-
Notifications
You must be signed in to change notification settings - Fork 152
/
SentenceMatchDataStream.py
180 lines (158 loc) · 8.86 KB
/
SentenceMatchDataStream.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
import numpy as np
import re
def make_batches(size, batch_size):
nb_batch = int(np.ceil(size/float(batch_size)))
return [(i*batch_size, min(size, (i+1)*batch_size)) for i in range(0, nb_batch)] # zgwang: starting point of each batch
def pad_2d_vals(in_vals, dim1_size, dim2_size, dtype=np.int32):
out_val = np.zeros((dim1_size, dim2_size), dtype=dtype)
if dim1_size > len(in_vals): dim1_size = len(in_vals)
for i in xrange(dim1_size):
cur_in_vals = in_vals[i]
cur_dim2_size = dim2_size
if cur_dim2_size > len(cur_in_vals): cur_dim2_size = len(cur_in_vals)
out_val[i,:cur_dim2_size] = cur_in_vals[:cur_dim2_size]
return out_val
def pad_3d_vals(in_vals, dim1_size, dim2_size, dim3_size, dtype=np.int32):
out_val = np.zeros((dim1_size, dim2_size, dim3_size), dtype=dtype)
if dim1_size > len(in_vals): dim1_size = len(in_vals)
for i in xrange(dim1_size):
in_vals_i = in_vals[i]
cur_dim2_size = dim2_size
if cur_dim2_size > len(in_vals_i): cur_dim2_size = len(in_vals_i)
for j in xrange(cur_dim2_size):
in_vals_ij = in_vals_i[j]
cur_dim3_size = dim3_size
if cur_dim3_size > len(in_vals_ij): cur_dim3_size = len(in_vals_ij)
out_val[i, j, :cur_dim3_size] = in_vals_ij[:cur_dim3_size]
return out_val
def read_all_instances(inpath, word_vocab=None, label_vocab=None, char_vocab=None, max_sent_length=100,
max_char_per_word=10, isLower=True):
instances = []
infile = open(inpath, 'rt')
idx = -1
for line in infile:
idx += 1
line = line.decode('utf-8').strip()
if line.startswith('-'): continue
items = re.split("\t", line)
label = items[0]
sentence1 = items[1].strip()
sentence2 = items[2].strip()
cur_ID = "{}".format(idx)
if len(items)>=4: cur_ID = items[3]
if isLower:
sentence1 = sentence1.lower()
sentence2 = sentence2.lower()
if label_vocab is not None:
label_id = label_vocab.getIndex(label)
if label_id >= label_vocab.vocab_size: label_id = 0
else:
label_id = int(label)
word_idx_1 = word_vocab.to_index_sequence(sentence1)
word_idx_2 = word_vocab.to_index_sequence(sentence2)
if char_vocab is not None:
char_matrix_idx_1 = char_vocab.to_character_matrix(sentence1, max_char_per_word=max_char_per_word)
char_matrix_idx_2 = char_vocab.to_character_matrix(sentence2, max_char_per_word=max_char_per_word)
else:
char_matrix_idx_1 = None
char_matrix_idx_2 = None
if len(word_idx_1) > max_sent_length:
word_idx_1 = word_idx_1[:max_sent_length]
if char_vocab is not None: char_matrix_idx_1 = char_matrix_idx_1[:max_sent_length]
if len(word_idx_2) > max_sent_length:
word_idx_2 = word_idx_2[:max_sent_length]
if char_vocab is not None: char_matrix_idx_2 = char_matrix_idx_2[:max_sent_length]
instances.append((label, sentence1, sentence2, label_id, word_idx_1, word_idx_2, char_matrix_idx_1, char_matrix_idx_2, cur_ID))
infile.close()
return instances
class SentenceMatchDataStream(object):
def __init__(self, inpath, word_vocab=None, char_vocab=None, label_vocab=None,
isShuffle=False, isLoop=False, isSort=True, options=None):
instances = read_all_instances(inpath, word_vocab=word_vocab, label_vocab=label_vocab,
char_vocab=char_vocab, max_sent_length=options.max_sent_length, max_char_per_word=options.max_char_per_word,
isLower=options.isLower)
# sort instances based on sentence length
if isSort: instances = sorted(instances, key=lambda instance: (len(instance[4]), len(instance[5]))) # sort instances based on length
self.num_instances = len(instances)
# distribute into different buckets
batch_spans = make_batches(self.num_instances, options.batch_size)
self.batches = []
for batch_index, (batch_start, batch_end) in enumerate(batch_spans):
cur_instances = []
for i in xrange(batch_start, batch_end):
cur_instances.append(instances[i])
cur_batch = InstanceBatch(cur_instances, with_char=options.with_char)
self.batches.append(cur_batch)
instances = None
self.num_batch = len(self.batches)
self.index_array = np.arange(self.num_batch)
self.isShuffle = isShuffle
if self.isShuffle: np.random.shuffle(self.index_array)
self.isLoop = isLoop
self.cur_pointer = 0
def nextBatch(self):
if self.cur_pointer>=self.num_batch:
if not self.isLoop: return None
self.cur_pointer = 0
if self.isShuffle: np.random.shuffle(self.index_array)
# print('{} '.format(self.index_array[self.cur_pointer]))
cur_batch = self.batches[self.index_array[self.cur_pointer]]
self.cur_pointer += 1
return cur_batch
def shuffle(self):
if self.isShuffle: np.random.shuffle(self.index_array)
def reset(self):
self.cur_pointer = 0
def get_num_batch(self):
return self.num_batch
def get_num_instance(self):
return self.num_instances
def get_batch(self, i):
if i >= self.num_batch: return None
return self.batches[self.index_array[i]]
class InstanceBatch(object):
def __init__(self, instances, with_char=False):
self.instances = instances
self.batch_size = len(instances)
self.question_len = 0
self.passage_len = 0
self.question_lengths = [] # tf.placeholder(tf.int32, [None])
self.in_question_words = [] # tf.placeholder(tf.int32, [None, None]) # [batch_size, question_len]
self.passage_lengths = [] # tf.placeholder(tf.int32, [None])
self.in_passage_words = [] # tf.placeholder(tf.int32, [None, None]) # [batch_size, passage_len]
self.label_truth = [] # [batch_size]
if with_char:
self.in_question_chars = [] # tf.placeholder(tf.int32, [None, None, None]) # [batch_size, question_len, q_char_len]
self.question_char_lengths = [] # tf.placeholder(tf.int32, [None, None]) # [batch_size, question_len]
self.in_passage_chars = [] # tf.placeholder(tf.int32, [None, None, None]) # [batch_size, passage_len, p_char_len]
self.passage_char_lengths = [] # tf.placeholder(tf.int32, [None, None]) # [batch_size, passage_len]
for (label, sentence1, sentence2, label_id, word_idx_1, word_idx_2, char_matrix_idx_1, char_matrix_idx_2, cur_ID) in instances:
cur_question_length = len(word_idx_1)
cur_passage_length = len(word_idx_2)
if self.question_len < cur_question_length: self.question_len = cur_question_length
if self.passage_len < cur_passage_length: self.passage_len = cur_passage_length
self.question_lengths.append(cur_question_length)
self.in_question_words.append(word_idx_1)
self.passage_lengths.append(cur_passage_length)
self.in_passage_words.append(word_idx_2)
self.label_truth.append(label_id)
if with_char:
self.in_question_chars.append(char_matrix_idx_1)
self.in_passage_chars.append(char_matrix_idx_2)
self.question_char_lengths.append([len(cur_char_idx) for cur_char_idx in char_matrix_idx_1])
self.passage_char_lengths.append([len(cur_char_idx) for cur_char_idx in char_matrix_idx_2])
# padding all value into np arrays
self.question_lengths = np.array(self.question_lengths, dtype=np.int32)
self.in_question_words = pad_2d_vals(self.in_question_words, self.batch_size, self.question_len, dtype=np.int32)
self.passage_lengths = np.array(self.passage_lengths, dtype=np.int32)
self.in_passage_words = pad_2d_vals(self.in_passage_words, self.batch_size, self.passage_len, dtype=np.int32)
self.label_truth = np.array(self.label_truth, dtype=np.int32)
if with_char:
max_char_length1 = np.max([np.max(aa) for aa in self.question_char_lengths])
self.in_question_chars = pad_3d_vals(self.in_question_chars, self.batch_size, self.question_len,
max_char_length1, dtype=np.int32)
max_char_length2 = np.max([np.max(aa) for aa in self.passage_char_lengths])
self.in_passage_chars = pad_3d_vals(self.in_passage_chars, self.batch_size, self.passage_len,
max_char_length2, dtype=np.int32)
self.question_char_lengths = pad_2d_vals(self.question_char_lengths, self.batch_size, self.question_len)
self.passage_char_lengths = pad_2d_vals(self.passage_char_lengths, self.batch_size, self.passage_len)