}'.format(np.array(f1_pair_list).reshape(-1,1)))
print('average : p {:.4f} r {:.4f} f1 {:.4f}\n'.format(p_pair, r_pair, f1_pair))
print_time()
def main(_):
# FLAGS.log_file_name = 'step2.log'
FLAGS.training_iter=20
for scope_name in ['Ind_BiLSTM', 'P_emotion', 'P_cause']:
FLAGS.scope= scope_name + '_1'
run()
FLAGS.scope= scope_name + '_2'
run()
if __name__ == '__main__':
tf.app.run()