Learning-Natural-Language-Inference-using-Bidirectional-LSTM-model-and-Inner-Attention/quora_batcher.py at master · cuteboydot/Learning-Natural-Language-Inference-using-Bidirectional-LSTM-model-and-Inner-Attention · GitHub

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
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

import numpy as np
import pandas as pd
import collections
import os
import json
import re
import collections
import datetime
import pickle
import random
from tqdm import trange
import time

class quora_data_batcher(object):
    def __init__(self, data_idx_list, data_idx_list_test, dic, seq_max):
        self.data_idx_list = data_idx_list
        self.data_idx_list_test = data_idx_list_test
        self.dic = dic
        self.voc_size = len(dic)
        self.seq_max = seq_max

        self.train_size = len(self.data_idx_list)
        self.test_size = len(self.data_idx_list_test)


    def get_train_batch_rand(self, size):
        np.random.seed(seed=int(time.time()))
        assert size <= len(self.data_idx_list)

        data_x1 = np.zeros((size, self.seq_max), dtype=np.int)
        data_x2 = np.zeros((size, self.seq_max), dtype=np.int)
        data_y = np.zeros(size, dtype=np.int)
        len_x1 = np.zeros(size, dtype=np.int)
        len_x2 = np.zeros(size, dtype=np.int)

        index = np.random.choice(range(len(self.data_idx_list)), size, replace=False)
        for a in range(len(index)):
            idx = index[a]

            s1 = self.data_idx_list[idx][0]
            s2 = self.data_idx_list[idx][1]
            label = self.data_idx_list[idx][2]

            x1 = s1 + [self.dic["<PAD>"]] * (self.seq_max - len(s1))
            x2 = s2 + [self.dic["<PAD>"]] * (self.seq_max - len(s2))
            y = label

            assert len(x1) == self.seq_max
            assert max(x1) < self.voc_size
            assert len(x2) == self.seq_max
            assert max(x2) < self.voc_size
            assert 0 <= y <= 1

            data_x1[a] = x1
            data_x2[a] = x2
            data_y[a] = y
            len_x1[a] = len(s1)
            len_x2[a] = len(s2)
        return data_x1, data_x2, data_y, len_x1, len_x2


    def get_test_batch_rand(self, size):
        np.random.seed(seed=int(time.time()))
        assert size <= len(self.data_idx_list_test)

        data_x1 = np.zeros((size, self.seq_max), dtype=np.int)
        data_x2 = np.zeros((size, self.seq_max), dtype=np.int)
        data_y = np.zeros(size, dtype=np.int)
        len_x1 = np.zeros(size, dtype=np.int)
        len_x2 = np.zeros(size, dtype=np.int)

        index = np.random.choice(range(len(self.data_idx_list_test)), size, replace=False)
        for a in range(len(index)):
            idx = index[a]

            s1 = self.data_idx_list_test[idx][0]
            s2 = self.data_idx_list_test[idx][1]
            label = self.data_idx_list_test[idx][2]

            x1 = s1 + [self.dic["<PAD>"]] * (self.seq_max - len(s1))
            x2 = s2 + [self.dic["<PAD>"]] * (self.seq_max - len(s2))
            y = label

            assert len(x1) == self.seq_max
            assert max(x1) < self.voc_size
            assert len(x2) == self.seq_max
            assert max(x2) < self.voc_size
            assert 0 <= y <= 1

            data_x1[a] = x1
            data_x2[a] = x2
            data_y[a] = y
            len_x1[a] = len(s1)
            len_x2[a] = len(s2)
        return data_x1, data_x2, data_y, len_x1, len_x2


    def get_train_batch_step(self, start, size):
        assert start+size <= len(self.data_idx_list)

        data_x1 = np.zeros((size, self.seq_max), dtype=np.int)
        data_x2 = np.zeros((size, self.seq_max), dtype=np.int)
        data_y = np.zeros(size, dtype=np.int)
        len_x1 = np.zeros(size, dtype=np.int)
        len_x2 = np.zeros(size, dtype=np.int)

        for a in range(size):
            s1 = self.data_idx_list[start+a][0]
            s2 = self.data_idx_list[start+a][1]
            label = self.data_idx_list[start+a][2]

            x1 = s1 + [self.dic["<PAD>"]] * (self.seq_max - len(s1))
            x2 = s2 + [self.dic["<PAD>"]] * (self.seq_max - len(s2))
            y = ans

            assert len(x1) == self.seq_max
            assert max(x1) < self.voc_size
            assert len(x2) == self.seq_max
            assert max(x2) < self.voc_size
            assert 0 <= y <= 1

            data_x1[a] = x1
            data_x2[a] = x2
            data_y[a] = y
            len_x1[a] = len(s1)
            len_x2[a] = len(s2)
        return data_x1, data_x2, data_y, len_x1, len_x2


    def get_test_batch_step(self, start, size):
        assert start+size <= len(self.data_idx_list_test)

        data_x1 = np.zeros((size, self.seq_max), dtype=np.int)
        data_x2 = np.zeros((size, self.seq_max), dtype=np.int)
        data_y = np.zeros(size, dtype=np.int)
        len_x1 = np.zeros(size, dtype=np.int)
        len_x2 = np.zeros(size, dtype=np.int)

        for a in range(size):
            s1 = self.data_idx_list_test[start+a][0]
            s2 = self.data_idx_list_test[start+a][1]
            label = self.data_idx_list_test[start+a][2]

            x1 = s1 + [self.dic["<PAD>"]] * (self.seq_max - len(s1))
            x2 = s2 + [self.dic["<PAD>"]] * (self.seq_max - len(s2))
            y = label

            assert len(x1) == self.seq_max
            assert max(x1) < self.voc_size
            assert len(x2) == self.seq_max
            assert max(x2) < self.voc_size
            assert 0 <= y <= 1

            data_x1[a] = x1
            data_x2[a] = x2
            data_y[a] = y
            len_x1[a] = len(s1)
            len_x2[a] = len(s2)
        return data_x1, data_x2, data_y, len_x1, len_x2