BERT源码注释(run_classifier.py) - 全网最详细

摘要：本文为BERT模型的run_classifier.py的详细注释，便于了解这个微调脚本究竟做了什么？模型可调参数有哪些？自己写一个processor的类，需要注意哪些？想要运行这个脚本，需要传入哪些参数？尽在本文详细注释！

# coding=utf-8

# bert注解版
# raw author: Google
# explain author：putdoor
"""BERT finetuning runner."""

from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

import collections
import csv
import os
import modeling
import random
import optimization  # 优化器
import tokenization  # 令牌化
import tensorflow as tf

flags = tf.flags

FLAGS = flags.FLAGS
# 用于支持命令行传递参数：如：python test_flags.py --model "My model"
## Required parameters  # 这5项参数为必填参数，把None改为自己所需要
flags.DEFINE_string(  
    "data_dir", "/home/work/my_model/bert/corpus/sentiment",  # 此sentiment下应有三个文件，名称分别为:train.tsv, eval.tsv, test.tsv
    "The input data dir. Should contain the .tsv files (or other data files) "
    "for the task.")

flags.DEFINE_string(
    "bert_config_file", os.path.join(os.path.dirname(os.path.abspath(__file__)), 'chinese_L-12_H-768_A-12/bert_config.json'),
    "The config json file corresponding to the pre-trained BERT model. "
    "This specifies the model architecture.")

flags.DEFINE_string("task_name", "sentiment", "The name of the task to train.")  # 自己随意起一个便于标识processor类的任务名,后续需填入main函数的字典

flags.DEFINE_string("vocab_file", os.path.join(os.path.dirname(os.path.abspath(__file__)), 'chinese_L-12_H-768_A-12/vocab.txt'),
                    "The vocabulary file that the BERT model was trained on.")

flags.DEFINE_string(
    "output_dir", "/home/work/my_model/output",
    "The output directory where the model checkpoints will be written.")

# Other parameters
flags.DEFINE_string(
    "init_checkpoint", None,
    "Initial checkpoint (usually from a pre-trained BERT model).")

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,
    "The maximum total input sequence length after WordPiece tokenization. "
    "Sequences longer than this will be truncated, and sequences shorter "
    "than this will be padded.")

flags.DEFINE_bool("do_train", False, "Whether to run training.")

flags.DEFINE_bool("do_eval", False, "Whether to run eval on the dev set.")

flags.DEFINE_bool(
    "do_predict", False,
    "Whether to run the model in inference mode on the test set.")

flags.DEFINE_integer("train_batch_size", 32, "Total batch size for training.")  # mini-batch方式的梯度下降，每批处理32个样本

flags.DEFINE_integer("eval_batch_size", 8, "Total batch size for eval.")  # batch为每批样本的数量，每批更新一次权重，使loss最小

flags.DEFINE_integer("predict_batch_size", 8, "Total batch size for predict.")

flags.DEFINE_float("learning_rate", 5e-5, "The initial learning rate for Adam.")

flags.DEFINE_float("num_train_epochs", 3.0,  # 所有样本轮一次，即为一个epoch，增大此值，计算量增大，一个20000条数据的二分类问题，epochs=4大概要10分钟(16G 单GPU)
                   "Total number of training epochs to perform.")

flags.DEFINE_float(
    "warmup_proportion", 0.1,  # 预热训练中，线性地增加学习率，详见注意力机制部分
    "Proportion of training to perform linear learning rate warmup for. "
    "E.g., 0.1 = 10% of training.")

flags.DEFINE_integer("save_checkpoints_steps", 1000,  # 保存检查点时的步数，达到1000时，保存一次模型
                     "How often to save the model checkpoint.")

flags.DEFINE_integer("iterations_per_loop", 1000,  # 在每个estimator调用中执行多少步骤
                     "How many steps to make in each estimator call.")

flags.DEFINE_bool("use_tpu", False, "Whether to use TPU or GPU/CPU.")

# TPU config：
tf.flags.DEFINE_string(
    "tpu_name", None,
    "The Cloud TPU to use for training. This should be either the name "
    "used when creating the Cloud TPU, or a grpc://ip.address.of.tpu:8470 "
    "url.")

tf.flags.DEFINE_string(
    "tpu_zone", None,
    "[Optional] GCE zone where the Cloud TPU is located in. If not "
    "specified, we will attempt to automatically detect the GCE project from "
    "metadata.")

tf.flags.DEFINE_string(
    "gcp_project", None,
    "[Optional] Project name for the Cloud TPU-enabled project. If not "
    "specified, we will attempt to automatically detect the GCE project from "
    "metadata.")

tf.flags.DEFINE_string("master", None, "[Optional] TensorFlow master URL.")

flags.DEFINE_integer(
    "num_tpu_cores", 8,
    "Only used if `use_tpu` is True. Total number of TPU cores to use.")


class InputExample(object):  # 每一行数据 to Inputexample对象
  """A single training/test example for simple sequence classification."""

  def __init__(self, guid, text_a, text_b=None, label=None):
    """Constructs a InputExample.

    Args:
      guid: Unique id for the example.
      text_a: string. The untokenized text of the first sequence. For single
        sequence tasks, only this sequence must be specified.
      text_b: (Optional) string. The untokenized text of the second sequence.
        Only must be specified for sequence pair tasks.
      label: (Optional) string. The label of the example. This should be
        specified for train and dev examples, but not for test examples.
    """
    self.guid = guid
    self.text_a = text_a
    self.text_b = text_b
    self.label = label


class PaddingInputExample(object):
"""当需要使用TPU训练时，eval和predict的数据需要是batch_size的整数倍，此类用于处理这类情况"""


class InputFeatures(object):
  """A single set of features of data."""

  def __init__(self,
               input_ids,  # 输入部分：token embedding：表示词向量，第一个词是CLS，分隔词有SEP，是单词本身
               input_mask,  # 输入部分：position embedding：为了令transformer感知词与词之间的位置关系
               segment_ids,  # 输入部分：segment embedding：text_a与text_b的句子关系
               label_id,  # 输出部分：标签，对应Y
               is_real_example=True):
    self.input_ids = input_ids
    self.input_mask = input_mask
    self.segment_ids = segment_ids
    self.label_id = label_id
    self.is_real_example = is_real_example


class DataProcessor(object):
  """Base class for data converters for sequence classification data sets."""

  def get_train_examples(self, data_dir):
    """Gets a collection of `InputExample`s for the train set."""
    raise NotImplementedError()

  def get_dev_examples(self, data_dir):
    """Gets a collection of `InputExample`s for the dev set."""
    raise NotImplementedError()

  def get_test_examples(self, data_dir):
    """Gets a collection of `InputExample`s for prediction."""
    raise NotImplementedError()

  def get_labels(self):
    """Gets the list of labels for this data set."""
    raise NotImplementedError()

  @classmethod
  def _read_tsv(cls, input_file, quotechar=None):
    """Reads a tab separated value file."""
    with tf.gfile.Open(input_file, "r") as f:
      reader = csv.reader(f, delimiter="\t", quotechar=quotechar)
      lines = []
      for line in reader:
        lines.append(line)
      return lines


class SentimentProcessor(DataProcessor):  # 自定义的类，用于处理二分类问题
    """Processor for the CoLA data set (GLUE version)."""
    def get_train_examples(self, data_dir):
        """See base class."""
        return self._create_examples(
            self._read_tsv(os.path.join(data_dir, "train.tsv")), "train")

    def get_dev_examples(self, data_dir):
        """See base class."""
        return self._create_examples(
            self._read_tsv(os.path.join(data_dir, "dev.tsv")), "dev")

    def get_test_examples(self, data_dir):
        """See base class."""
        return self._create_examples(
            self._read_tsv(os.path.join(data_dir, "test.tsv")), "test")

    def get_labels(self):  # 二分类问题返回的标签值为0，1
        """See base class."""
        return ["0", "1"]

    def _create_examples(self, lines, set_type):
        """Creates examples for the training and dev sets."""
        examples = []
        for (i, line) in enumerate(lines):
            # Only the test set has a header
            guid = "%s-%s" % (set_type, i)
            text_a = tokenization.convert_to_unicode(line[1])
            label = tokenization.convert_to_unicode(line[0])
            examples.append(
                InputExample(guid=guid, text_a=text_a, text_b=None, label=label))
        return examples


def convert_single_example(ex_index, example, label_list, max_seq_length,
                           tokenizer):
  """Converts a single `InputExample` into a single `InputFeatures`."""

  if isinstance(example, PaddingInputExample):  # 相当于实例example为空，返回的数据
    return InputFeatures(
        input_ids=[0] * max_seq_length,
        input_mask=[0] * max_seq_length,
        segment_ids=[0] * max_seq_length,
        label_id=0,
        is_real_example=False)

  label_map = {}
  for (i, label) in enumerate(label_list):  # 标签映射
    label_map[label] = i

  tokens_a = tokenizer.tokenize(example.text_a)
  tokens_b = None
  
  if example.text_b:
    tokens_b = tokenizer.tokenize(example.text_b)

  if tokens_b:
    # Modifies `tokens_a` and `tokens_b` in place so that the total
    # length is less than the specified length.
    # Account for [CLS], [SEP], [SEP] with "- 3"
    _truncate_seq_pair(tokens_a, tokens_b, max_seq_length - 3)  # 截断序列对：将序列截断至最大允许长度
  else:
    # Account for [CLS] and [SEP] with "- 2"
    if len(tokens_a) > max_seq_length - 2:
      tokens_a = tokens_a[0:(max_seq_length - 2)]

  # The convention in BERT is:
  # (a) For sequence pairs:
  #  tokens:   [CLS] is this jack ##son ##ville ? [SEP] no it is not . [SEP]
  #  type_ids: 0     0  0    0    0     0       0 0     1  1  1  1   1 1
  # (b) For single sequences:
  #  tokens:   [CLS] the dog is hairy . [SEP]
  #  type_ids: 0     0   0   0  0     0 0
  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)

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

  # tokenizer ：是bert源码中提供的模块，其实主要作用就是将句子拆分成字，并且将字映射成id
  input_ids = tokenizer.convert_tokens_to_ids(tokens)  

  input_mask = [1] * len(input_ids)  # 暂时认为词与词之间的位置关系由索引决定就可以了[1,1,1...] --> index: 0,1,2...

  # Zero-pad up to the sequence 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

  label_id = label_map[example.label]
  if ex_index < 5:
    tf.logging.info("*** Example ***")
    tf.logging.info("guid: %s" % (example.guid))
    tf.logging.info("tokens: %s" % " ".join(
        [tokenization.printable_text(x) for x in tokens]))
    tf.logging.info("input_ids: %s" % " ".join([str(x) for x in input_ids]))
    tf.logging.info("input_mask: %s" % " ".join([str(x) for x in input_mask]))
    tf.logging.info("segment_ids: %s" % " ".join([str(x) for x in segment_ids]))
    tf.logging.info("label: %s (id = %d)" % (example.label, label_id))

  feature = InputFeatures(
      input_ids=input_ids,
      input_mask=input_mask,
      segment_ids=segment_ids,
      label_id=label_id,
      is_real_example=True)
  return feature


# 实现了俩功能：1.调用convert_single_example转化Input_example为Feature_example 2.转换为TFRecord格式，便于大型数据处理
def file_based_convert_examples_to_features(
    examples, label_list, max_seq_length, tokenizer, output_file):
  """Convert a set of `InputExample`s to a TFRecord file."""   # 写'train.tf_record'文件到output_dir下
                                                               # TFRecord内部采用二进制编码，加载快，对大型数据转换友好

# 此模块主要分为两个部分：1.TFRecord生成器, 2.Example模块
  writer = tf.python_io.TFRecordWriter(output_file)  # 最外层：是TFRecord生成器部分，内部需要传入tf_example

  for (ex_index, example) in enumerate(examples):
    if ex_index % 10000 == 0:
      tf.logging.info("Writing example %d of %d" % (ex_index, len(examples)))  # 跟踪examples转换进度
	# 把examples数据转化为features，用到前面的单次转换函数
    feature = convert_single_example(ex_index, example, label_list,
                                     max_seq_length, tokenizer)
    # 这是Example模块
    def create_int_feature(values):
      f = tf.train.Feature(int64_list=tf.train.Int64List(value=list(values)))
      return f
      
    features = collections.OrderedDict()  # 其实就是一个字典
    
    # 以下五句都是调用create_int_feature生成value往features字典中填充
    features["input_ids"] = create_int_feature(feature.input_ids)
    features["input_mask"] = create_int_feature(feature.input_mask)
    features["segment_ids"] = create_int_feature(feature.segment_ids)
    features["label_ids"] = create_int_feature([feature.label_id])
    features["is_real_example"] = create_int_feature(
        [int(feature.is_real_example)])

    tf_example = tf.train.Example(features=tf.train.Features(feature=features))  # 最外层是tf.train.Features()的实例，内层是feature的字典
    writer.write(tf_example.SerializeToString())
  writer.close()


# 这是一个闭包，外层函数返回内层函数的引用，内层函数使用外层函数的参数
def file_based_input_fn_builder(input_file, seq_length, is_training,  # 此input_file是TFRecord文件
                                drop_remainder):
  """Creates an `input_fn` closure to be passed to TPUEstimator."""  # 生成一个input_fn闭包传递给TPUEstimator

  name_to_features = {
      # 是tensorflow example协议中的一种解析，这里面，传入了shape和dtype
      "input_ids": tf.FixedLenFeature([seq_length], tf.int64),
      "input_mask": tf.FixedLenFeature([seq_length], tf.int64),
      "segment_ids": tf.FixedLenFeature([seq_length], tf.int64),
      "label_ids": tf.FixedLenFeature([], tf.int64),
      "is_real_example": tf.FixedLenFeature([], tf.int64),
  }

  def _decode_record(record, name_to_features):
    """Decodes a record to a TensorFlow example."""
    example = tf.parse_single_example(record, name_to_features)

    # tf.Example only supports tf.int64, but the TPU only supports tf.int32.
    # So cast all int64 to int32.
    for name in list(example.keys()):
      t = example[name]
      if t.dtype == tf.int64:
        t = tf.to_int32(t)
      example[name] = t

    return example

  def input_fn(params):
    """The actual input function."""
    batch_size = params["batch_size"]

    # 对于训练，我们需要大量的并行的读取和洗牌
    # 对于评估，我们不需要洗牌，并行的读取也无关紧要
    d = tf.data.TFRecordDataset(input_file)
    if is_training:
      d = d.repeat()  # 重复
      d = d.shuffle(buffer_size=100)  # 洗牌，缓冲区=100

    d = d.apply(
        tf.contrib.data.map_and_batch(
            # 调用_decode_record函数：1.解析TFRecord为example 2.int64 to int32
            lambda record: _decode_record(record, name_to_features),  
            batch_size=batch_size,
            drop_remainder=drop_remainder))

    return d

  return input_fn


def _truncate_seq_pair(tokens_a, tokens_b, max_length):
  """Truncates a sequence pair in place to the maximum length."""  # 将序列对截断到最大长度max_length

  # This is a simple heuristic which will always truncate the longer sequence
  # one token at a time. This makes more sense than truncating an equal percent
  # of tokens from each, since if one sequence is very short then each token
  # that's truncated likely contains more information than a longer sequence.
  while True:
    total_length = len(tokens_a) + len(tokens_b)
    if total_length <= max_length:
      break
    if len(tokens_a) > len(tokens_b):
      tokens_a.pop()
    else:
      tokens_b.pop()


# 做了两件事:1.使用modeling.py中的BerModel类创建模型 2.计算交叉熵损失loss
def create_model(bert_config, is_training, input_ids, input_mask, segment_ids,  # 创建分类器模型
                 labels, num_labels, use_one_hot_embeddings):
  """Creates a classification model."""
  model = modeling.BertModel(
      config=bert_config,
      is_training=is_training,
      input_ids=input_ids,
      input_mask=input_mask,
      token_type_ids=segment_ids,
      use_one_hot_embeddings=use_one_hot_embeddings)

  # In the demo, we are doing a simple classification task on the entire
  # segment.
  #
  # If you want to use the token-level output, use model.get_sequence_output()
  # instead.
  output_layer = model.get_pooled_output()  # 输出层

  hidden_size = output_layer.shape[-1].value  # 隐藏层大小

  output_weights = tf.get_variable(  # 输出层权重
      "output_weights", [num_labels, hidden_size],
      initializer=tf.truncated_normal_initializer(stddev=0.02))

  output_bias = tf.get_variable(  # 输出层偏置
      "output_bias", [num_labels], initializer=tf.zeros_initializer())

  with tf.variable_scope("loss"):
    if is_training:
      # I.e., 0.1 dropout
      output_layer = tf.nn.dropout(output_layer, keep_prob=0.9)  # dropout：减小模型过拟合，保留90%的网络连接，随机drop 10%

    logits = tf.matmul(output_layer, output_weights, transpose_b=True)  # w*x的矩阵乘法
    logits = tf.nn.bias_add(logits, output_bias)  # w*x + b
    probabilities = tf.nn.softmax(logits, axis=-1)  # 把输出结果指数归一化映射到（0，1）区间
    log_probs = tf.nn.log_softmax(logits, axis=-1)  # 相当于对上式的每个值求log，落在 负无穷到0之间

	# 标签one_hot编码，相当于增加标签维度，变稀疏化
    one_hot_labels = tf.one_hot(labels, depth=num_labels, dtype=tf.float32)
    per_example_loss = -tf.reduce_sum(one_hot_labels * log_probs, axis=-1)
    # 交叉熵损失：交叉熵的值越小，两个概率分布就越接近
    loss = tf.reduce_mean(per_example_loss)

    return (loss, per_example_loss, logits, probabilities)


def model_fn_builder(bert_config, num_labels, init_checkpoint, learning_rate,
                     num_train_steps, num_warmup_steps, use_tpu,
                     use_one_hot_embeddings):
  """Returns `model_fn` closure for TPUEstimator."""  # 返回给TPUEstimator闭包model_fn

  def model_fn(features, labels, mode, params):  # pylint: disable=unused-argument
    """The `model_fn` for TPUEstimator."""

    tf.logging.info("*** Features ***")
    for name in sorted(features.keys()):
      tf.logging.info("  name = %s, shape = %s" % (name, features[name].shape))

    input_ids = features["input_ids"]
    input_mask = features["input_mask"]
    segment_ids = features["segment_ids"]
    label_ids = features["label_ids"]
    is_real_example = None
    if "is_real_example" in features:
      is_real_example = tf.cast(features["is_real_example"], dtype=tf.float32)  # tf.cast()数据类型的转换
    else:
      is_real_example = tf.ones(tf.shape(label_ids), dtype=tf.float32)  # 生成所有数字为1的tensor

    is_training = (mode == tf.estimator.ModeKeys.TRAIN)

	# 1.创建bert的model 2.计算loss
    (total_loss, per_example_loss, logits, probabilities) = create_model( 
        bert_config, is_training, input_ids, input_mask, segment_ids, label_ids,
        num_labels, use_one_hot_embeddings)

    tvars = tf.trainable_variables()
    initialized_variable_names = {}
    scaffold_fn = None
    if init_checkpoint:
      (assignment_map, initialized_variable_names
      ) = modeling.get_assignment_map_from_checkpoint(tvars, init_checkpoint)
      if use_tpu:

        def tpu_scaffold():  # 用于创建或收集训练模型通常需要的部件
          tf.train.init_from_checkpoint(init_checkpoint, assignment_map)
          return tf.train.Scaffold()

        scaffold_fn = tpu_scaffold
      else:
        tf.train.init_from_checkpoint(init_checkpoint, assignment_map)

    tf.logging.info("**** Trainable Variables ****")
    for var in tvars:
      init_string = ""
      if var.name in initialized_variable_names:
        init_string = ", *INIT_FROM_CKPT*"
      tf.logging.info("  name = %s, shape = %s%s", var.name, var.shape,  # 得到可用于训练的变量名
                      init_string)

    output_spec = None
    if mode == tf.estimator.ModeKeys.TRAIN:

      train_op = optimization.create_optimizer(
          total_loss, learning_rate, num_train_steps, num_warmup_steps, use_tpu)

      output_spec = tf.contrib.tpu.TPUEstimatorSpec(
          mode=mode,  # 模式为训练模式
          loss=total_loss,  # 损失
          train_op=train_op,  # 操作
          scaffold_fn=scaffold_fn)  # 训练模型需要的部件
    elif mode == tf.estimator.ModeKeys.EVAL:

      def metric_fn(per_example_loss, label_ids, logits, is_real_example):
        predictions = tf.argmax(logits, axis=-1, output_type=tf.int32)  # 返回array中，最大值的索引
        accuracy = tf.metrics.accuracy(
            labels=label_ids, predictions=predictions, weights=is_real_example)
        loss = tf.metrics.mean(values=per_example_loss, weights=is_real_example)
        return {
            "eval_accuracy": accuracy,
            "eval_loss": loss,
        }

      eval_metrics = (metric_fn,
                      [per_example_loss, label_ids, logits, is_real_example])
      output_spec = tf.contrib.tpu.TPUEstimatorSpec(
          mode=mode,  # 模式为评估模式
          loss=total_loss,
          eval_metrics=eval_metrics,  # 此处内含：accuracy和loss
          scaffold_fn=scaffold_fn)
    else:
      output_spec = tf.contrib.tpu.TPUEstimatorSpec(
          mode=mode,
          # 此probabilities为softmax的计算的概率
          predictions={"probabilities": probabilities},
          scaffold_fn=scaffold_fn)
    return output_spec

  return model_fn


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

  processors = {
      "cola": ColaProcessor,  # 填入自定义处理数据的类，必填项
      "mnli": MnliProcessor,
      "mrpc": MrpcProcessor,
      "xnli": XnliProcessor,
      "sentiment": SentimentProcessor
  }

  tokenization.validate_case_matches_checkpoint(FLAGS.do_lower_case,  # 验证实例，匹配检查点
                                                FLAGS.init_checkpoint)

  if not FLAGS.do_train and not FLAGS.do_eval and not FLAGS.do_predict:
    raise ValueError(
        "At least one of `do_train`, `do_eval` or `do_predict' must be True.")

  bert_config = modeling.BertConfig.from_json_file(FLAGS.bert_config_file)
  # 上面三个判断至少有一个为真执行下面语句
  if FLAGS.max_seq_length > bert_config.max_position_embeddings:  # 语句长度设置的比如为128，不应大于bert官方训练时的长度
    raise ValueError(
        "Cannot use sequence length %d because the BERT model "
        "was only trained up to sequence length %d" %
        (FLAGS.max_seq_length, bert_config.max_position_embeddings))

  tf.gfile.MakeDirs(FLAGS.output_dir)  # 此语句会创建output_dir目录，所以，只要指定，无需再单独创建

  task_name = FLAGS.task_name.lower()  # 传入的自定义类Sentiment名称小写化为sentiment

  if task_name not in processors:
    raise ValueError("Task not found: %s" % (task_name))

  processor = processors[task_name]()  # 类()  -->  实例化

  label_list = processor.get_labels()  # 类.方法，此label_list为['0','1']类别标签集

  tokenizer = tokenization.FullTokenizer(
      vocab_file=FLAGS.vocab_file, do_lower_case=FLAGS.do_lower_case)  # 加载训练的中文词典，输入数据是否小写化

  tpu_cluster_resolver = None  # tpu集群处理
  if FLAGS.use_tpu and FLAGS.tpu_name:  # 使用tpu时创建tpu集群处理实例
    tpu_cluster_resolver = tf.contrib.cluster_resolver.TPUClusterResolver(
        FLAGS.tpu_name, zone=FLAGS.tpu_zone, project=FLAGS.gcp_project)

  is_per_host = tf.contrib.tpu.InputPipelineConfig.PER_HOST_V2  # per_host：每主机，XLnet中num_core_per_host指的是每主机核数
  run_config = tf.contrib.tpu.RunConfig(  # tpu的运行配置
      cluster=tpu_cluster_resolver,
      master=FLAGS.master,
      model_dir=FLAGS.output_dir,
      save_checkpoints_steps=FLAGS.save_checkpoints_steps,
      tpu_config=tf.contrib.tpu.TPUConfig(  # tf.contrib模块是tf.nn的上一层的tf.layer的上层,主要提供一些图上的操作：如正则化，摘要操作。。。
          iterations_per_loop=FLAGS.iterations_per_loop,  # 在每个estimator调用中执行多少步，default中为1000步
          num_shards=FLAGS.num_tpu_cores,  # tpu核数，default为8
          per_host_input_for_training=is_per_host))

  train_examples = None
  num_train_steps = None
  num_warmup_steps = None



  if FLAGS.do_train:
    # 返回列表是一行一行的Inputexample对象，每行包括了guid，train_a,label..
    train_examples = processor.get_train_examples(FLAGS.data_dir)
    # 训练的次数：(训练集的样本数/每批次大小)*训练几轮
    num_train_steps = int(
        len(train_examples) / FLAGS.train_batch_size * FLAGS.num_train_epochs)
        # 举例：num_train_steps = (20000/32)*3 = 1800次，
        # 也就是权重更新，loss下降1800次                                                            
            
    # 在预热学习中，线性地增加学习率                                               
    num_warmup_steps = int(num_train_steps * FLAGS.warmup_proportion)  

  model_fn = model_fn_builder(
      bert_config=bert_config,
      num_labels=len(label_list),
      init_checkpoint=FLAGS.init_checkpoint,
      learning_rate=FLAGS.learning_rate,
      num_train_steps=num_train_steps,
      num_warmup_steps=num_warmup_steps,
      use_tpu=FLAGS.use_tpu,
      use_one_hot_embeddings=FLAGS.use_tpu)

  # If TPU is not available, this will fall back to normal Estimator on CPU
  # or GPU.
  estimator = tf.contrib.tpu.TPUEstimator(  # 构建估计器对象
      use_tpu=FLAGS.use_tpu,
      model_fn=model_fn,
      config=run_config,
      train_batch_size=FLAGS.train_batch_size,
      eval_batch_size=FLAGS.eval_batch_size,
      predict_batch_size=FLAGS.predict_batch_size)

  if FLAGS.do_train:
    train_file = os.path.join(FLAGS.output_dir, "train.tf_record")  # record：记录，档案
    # 实现Input_example到Feature_example, TFRecord化
    file_based_convert_examples_to_features(
        train_examples, label_list, FLAGS.max_seq_length, tokenizer, train_file)
    tf.logging.info("***** Running training *****")
    tf.logging.info("  Num examples = %d", len(train_examples))
    tf.logging.info("  Batch size = %d", FLAGS.train_batch_size)
    tf.logging.info("  Num steps = %d", num_train_steps)
    # 调用此函数，完成：1.TFRecord to example 2.int64 to int32
    train_input_fn = file_based_input_fn_builder(
        input_file=train_file,
        seq_length=FLAGS.max_seq_length,
        is_training=True,
        drop_remainder=True)
    estimator.train(input_fn=train_input_fn, max_steps=num_train_steps)



  if FLAGS.do_eval:
    eval_examples = processor.get_dev_examples(FLAGS.data_dir)
    num_actual_eval_examples = len(eval_examples)
    if FLAGS.use_tpu:
      # 不为整数倍时，填充
      while len(eval_examples) % FLAGS.eval_batch_size != 0:
        eval_examples.append(PaddingInputExample())

    eval_file = os.path.join(FLAGS.output_dir, "eval.tf_record")
    file_based_convert_examples_to_features(
        eval_examples, label_list, FLAGS.max_seq_length, tokenizer, eval_file)

    tf.logging.info("***** Running evaluation *****")
    # 数量 =（实际的数量，填充的数量）
    tf.logging.info("  Num examples = %d (%d actual, %d padding)",
                    len(eval_examples), num_actual_eval_examples,
                    len(eval_examples) - num_actual_eval_examples)
    tf.logging.info("  Batch size = %d", FLAGS.eval_batch_size)

    # This tells the estimator to run through the entire set.
    eval_steps = None  # 遍历整个集合
    # However, if running eval on the TPU, you will need to specify the
    # number of steps.
    if FLAGS.use_tpu:
      # 假定使用TPU时，前面整除处理已经成功，将得到eval_steps为整数值
      assert len(eval_examples) % FLAGS.eval_batch_size == 0
      eval_steps = int(len(eval_examples) // FLAGS.eval_batch_size)
    # 如果使用tpu的话，删除剩余的部分（可能是无法整除的部分）
    eval_drop_remainder = True if FLAGS.use_tpu else False
    # 1.TFRecord to example 2.int64 to int32 为estimator提供输入
    eval_input_fn = file_based_input_fn_builder(
        input_file=eval_file,
        seq_length=FLAGS.max_seq_length,
        is_training=False,
        drop_remainder=eval_drop_remainder)

    result = estimator.evaluate(input_fn=eval_input_fn, steps=eval_steps)
    
	# 生成验证集评估指标数据
    output_eval_file = os.path.join(FLAGS.output_dir, "eval_results.txt")
    with tf.gfile.GFile(output_eval_file, "w") as writer:
      tf.logging.info("***** Eval results *****")
      for key in sorted(result.keys()):
        tf.logging.info("  %s = %s", key, str(result[key]))
        writer.write("%s = %s\n" % (key, str(result[key])))

  if FLAGS.do_predict:
    predict_examples = processor.get_test_examples(FLAGS.data_dir)
    num_actual_predict_examples = len(predict_examples)
    if FLAGS.use_tpu:
      while len(predict_examples) % FLAGS.predict_batch_size != 0:
        predict_examples.append(PaddingInputExample())

    predict_file = os.path.join(FLAGS.output_dir, "predict.tf_record")
    # 1.调用convert_single_example转化Input_example为Feature_example
    # 2.转换为TFRecord格式，便于大型数据处理
    file_based_convert_examples_to_features(predict_examples, label_list,
                                                                           
                                            FLAGS.max_seq_length, tokenizer,
                                            predict_file)

    tf.logging.info("***** Running prediction*****")
    tf.logging.info("  Num examples = %d (%d actual, %d padding)",
                    len(predict_examples), num_actual_predict_examples,
                    len(predict_examples) - num_actual_predict_examples)
    tf.logging.info("  Batch size = %d", FLAGS.predict_batch_size)

    predict_drop_remainder = True if FLAGS.use_tpu else False
    # 1.TFRecord to example 2.int64 to int32 为estimator提供输入
    predict_input_fn = file_based_input_fn_builder(
        input_file=predict_file,
        seq_length=FLAGS.max_seq_length,
        is_training=False,
        drop_remainder=predict_drop_remainder)

    result = estimator.predict(input_fn=predict_input_fn)
	# 写测试集预测结果文件
    output_predict_file = os.path.join(FLAGS.output_dir, "test_results.tsv")
    with tf.gfile.GFile(output_predict_file, "w") as writer:
      num_written_lines = 0
      tf.logging.info("***** Predict results *****")
      for (i, prediction) in enumerate(result):
        # 写入probabilities的键值对，比如二分类：有预测为0的一列，预测为1的一列
        probabilities = prediction["probabilities"]
        if i >= num_actual_predict_examples:
          break
        output_line = "\t".join(
            str(class_probability)
            # 例如：某一行class_probability为(0.98,0.02)->('0.98','0.02')
            for class_probability in probabilities) + "\n"
        writer.write(output_line)
        num_written_lines += 1
    assert num_written_lines == num_actual_predict_examples


if __name__ == "__main__":
  flags.mark_flag_as_required("data_dir")
  flags.mark_flag_as_required("task_name")
  flags.mark_flag_as_required("vocab_file")
  flags.mark_flag_as_required("bert_config_file")
  flags.mark_flag_as_required("output_dir")
  tf.app.run()