-
Notifications
You must be signed in to change notification settings - Fork 46
/
data_iterator.py
206 lines (176 loc) · 6.65 KB
/
data_iterator.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
import nltk
import numpy
import os
import random
import cPickle
import gzip
import codecs
from tempfile import mkstemp
def fopen(filename, mode='r'):
if filename.endswith('.gz'):
return gzip.open(filename, mode)
return open(filename, mode)
class TextIterator:
"""Simple Bitext iterator."""
def __init__(self,
source, source_dict,
target=None, target_dict=None,
source_word_level=0,
target_word_level=0,
batch_size=128,
job_id=0,
sort_size=20,
n_words_source=-1,
n_words_target=-1,
shuffle_per_epoch=False):
self.source_file = source
self.target_file = target
self.source = fopen(source, 'r')
with open(source_dict, 'rb') as f:
self.source_dict = cPickle.load(f)
if target is not None:
self.target = fopen(target, 'r')
if target_dict is not None:
with open(target_dict, 'rb') as f:
self.target_dict = cPickle.load(f)
else:
self.target = None
self.source_word_level = source_word_level
self.target_word_level = target_word_level
self.batch_size = batch_size
self.n_words_source = n_words_source
self.n_words_target = n_words_target
self.shuffle_per_epoch = shuffle_per_epoch
self.source_buffer = []
self.target_buffer = []
self.k = batch_size * sort_size
self.end_of_data = False
self.job_id = job_id
def __iter__(self):
return self
def reset(self):
if self.shuffle_per_epoch:
# close current files
self.source.close()
if self.target is None:
self.shuffle([self.source_file])
self.source = fopen(self.source_file + '.reshuf_%d' % self.job_id, 'r')
else:
self.target.close()
# shuffle *original* source files,
self.shuffle([self.source_file, self.target_file])
# open newly 're-shuffled' file as input
self.source = fopen(self.source_file + '.reshuf_%d' % self.job_id, 'r')
self.target = fopen(self.target_file + '.reshuf_%d' % self.job_id, 'r')
else:
self.source.seek(0)
if self.target is not None:
self.target.seek(0)
@staticmethod
def shuffle(files):
tf_os, tpath = mkstemp()
tf = open(tpath, 'w')
fds = [open(ff) for ff in files]
for l in fds[0]:
lines = [l.strip()] + [ff.readline().strip() for ff in fds[1:]]
print >>tf, "|||".join(lines)
[ff.close() for ff in fds]
tf.close()
tf = open(tpath, 'r')
lines = tf.readlines()
random.shuffle(lines)
fds = [open(ff+'.reshuf','w') for ff in files]
for l in lines:
s = l.strip().split('|||')
for ii, fd in enumerate(fds):
print >>fd, s[ii]
[ff.close() for ff in fds]
os.remove(tpath)
return
def next(self):
if self.end_of_data:
self.end_of_data = False
self.reset()
raise StopIteration
source = []
target = []
# fill buffer, if it's empty
if self.target is not None:
assert len(self.source_buffer) == len(self.target_buffer), 'Buffer size mismatch!'
if len(self.source_buffer) == 0:
for k_ in xrange(self.k):
ss = self.source.readline()
if ss == "":
break
if self.source_word_level:
ss = ss.strip().split()
else:
ss = ss.strip()
ss = list(ss.decode('utf8'))
self.source_buffer.append(ss)
if self.target is not None:
tt = self.target.readline()
if tt == "":
break
if self.target_word_level:
tt = tt.strip().split()
else:
tt = tt.strip()
tt = list(tt.decode('utf8'))
self.target_buffer.append(tt)
if self.target is not None:
# sort by target buffer
tlen = numpy.array([len(t) for t in self.target_buffer])
tidx = tlen.argsort()
_sbuf = [self.source_buffer[i] for i in tidx]
_tbuf = [self.target_buffer[i] for i in tidx]
self.target_buffer = _tbuf
else:
slen = numpy.array([len(s) for s in self.source_buffer])
sidx = slen.argsort()
_sbuf = [self.source_buffer[i] for i in sidx]
self.source_buffer = _sbuf
if self.target is not None:
if len(self.source_buffer) == 0 or len(self.target_buffer) == 0:
self.end_of_data = False
self.reset()
raise StopIteration
elif len(self.source_buffer) == 0:
self.end_of_data = False
self.reset()
raise StopIteration
try:
# actual work here
while True:
# read from source file and map to word index
try:
ss_ = self.source_buffer.pop()
except IndexError:
break
ss = [self.source_dict[w] if w in self.source_dict else 1 for w in ss_]
if self.n_words_source > 0:
ss = [w if w < self.n_words_source else 1 for w in ss]
source.append(ss)
if self.target is not None:
# read from target file and map to word index
tt_ = self.target_buffer.pop()
tt = [self.target_dict[w] if w in self.target_dict else 1 for w in tt_]
if self.n_words_target > 0:
tt = [w if w < self.n_words_target else 1 for w in tt]
target.append(tt)
if len(source) >= self.batch_size:
break
except IOError:
self.end_of_data = True
if self.target is not None:
if len(source) <= 0 or len(target) <= 0:
self.end_of_data = False
self.reset()
raise StopIteration
return source, target
else:
if len(source) <= 0:
self.end_of_data = False
self.reset()
raise StopIteration
return source