AI_深度学习_NLP_BERT基于词向量的文本分析模型构建和特征提取
作者:田超凡
时间:20190823
实现原理概述:
对已经构建的模型做模型训练前的准备工作,包括但不限于
1.文本数据的提取,编码转换和清洗
2.对提取的文本数据进行模型参数封装,基于CNN的卷积核构建原理来构建模型训练参数
3.定义模型训练的机器学习算法,此处只是做文本分析和文本预测,涉及到小范围=>大范围迭代检索,所以采用算法是线性回归+SoftMax聚合算法,采用的训练网络是CNN和BP/RBF
4.代码封装线性回归算法和CNN-BP/RBF网络结构层
5.代入已经封装好的模型训练测试数据并进行池化和维度转换,输出提取的符合模型训练条件的模型参数结果集
6.接下来重点还是模型数据的提取和转换、算法和网络的定义和封装、卷积核大小的定义和迭代输出、文本特征提取(基于词向量)
7.在网络层,使用LSTM或者BP都可以实现,但是推荐BP,理由是LSTM是链式执行过程,必须基于线性回归逐一迭代,影响训练性能
8.在训练时的训练方式可以选择CPU默认/GPU,推荐使用GPU+CUNA10的方式作为训练方式
# coding=utf-8
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
# nlp-bert 进行文本分析预测训练测试数据集的读取和清洗
import os
import modeling
import optimization
import tensorflow as tf
from tensorflow.estimator import RunConfig, Estimator, EstimatorSpec
os.environ["CUDA_VISIBLE_DEVICES"] = "0"
flags = tf.flags
FLAGS = flags.FLAGS
## Required parameters
flags.DEFINE_string(
"bert_config_file", 'pretrain_model/bert_config.json',
"The config json file corresponding to the pre-trained BERT model. "
"This specifies the model architecture.")
flags.DEFINE_string(
"input_file", "data_ch/dev.tf_record",
"Input TF example files (can be a glob or comma separated).")
flags.DEFINE_string(
"dev_file", "data_ch/dev.tf_record",
"Input TF example files (can be a glob or comma separated).")
flags.DEFINE_string(
"output_dir", 'output',
"The output directory where the model checkpoints will be written.")
# Other parameters
flags.DEFINE_string(
"init_checkpoint", 'pretrain_model/bert_model.ckpt',
"Initial checkpoint (usually from a pre-trained BERT model).")
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. Must match data generation.")
flags.DEFINE_integer(
"max_predictions_per_seq", 20,
"Maximum number of masked LM predictions per sequence. "
"Must match data generation.")
flags.DEFINE_bool("do_train", True, "Whether to run training.")
flags.DEFINE_bool("do_eval", True, "Whether to run eval on the dev set.")
flags.DEFINE_integer("train_batch_size", 1, "Total batch size for training.")
flags.DEFINE_integer("eval_batch_size", 1, "Total batch size for eval.")
flags.DEFINE_float("learning_rate", 5e-5, "The initial learning rate for Adam.")
flags.DEFINE_integer("num_train_steps", 100000, "Number of training steps.")
flags.DEFINE_integer("num_warmup_steps", 10000, "Number of warmup steps.")
flags.DEFINE_integer("save_checkpoints_steps", 10000,
"How often to save the model checkpoint.")
flags.DEFINE_integer("iterations_per_loop", 1000,
"How many steps to make in each estimator call.")
flags.DEFINE_integer("max_eval_steps", 1000, "Maximum number of eval steps.")
flags.DEFINE_bool("use_tpu", False, "Whether to use TPU or GPU/CPU.")
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.")
# TODO TCF 定义一个模型构造器
def model_fn_builder(bert_config, init_checkpoint: object, learning_rate,
num_train_steps, num_warmup_steps, use_tpu,
use_one_hot_embeddings):
"""Returns `model_fn` closure for TPUEstimator."""
# TODO TCF 定义一个模型虚拟化对象
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"] # 经过遮蔽处理后的字的id
input_mask = features["input_mask"] # 输入字的权重mask,有字的地方对应1,没字的地方对应0,0的位置处不计算损失
segment_ids = features["segment_ids"] # 每个字属于哪句话
masked_lm_positions = features["masked_lm_positions"] # 被遮蔽掉的字的位置
masked_lm_ids = features["masked_lm_ids"] # 被遮蔽掉的字的id
masked_lm_weights = features["masked_lm_weights"] # 被遮蔽掉的字的权重mask,有遮蔽的地方为1,padding的地方为0,0的位置不计算损失
next_sentence_labels = features["next_sentence_labels"] # 连续还是不连续标签(1表示不连续,0表示连续)
is_training = (mode == tf.estimator.ModeKeys.TRAIN)
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)
(masked_lm_loss,
masked_lm_example_loss, masked_lm_log_probs) = get_masked_lm_output(
bert_config, model.get_sequence_output(), model.get_embedding_table(),
masked_lm_positions, masked_lm_ids, masked_lm_weights) # 预测出被遮蔽掉的字的id,这里有一个损失
(next_sentence_loss, next_sentence_example_loss,
next_sentence_log_probs) = get_next_sentence_output(
bert_config, model.get_pooled_output(), next_sentence_labels) # 预测出句子是否相邻,这里有一个损失
total_loss = masked_lm_loss + next_sentence_loss # 将两个损失相加
tvars = tf.trainable_variables() # 把所有可训练的参数拿出来
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():
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,
init_string)
output_spec = None
if mode == tf.estimator.ModeKeys.TRAIN:
# TODO TCF 基于GPU的模型训练
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,
# loss=total_loss,
# train_op=train_op,
# scaffold_fn=scaffold_fn)
# TODO TCF TensorFlow内置函数,基于LSTM将模型训练内核转换为GPU
output_spec = EstimatorSpec(mode=mode, loss=total_loss, train_op=train_op) # 替换成gpu
elif mode == tf.estimator.ModeKeys.EVAL:
# 模型训练结果不准确,可以使用tf.densor调整卷积核大小实现类似效果
def metric_fn(masked_lm_example_loss, masked_lm_log_probs, masked_lm_ids,
masked_lm_weights, next_sentence_example_loss,
next_sentence_log_probs, next_sentence_labels):
"""Computes the loss and accuracy of the model."""
masked_lm_log_probs = tf.reshape(masked_lm_log_probs,
[-1, masked_lm_log_probs.shape[-1]])
masked_lm_predictions = tf.argmax(
masked_lm_log_probs, axis=-1, output_type=tf.int32)
masked_lm_example_loss = tf.reshape(masked_lm_example_loss, [-1])
masked_lm_ids = tf.reshape(masked_lm_ids, [-1])
masked_lm_weights = tf.reshape(masked_lm_weights, [-1])
masked_lm_accuracy = tf.metrics.accuracy(
labels=masked_lm_ids,
predictions=masked_lm_predictions,
weights=masked_lm_weights) # 得到预测被屏蔽的字的准确率
masked_lm_mean_loss = tf.metrics.mean(
values=masked_lm_example_loss, weights=masked_lm_weights) # 得到平均每个样本预测被屏蔽的字的损失
next_sentence_log_probs = tf.reshape(
next_sentence_log_probs, [-1, next_sentence_log_probs.shape[-1]])
next_sentence_predictions = tf.argmax(
next_sentence_log_probs, axis=-1, output_type=tf.int32)
next_sentence_labels = tf.reshape(next_sentence_labels, [-1])
next_sentence_accuracy = tf.metrics.accuracy(
labels=next_sentence_labels, predictions=next_sentence_predictions) # 得到预测句子是否相邻的准确率
next_sentence_mean_loss = tf.metrics.mean(
values=next_sentence_example_loss) # 得到平均每个样本预测句子是否相邻的损失
return {
"masked_lm_accuracy": masked_lm_accuracy,
"masked_lm_loss": masked_lm_mean_loss,
"next_sentence_accuracy": next_sentence_accuracy,
"next_sentence_loss": next_sentence_mean_loss,
}
eval_metrics = (metric_fn, [
masked_lm_example_loss, masked_lm_log_probs, masked_lm_ids,
masked_lm_weights, next_sentence_example_loss,
next_sentence_log_probs, next_sentence_labels
])
# output_spec = tf.contrib.tpu.TPUEstimatorSpec(
# mode=mode,
# loss=total_loss,
# eval_metrics=eval_metrics,
# scaffold_fn=scaffold_fn)
output_spec = EstimatorSpec(
mode=mode,
loss=total_loss,
eval_metric_ops=eval_metrics)
else:
raise ValueError("Only TRAIN and EVAL modes are supported: %s" % (mode))
return output_spec
return model_fn
# TODO TCF 输出卷积层的模型训练结果
def get_masked_lm_output(bert_config, input_tensor, output_weights, positions,
label_ids, label_weights):
"""Get loss and log probs for the masked LM."""
input_tensor = gather_indexes(input_tensor, positions) # 得到被屏蔽的字抽取出来的特征向量
with tf.variable_scope("cls/predictions"):
# We apply one more non-linear transformation before the output layer.
# This matrix is not used after pre-training.
with tf.variable_scope("transform"):
input_tensor = tf.layers.dense(
input_tensor,
units=bert_config.hidden_size,
activation=modeling.get_activation(bert_config.hidden_act),
kernel_initializer=modeling.create_initializer(
bert_config.initializer_range)) # 将被屏蔽的字抽取出来的特征向量接一层全连接,激活函数用的是gelu激活函数
input_tensor = modeling.layer_norm(input_tensor) # 进行层归一化
# The output weights are the same as the input embeddings, but there is
# an output-only bias for each token.
output_bias = tf.get_variable(
"output_bias",
shape=[bert_config.vocab_size],
initializer=tf.zeros_initializer())
logits = tf.matmul(input_tensor, output_weights, transpose_b=True)
logits = tf.nn.bias_add(logits, output_bias)
log_probs = tf.nn.log_softmax(logits, axis=-1) # 得到被屏蔽的字预测出来属于字典中每个字的log概率
label_ids = tf.reshape(label_ids, [-1])
label_weights = tf.reshape(label_weights, [-1]) # label_weights指的是权重mask,被屏蔽掉的字中用padding的为0,不计算损失
one_hot_labels = tf.one_hot(
label_ids, depth=bert_config.vocab_size, dtype=tf.float32)
# The `positions` tensor might be zero-padded (if the sequence is too
# short to have the maximum number of predictions). The `label_weights`
# tensor has a value of 1.0 for every real prediction and 0.0 for the
# padding predictions.
per_example_loss = -tf.reduce_sum(log_probs * one_hot_labels, axis=[-1]) # 得到每个预测被屏蔽的字的损失
numerator = tf.reduce_sum(label_weights * per_example_loss) # 被屏蔽掉的字中用padding的地方不计算损失,最后得到预测屏蔽字的总损失
denominator = tf.reduce_sum(label_weights) + 1e-5 # 求被屏蔽掉的字中不是用padding的一共多少字
loss = numerator / denominator # 得到预测屏蔽字的平均损失
return (loss, per_example_loss, log_probs)
def get_next_sentence_output(bert_config, input_tensor, labels):
"""Get loss and log probs for the next sentence prediction."""
# Simple binary classification. Note that 0 is "next sentence" and 1 is
# "random sentence". This weight matrix is not used after pre-training.
with tf.variable_scope("cls/seq_relationship"):
output_weights = tf.get_variable(
"output_weights",
shape=[2, bert_config.hidden_size],
initializer=modeling.create_initializer(bert_config.initializer_range))
output_bias = tf.get_variable(
"output_bias", shape=[2], initializer=tf.zeros_initializer())
logits = tf.matmul(input_tensor, output_weights, transpose_b=True)
# TODO TCF 为当前训练网络层加入Logitic回归函数去构建BP网络
logits = tf.nn.bias_add(logits, output_bias)
log_probs = tf.nn.log_softmax(logits, axis=-1) # 得到预测句子属于相邻还是不相邻的log概率分布
labels = tf.reshape(labels, [-1])
# TODO TCF RNN OneHot网络构建概率泊松分布
one_hot_labels = tf.one_hot(labels, depth=2, dtype=tf.float32)
per_example_loss = -tf.reduce_sum(one_hot_labels * log_probs, axis=-1) # 得到预测句子是否相邻的每个序列的损失
loss = tf.reduce_mean(per_example_loss) # 得到预测句子是否相邻的平均损失
return (loss, per_example_loss, log_probs)
def gather_indexes(sequence_tensor, positions):
"""Gathers the vectors at the specific positions over a minibatch."""
sequence_shape = modeling.get_shape_list(sequence_tensor, expected_rank=3)
batch_size = sequence_shape[0]
seq_length = sequence_shape[1]
width = sequence_shape[2]
# TODO TCF 模型训练率约为22.91%
flat_offsets = tf.reshape(
tf.range(0, batch_size, dtype=tf.int32) * seq_length, [-1, 1])
flat_positions = tf.reshape(positions + flat_offsets, [-1])
flat_sequence_tensor = tf.reshape(sequence_tensor,
[batch_size * seq_length, width])
output_tensor = tf.gather(flat_sequence_tensor, flat_positions)
return output_tensor
def input_fn_builder(input_files,
max_seq_length,
max_predictions_per_seq,
is_training,
num_cpu_threads=4):
"""Creates an `input_fn` closure to be passed to TPUEstimator."""
# TODO TCF 固定模型卷积层的卷积核通道数
def input_fn(params):
"""The actual input function."""
if is_training:
batch_size = params["train_batch_size"]
else:
batch_size = params['eval_batch_size']
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),
"masked_lm_positions":
tf.FixedLenFeature([max_predictions_per_seq], tf.int64),
"masked_lm_ids":
tf.FixedLenFeature([max_predictions_per_seq], tf.int64),
"masked_lm_weights":
tf.FixedLenFeature([max_predictions_per_seq], tf.float32),
"next_sentence_labels":
tf.FixedLenFeature([1], tf.int64),
}
# For training, we want a lot of parallel reading and shuffling.
# For eval, we want no shuffling and parallel reading doesn't matter.
# TODO TCF 如果是模型训练阶段,则需要执行一些特征处理
# TODO TCF 如果是模型训练结束阶段,则停止和模型训练结果不匹配部分的数据集的特征处理
if is_training:
d = tf.data.Dataset.from_tensor_slices(tf.constant(input_files))
d = d.repeat()
# TODO TCF 模型特征提取和转换,特征提取数量上限即为读取的测试数据集文件字节
d = d.shuffle(buffer_size=len(input_files))
# `cycle_length` is the number of parallel files that get read.
# TODO TCF 指定需要迭代卷积层的次数
cycle_length = min(num_cpu_threads, len(input_files))
# `sloppy` mode means that the interleaving is not exact. This adds
# even more randomness to the training pipeline.
# TODO TCF 应用卷积层的卷积核、模型训练集和特征数进行模型转换
d = d.apply(
tf.contrib.data.parallel_interleave(
tf.data.TFRecordDataset,
sloppy=is_training,
cycle_length=cycle_length))
d = d.shuffle(buffer_size=100)
else:
# TODO TCF 非模型训练阶段,记录测试数据集即可,不再进行迭代特征提取和转换
d = tf.data.TFRecordDataset(input_files)
# Since we evaluate for a fixed number of steps we don't want to encounter
# out-of-range exceptions.
d = d.repeat()
# We must `drop_remainder` on training because the TPU requires fixed
# size dimensions. For eval, we assume we are evaluating on the CPU or GPU
# and we *don't* want to drop the remainder, otherwise we wont cover
# every sample.
d = d.apply(
tf.contrib.data.map_and_batch(
lambda record: _decode_record(record, name_to_features),
batch_size=batch_size,
num_parallel_batches=num_cpu_threads,
drop_remainder=True))
return d
return input_fn
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.
# TODO TCF 如果是调用Example模块进行数据提取,则必须是INT64位编码,如果是调用TPU进行数据提取,必须是INT32位编码
for name in list(example.keys()):
t = example[name]
if t.dtype == tf.int64:
t = tf.to_int32(t)
example[name] = t
return example
# TODO TCF 入口,主要做以下几件事:
# TODO TCF 对已经构建的模型做模型训练前的准备工作,包括但不限于
# TODO TCF 1.文本数据的提取,编码转换和清洗
# TODO TCF 2.对提取的文本数据进行模型参数封装,基于CNN的卷积核构建原理来构建模型训练参数
# TODO TCF 3.定义模型训练的机器学习算法,此处只是做文本分析和文本预测,涉及到小范围=>大范围迭代检索,
# TODO TCF 所以采用算法是线性回归+SoftMax聚合算法,采用的训练网络是CNN和BP/RBF
# TODO TCF 4.代码封装线性回归算法和CNN-BP/RBF网络结构层
# TODO TCF 5.代入已经封装好的模型训练测试数据并进行池化和维度转换,输出提取的符合模型训练条件的模型参数结果集
# TODO TCF 重点还是模型数据的提取和转换、算法和网络的定义和封装、卷积核大小的定义和迭代输出、文本特征提取(基于词向量)
# TODO TCF 在网络层,使用LSTM或者BP都可以实现,但是推荐BP,理由是LSTM是链式执行过程,必须基于线性回归逐一迭代,影响训练性能
# TODO TCF 在训练时的训练方式可以选择CPU默认/GPU,推荐使用GPU+CUNA10的方式作为训练方式
def main(_):
tf.logging.set_verbosity(tf.logging.INFO)
# TODO TCF 如果模型训练中断,则直接抛出异常结束
if not FLAGS.do_train and not FLAGS.do_eval:
raise ValueError("At least one of `do_train` or `do_eval` must be True.")
# TODO TCF 基于NLP-BERT去读取训练数据文本并进行训练数据转换和格式化工作
bert_config = modeling.BertConfig.from_json_file(FLAGS.bert_config_file)
# TODO TCF 如果定义的输出目录不存在,则先创建输出文件存放目录
tf.gfile.MakeDirs(FLAGS.output_dir)
input_files, dev_files = [], []
for input_pattern in FLAGS.input_file.split(","):
input_files.extend(tf.gfile.Glob(input_pattern))
for input_pattern in FLAGS.dev_file.split(","):
dev_files.extend(tf.gfile.Glob(input_pattern))
tf.logging.info("*** Input Files ***")
for input_file in input_files:
tf.logging.info(" %s" % input_file)
tpu_cluster_resolver = None
if FLAGS.use_tpu and FLAGS.tpu_name:
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
# run_config = tf.contrib.tpu.RunConfig(
# 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(
# iterations_per_loop=FLAGS.iterations_per_loop,
# num_shards=FLAGS.num_tpu_cores,
# per_host_input_for_training=is_per_host))
model_fn = model_fn_builder(
bert_config=bert_config,
init_checkpoint=FLAGS.init_checkpoint,
learning_rate=FLAGS.learning_rate,
num_train_steps=FLAGS.num_train_steps,
num_warmup_steps=FLAGS.num_warmup_steps,
use_tpu=FLAGS.use_tpu,
use_one_hot_embeddings=FLAGS.use_tpu) # 返回的是一个函数
# TODO TCF 以下部分是当训练过程中如果TPU方式出现问题就切换回默认的CPU处理方式,相当于容灾操作。
# If TPU is not available, this will fall back to normal Estimator on CPU
# or GPU.
# estimator = tf.contrib.tpu.TPUEstimator(
# 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)# 使用优化器,可以考虑换成GPU格式的优化器
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
config.gpu_options.per_process_gpu_memory_fraction = 0.8
config.log_device_placement = False
# TODO TCF model_fn作为构建Elastic的核心处理函数,需要单独封装
estimator = Estimator(model_fn=model_fn, config=RunConfig(session_config=config, model_dir=FLAGS.output_dir,
save_checkpoints_steps=FLAGS.save_checkpoints_steps),
params={'train_batch_size': FLAGS.train_batch_size,
'eval_batch_size': FLAGS.eval_batch_size}) # GPU版优化器
if FLAGS.do_train:
tf.logging.info("***** Running training *****")
tf.logging.info(" Batch size = %d", FLAGS.train_batch_size)
train_input_fn = input_fn_builder(
input_files=input_files,
max_seq_length=FLAGS.max_seq_length,
max_predictions_per_seq=FLAGS.max_predictions_per_seq,
is_training=True) # 返回一个函数
# TODO TCF 通知模型训练层模型参数已构建完毕
estimator.train(input_fn=train_input_fn, max_steps=FLAGS.num_train_steps)
if FLAGS.do_eval:
tf.logging.info("***** Running evaluation *****")
tf.logging.info(" Batch size = %d", FLAGS.eval_batch_size)
eval_input_fn = input_fn_builder(
input_files=dev_files,
max_seq_length=FLAGS.max_seq_length,
max_predictions_per_seq=FLAGS.max_predictions_per_seq,
is_training=False)
result = estimator.evaluate(
input_fn=eval_input_fn, steps=FLAGS.max_eval_steps)
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 __name__ == "__main__":
# flags.mark_flag_as_required("input_file")
# flags.mark_flag_as_required("bert_config_file")
# flags.mark_flag_as_required("output_dir")
tf.app.run()