forked from StevenWD/HR-BiLSTM
-
Notifications
You must be signed in to change notification settings - Fork 0
/
eval.py
140 lines (120 loc) · 5.82 KB
/
eval.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
# coding: utf-8
import json
import numpy as np
from keras.layers import Input, Embedding, LSTM, Dense, Bidirectional, Add, MaxPooling1D, Concatenate, Dot, Flatten
from keras.models import Model
from keras.preprocessing import sequence
from keras import backend as K
import os
import tensorflow as tf
from configparser import ConfigParser
from keras.optimizers import Adam
from keras import losses
from sklearn.utils.class_weight import compute_class_weight
from preprocess import readData
from preprocess import readRelation
def ranking_loss(y_true, y_pred):
return K.maximum(0.0, 0.1 + K.sum(y_pred*y_true,axis=-1))
def model_construct():
# CONFIG
config = ConfigParser()
config.read('./config.ini')
question_input = Input(shape=(config.getint('pre', 'question_maximum_length'), ), dtype='int32',name="question_input")
relation_all_input = Input(shape=(config.getint('pre', 'relation_word_maximum_length'), ), dtype='int32',name="relation_all_input")
relation_input = Input(shape=(config.getint('pre', 'relation_maximum_length'), ), dtype='int32',name="relation_input")
question_emd = np.load('/home/stevenwd/HR-BiLSTM/question_emd_matrix.npy')
relation_emd = np.load('/home/stevenwd/HR-BiLSTM/relation_emd_matrix.npy')
relation_all_emd = np.load('/home/stevenwd/HR-BiLSTM/relation_all_emd_matrix.npy')
question_emd = Embedding(question_emd.shape[0],
config.getint('pre', 'word_emd_length'),
weights=[question_emd],
input_length=config.getint('pre', 'question_maximum_length'),
trainable=False,name="question_emd")(question_input)
sharedEmbd_r_w = Embedding(relation_all_emd.shape[0],
config.getint('pre', 'word_emd_length'),
weights=[relation_all_emd],
input_length=config.getint('pre', 'relation_word_maximum_length'),
trainable=False,name="sharedEmbd_r_w")
relation_word_emd = sharedEmbd_r_w(relation_all_input)
sharedEmbd_r = Embedding(relation_emd.shape[0],
config.getint('pre', 'word_emd_length'),
weights=[relation_emd],
input_length=config.getint('pre', 'relation_maximum_length'),
trainable=True,name="sharedEmbd_r")
relation_emd = sharedEmbd_r(relation_input)
bilstem_layer = Bidirectional(LSTM(units=40, return_sequences=True, implementation=2),name="bilstem_layer")
question_bilstm_1 = bilstem_layer(question_emd)
question_bilstm_2 = Bidirectional(LSTM(units=40, return_sequences=True, implementation=2),name="question_bilstm_2")(question_bilstm_1)
relation_word_bilstm = bilstem_layer(relation_word_emd)
relation_bilstm = bilstem_layer(relation_emd)
question_res = Add()([question_bilstm_1, question_bilstm_2])
question_maxpool = MaxPooling1D(80, padding='same')(question_res)
question_flatten = Flatten()(question_maxpool)
relation_maxpool = MaxPooling1D(80, padding='same')(relation_bilstm)
relation_word_maxpool = MaxPooling1D(80, padding='same')(relation_word_bilstm)
relation_res = Add()([relation_maxpool, relation_word_maxpool])
relation_flatten = Flatten()(relation_res)
result = Dot(axes=-1, normalize=True)([question_flatten, relation_flatten])
model = Model(inputs=[question_input, relation_input, relation_all_input,], outputs=result)
model.compile(optimizer=Adam(), loss=ranking_loss)
return model
if __name__ == '__main__':
# GPU settings
os.environ["CUDA_VISIBLE_DEVICES"] = "6"
neg_num = json.load(open('./neg_number.json', 'r'))
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
sess = tf.Session(config=config)
K.set_session(sess)
model = model_construct()
model.load_weights('./my_model_weights_06_2048.h5')
print(model.summary())
question_feature = np.load('./test_question_feature.npy')
relation_feature = np.load('./test_relation_feature.npy')
relation_all_feature = np.load('./test_relation_all_feature.npy')
print('positive data loaded...')
simi_pos = model.predict([question_feature, relation_feature, relation_all_feature], batch_size=1024)
print('positive similarity computed...')
np.save('test_pre_pos.npy', simi_pos)
relation_feature_neg = np.load('./test_relation_feature_neg.npy')
relation_all_feature_neg = np.load('./test_relation_all_feature_neg.npy')
print('negtive data loaded...')
simi_neg = model.predict([question_feature, relation_feature_neg, relation_all_feature_neg], batch_size=1024)
print('negtive similarity computed...')
np.save('test_pre_neg.npy', simi_neg)
acc = np.sum(simi_pos>simi_neg) / simi_pos.shape[0]
print("relation accurcy: " + str(acc))
index = 0
false_list = list()
true_list = list()
all_set = set()
config = ConfigParser()
config.read('./config.ini')
data = readData(config.get('pre', 'test_filepath'))
relation = readRelation(config.get('pre', 'relation_filepath'))
for num,neg_index in neg_num:
if np.sum(simi_pos[index: index+num]-simi_neg[index: index+num]<0) > 0:
false_list.append(neg_index)
print (simi_pos[index])
print (np.max(simi_neg[index: index+num]))
print (len(simi_neg[index: index+num]))
print (np.argmax(simi_neg[index: index+num]))
print (simi_neg[index: index+num][np.argmax(simi_neg[index: index+num])])
print (neg_index)
print ("")
pass
else:
true_list.append(neg_index)
index += num
all_set.add(neg_index)
print (max(true_list))
true_list = set([i for i in true_list if i in all_set and i not in false_list])
print (len(all_set))
print (len(true_list))
for i in all_set:
if i not in true_list:
print (i)
print (data[0])
print (relation[0][1])
print (relation[1][1])
print('sentence accurcy: '+str(len(true_list)/len(all_set)))