-
Notifications
You must be signed in to change notification settings - Fork 1
/
build_tfrecords.py
187 lines (148 loc) · 5.92 KB
/
build_tfrecords.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
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
# File : build_tfrecords.py
# Author : Yan <yanwong@126.com>
# Date : 07.04.2020
# Last Modified Date: 26.04.2020
# Last Modified By : Yan <yanwong@126.com>
import os
import collections
import argparse
import logging
import json
import numpy as np
import tensorflow as tf
import special_words
logging.basicConfig(level=logging.INFO)
def _build_vocab(input_file, output_dir):
""" Load the vocab file created by word2vec to build the vocab dict.
Args:
input_file: The processed corpus. Each line contains a pair of word and tag.
Returns:
An ordered dict mapping each character to its Id.
"""
word_cnt = collections.Counter()
with tf.io.gfile.GFile(input_file, mode='r') as f:
for line in f:
line = line.strip()
if not line:
continue
word_cnt.update(line.split()[0])
sorted_items = word_cnt.most_common()
vocab = collections.OrderedDict()
vocab[special_words.PAD] = special_words.PAD_ID
vocab[special_words.UNK] = special_words.UNK_ID
for i, item in enumerate(sorted_items):
vocab[item[0]] = i + 2 # 0: PAD, 1: UNK
logging.info('Create vocab with %d words.', len(vocab))
vocab_file = os.path.join(output_dir, 'vocab.txt')
with tf.io.gfile.GFile(vocab_file, mode='w') as f:
f.write('\n'.join(vocab.keys()))
logging.info('Wrote vocab file to %s', vocab_file)
return vocab
def _int64_feature(value):
"""Returns an int64_list from a bool / enum / int / uint."""
return tf.train.Feature(
int64_list=tf.train.Int64List(value=[int(v) for v in value]))
def _sentence_to_ids(sent, vocab):
"""Helper for converting a sentence (list of words) to a list of ids."""
ids = [vocab.get(w, special_words.UNK_ID) for w in sent]
return ids
def _create_serialized_example(sent, tags, vocab):
"""Helper for creating a serialized Example proto."""
example = tf.train.Example(features=tf.train.Features(feature={
"sentence": _int64_feature(_sentence_to_ids(sent, vocab)),
"tags": _int64_feature(tags)
}))
return example.SerializeToString()
def _build_dataset(filename, vocab, tag_dict):
""" Build dataset from the file in Co-NLL NER task 2002 format
(check corpus/pku_training.txt).
Args:
filename: The file contains sentences of which each character has been tagged.
vocab: A dict mapping each character to Id.
tag_dict: A dict mapping each tag to Id.
Returns:
A list containing serialized examples.
"""
serialized = []
with tf.io.gfile.GFile(filename, 'r') as f:
sent = []
tags = []
# tag_dict = {'S': 0, 'B': 1, 'M': 2, 'E': 3}
for line in f:
line = line.strip()
if not line:
if sent and tags:
serialized.append(
_create_serialized_example(sent, tags, vocab))
sent = []
tags = []
else:
toks = line.split()
assert(len(toks) >= 2 and toks[1] in tag_dict)
sent.append(toks[0])
tags.append(tag_dict[toks[1]])
return serialized
def _write_shard(filename, dataset, indices):
"""Writes a TFRecord shard."""
with tf.io.TFRecordWriter(filename) as writer:
for j in indices:
writer.write(dataset[j])
def _write_dataset(name, dataset, indices, num_shards, output_dir):
"""Writes a sharded TFRecord dataset.
Args:
name: Name of the dataset (e.g. "train").
dataset: List of serialized Example protos.
indices: List of indices of 'dataset' to be written.
num_shards: The number of output shards.
"""
borders = np.int32(np.linspace(0, len(indices), num_shards + 1))
for i in range(num_shards):
filename = os.path.join(
output_dir, '%s-%.5d-of-%.5d' % (name, i, num_shards))
shard_indices = indices[borders[i]:borders[i + 1]]
_write_shard(filename, dataset, shard_indices)
logging.info('Wrote dataset indices [%d, %d) to output shard %s',
borders[i], borders[i + 1], filename)
def main():
parser = argparse.ArgumentParser(
description='Make processed corpus datasets.')
parser.add_argument(
'input_file',
help='Each character and if tag appear on their own line.')
parser.add_argument('output_dir', help='The output directory.')
parser.add_argument('tag_dict',
help='Json-format tag dict file. For example,'
' for word segmentation task, the dict'
' contains "S", "B", "M", "E"; for NER task,'
' the dict contains "O", "B-PER", "I-PER",'
' "B-ORG", "I-ORG", "B-LOC", "I-LOC".'
' Check corpus/seg_tags.json and'
' corpus/ner_tags.json.')
parser.add_argument(
'-validation_percentage', type=float, default=0.1,
help='Percentage of the training data used for validation.')
parser.add_argument('-train_shards', type=int, default=100,
help='Number of output shards for the training set.')
parser.add_argument('-validation_shards', type=int, default=1,
help='Number of output shards for the validation set.')
args = parser.parse_args()
if not tf.io.gfile.isdir(args.output_dir):
tf.io.gfile.makedirs(args.output_dir)
with open(args.tag_dict, 'r') as f:
tag_dict = json.load(f)
vocab = _build_vocab(args.input_file, args.output_dir)
dataset = _build_dataset(args.input_file, vocab, tag_dict)
logging.info('Shuffling dataset.')
np.random.seed(123)
shuffled_indices = np.random.permutation(len(dataset))
num_validation_sentences = int(args.validation_percentage * len(dataset))
val_indices = shuffled_indices[:num_validation_sentences]
train_indices = shuffled_indices[num_validation_sentences:]
_write_dataset('train', dataset, train_indices,
args.train_shards, args.output_dir)
_write_dataset('valid', dataset, val_indices,
args.validation_shards, args.output_dir)
if __name__ == '__main__':
main()