run_pretraining_gpu.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.
"""Run masked LM/next sentence masked_lm pre-training for BERT."""

from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import time
import os
import modeling
# import optimization_gpu
import tensorflow as tf

flags = tf.flags

FLAGS = flags.FLAGS

## Required parameters
flags.DEFINE_integer(
    "n_gpus", 10,
    "GPU number")

flags.DEFINE_string(
    "bert_config_file", "bert_config.json",
    "The config json file corresponding to the pre-trained BERT model. "
    "This specifies the model architecture.")

flags.DEFINE_string(
    "input_file", "tmp_data_128/sample1.tfrecords,tmp_data_128/sample2.tfrecords,tmp_data_128/sample3.tfrecords,tmp_data_128/sample4.tfrecords,tmp_data_128/sample5.tfrecords,tmp_data_128/sample6.tfrecords,tmp_data_128/sample7.tfrecords,tmp_data_128/sample8.tfrecords,tmp_data_128/sample9.tfrecords,tmp_data_128/sample10.tfrecords",
    "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", None,
    "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", False, "Whether to run eval on the dev 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_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", 1000,
                     "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", 100, "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.")



def _deduplicate_indexed_slices(values, indices):
    """Sums `values` associated with any non-unique `indices`.
    Args:
      values: A `Tensor` with rank >= 1.
      indices: A one-dimensional integer `Tensor`, indexing into the first
      dimension of `values` (as in an IndexedSlices object).
    Returns:
      A tuple of (`summed_values`, `unique_indices`) where `unique_indices` is a
      de-duplicated version of `indices` and `summed_values` contains the sum of
      `values` slices associated with each unique index.
    """
    unique_indices, new_index_positions = tf.unique(indices)
    summed_values = tf.unsorted_segment_sum(
        values, new_index_positions,
        tf.shape(unique_indices)[0])
    return (summed_values, unique_indices)


def average_gradients(tower_grads, batch_size, options):
    # calculate average gradient for each shared variable across all GPUs
    average_grads = []
    count = 0
    for grad_and_vars in zip(*tower_grads):
        # Note that each grad_and_vars looks like the following:
        #   ((grad0_gpu0, var0_gpu0), ... , (grad0_gpuN, var0_gpuN))
        # We need to average the gradients across each GPU.
        count += 1
        g0, v0 = grad_and_vars[0]

        if g0 is None:
            # no gradient for this variable, skip it
            average_grads.append((g0, v0))
            continue

        if isinstance(g0, tf.IndexedSlices):
            # If the gradient is type IndexedSlices then this is a sparse
            #   gradient with attributes indices and values.
            # To average, need to concat them individually then create
            #   a new IndexedSlices object.
            indices = []
            values = []
            for g, v in grad_and_vars:
                indices.append(g.indices)
                values.append(g.values)
            all_indices = tf.concat(indices, 0)
            avg_values = tf.concat(values, 0) / len(grad_and_vars)
            # deduplicate across indices
            av, ai = _deduplicate_indexed_slices(avg_values, all_indices)
            grad = tf.IndexedSlices(av, ai, dense_shape=g0.dense_shape)

        else:
            # a normal tensor can just do a simple average
            grads = []
            for g, v in grad_and_vars:
                # Add 0 dimension to the gradients to represent the tower.
                expanded_g = tf.expand_dims(g, 0)
                # Append on a 'tower' dimension which we will average over
                grads.append(expanded_g)

            # Average over the 'tower' dimension.
            grad = tf.concat(grads, 0)
            grad = tf.reduce_mean(grad, 0)

        # the Variables are redundant because they are shared
        # across towers. So.. just return the first tower's pointer to
        # the Variable.
        v = grad_and_vars[0][1]
        grad_and_var = (grad, v)

        average_grads.append(grad_and_var)

    assert len(average_grads) == len(list(zip(*tower_grads)))

    return average_grads


def clip_by_global_norm_summary(t_list, clip_norm, norm_name, variables):
    # wrapper around tf.clip_by_global_norm that also does summary ops of norms

    # compute norms
    # use global_norm with one element to handle IndexedSlices vs dense
    norms = [tf.global_norm([t]) for t in t_list]

    # summary ops before clipping
    summary_ops = []
    for ns, v in zip(norms, variables):
        name = 'norm_pre_clip/' + v.name.replace(":", "_")
        summary_ops.append(tf.summary.scalar(name, ns))

    # clip
    clipped_t_list, tf_norm = tf.clip_by_global_norm(t_list, clip_norm)

    # summary ops after clipping
    norms_post = [tf.global_norm([t]) for t in clipped_t_list]
    for ns, v in zip(norms_post, variables):
        name = 'norm_post_clip/' + v.name.replace(":", "_")
        summary_ops.append(tf.summary.scalar(name, ns))

    summary_ops.append(tf.summary.scalar(norm_name, tf_norm))

    return clipped_t_list, tf_norm, summary_ops


def clip_grads(grads, all_clip_norm_val, do_summaries):
    # grads = [(grad1, var1), (grad2, var2), ...]
    def _clip_norms(grad_and_vars, val, name):
        # grad_and_vars is a list of (g, v) pairs
        grad_tensors = [g for g, v in grad_and_vars]
        vv = [v for g, v in grad_and_vars]
        scaled_val = val
        if do_summaries:
            clipped_tensors, g_norm, so = clip_by_global_norm_summary(
                grad_tensors, scaled_val, name, vv)
        else:
            so = []
            clipped_tensors, g_norm = tf.clip_by_global_norm(
                grad_tensors, scaled_val)

        ret = []
        for t, (g, v) in zip(clipped_tensors, grad_and_vars):
            ret.append((t, v))

        return ret, so

    ret, summary_ops = _clip_norms(grads, all_clip_norm_val, 'norm_grad')

    assert len(ret) == len(grads)

    return ret, summary_ops

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

    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"]
        masked_lm_positions = features["masked_lm_positions"]
        masked_lm_ids = features["masked_lm_ids"]
        masked_lm_weights = features["masked_lm_weights"]
        next_sentence_labels = features["next_sentence_labels"]

        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)

        (next_sentence_loss, next_sentence_example_loss,
         next_sentence_log_probs) = get_next_sentence_output(
            bert_config, model.get_pooled_output(), next_sentence_labels)

        total_loss = masked_lm_loss + next_sentence_loss

        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
        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,
                loss=total_loss,
                train_op=train_op,
                scaffold_fn=scaffold_fn)
        elif mode == tf.estimator.ModeKeys.EVAL:

            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)
                masked_lm_mean_loss = tf.metrics.mean(
                    values=masked_lm_example_loss, weights=masked_lm_weights)

                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)
                next_sentence_mean_loss = tf.metrics.mean(
                    values=next_sentence_example_loss)

                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)
        else:
            raise ValueError("Only TRAIN and EVAL modes are supported: %s" % (mode))

        return output_spec

    return model_fn


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)

    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))
            input_tensor = modeling.layer_norm(input_tensor)

        # 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)

        label_ids = tf.reshape(label_ids, [-1])
        label_weights = tf.reshape(label_weights, [-1])

        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])
        numerator = tf.reduce_sum(label_weights * per_example_loss)
        denominator = tf.reduce_sum(label_weights) + 1e-5
        loss = numerator / denominator

    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)
        logits = tf.nn.bias_add(logits, output_bias)
        log_probs = tf.nn.log_softmax(logits, axis=-1)
        labels = tf.reshape(labels, [-1])
        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)
        loss = tf.reduce_mean(per_example_loss)
        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]

    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 parse_input_fn_result(result):
    """Gets features, labels, and hooks from the result of an Estimator input_fn.

    Args:
      result: output of an input_fn to an estimator, which should be one of:

        * A 'tf.data.Dataset' object: Outputs of `Dataset` object must be a
            tuple (features, labels) with same constraints as below.
        * A tuple (features, labels): Where `features` is a `Tensor` or a
          dictionary of string feature name to `Tensor` and `labels` is a
          `Tensor` or a dictionary of string label name to `Tensor`. Both
          `features` and `labels` are consumed by `model_fn`. They should
          satisfy the expectation of `model_fn` from inputs.

    Returns:
      Tuple of features, labels, and input_hooks, where features are as described
      above, labels are as described above or None, and input_hooks are a list
      of SessionRunHooks to be included when running.

    Raises:
      ValueError: if the result is a list or tuple of length != 2.
    """
    try:
        # We can't just check whether this is a tf.data.Dataset instance here,
        # as this is plausibly a PerDeviceDataset. Try treating as a dataset first.
        iterator = result.make_initializable_iterator()
    except AttributeError:
        # Not a dataset or dataset-like-object. Move along.
        pass
    else:
        result = iterator.get_next()
    return result,iterator


def input_fn(input_files,
             batch_size,
             max_seq_length,
             max_predictions_per_seq,
             is_training,
             num_cpu_threads=4):
    """The actual input function."""
    # batch_size = params["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.
    if is_training:
        d = tf.data.Dataset.from_tensor_slices(tf.constant(input_files))
        d = d.repeat()
        d = d.shuffle(buffer_size=len(input_files))

        # `cycle_length` is the number of parallel files that get read.
        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.
        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:
        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


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 main(_):
    mode = tf.estimator.ModeKeys.TRAIN
    use_one_hot_embeddings = FLAGS.use_tpu

    tf.logging.set_verbosity(tf.logging.INFO)

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

    bert_config = modeling.BertConfig.from_json_file(FLAGS.bert_config_file)

    #tf.gfile.MakeDirs(FLAGS.output_dir)

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

    tf.logging.info("*** Input Files ***")
    for input_file in input_files:
        tf.logging.info("  %s" % input_file)

    if FLAGS.do_train:
        tf.logging.info("***** Running training *****")
        tf.logging.info("  Batch size = %d", FLAGS.train_batch_size)
        n_gpus = FLAGS.n_gpus
        batch_size = FLAGS.train_batch_size
        d = input_fn(input_files,FLAGS.train_batch_size*n_gpus,FLAGS.max_seq_length,
                     FLAGS.max_predictions_per_seq,True)
        features,iterator = parse_input_fn_result(d)
        # 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)
        # estimator.train(input_fn=train_input_fn, max_steps=FLAGS.num_train_steps)

        input_ids_list = tf.split(features["input_ids"], n_gpus, axis=0)
        input_mask_list = tf.split(features["input_mask"], n_gpus, axis=0)
        segment_ids_list = tf.split(features["segment_ids"], n_gpus, axis=0)
        masked_lm_positions_list = tf.split(features["masked_lm_positions"], n_gpus, axis=0)
        masked_lm_ids_list = tf.split(features["masked_lm_ids"], n_gpus, axis=0)
        masked_lm_weights_list = tf.split(features["masked_lm_weights"], n_gpus, axis=0)
        next_sentence_labels_list = tf.split(features["next_sentence_labels"], n_gpus, axis=0)

        # multi-gpu train

        # optimizer = optimization_gpu.create_optimizer(
        #     None, FLAGS.learning_rate, FLAGS.num_train_steps, FLAGS.num_warmup_steps, False)
        optimizer = tf.train.AdamOptimizer(learning_rate=FLAGS.learning_rate)

        # calculate the gradients on each GPU
        tower_grads = []
        models = []
        loss_print = tf.get_variable(
            'train_perplexity', [],
            initializer=tf.constant_initializer(0.0), trainable=False)
        for k in range(n_gpus):
            with tf.device('/gpu:%d' % k):
                with tf.variable_scope('lm', reuse=k > 0):
                    # calculate the loss for one model replica and get
                    #   lstm states

                    input_ids = input_ids_list[k]
                    input_mask = input_mask_list[k]
                    segment_ids = segment_ids_list[k]
                    masked_lm_positions = masked_lm_positions_list[k]
                    masked_lm_ids = masked_lm_ids_list[k]
                    masked_lm_weights = masked_lm_weights_list[k]
                    next_sentence_labels = next_sentence_labels_list[k]

                    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)

                    (next_sentence_loss, next_sentence_example_loss,
                     next_sentence_log_probs) = get_next_sentence_output(
                        bert_config, model.get_pooled_output(), next_sentence_labels)

                    total_loss = masked_lm_loss + next_sentence_loss


                    loss = total_loss
                    models.append(model)
                    # get gradients
                    grads = optimizer.compute_gradients(
                        loss,
                        aggregation_method=tf.AggregationMethod.EXPERIMENTAL_TREE,
                    )
                    tower_grads.append(grads)
                    # keep track of loss across all GPUs
                    loss_print += loss

        average_grads = average_gradients(tower_grads, None, None)
        average_grads, norm_summary_ops = clip_grads(average_grads, 10.0, True)
        loss_print = loss_print / n_gpus
        train_op = optimizer.apply_gradients(average_grads)
        init = tf.global_variables_initializer()
        saver = tf.train.Saver(tf.global_variables(), max_to_keep=2)
    with tf.Session(config=tf.ConfigProto(
            allow_soft_placement=True)) as sess:
        sess.run(init)
        sess.run(iterator.initializer)
        checkpoint_path = os.path.join(FLAGS.output_dir, 'model.ckpt')
        if os.path.exists(FLAGS.output_dir):
            saver.restore(sess, checkpoint_path)

        count = 0
        t0 = time.time()
        sum = 0
        while True:

            _, loss_print_ = sess.run([train_op, loss_print])
            # optimistic_restore(sess, checkpoint_path + "-0")
            # loss_print_2 = sess.run([loss_print])
            sum+=loss_print_
            count += 1
            if count%300==0:
                print("------------")
                print(time.time() - t0," s")
                t0 = time.time()
                print("loss ",sum/count)
                sum=0
                count=0
                checkpoint_path = os.path.join(FLAGS.output_dir, 'model.ckpt')
                saver.save(sess, checkpoint_path)


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=input_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()