tf.estimator.Estimator

1.定义

tf.estimator.Estimator(model_fn=model_fn) #model_fn是一个方法

2.定义model_fn:

    def model_fn_builder(self, bert_config, num_labels, init_checkpoint):
        """
        :param bert_config:
        :param num_labels:
        :param init_checkpoint:
        :param learning_rate:
        :param num_train_steps:
        :param num_warmup_steps:
        :return:
        """
        def model_fn(features, labels, mode, params):
            """
　　　　　　　这4个参数必须这样定义，就算是不用某个参数，也要把它定义出来
            :param features: 是estimator传过来的feature
            :param labels: 数据标签
            :param mode: tf.estimator.TRAIN/tf.estimator.EVAL/tf.estimator.PREDICTION
            :param params:这个暂时没弄懂
            :return:
            """
            input_ids = features['input_ids']
            input_mask = features['input_mask']
            segment_ids = features['segment_ids']
            probabilities = self.creat_model(bert_config, input_ids, input_mask, segment_ids, num_labels) # 这里是重点，这里要定义模型和要取模型的什么值

            tvars = tf.trainable_variables()
            (assignment_map, initialized_variable_names) = modeling.get_assignment_map_from_checkpoint(tvars, init_checkpoint) # assignment_map是模型所有的变量字典，init_checkpoint为模型文件
            tf.train.init_from_checkpoint(init_checkpoint, assignment_map) # 加载模型

            output_spec = tf.estimator.EstimatorSpec(mode=mode, predictions=probabilities) # 应为上面已经从create_model中获取了我们要做什么op，获取什么值，prediction为op或值
            return output_spec

        return model_fn

def get_assignment_map_from_checkpoint(tvars, init_checkpoint):
  """Compute the union of the current variables and checkpoint variables."""
  assignment_map = {}
  initialized_variable_names = {}

  name_to_variable = collections.OrderedDict()
  for var in tvars:
    name = var.name
    m = re.match("^(.*):\d+$", name)
    if m is not None:
      name = m.group(1)
    name_to_variable[name] = var

  init_vars = tf.train.list_variables(init_checkpoint)

  assignment_map = collections.OrderedDict()
  for x in init_vars:
    (name, var) = (x[0], x[1])
    if name not in name_to_variable:
      continue
    assignment_map[name] = name
    initialized_variable_names[name] = 1
    initialized_variable_names[name + ":0"] = 1

  return (assignment_map, initialized_variable_names)

    def creat_model(self, bert_config, input_ids, input_mask, segment_ids, num_labels):
        """

        :param bert_config:
        :param input_ids:
        :param input_mask:
        :param segment_ids:
        :param num_labels:
        :return:
        """
        model = modeling.BertModel(
            config=bert_config,
            is_training=False,
            input_ids=input_ids,
            input_mask=input_mask,
            token_type_ids=segment_ids,
            use_one_hot_embeddings=False)

        output_layer = model.get_pooled_output()

        hidden_size = output_layer.shape[-1].value
　　　　
　　　　
　　　　# 获得已经训练好的值　　
        output_weights = tf.get_variable(
            "output_weights", [num_labels, hidden_size],
            initializer=tf.truncated_normal_initializer(stddev=0.02))

        output_bias = tf.get_variable(
            "output_bias", [num_labels], initializer=tf.zeros_initializer())

        logits = tf.matmul(output_layer, output_weights, transpose_b=True)
        logits = tf.nn.bias_add(logits, output_bias)
        probabilities = tf.nn.softmax(logits, axis=-1)

        return probabilities

2.使用estimator.predict

def predict(self, text_a, text_b):
    """

    :param text_a:
    :param text_b:
    :return:
    """

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

    input_ids, input_mask, segment_ids = self.convert_single_example(text_a, text_b)

    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)

    tf_example = tf.train.Example(features=tf.train.Features(feature=features)) # 将feature转换为example

    self.writer.write(tf_example.SerializeToString())# 序列化example，写入tfrecord文件

    result = self.estimator.predict(input_fn=self.predict_input_fn)

    def file_based_input_fn_builder(self):
        """

        :param examples:
        :return:
        """
        name_to_features = {
            "input_ids": tf.FixedLenFeature([MAX_SEQ_LENGTH], tf.int64),
            "input_mask": tf.FixedLenFeature([MAX_SEQ_LENGTH], tf.int64),
            "segment_ids": tf.FixedLenFeature([MAX_SEQ_LENGTH], tf.int64),
        }

        def decode_record(_examples, _name_to_feature):
            """

            :param _examples:
            :param _name_to_feature:
            :return:
            """

            return tf.parse_single_example(_examples, _name_to_feature)

        def input_fn():
            """

            :param params:
            :return:
            """
            d = tf.data.TFRecordDataset(self.predict_file) # 读取TFRecord文件
            d = d.apply(
                tf.data.experimental.map_and_batch(
                    lambda record: decode_record(record, name_to_features), # 将序列化的feature映射到字典上
                    batch_size=1,
                    drop_remainder=False))

            return d # 这里返回的值会进入到定义estimator时的model_fn中，model_fn中的feature是d.get_next()的结果

        return input_fn

1