run_classifier_imbalance.py

# coding=utf-8
# Copyright 2018 The Google AI Language Team Authors.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""BERT finetuning runner."""

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

import os
import collections
import csv
import pandas as pd
from sklearn.metrics import classification_report
import numpy as np
import tensorflow as tf
import tf_metrics
from sklearn.utils.class_weight import compute_class_weight
from tensorflow.contrib import *
from tensorflow.contrib.distribute import AllReduceCrossDeviceOps

import custom_optimization
import modeling
import optimization
import tokenization

flags = tf.flags

FLAGS = flags.FLAGS

## Required parameters
flags.DEFINE_string(
    "data_dir", None,
    "The input data dir. Should contain the .tsv files (or other data files) "
    "for the task.")

flags.DEFINE_string(
    "bert_config_file", None,
    "The config json file corresponding to the pre-trained BERT model. "
    "This specifies the model architecture.")

flags.DEFINE_string("task_name", None, "The name of the task to train.")

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

flags.DEFINE_string(
    "output_dir", None,
    "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).")

# 是否进行字符串转小写操作
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_early_stopping", True, "Whether to do early stopping")

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

flags.DEFINE_string("train_file", "train.tsv", "train file name in data_dir.")
flags.DEFINE_string("eval_file", "dev.tsv", "evaluate file name in data_dir.")
flags.DEFINE_string("predict_file", "test.tsv", "test file name in data_dir.")

flags.DEFINE_string("classifier_mode", "binary", "binary or multi-class for classifier.")

# 初始学习率 for Adam 优化器
flags.DEFINE_float("learning_rate", 5e-5, "The initial learning rate for Adam.")

# 总训练轮数
flags.DEFINE_float("num_train_epochs", 3.0,
                   "Total number of training epochs to perform.")

flags.DEFINE_float(
    "warmup_proportion", 0.1,
    "Proportion of training to perform linear learning rate warmup for. "
    "E.g., 0.1 = 10% of training.")

flags.DEFINE_integer("save_checkpoints_steps", 1000,
                     "How often to save the model checkpoint.")

flags.DEFINE_integer("iterations_per_loop", 1000,
                     "How many steps to make in each estimator call.")

# early stopping parameters
flags.DEFINE_integer("min_steps", 1000, "The minimum steps for early stopping")
flags.DEFINE_integer("max_steps_without_increase", 5000,
                     "How many steps to make in each estimator call.")
flags.DEFINE_integer("max_steps_without_decrease", 5000,
                     "How many steps to make in each estimator call.")

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

flags.DEFINE_string("class_weight", None, "Whether to use class weight(balanced or None).")

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.")

flags.DEFINE_integer(
    "num_gpu_cores", 2,
    "OTotal number of GPU cores to use.")
flags.DEFINE_bool("use_fp16", False, "Whether to use fp16.")


class InputExample(object):
    """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):
    """Fake example so the num input examples is a multiple of the batch size.

    When running eval/predict on the TPU, we need to pad the number of examples
    to be a multiple of the batch size, because the TPU requires a fixed batch
    size. The alternative is to drop the last batch, which is bad because it means
    the entire output data won't be generated.

    We use this class instead of `None` because treating `None` as padding
    battches could cause silent errors.
    """


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

    def __init__(self,
                 input_ids,
                 input_mask,
                 segment_ids,
                 label_id,
                 is_real_example=True,
                 label_weights=None):
        if label_weights is None:
            label_weights = [1.0]
        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
        self.label_weights = label_weights


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='"'):
        """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 XnliProcessor(DataProcessor):
    """Processor for the XNLI data set."""

    def __init__(self):
        self.language = "zh"

    def get_train_examples(self, data_dir):
        """See base class."""
        lines = self._read_tsv(
            os.path.join(data_dir, "multinli",
                         "multinli.train.%s.tsv" % self.language))
        examples = []
        for (i, line) in enumerate(lines):
            if i == 0:
                continue
            guid = "train-%d" % (i)
            text_a = tokenization.convert_to_unicode(line[0])
            text_b = tokenization.convert_to_unicode(line[1])
            label = tokenization.convert_to_unicode(line[2])
            if label == tokenization.convert_to_unicode("contradictory"):
                label = tokenization.convert_to_unicode("contradiction")
            examples.append(
                InputExample(guid=guid, text_a=text_a, text_b=text_b, label=label))
        return examples

    def get_dev_examples(self, data_dir):
        """See base class."""
        lines = self._read_tsv(os.path.join(data_dir, "xnli.dev.tsv"))
        examples = []
        for (i, line) in enumerate(lines):
            if i == 0:
                continue
            guid = "dev-%d" % (i)
            language = tokenization.convert_to_unicode(line[0])
            if language != tokenization.convert_to_unicode(self.language):
                continue
            text_a = tokenization.convert_to_unicode(line[6])
            text_b = tokenization.convert_to_unicode(line[7])
            label = tokenization.convert_to_unicode(line[1])
            examples.append(
                InputExample(guid=guid, text_a=text_a, text_b=text_b, label=label))
        return examples

    def get_labels(self):
        """See base class."""
        return ["contradiction", "entailment", "neutral"]


class MnliProcessor(DataProcessor):
    """Processor for the MultiNLI 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_matched.tsv")),
            "dev_matched")

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

    def get_labels(self):
        """See base class."""
        return ["contradiction", "entailment", "neutral"]

    def _create_examples(self, lines, set_type):
        """Creates examples for the training and dev sets."""
        examples = []
        for (i, line) in enumerate(lines):
            if i == 0:
                continue
            guid = "%s-%s" % (set_type, tokenization.convert_to_unicode(line[0]))
            text_a = tokenization.convert_to_unicode(line[8])
            text_b = tokenization.convert_to_unicode(line[9])
            if set_type == "test":
                label = "contradiction"
            else:
                label = tokenization.convert_to_unicode(line[-1])
            examples.append(
                InputExample(guid=guid, text_a=text_a, text_b=text_b, label=label))
        return examples


class MrpcProcessor(DataProcessor):
    """Processor for the MRPC 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, FLAGS.train_file)), "train")

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

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

    def get_labels(self):
        """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):
            if i == 0:
                continue
            guid = "%s-%s" % (set_type, i)
            text_a = tokenization.convert_to_unicode(line[3])
            text_b = tokenization.convert_to_unicode(line[4])
            if set_type == "test":
                label = "0"
            else:
                label = tokenization.convert_to_unicode(line[0])
            examples.append(
                InputExample(guid=guid, text_a=text_a, text_b=text_b, label=label))
        return examples


class ColaProcessor(DataProcessor):
    """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, FLAGS.train_file)), "train")

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

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

    def get_labels(self):
        """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
            if set_type == "test" and i == 0:
                continue
            guid = "%s-%s" % (set_type, i)
            if set_type == "test":
                text_a = tokenization.convert_to_unicode(line[1])
                label = "0"
            else:
                text_a = tokenization.convert_to_unicode(line[3])
                label = tokenization.convert_to_unicode(line[1])
            examples.append(
                InputExample(guid=guid, text_a=text_a, text_b=None, label=label))
        return examples


class SedProcessor(DataProcessor):
    """Processor for the sentence-level error detection(SED) 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, FLAGS.train_file)), "train")

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

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

    def get_labels(self):
        """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
            if set_type == "test" and i == 0:
                continue
            guid = "%s-%s" % (set_type, i)
            if set_type == "test":
                text_a = tokenization.convert_to_unicode(line[1])
                label = "0"
            else:
                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


class TopicRelProcessor(DataProcessor):
    """Processor for the sentence-level error detection(SED) 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, FLAGS.train_file)), "train")

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

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

    def get_labels(self):
        """See base class."""
        return ["0.0", "1.0", "2.0", "3.0"]

    def _create_examples(self, lines, set_type):
        """Creates examples for the training and dev sets."""
        examples = []
        for (i, line) in enumerate(lines):
            # ignore first header line
            if i == 0:
                continue
            if i <= 3:
                print("i={}, line={}".format(i, line))
            guid = "%s-%s" % (set_type, i)
            if set_type == "test":
                text_a = tokenization.convert_to_unicode(line[0])
                label = "3.0"
            else:
                text_a = tokenization.convert_to_unicode(line[0])
                label = tokenization.convert_to_unicode(line[1])
            examples.append(InputExample(guid=guid, text_a=text_a, text_b=None, label=label))
        return examples


class GarbledSentsProcessor(DataProcessor):
    """Processor for the garbled sentence data set."""

    def get_train_examples(self, data_dir):
        """See base class."""
        return self._create_examples(
            self._read_tsv(os.path.join(data_dir, FLAGS.train_file)), "train")

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

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

    def get_labels(self):
        """See base class."""
        return ["0", "1"]

    def get_class_weight(self, y, class_weight):
        classes = self.get_labels()
        return compute_class_weight(class_weight, np.array(classes), y)

    def creat_example(self, text, label="0", set_type="test", i=0):
        guid = "%s-%s" % (set_type, i)
        if set_type == "test":
            text_a = tokenization.convert_to_unicode(text)
            label = "0"
        else:
            text_a = tokenization.convert_to_unicode(text)
            label = tokenization.convert_to_unicode(label)
        return InputExample(guid=guid, text_a=text_a, text_b=None, label=label)

    def extract_features(self, sentences, max_seq_length, tokenizer, class_weight):
        label_list = self.get_labels()
        examples = []
        for sent in sentences:
            example = self.creat_example(sent)
            examples.append(example)
        features = convert_examples_to_features(examples, label_list, max_seq_length, tokenizer, class_weight)
        return features

    def _create_examples(self, lines, set_type):
        """Creates examples for the training and dev sets."""
        examples = []
        # labels = []
        for (i, line) in enumerate(lines):
            # ignore first header line
            if i == 0:
                continue
            if i <= 3:
                print("i={}, line={}".format(i, line))
            guid = "%s-%s" % (set_type, i)
            if set_type == "test":
                text_a = tokenization.convert_to_unicode(line[0])
                label = "0"
            else:
                text_a = tokenization.convert_to_unicode(line[0])
                label = tokenization.convert_to_unicode(line[1])
            examples.append(InputExample(guid=guid, text_a=text_a, text_b=None, label=label))
            # labels.append(label)
        # if set_type == "train":
        #     self.class_weight_ = self.get_class_weight(labels, "balanced")
        # else:
        #     self.class_weight_ = self.get_class_weight(labels, None)
        # print("set_type={}, class_weight_={}".format(set_type, self.class_weight_))
        return examples


class DiscProcessor(DataProcessor):
    """Processor for the Discourse Identifier data set."""

    def get_train_examples(self, data_dir):
        """See base class."""
        return self._create_examples(
            self._read_tsv(os.path.join(data_dir, FLAGS.train_file)), "train")

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

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

    def get_labels(self):
        """See base class."""
        return [str(i) for i in range(13)]

    def _create_examples(self, lines, set_type):
        """Creates examples for the training and dev sets."""
        examples = []
        for (i, line) in enumerate(lines):
            if i == 0:
                continue
            if i <= 3:
                print("i={}, line={}".format(i, line))
            guid = "%s-%s" % (set_type, i)
            if set_type == "test":
                text_a = tokenization.convert_to_unicode(line[1])
                label = "0"
            else:
                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 get_examples_class_weight(examples, class_weight, classes):
    """

    :param examples:
    :param class_weight: dict or "balanced"
    :param classes: list
    :return: list
    """
    y = [i.label for i in examples if i.label and not isinstance(i, PaddingInputExample)]
    # print("class_weight={}\nclasses={}\ny={}".format(class_weight, classes, y))
    return compute_class_weight(class_weight, np.unique(y), np.array(y))


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

    if isinstance(example, PaddingInputExample):
        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_weights=[1.0] * len(class_weight),
        )
    # 构建标签文本-索引id
    label_map = {}
    for (i, label) in enumerate(label_list):
        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)
    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
    #
    # Where "type_ids" are used to indicate whether this is the first
    # sequence or the second sequence. The embedding vectors for `type=0` and
    # `type=1` were learned during pre-training and are added to the wordpiece
    # embedding vector (and position vector). This is not *strictly* necessary
    # since the [SEP] token unambiguously separates the sequences, but it makes
    # it easier for the model to learn the concept of sequences.
    #
    # For classification tasks, the first vector (corresponding to [CLS]) is
    # used as the "sentence vector". Note that this only makes sense because
    # the entire model is fine-tuned.
    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)

    input_ids = tokenizer.convert_tokens_to_ids(tokens)

    # The mask has 1 for real tokens and 0 for padding tokens. Only real
    # tokens are attended to.
    input_mask = [1] * len(input_ids)

    # 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

    # 根据标签文本转换为标签索引id， 所以输入的都是标签列表的(get_labels()函数)索引id，最后预测argmax也是索引id
    label_id = label_map[example.label]
    label_weights = class_weight

    # 打印日志
    if ex_index < 3:
        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))
        tf.logging.info("label_weights={}".format(label_weights))

    # 转换成InputFeatures对象
    feature = InputFeatures(
        input_ids=input_ids,
        input_mask=input_mask,
        segment_ids=segment_ids,
        label_id=label_id,
        is_real_example=True,
        label_weights=label_weights,
    )
    return feature


def file_based_convert_examples_to_features(
        examples, label_list, max_seq_length, tokenizer, output_file, is_training, class_weight):
    """Convert a set of `InputExample`s to a TFRecord file."""

    writer = tf.python_io.TFRecordWriter(output_file)

    if is_training and class_weight == "balanced":
        class_weight = get_examples_class_weight(examples, class_weight, label_list)
    else:
        class_weight = [1.0] * len(label_list)

    for (ex_index, example) in enumerate(examples):
        if ex_index % 10000 == 0:
            tf.logging.info("Writing example %d of %d" % (ex_index, len(examples)))

        feature = convert_single_example(ex_index, example, label_list,
                                         max_seq_length, tokenizer, class_weight)

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

        features = collections.OrderedDict()
        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["label_weights"] = tf.train.Feature(
            float_list=tf.train.FloatList(value=feature.label_weights))
        features["is_real_example"] = create_int_feature(
            [int(feature.is_real_example)])

        tf_example = tf.train.Example(features=tf.train.Features(feature=features))
        writer.write(tf_example.SerializeToString())
    writer.close()


def file_based_input_fn_builder(input_file, seq_length, is_training,
                                drop_remainder, num_labels):
    """
    Creates an `input_fn` closure to be passed to TPUEstimator.
    基于文件的数据输入函数.

    """

    # 特征列匹配
    name_to_features = {
        "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),
        "label_weights": tf.FixedLenFeature([num_labels], tf.float32),
        "is_real_example": tf.FixedLenFeature([], tf.int64),
    }

    def _decode_record(record, name_to_features):
        """
        Decodes a record to a TensorFlow example.
        解析tf.record文件数据
        """
        # 解析单个样本
        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.
        # 将样本中int64数据类型转换为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.
        生成batch样本.

        """
        batch_size = params["batch_size"]

        # For training, we want a lot of parallel reading and shuffling.
        # For eval, we want no shuffling and parallel reading doesn't matter.
        d = tf.data.TFRecordDataset(input_file)
        if is_training:
            d = d.repeat()
            d = d.shuffle(buffer_size=100)

        # 应用批处理batch size
        d = d.apply(
            tf.contrib.data.map_and_batch(
                lambda record: _decode_record(record, name_to_features),
                batch_size=batch_size,
                drop_remainder=drop_remainder))

        return d

    return input_fn


def serving_input_receiver_fn():
    """
    Serving input_fn that builds features from placeholders.

    Returns
    -------
    tf.estimator.export.ServingInputReceiver
    """
    input_ids = tf.placeholder(dtype=tf.int64, shape=[None, None], name='input_ids')
    input_mask = tf.placeholder(dtype=tf.int64, shape=[None, None], name='input_mask')
    segment_ids = tf.placeholder(dtype=tf.int64, shape=[None, None], name='segment_ids')
    label_ids = tf.placeholder(dtype=tf.int64, shape=[None, None], name='label_ids')
    label_weights = tf.placeholder(dtype=tf.float32, shape=[None, None], name='label_weights')
    is_real_example = tf.placeholder(dtype=tf.int64, shape=[None, None], name='is_real_example')

    features = {'input_ids': input_ids,
                'input_mask': input_mask,
                'segment_ids': segment_ids,
                "label_ids": label_ids,
                "label_weights": label_weights,
                "is_real_example": is_real_example,
                }
    receiver_tensors = {'input_ids': input_ids,
                        'input_mask': input_mask,
                        'segment_ids': segment_ids,
                        "label_ids": label_ids,
                        "label_weights": label_weights,
                        "is_real_example": is_real_example,
                        }
    return tf.estimator.export.ServingInputReceiver(features, receiver_tensors)


def serving_input_fn():
    """

    :return:
    """
    input_ids = tf.placeholder(dtype=tf.int64, shape=[None, None], name='input_ids')
    input_mask = tf.placeholder(dtype=tf.int64, shape=[None, None], name='input_mask')
    segment_ids = tf.placeholder(dtype=tf.int64, shape=[None, None], name='segment_ids')
    label_ids = tf.placeholder(dtype=tf.int64, shape=[None, None], name='label_ids')
    label_weights = tf.placeholder(dtype=tf.float32, shape=[None, None], name='label_weights')
    is_real_example = tf.placeholder(dtype=tf.int64, shape=[None, None], name='is_real_example')
    input_fn = tf.estimator.export.build_raw_serving_input_receiver_fn({'input_ids': input_ids,
                                                                        'input_mask': input_mask,
                                                                        'segment_ids': segment_ids,
                                                                        "label_ids": label_ids,
                                                                        "label_weights": label_weights,
                                                                        "is_real_example": is_real_example,
                                                                        })()
    return input_fn


def _truncate_seq_pair(tokens_a, tokens_b, max_length):
    """Truncates a sequence pair in place to the maximum 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()


def create_model(bert_config, is_training, input_ids, input_mask, segment_ids,
                 label_ids, num_labels, use_one_hot_embeddings, label_weights):
    """
    Creates a classification model.

    Args:
        bert_config:
        is_training:
        input_ids:
        input_mask:
        segment_ids:
        label_ids: [batch_size=32, ]
        num_labels: int, 类别个数
        use_one_hot_embeddings:
        label_weights: [batch_size=32, num_labels=2]

    """
    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: [batch_size=32, hidden_size=768]
    output_layer = model.get_pooled_output()

    # hidden_size: 768
    hidden_size = output_layer.shape[-1].value

    # output_weights: [num_labels=2, hidden_size=768]
    output_weights = tf.get_variable(
        "output_weights", [num_labels, hidden_size],
        initializer=tf.truncated_normal_initializer(stddev=0.02))

    # output_bias: [num_labels=2, ]
    output_bias = tf.get_variable(
        "output_bias", [num_labels], initializer=tf.zeros_initializer())

    # 计算损失
    with tf.variable_scope("loss"):
        if is_training:
            # I.e., 0.1 dropout
            # 在训练阶段，为了防止过拟合，在全连接层随机扔掉一部分神经元。也就是让某个神经元的激活值以一定的概率p让其停止工作，
            # 在这次训练过程中不更新权值，也不参加神经网络的计算。 但是它的权重得保留下来（只是暂时不更新而已），因为下次样本输入时它可能又得工作了
            # dropout作用：使输入tensor中某些元素变为0，其它没变0的元素值乘以1/keep_prob
            # train的时候才是dropout起作用的时候，eval/test的时候不应该让dropout起作用
            output_layer = tf.nn.dropout(output_layer, keep_prob=0.9)

        # logits类别概率: [batch_size=32, num_labels=2]
        # 矩阵相乘 a*b, 其中b在进行乘法计算前进行转置
        logits = tf.matmul(output_layer, output_weights, transpose_b=True)
        logits = tf.nn.bias_add(logits, output_bias)

        # 针对类别概率进行归一化处理probabilities: [batch_size=32, class_size=2], class_size=num_labels
        probabilities = tf.nn.softmax(logits, axis=-1)

        # 针对类别概率进行对数归一化处理log_probs: [batch_size=32, class_size=2]
        log_probs = tf.nn.log_softmax(logits, axis=-1)

        # one_hot_labels: [batch_size, class_size=2]
        one_hot_labels = tf.one_hot(label_ids, depth=num_labels, dtype=tf.float32)

        # 计算每个样本的NNL负对数似然损失(nagative log likelihood)
        per_example_loss = -tf.reduce_sum(one_hot_labels * log_probs * label_weights, axis=-1)

        # 计算batch样本的平均损失
        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, fp16=FLAGS.use_fp16):
    """
    Returns `model_fn` closure for TPUEstimator.

    Args:
        num_labels: int, 类别个数

    """

    def model_fn(features, labels, mode, params, config=None):  # pylint: disable=unused-argument
        """
        The `model_fn` for TPUEstimator.
        模型有训练，验证和测试三种阶段，而且对于不同模式，对数据有不同的处理方式。例如在训练阶段，我们需要将数据喂给模型，
        模型基于输入数据给出预测值，然后我们在通过预测值和真实值计算出loss，最后用loss更新网络参数，
        而在评估阶段，我们则不需要反向传播更新网络参数，换句话说，mdoel_fn需要对三种模式设置三套代码。

        Args:
            features: dict of Tensor, This is batch_features from input_fn,`Tensor` or dict of `Tensor` (depends on data passed to `fit`
            labels: This is batch_labels from input_fn. features, labels是从输入函数input_fn中返回的特征和标签batch
            mode: An instance of tf.estimator.ModeKeys
            params: Additional configuration for hyper-parameters. 是一个字典，它可以传入许多参数用来构建网络或者定义训练方式等


        Return:
            tf.estimator.EstimatorSpec

        """

        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"]
        label_weights = features["label_weights"]
        is_real_example = None
        if "is_real_example" in features:
            # 类型强制转换为tf.float32
            is_real_example = tf.cast(features["is_real_example"], dtype=tf.float32)
        else:
            # 创建一个将所有元素都设置为1的张量Tensor.
            is_real_example = tf.ones(tf.shape(label_ids), dtype=tf.float32)

        # 根据mode判断是否为训练模式
        is_training = (mode == tf.estimator.ModeKeys.TRAIN)

        # 基于特征数据创建模型，并计算loss等
        (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, label_weights)

        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():
                    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
        # 训练模式
        if mode == tf.estimator.ModeKeys.TRAIN:
            if FLAGS.num_gpu_cores > 1:
                train_op = custom_optimization.create_optimizer(
                    total_loss, learning_rate, num_train_steps, num_warmup_steps, fp16=fp16)

                output_spec = tf.estimator.EstimatorSpec(
                    mode=mode,
                    loss=total_loss,
                    train_op=train_op,
                    scaffold=scaffold_fn)
            else:
                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)
        # 评估模式
        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)
                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)
                # add more metrics
                pr, pr_op = tf.metrics.precision(labels=label_ids, predictions=predictions, weights=is_real_example)
                re, re_op = tf.metrics.recall(labels=label_ids, predictions=predictions, weights=is_real_example)
                # if FLAGS.classifier_mode == "multi-class":
                #     # multi-class
                #     # pr, pr_op = tf_metrics.precision(label_ids, predictions, num_labels, average="macro")
                #     # re, re_op = tf_metrics.recall(label_ids, predictions, num_labels, average="macro")
                #     f1 = tf_metrics.f1(label_ids, predictions, num_labels, average="macro")
                # else:
                #     # binary classifier
                #     f1 = tf.contrib.metrics.f1_score(label_ids, predictions)
                #     # f1, f1_op = (2 * pr * re) / (pr + re)  # f1-score for binary classification
                # 返回结果：dict: {key: value(tuple: (metric_tensor, update_op)) }
                return {
                    "eval_accuracy": accuracy,
                    "eval_loss": loss,
                    "eval_precision": (pr, pr_op),
                    "eval_recall": (re, re_op),
                    # "eval_f1": f1,
                }

            eval_metrics = (metric_fn, [per_example_loss, label_ids, logits, is_real_example])
            if FLAGS.num_gpu_cores > 1:
                output_spec = tf.estimator.EstimatorSpec(
                    mode=mode,
                    loss=total_loss,
                    eval_metric_ops=metric_fn(per_example_loss, label_ids, logits, is_real_example),
                    scaffold=scaffold_fn,
                )
            else:
                # eval on single-gpu only
                output_spec = tf.contrib.tpu.TPUEstimatorSpec(
                    mode=mode,
                    loss=total_loss,
                    eval_metrics=eval_metrics,
                    scaffold_fn=scaffold_fn)
        else:
            # tf.estimator.ModeKeys.PREDICT 预测模式
            # 基于logits计算最大的概率所在索引的label
            predictions = tf.argmax(logits, axis=-1, output_type=tf.int32)
            if FLAGS.num_gpu_cores > 1:
                # 多GPUs
                output_spec = tf.estimator.EstimatorSpec(
                    mode=mode,
                    predictions={"probabilities": probabilities, "predictions": predictions})
            else:
                output_spec = tf.contrib.tpu.TPUEstimatorSpec(
                    mode=mode,
                    predictions={"probabilities": probabilities, "predictions": predictions},
                    scaffold_fn=scaffold_fn)
        return output_spec

    return model_fn


# This function is not used by this file but is still used by the Colab and
# people who depend on it.
def input_fn_builder(features, seq_length, is_training, drop_remainder):
    """Creates an `input_fn` closure to be passed to TPUEstimator."""

    all_input_ids = []
    all_input_mask = []
    all_segment_ids = []
    all_label_ids = []

    for feature in features:
        all_input_ids.append(feature.input_ids)
        all_input_mask.append(feature.input_mask)
        all_segment_ids.append(feature.segment_ids)
        all_label_ids.append(feature.label_id)

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

        num_examples = len(features)

        # This is for demo purposes and does NOT scale to large data sets. We do
        # not use Dataset.from_generator() because that uses tf.py_func which is
        # not TPU compatible. The right way to load data is with TFRecordReader.
        d = tf.data.Dataset.from_tensor_slices({
            "input_ids":
                tf.constant(
                    all_input_ids, shape=[num_examples, seq_length],
                    dtype=tf.int32),
            "input_mask":
                tf.constant(
                    all_input_mask,
                    shape=[num_examples, seq_length],
                    dtype=tf.int32),
            "segment_ids":
                tf.constant(
                    all_segment_ids,
                    shape=[num_examples, seq_length],
                    dtype=tf.int32),
            "label_ids":
                tf.constant(all_label_ids, shape=[num_examples], dtype=tf.int32),
        })

        if is_training:
            d = d.repeat()
            d = d.shuffle(buffer_size=100)

        d = d.batch(batch_size=batch_size, drop_remainder=drop_remainder)
        return d

    return input_fn


# This function is not used by this file but is still used by the Colab and
# people who depend on it.
def convert_examples_to_features(examples, label_list, max_seq_length,
                                 tokenizer, class_weight):
    """Convert a set of `InputExample`s to a list of `InputFeatures`."""

    feats = []
    class_weight = get_examples_class_weight(examples, class_weight, label_list)
    for (ex_index, example) in enumerate(examples):
        if ex_index % 10000 == 0:
            tf.logging.info("Writing example %d of %d" % (ex_index, len(examples)))

        feature = convert_single_example(ex_index, example, label_list,
                                         max_seq_length, tokenizer, class_weight)
        features = {
            "input_ids": feature.input_ids,
            "input_mask": feature.input_mask,
            "segment_ids": feature.segment_ids,
            "label_ids": [feature.label_id],
            "label_weights": list(feature.label_weights),
            "is_real_example": int(feature.is_real_example),
        }
        feats.append(features)
    return feats


def print_cls_report(fileTest=os.path.join(FLAGS.data_dir, FLAGS.train_file),
                     fileResult=os.path.join(FLAGS.output_dir, "test_results.tsv"),
                     sep="\t",
                     header="label"):
    try:
        y_true = pd.read_csv(fileTest, sep=sep)[header]
        y_pred = [float(i.strip()) for i in open(fileResult, "r").readlines()]
        rp = classification_report(y_true, y_pred, digits=4)
        print("report:\n{}".format(rp))
    except Exception as e:
        print("Error: {}".format(repr(e)))
    return


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

    processors = {
        "cola": ColaProcessor,
        "mnli": MnliProcessor,
        "mrpc": MrpcProcessor,
        "xnli": XnliProcessor,
        "disc": DiscProcessor,
        "sed": SedProcessor,
        "topicrel": TopicRelProcessor,
        "garbledsents": GarbledSentsProcessor
    }

    tokenization.validate_case_matches_checkpoint(FLAGS.do_lower_case,
                                                  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)

    if FLAGS.max_seq_length > bert_config.max_position_embeddings:
        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))

    # local_device_protos = device_lib.list_local_devices()
    # FLAGS.num_gpu_cores = min(sum([1 for d in local_device_protos if d.device_type == 'GPU']), FLAGS.num_gpu_cores)
    tf.logging.info(f"Info: {FLAGS.num_gpu_cores} GPUs found")

    tf.gfile.MakeDirs(FLAGS.output_dir)

    task_name = FLAGS.task_name.lower()

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

    processor = processors[task_name]()
    # 获取标签列表
    label_list = processor.get_labels()
    num_labels = len(label_list)

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

    train_examples = None
    num_train_steps = None
    num_warmup_steps = None
    if FLAGS.do_train:
        train_examples = processor.get_train_examples(FLAGS.data_dir)
        num_train_steps = int(
            len(train_examples) / FLAGS.train_batch_size * FLAGS.num_train_epochs)
        num_warmup_steps = int(num_train_steps * FLAGS.warmup_proportion)

    model_fn = model_fn_builder(
        bert_config=bert_config,
        num_labels=num_labels,
        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 FLAGS.use_tpu and FLAGS.tpu_name:
        tf.logging.info(f"***** {FLAGS.tpu_name} TPU Running *****")
        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))

        # 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,
            predict_batch_size=FLAGS.predict_batch_size)

    elif FLAGS.num_gpu_cores > 1:
        # multi-gpus 训练
        tf.logging.info(f"***** {FLAGS.num_gpu_cores} GPUs Running *****")
        # 1.先定义分布式训练的镜像策略：MirroredStrategy
        dist_strategy = tf.contrib.distribute.MirroredStrategy(
            num_gpus=FLAGS.num_gpu_cores,  # 使用gpu的个数
            cross_device_ops=AllReduceCrossDeviceOps('nccl', num_packs=FLAGS.num_gpu_cores),  # 各设备之间的数据操作方式
            # cross_device_ops=AllReduceCrossDeviceOps('hierarchical_copy'),
        )
        # 2.设置会话session配置
        session_config = tf.ConfigProto(
            inter_op_parallelism_threads=0,
            intra_op_parallelism_threads=0,
            allow_soft_placement=True,
            gpu_options=tf.GPUOptions(allow_growth=True))

        # 3.设置运行配置run_config
        run_config = tf.contrib.tpu.RunConfig(
            train_distribute=dist_strategy,
            eval_distribute=dist_strategy,
            model_dir=FLAGS.output_dir,
            session_config=session_config,
            save_checkpoints_steps=FLAGS.save_checkpoints_steps,
            keep_checkpoint_max=15)
        # 4.构造CPU、GPU评估器对象
        estimator = tf.estimator.Estimator(
            model_fn=model_fn,
            config=run_config,
            params={"batch_size": FLAGS.train_batch_size})
    else:
        # 单GPU或者CPU训练
        tf.logging.info("***** Single GPU/CPU Running *****")
        tpu_cluster_resolver = None
        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))

        # 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,
            predict_batch_size=FLAGS.predict_batch_size)

    if FLAGS.do_train and FLAGS.do_eval:
        '''
        pipeline: train examples -> (write tf_record) -> (read tf_record) -> train 
        '''
        # ------------- 1.Train ------------- #
        # 1.1 将文本预处理，并存储为tf.record格式文件
        train_file = os.path.join(FLAGS.output_dir, "train.tf_record")
        file_based_convert_examples_to_features(train_examples, label_list, FLAGS.max_seq_length, tokenizer, train_file,
                                                True, FLAGS.class_weight)
        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)
        # 1.2 基于tf.record格式文件进行迭代batch处理样本
        train_input_fn = file_based_input_fn_builder(
            input_file=train_file,
            seq_length=FLAGS.max_seq_length,
            is_training=True,
            drop_remainder=True, num_labels=len(label_list))
        # ------------- 2.Eval ------------- #
        eval_examples = processor.get_dev_examples(FLAGS.data_dir)
        num_actual_eval_examples = len(eval_examples)
        if FLAGS.use_tpu:
            # TPU requires a fixed batch size for all batches, therefore the number
            # of examples must be a multiple of the batch size, or else examples
            # will get dropped. So we pad with fake examples which are ignored
            # later on. These do NOT count towards the metric (all tf.metrics
            # support a per-instance weight, and these get a weight of 0.0).
            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, False, FLAGS.class_weight)

        tf.logging.info("***** Running evaluation *****")
        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
        # However, if running eval on the TPU, you will need to specify the
        # number of steps.
        if FLAGS.use_tpu:
            assert len(eval_examples) % FLAGS.eval_batch_size == 0
            eval_steps = int(len(eval_examples) // FLAGS.eval_batch_size)

        eval_drop_remainder = True if FLAGS.use_tpu else False
        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, num_labels=len(label_list))

        if FLAGS.do_early_stopping:
            # 1.3 early stopping
            early_stopping_hook = tf.estimator.experimental.stop_if_no_decrease_hook(
                estimator=estimator,
                metric_name='eval_loss',
                max_steps_without_decrease=FLAGS.max_steps_without_decrease,
                eval_dir=None,
                min_steps=FLAGS.min_steps,
                run_every_secs=None,
                run_every_steps=FLAGS.save_checkpoints_steps)
            train_spec = tf.estimator.TrainSpec(input_fn=train_input_fn, max_steps=num_train_steps,
                                                hooks=[early_stopping_hook])
            eval_spec = tf.estimator.EvalSpec(input_fn=eval_input_fn, steps=eval_steps, throttle_secs=0)

            tf.estimator.train_and_evaluate(estimator, train_spec, eval_spec)
            tf.logging.info("***** Training & Evaluating completed*****")
        else:
            estimator.train(input_fn=train_input_fn, max_steps=num_train_steps)
            tf.logging.info("***** Training completed*****")

        # export SavedModel format for TF serving
        export_dir_base = os.path.join(FLAGS.output_dir, 'saved_model')
        estimator.export_saved_model(export_dir_base, serving_input_receiver_fn)
        tf.logging.info("***** SavedModel export completed*****")

    elif FLAGS.do_eval:
        eval_examples = processor.get_dev_examples(FLAGS.data_dir)
        num_actual_eval_examples = len(eval_examples)
        if FLAGS.use_tpu:
            # TPU requires a fixed batch size for all batches, therefore the number
            # of examples must be a multiple of the batch size, or else examples
            # will get dropped. So we pad with fake examples which are ignored
            # later on. These do NOT count towards the metric (all tf.metrics
            # support a per-instance weight, and these get a weight of 0.0).
            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, False, FLAGS.class_weight)

        tf.logging.info("***** Running evaluation *****")
        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
        # However, if running eval on the TPU, you will need to specify the
        # number of steps.
        if FLAGS.use_tpu:
            assert len(eval_examples) % FLAGS.eval_batch_size == 0
            eval_steps = int(len(eval_examples) // FLAGS.eval_batch_size)

        eval_drop_remainder = True if FLAGS.use_tpu else False
        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, num_labels=len(label_list))

        result = estimator.evaluate(input_fn=eval_input_fn, steps=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])))
        tf.logging.info("***** Evaluation completed*****")

    if FLAGS.do_predict:
        predict_examples = processor.get_test_examples(FLAGS.data_dir)
        num_actual_predict_examples = len(predict_examples)
        if FLAGS.use_tpu:
            # TPU requires a fixed batch size for all batches, therefore the number
            # of examples must be a multiple of the batch size, or else examples
            # will get dropped. So we pad with fake examples which are ignored
            # later on.
            while len(predict_examples) % FLAGS.predict_batch_size != 0:
                predict_examples.append(PaddingInputExample())

        predict_file = os.path.join(FLAGS.output_dir, "predict.tf_record")
        file_based_convert_examples_to_features(predict_examples, label_list,
                                                FLAGS.max_seq_length, tokenizer,
                                                predict_file, False, FLAGS.class_weight)

        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
        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, num_labels=len(label_list))

        result = estimator.predict(input_fn=predict_input_fn)

        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):
                probabilities = prediction["probabilities"]
                # 根据输出的索引id找到对应的label
                y_pred = label_list[prediction["predictions"]]
                if i <= 3:
                    print(f"probabilities={probabilities}")
                    print("y_pred={}".format(y_pred))
                if i >= num_actual_predict_examples:
                    break
                # output_line = "\t".join(
                #     [str(class_probability) for class_probability in probabilities] + [str(y_pred)]) + "\n"
                writer.write(str(y_pred) + "\n")
                num_written_lines += 1
        assert num_written_lines == num_actual_predict_examples
        tf.logging.info("***** Prediction completed*****")
        # 打印输出分类任务的性能指标
        print_cls_report(fileTest=os.path.join(FLAGS.data_dir, FLAGS.predict_file),
                         fileResult=os.path.join(FLAGS.output_dir, "test_results.tsv"),
                         sep="\t",
                         header="label")


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()