create_pretraining_data.py

#!/usr/bin/python3
# -*- coding: utf-8 -*-
'''
date: 2019/12/3
mail: cally.maxiong@gmail.com
blog: http://www.cnblogs.com/callyblog/
'''
"""Create masked LM/next sentence masked_lm TF examples for BERT."""

import collections
import json
import os
import random
import re

import tensorflow as tf
from bert import tokenization

flags = tf.flags

FLAGS = flags.FLAGS

flags.DEFINE_string("input_dir", None,
                    "Input directory of raw text files")

flags.DEFINE_string("output_dir", None,
                    "Output directory for created tfrecord files")

flags.DEFINE_string("vocab_file", None,
                    "The vocabulary file that the BERT model was trained on.")

flags.DEFINE_list("params", ['title', 'content'],
                    "the params in dataset, like label sentences etc")

flags.DEFINE_bool(
    "do_lower_case", True,
    "Whether to lower case the input text. Should be True for uncased "
    "models and False for cased models.")

flags.DEFINE_bool(
    "do_whole_word_mask", True,
    "Whether to do whole word mask")

flags.DEFINE_bool(
    "ramdom_next", True,
    "Whether to use next sentence loss, it's for fitting Roberta, if False, the training data will like Bert or AlBert, "
    "if True, it will like Roberta training data")

flags.DEFINE_integer("max_seq_length", 512, "Maximum sequence length.")

flags.DEFINE_integer("max_predictions_per_seq", 20,
                     "Maximum number of masked LM predictions per sequence.")

flags.DEFINE_integer("random_seed", 12345, "Random seed for data generation.")

flags.DEFINE_integer(
    "dupe_factor", 5,
    "Number of times to duplicate the input data (with different masks).")

flags.DEFINE_float("masked_lm_prob", 0.15, "Masked LM probability.")

flags.DEFINE_integer("max_workers", 2, "max workers")

flags.DEFINE_float(
    "short_seq_prob", 0.1,
    "Probability of creating sequences which are shorter than the "
    "maximum length.")

flags.DEFINE_integer(
    "doc_stride", 256,
    "When splitting up a long document into chunks, how much stride to "
    "take between chunks.")

class TrainingInstance(object):
  """A single training instance (sentence pair)."""

  def __init__(self, tokens, segment_ids, masked_lm_positions, masked_lm_labels,
               is_random_next):
    self.tokens = tokens
    self.segment_ids = segment_ids
    self.is_random_next = is_random_next
    self.masked_lm_positions = masked_lm_positions
    self.masked_lm_labels = masked_lm_labels

  def __str__(self):
    s = ""
    s += "tokens: %s\n" % (" ".join(
        [tokenization.printable_text(x) for x in self.tokens]))
    s += "segment_ids: %s\n" % (" ".join([str(x) for x in self.segment_ids]))
    s += "is_random_next: %s\n" % self.is_random_next
    s += "masked_lm_positions: %s\n" % (" ".join(
        [str(x) for x in self.masked_lm_positions]))
    s += "masked_lm_labels: %s\n" % (" ".join(
        [tokenization.printable_text(x) for x in self.masked_lm_labels]))
    s += "\n"
    return s

  def __repr__(self):
    return self.__str__()


def write_instance_to_example_files(instances, tokenizer, max_seq_length,
                                    max_predictions_per_seq, output_files):
  """Create TF example files from `TrainingInstance`s."""
  writers = []
  for output_file in output_files:
    writers.append(tf.python_io.TFRecordWriter(output_file))

  writer_index = 0

  total_written = 0
  for (inst_index, instance) in enumerate(instances):
    input_ids = tokenizer.convert_tokens_to_ids(instance.tokens)
    input_mask = [1] * len(input_ids)
    segment_ids = list(instance.segment_ids)
    assert len(input_ids) <= max_seq_length

    while len(input_ids) < max_seq_length:
      input_ids.append(0)
      input_mask.append(0)
      segment_ids.append(0)

    assert len(input_ids) == max_seq_length
    assert len(input_mask) == max_seq_length
    assert len(segment_ids) == max_seq_length

    masked_lm_positions = list(instance.masked_lm_positions)
    masked_lm_ids = tokenizer.convert_tokens_to_ids(instance.masked_lm_labels)
    masked_lm_weights = [1.0] * len(masked_lm_ids)

    while len(masked_lm_positions) < max_predictions_per_seq:
      masked_lm_positions.append(0)
      masked_lm_ids.append(0)
      masked_lm_weights.append(0.0)

    next_sentence_label = 1 if instance.is_random_next else 0

    features = collections.OrderedDict()
    features["input_ids"] = create_int_feature(input_ids)
    features["input_mask"] = create_int_feature(input_mask)
    features["segment_ids"] = create_int_feature(segment_ids)
    features["masked_lm_positions"] = create_int_feature(masked_lm_positions)
    features["masked_lm_ids"] = create_int_feature(masked_lm_ids)
    features["masked_lm_weights"] = create_float_feature(masked_lm_weights)
    features["next_sentence_labels"] = create_int_feature([next_sentence_label])

    tf_example = tf.train.Example(features=tf.train.Features(feature=features))

    writers[writer_index].write(tf_example.SerializeToString())
    writer_index = (writer_index + 1) % len(writers)

    total_written += 1

  for writer in writers:
    writer.close()

  tf.logging.info("Wrote %d total instances", total_written)


def create_int_feature(values):
  feature = tf.train.Feature(int64_list=tf.train.Int64List(value=list(values)))
  return feature


def create_float_feature(values):
  feature = tf.train.Feature(float_list=tf.train.FloatList(value=list(values)))
  return feature


def create_training_instances(file, tokenizer, max_seq_length,
                              dupe_factor, short_seq_prob, masked_lm_prob,
                              max_predictions_per_seq, rng, params):
  all_documents = []

  tf.logging.info('start convert %s to tokens' % file)
  f = open(file, 'r', encoding='utf-8')
  lines = [line for line in f]
  from tqdm import tqdm
  for line in tqdm(lines):
    line = json.loads(line)

    for param in params:
        content = line[param]
        if not len(content) < 20:
          content = tokenization.convert_to_unicode(content.replace("<eop>", ""))
          tokens = tokenizer.tokenize(content)
          tokens = [token for token in tokens if token != '[UNK]']
          if len(tokens) == 0:
            continue
          all_documents.append(tokens)

  tf.logging.info('end convert %s to tokens' % file)

  vocab_words = list(tokenizer.vocab.keys())
  instances = []

  tf.logging.info('start convert %s tokens to instance' % file)
  for document_index in tqdm(range(len(all_documents))):
    for _ in range(dupe_factor):
      instances.extend(
        create_instances_from_document(
          all_documents, document_index, max_seq_length, short_seq_prob,
          masked_lm_prob, max_predictions_per_seq, vocab_words, rng))

  tf.logging.info('end convert %s tokens to instance' % file)
  rng.shuffle(instances)

  return instances

def create_instances_from_document1(
        all_documents, document_index, max_seq_length, short_seq_prob,
        masked_lm_prob, max_predictions_per_seq, vocab_words, rng):
  """Creates `TrainingInstance`s for a single document."""
  document = all_documents[document_index]

  # Account for [CLS], [SEP], [SEP]
  max_num_tokens = max_seq_length - 3

  target_seq_length = max_num_tokens
  if rng.random() < short_seq_prob:
    target_seq_length = rng.randint(2, max_num_tokens)

  instances = []
  current_chunk = []
  current_length = 0
  i = 0
  while i < len(document):
    segment = document[i]
    current_chunk.append(segment)
    current_length += len(segment)
    if i == len(document) - 1 or current_length >= target_seq_length:
      if current_chunk:
        a_end = 1
        if len(current_chunk) >= 2:
          a_end = rng.randint(1, len(current_chunk) - 1)

        tokens_a = []
        for j in range(a_end):
          tokens_a.extend(current_chunk[j])

        tokens_b = []
        # Random next
        if len(current_chunk) == 1 or rng.random() < 0.5:
          is_random_next = True
          target_b_length = target_seq_length - len(tokens_a)

          for _ in range(10):
            random_document_index = rng.randint(0, len(all_documents) - 1)
            if random_document_index != document_index:
              break

          random_document = all_documents[random_document_index]

          random_start = rng.randint(0, len(random_document) - 1)
          for j in range(random_start, len(random_document)):
            tokens_b.extend(random_document[j])
            if len(tokens_b) >= target_b_length:
              break

          num_unused_segments = len(current_chunk) - a_end
          i -= num_unused_segments
        # Actual next
        else:
          is_random_next = False
          for j in range(a_end, len(current_chunk)):
            tokens_b.extend(current_chunk[j])
        truncate_seq_pair(tokens_a, tokens_b, max_num_tokens, rng)

        assert len(tokens_a) >= 1
        assert len(tokens_b) >= 1

        tokens = []
        segment_ids = []
        tokens_a = get_new_segment(''.join(tokens_a))
        tokens_b = get_new_segment(''.join(tokens_b))
        tokens.append("[CLS]")
        segment_ids.append(0)
        for token in tokens_a:
          tokens.append(token)
          segment_ids.append(0)

        tokens.append("[SEP]")
        segment_ids.append(0)

        for token in tokens_b:
          tokens.append(token)
          segment_ids.append(1)
        tokens.append("[SEP]")
        segment_ids.append(1)

        (tokens, masked_lm_positions,
         masked_lm_labels) = create_masked_lm_predictions(
             tokens, masked_lm_prob, max_predictions_per_seq, vocab_words, rng)
        instance = TrainingInstance(
            tokens=tokens,
            segment_ids=segment_ids,
            is_random_next=is_random_next,
            masked_lm_positions=masked_lm_positions,
            masked_lm_labels=masked_lm_labels)
        instances.append(instance)
      current_chunk = []
      current_length = 0
    i += 1

  return instances

def get_raw_instance(document, max_sequence_length):
  """
  获取初步的训练实例，将整段按照max_sequence_length切分成多个部分,并以多个处理好的实例的形式返回。
  :param document: 一整段
  :param max_sequence_length:
  :return: a list. each element is a sequence of text
  """
  max_sequence_length_allowed = max_sequence_length - 3  # [CLS] [SEP] [SEP]

  doc_spans = []
  offset = 0
  _DocSpan = collections.namedtuple(  # pylint: disable=invalid-name
    "DocSpan", ["start", "length", 'text'])
  while offset < len(document):
    length = len(document) - offset
    if length > max_sequence_length_allowed:
      length = max_sequence_length_allowed

    doc_spans.append(_DocSpan(start=offset, length=length, text=''.join(document[offset: offset + length])))

    if offset + length == len(document):
      break
    offset += min(length, FLAGS.doc_stride)

  if len(doc_spans) > 0 and doc_spans[-1].length / max_sequence_length_allowed < 0.1:
    doc_spans = doc_spans[:-1]

  return doc_spans

def create_instances_from_document(
        all_documents, document_index, max_seq_length, short_seq_prob,
        masked_lm_prob, max_predictions_per_seq, vocab_words, rng):
  """Creates `TrainingInstance`s for a single document."""
  document = all_documents[document_index]

  target_seq_length = rng.randint(int(max_seq_length * short_seq_prob), max_seq_length)

  instances = []
  doc_spans = get_raw_instance(document, target_seq_length)  # document即一整段话，包含多个句子。每个句子叫做segment.
  for j, span in enumerate(doc_spans):
    raw_text_list = get_new_segment(span, document)  # 结合分词的中文的whole mask设置即在需要的地方加上“##”

    split_num = rng.randint(int(len(raw_text_list) * 0.3), int(len(raw_text_list) * 0.7))

    while split_num < len(raw_text_list) and len(re.findall('##[\u4E00-\u9FA5]', raw_text_list[split_num])) > 0:
      split_num += 1

    if split_num == len(raw_text_list) - 1:
      continue

    tokens_a = raw_text_list[:split_num]
    tokens_b = raw_text_list[split_num:]

    # 1、设置token, segment_ids
    is_random_next = True  # this will not be used, so it's value doesn't matter
    tokens = []
    segment_ids = []
    tokens.append("[CLS]")
    segment_ids.append(0)
    for token in tokens_a:
      tokens.append(token)
      segment_ids.append(0)
    tokens.append("[SEP]")
    segment_ids.append(0)

    for token in tokens_b:
      tokens.append(token)
      segment_ids.append(1)
    tokens.append('[SEP]')
    segment_ids.append(1)

    # 2、调用原有的方法
    (tokens, masked_lm_positions,
     masked_lm_labels) = create_masked_lm_predictions(
      tokens, masked_lm_prob, max_predictions_per_seq, vocab_words, rng)
    instance = TrainingInstance(
      tokens=tokens,
      segment_ids=segment_ids,
      is_random_next=is_random_next,
      masked_lm_positions=masked_lm_positions,
      masked_lm_labels=masked_lm_labels)
    instances.append(instance)

  return instances

def get_new_segment(span, document):  # 新增的方法 ####
  """
  输入一句话，返回一句经过处理的话: 为了支持中文全称mask，将被分开的词，将上特殊标记("#")，使得后续处理模块，能够知道哪些字是属于同一个词的。
  :param segment: 一句话
  :return: 一句处理过的话
  """
  import jieba
  seq_cws = jieba.cut(span.text)

  segment = document[span.start: span.start + span.length]
  seq_cws_dict = {x: 1 for x in seq_cws}
  new_segment = []
  i = 0
  while i < span.length:
    if len(re.findall('[\u4E00-\u9FA5]', segment[i])) == 0:  # 不是中文的，原文加进去。
      new_segment.append(segment[i])
      i += 1
      continue

    has_add = False
    for length in range(3, 0, -1):
      if i + length > len(segment):
        continue
      if ''.join(segment[i:i + length]) in seq_cws_dict:
        new_segment.append(segment[i])
        for l in range(1, length):
          new_segment.append('##' + segment[i + l])
        i += length
        has_add = True
        break
    if not has_add:
      new_segment.append(segment[i])
      i += 1

  return new_segment

MaskedLmInstance = collections.namedtuple("MaskedLmInstance",
                                          ["index", "label"])


def create_masked_lm_predictions(tokens, masked_lm_prob,
                                 max_predictions_per_seq, vocab_words, rng):
  """Creates the predictions for the masked LM objective."""

  cand_indexes = []
  for (i, token) in enumerate(tokens):
    if token == "[CLS]" or token == "[SEP]":
      continue

    if (FLAGS.do_whole_word_mask and len(cand_indexes) >= 1 and
            token.startswith("##")):
      cand_indexes[-1].append(i)
    else:
      cand_indexes.append([i])

  rng.shuffle(cand_indexes)

  output_tokens = [t[2:] if len(re.findall('##[\u4E00-\u9FA5]', t)) > 0 else t for t in tokens]

  num_to_predict = min(max_predictions_per_seq,
                       max(1, int(round(len(tokens) * masked_lm_prob))))

  masked_lms = []
  covered_indexes = set()
  for index_set in cand_indexes:
    if len(masked_lms) >= num_to_predict:
      break
    # If adding a whole-word mask would exceed the maximum number of
    # predictions, then just skip this candidate.
    if len(masked_lms) + len(index_set) > num_to_predict:
      continue
    is_any_index_covered = False
    for index in index_set:
      if index in covered_indexes:
        is_any_index_covered = True
        break
    if is_any_index_covered:
      continue
    for index in index_set:
      covered_indexes.add(index)

      # 80% of the time, replace with [MASK]
      if rng.random() < 0.8:
        masked_token = "[MASK]"
      else:
        # 10% of the time, keep original
        if rng.random() < 0.5:
          masked_token = tokens[index][2:] if len(re.findall('##[\u4E00-\u9FA5]', tokens[index])) > 0 else tokens[index]
        # 10% of the time, replace with random word
        else:
          masked_token = vocab_words[rng.randint(0, len(vocab_words) - 1)]

      output_tokens[index] = masked_token

      masked_lms.append(MaskedLmInstance(index=index, label=tokens[index]))
  assert len(masked_lms) <= num_to_predict
  masked_lms = sorted(masked_lms, key=lambda x: x.index)

  masked_lm_positions = []
  masked_lm_labels = []
  for p in masked_lms:
    masked_lm_positions.append(p.index)
    if len(re.findall('##[\u4E00-\u9FA5]', p.label)) > 0:
      masked_lm_labels.append(p.label[2:])
    else:
      masked_lm_labels.append(p.label)

  return (output_tokens, masked_lm_positions, masked_lm_labels)


def truncate_seq_pair(tokens_a, tokens_b, max_num_tokens, rng):
  """Truncates a pair of sequences to a maximum sequence length."""
  while True:
    total_length = len(tokens_a) + len(tokens_b)
    if total_length <= max_num_tokens:
      break

    trunc_tokens = tokens_a if len(tokens_a) > len(tokens_b) else tokens_b
    assert len(trunc_tokens) >= 1

    # We want to sometimes truncate from the front and sometimes from the
    # back to add more randomness and avoid biases.
    if rng.random() < 0.5:
      del trunc_tokens[0]
    else:
      trunc_tokens.pop()


def main(_):
  tf.logging.set_verbosity(tf.logging.INFO)

  input_files = []
  for file in os.listdir(FLAGS.input_dir):
    input_files.append(os.path.join(FLAGS.input_dir, file))

  for file in input_files:
    tf.logging.info('input file %s' % file)

  tokenizer = tokenization.FullTokenizer(
    vocab_file=FLAGS.vocab_file, do_lower_case=FLAGS.do_lower_case)

  for file in input_files:
    tf.logging.info('start file %s' % file)
    rng = random.Random(FLAGS.random_seed)
    instances = create_training_instances(
      file, tokenizer, FLAGS.max_seq_length, FLAGS.dupe_factor,
      FLAGS.short_seq_prob, FLAGS.masked_lm_prob, FLAGS.max_predictions_per_seq,
      rng, params=FLAGS.params)

    tf.logging.info('start convert %s instance to tfrecords' % file)
    paths = file.split('/')
    output_files = [os.path.join(FLAGS.output_dir, '%s_%s.tfrecord' % (paths[-2], paths[-1][:-5]))]
    write_instance_to_example_files(instances, tokenizer, FLAGS.max_seq_length,
                                    FLAGS.max_predictions_per_seq, output_files)
    tf.logging.info('end convert %s instance to tfrecords' % file)

    del instances
  tf.logging.info('all were done')

if __name__ == "__main__":
  tf.app.run()