BERT学习笔记：create_pretraining_data.py

BERT 源码初探之 create_pretraining_data.py

本文源码来源于 Github上的BERT 项目中的 run_pretraining.py 文件。阅读本文需要对Attention Is All You Need以及BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding两篇论文有所了解，以及部分关于深度学习、自然语言处理和Tensorflow的储备知识。

0 前言

• 关于Tensorflow：本文基于谷歌官方在GitHub上公布的BERT预训练模型，基于Tensorflow 1.13.1 运行。有关Tensorflow的部分建议参照官方网站。
• 关于Transformer：Transformer是Google提出的一种完全基于注意力机制的模型，想要对齐进行了解请参照官方论文Attention Is All You Need或者我的另一篇博客Transformer 学习笔记。
• 关于BERT：BERT也是Google提出的一个基于Transformer的预训练网络模型，更多和该模型有关的内容请参照官方论文BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding、官方代码实现Github上的BERT以及我的另一篇博客BERT 学习笔记。

1 简介

要使用时才发现 BERT 提供了把文本数据转化为预训练模型所需的数据的代码，因此本文就来阅读这一部分代码吧。

2 源码解释

2.1 参数定义

2.1.1 必须参数

flags.DEFINE_string("input_file", None,
                    "Input raw text file (or comma-separated list of files).")

flags.DEFINE_string(
    "output_file", None,
    "Output TF example file (or comma-separated list of files).")

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

• 文件输入路径
• 输出文件路径
• 词典文件路径

2.2.2 可选参数

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_integer("max_seq_length", 128, "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.")

• 是否小写输入
• 最大句子的长度
• 每一句MLM预测的百分比
• 随机数种子（用于数据生成）

flags.DEFINE_integer(
    "dupe_factor", 10,
    "Number of times to duplicate the input data (with different masks).")

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

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

• 复制输入数据的次数（采用不同的masks）
• MLM的比例
• 生成小于最大长度的句子的概率

2.2 训练实例

2.2.1 一个单独的训练实例（TrainingInstance）

class TrainingInstance(object):

  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

类定义以及初始化

  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__()

将自己字符串化的方法

2.2.2 训练实例相关方法

def write_instance_to_example_files(instances, tokenizer, max_seq_length,
                                    max_predictions_per_seq, output_files):
  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

    if inst_index < 20:
      tf.logging.info("*** Example ***")
      tf.logging.info("tokens: %s" % " ".join(
          [tokenization.printable_text(x) for x in instance.tokens]))

      for feature_name in features.keys():
        feature = features[feature_name]
        values = []
        if feature.int64_list.value:
          values = feature.int64_list.value
        elif feature.float_list.value:
          values = feature.float_list.value
        tf.logging.info(
            "%s: %s" % (feature_name, " ".join([str(x) for x in values])))

  for writer in writers:
    writer.close()

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

把格式好的用于训练的数据写入tfrecord文件中，这个方法的主要作用就是把已经处理好的数据写入二进制文件中

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

创建不同的训练特征，创建int类型和float类型的训练特征

2.2.3 创建训练实例（create_training_instances）

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

方法的定义，从行组合的输入文件中创建训练实例。输入数据的格式如下：

• 每行是一个文档中的一句话，凭借这个来进行下一句预测的任务
• 文档之间隔一个空行，因为下一句预测不跨文档

  for input_file in input_files:
    with tf.gfile.GFile(input_file, "r") as reader:
      while True:
        line = tokenization.convert_to_unicode(reader.readline())
        if not line:
          break
        line = line.strip()

        # Empty lines are used as document delimiters
        if not line:
          all_documents.append([])
        tokens = tokenizer.tokenize(line)
        if tokens:
          all_documents[-1].append(tokens)

从文档中读入数据，token化后存入all_documents中

  all_documents = [x for x in all_documents if x]
  rng.shuffle(all_documents)

把空文档从其中中删除掉

  vocab_words = list(tokenizer.vocab.keys())
  instances = []
  for _ in range(dupe_factor):
    for document_index in range(len(all_documents)):
      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))

  rng.shuffle(instances)
  return instances

加载词典，调用 create_instances_from_document 方法根据单文档创建训练实例

2.2.4 根据单文档创建训练实例（create_instances_from_document）

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]

  max_num_tokens = max_seq_length - 3

• 定义方法头，获取要处理的文档
• 为了加入 [CLS], [SEP], [SEP] ，最大长度应该减三

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

大部分情况下，我们希望把长度填充到最大长度，但是一少部分情况下我们希望采用短句来最小化预训练和微调的差异。总的来说 target_seq_length 是一个粗略的目标，而 max_seq_length是一个强制的限制。

    if i == len(document) - 1 or current_length >= target_seq_length:
      if current_chunk:
        # `a_end` is how many segments from `current_chunk` go into the `A`
        # (first) sentence.
        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
        is_random_next = False
        if len(current_chunk) == 1 or rng.random() < 0.5:
          is_random_next = True
          target_b_length = target_seq_length - len(tokens_a)

          # This should rarely go for more than one iteration for large
          # corpora. However, just to be careful, we try to make sure that
          # the random document is not the same as the document
          # we're processing.
          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
          # We didn't actually use these segments so we "put them back" so
          # they don't go to waste.
          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

如果是文档中的最后一句话或者长度已经超过了最大长度，那么把这一句从头开始随机切一段作为第一句，剩下的部分作为第二句，而随机的第二句就从其他文档中随机选择一句同样长度超过的切割相同位置开始相同长度的一段作为第二句。

        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)

添加标记符号

        (tokens, masked_lm_positions,
         masked_lm_labels) = create_masked_lm_predictions(
             tokens, masked_lm_prob, max_predictions_per_seq, vocab_words, rng)

调用 create_masked_lm_predictions 方法来添加MLM数据

        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

创建 TrainingInstance 对象返回。

2.2.5 创建MLM的预测（create_masked_lm_predictions）

def create_masked_lm_predictions(tokens, masked_lm_prob,
                                 max_predictions_per_seq, vocab_words, rng):

定义方法头

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

  rng.shuffle(cand_indexes)

获取输入的字符，打乱。

  output_tokens = list(tokens)

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

获取输出字符，概率大小

  masked_lms = []
  covered_indexes = set()
  for index in cand_indexes:
    if len(masked_lms) >= num_to_predict:
      break
    if index in covered_indexes:
      continue
    covered_indexes.add(index)

    masked_token = None
    # 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]
      # 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]))

选择部分字符用 [MASK] 进行替换，详细内容参加论文原文。

  masked_lm_positions = []
  masked_lm_labels = []
  for p in masked_lms:
    masked_lm_positions.append(p.index)
    masked_lm_labels.append(p.label)

  return (output_tokens, masked_lm_positions, masked_lm_labels)

调整数据格式，返回。

2.2.6 截断序列长度（truncate_seq_pair）

def truncate_seq_pair(tokens_a, tokens_b, max_num_tokens, rng):
  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

    if rng.random() < 0.5:
      del trunc_tokens[0]
    else:
      trunc_tokens.pop()

把一对句子截断到最大序列长度。

2.3 main(_) 方法

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

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

  input_files = []
  for input_pattern in FLAGS.input_file.split(","):
    input_files.extend(tf.gfile.Glob(input_pattern))

  tf.logging.info("*** Reading from input files ***")
  for input_file in input_files:
    tf.logging.info("  %s", input_file)

  rng = random.Random(FLAGS.random_seed)
  instances = create_training_instances(
      input_files, tokenizer, FLAGS.max_seq_length, FLAGS.dupe_factor,
      FLAGS.short_seq_prob, FLAGS.masked_lm_prob, FLAGS.max_predictions_per_seq,
      rng)

  output_files = FLAGS.output_file.split(",")
  tf.logging.info("*** Writing to output files ***")
  for output_file in output_files:
    tf.logging.info("  %s", output_file)

  write_instance_to_example_files(instances, tokenizer, FLAGS.max_seq_length,
                                  FLAGS.max_predictions_per_seq, output_files)

• 加载数据
• 处理成指定格式
• 存入二进制输出文件中

2.4 主函数入口

  flags.mark_flag_as_required("input_file")
  flags.mark_flag_as_required("output_file")
  flags.mark_flag_as_required("vocab_file")
  tf.app.run()

定义必须的函数并运行。

3 结论

这篇源码里对 BERT 是如果创建预训练所需数据有一个比较清晰的解释了，其中最有趣的是其中的根据单文档创建训练实例的方法，并不是简单的把文档拼接起来然后随机截取一个位置作为上下文分割，而是尽可能的采用用户在源文件里给出的断句方式。