run_squad2.py

# coding=utf-8
# Copyright 2018 The Google AI Language Team Authors and The HugginFace Inc. team.
# Copyright (c) 2018, NVIDIA CORPORATION.  All rights reserved.
#
# 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 BERT on SQuAD 2.0"""

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

import argparse
import collections
import logging
import json
import math
import spacy
import os
import random
import pickle
from tqdm import tqdm, trange
import time
import numpy as np
import torch
from torch.utils.data import TensorDataset, DataLoader, RandomSampler, SequentialSampler
from torch.utils.data.distributed import DistributedSampler
from pytorch_pretrained_bert.tokenization import whitespace_tokenize, BasicTokenizer, BertTokenizer
from pytorch_pretrained_bert.modeling import BertForQuestionAnsweringSpanMask
from pytorch_pretrained_bert.optimization import BertAdam
from pytorch_pretrained_bert.file_utils import PYTORCH_PRETRAINED_BERT_CACHE

logging.basicConfig(format='%(asctime)s - %(levelname)s - %(name)s -   %(message)s',
                    datefmt='%m/%d/%Y %H:%M:%S',
                    level=logging.INFO)
logger = logging.getLogger(__name__)


class SquadExample(object):
    """
    A single training/test example for the Squad dataset.
    For examples without an answer, the start and end position are -1.
    """

    def __init__(self,
                 qas_id,
                 question_text,
                 doc_tokens,
                 orig_answer_text=None,
                 start_position=None,
                 end_position=None,
                 is_impossible=None,
                 que_heads=None,
                 que_types=None,
                 doc_heads=None,
                 doc_types=None,
                 que_span=None,
                 doc_span=None,
                 token_doc=None,
                 token_que=None
                 ):
        self.qas_id = qas_id
        self.question_text = question_text
        self.doc_tokens = doc_tokens
        self.orig_answer_text = orig_answer_text
        self.start_position = start_position
        self.end_position = end_position
        self.is_impossible = is_impossible
        self.que_heads = que_heads
        self.que_types = que_types
        self.doc_heads = doc_heads
        self.doc_types = doc_types
        self.que_span = que_span
        self.doc_span = doc_span
        self.token_doc = token_doc
        self.token_que = token_que

    def __str__(self):
        return self.__repr__()

    def __repr__(self):
        s = ""
        s += "qas_id: %s" % (self.qas_id)
        s += ", question_text: %s" % (
            self.question_text)
        s += ", doc_tokens: [%s]" % (" ".join(self.doc_tokens))
        if self.start_position:
            s += ", start_position: %d" % (self.start_position)
        if self.start_position:
            s += ", end_position: %d" % (self.end_position)
        if self.start_position:
            s += ", is_impossible: %r" % (self.is_impossible)
        return s


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

    def __init__(self,
                 unique_id,
                 example_index,
                 doc_span_index,
                 tokens,
                 token_to_orig_map,
                 token_is_max_context,
                 input_ids,
                 input_mask,
                 segment_ids,
                 start_position=None,
                 end_position=None,
                 is_impossible=None,
                 input_span_mask=None):
        self.unique_id = unique_id
        self.example_index = example_index
        self.doc_span_index = doc_span_index
        self.tokens = tokens
        self.token_to_orig_map = token_to_orig_map
        self.token_is_max_context = token_is_max_context
        self.input_ids = input_ids
        self.input_mask = input_mask
        self.segment_ids = segment_ids
        self.start_position = start_position
        self.end_position = end_position
        self.is_impossible = is_impossible
        self.input_span_mask = input_span_mask


def is_whitespace(c):
    if c == " " or c == "\t" or c == "\r" or c == "\n" or ord(c) == 0x202F:
        return True
    return False


class SimpleNlp(object):
    def __init__(self):
        self.nlp = spacy.load('en', disable=['parser', 'tagger', 'ner', 'textcat'])
        self.nlp.add_pipe(self.nlp.create_pipe('sentencizer'))
        # self.nlp = nltk.data.load('tokenizers/punkt/english.pickle').tokenize

    def nlp(self, text):
        return self.nlp(text)


def read_squad_examples(input_file, input_tag_file, is_training):
    simple_nlp = SimpleNlp()
    """Read a SQuAD json file into a list of SquadExample."""
    with open(input_file, "r", encoding='utf-8') as reader:
        source = json.load(reader)
        input_data = source["data"]
        version = source["version"]
    input_tag_data = []

    with open(input_tag_file, "r", encoding='utf-8') as reader:
        for line in reader:
            input_tag_data.append(json.loads(line))

    qas_id_to_tag_idx_map = {}
    all_dqtag_data = []
    for idx, tag_data in enumerate(tqdm(input_tag_data,ncols=50)):
        qas_id = tag_data["qas_id"]
        qas_id_to_tag_idx_map[qas_id] = idx
        tag_rep = tag_data["tag_rep"]
        dqtag_data = {
            "qas_id": qas_id,
            "head_que": [int(i) for i in tag_rep["pred_head_que"]],
            "span_que": [eval(i) for i in tag_rep["hpsg_list_que"]],
            "type_que": tag_rep["pred_type_que"],
            "span_doc": [eval(i) for sen_span in tag_rep["hpsg_list_doc"] for i in sen_span],
            "type_doc": [i for sen in tag_rep["pred_type_doc"] for i in sen],
            "head_doc": [int(i) for sen_head in tag_rep["pred_head_doc"] for i in sen_head],
            "token_doc": [token for sen_token in tag_rep['doc_tokens'] for token in sen_token],
            "token_que": tag_rep['que_tokens']
        }
        all_dqtag_data.append(dqtag_data)

    examples = []
    for entry in tqdm(input_data,ncols=50,desc="reading examples:"):
        for paragraph in entry["paragraphs"]:
            paragraph_text = paragraph["context"]

            sen_texts = simple_nlp.nlp(paragraph_text)
            sen_list = []

            for sen_ix, sent in enumerate(sen_texts.sents):
                sent_tokens = []
                prev_is_whitespace = True
                for c in sent.string:
                    if is_whitespace(c):
                        prev_is_whitespace = True
                    else:
                        if prev_is_whitespace:
                            sent_tokens.append(c)
                        else:
                            sent_tokens[-1] += c
                        prev_is_whitespace = False
                sen_list.append((sen_ix, sent_tokens))

            doc_tokens = []
            char_to_word_offset = []
            prev_is_whitespace = True
            for c in paragraph_text:
                if is_whitespace(c):
                    prev_is_whitespace = True
                else:
                    if prev_is_whitespace:
                        doc_tokens.append(c)
                    else:
                        doc_tokens[-1] += c
                    prev_is_whitespace = False
                char_to_word_offset.append(len(doc_tokens) - 1)

            cnt_token = 0
            new_sen_list = []
            flag = False
            tmp_token = ""
            for sen_ix, sent_tokens in sen_list:
                new_sent_tokens = []
                for tok_ix, token in enumerate(sent_tokens):
                    if tok_ix == 0 and flag:
                        token = tmp_token+token
                        flag = False
                        tmp_token = ""
                        assert token == doc_tokens[cnt_token]

                    if token != doc_tokens[cnt_token]:
                        assert tok_ix == len(sent_tokens) - 1
                        tmp_token = token
                        flag = True
                    else:
                        assert token == doc_tokens[cnt_token]
                        new_sent_tokens.append(token)

                        cnt_token += 1
                new_sen_list.append(new_sent_tokens)

            # double check
            cnt_token = 0
            for sent_tokens in new_sen_list:
                new_sent_tokens = []
                for tok_ix, token in enumerate(sent_tokens):
                    assert token == doc_tokens[cnt_token]
                    new_sent_tokens.append(token)
                    cnt_token += 1

            for qa in paragraph["qas"]:
                qas_id = qa["id"]
                dqtag = all_dqtag_data[qas_id_to_tag_idx_map[qas_id]]
                assert dqtag["qas_id"] == qas_id

                span_doc = dqtag["span_doc"]
                head_doc = dqtag["head_doc"]
                type_doc = dqtag["type_doc"]
                assert len(span_doc) == len(head_doc) == len(type_doc) == cnt_token
                # reconstruct into sentences
                new_span_doc = []
                new_head_doc = []
                new_type_doc = []
                cnt = 0
                for sent_tokens in new_sen_list:
                    new_span_sen = []
                    new_head_sen = []
                    new_type_sen = []
                    for _ in sent_tokens:
                        new_span_sen.append(span_doc[cnt])
                        new_head_sen.append(head_doc[cnt])
                        new_type_sen.append(type_doc[cnt])
                        cnt += 1
                    new_span_doc.append(new_span_sen)
                    new_head_doc.append(new_head_sen)
                    new_type_doc.append(new_type_sen)

                question_text = qa["question"]
                start_position = None
                end_position = None
                orig_answer_text = None
                is_impossible = False
                if is_training:
                    if version == "v2.0":
                        is_impossible = qa["is_impossible"]

                    if not is_impossible:
                        answer = qa["answers"][0]
                        orig_answer_text = answer["text"]
                        answer_offset = answer["answer_start"]
                        answer_length = len(orig_answer_text)
                        start_position = char_to_word_offset[answer_offset]
                        end_position = char_to_word_offset[answer_offset + answer_length - 1]
                        # Only add answers where the text can be exactly recovered from the
                        # document. If this CAN'T happen it's likely due to weird Unicode
                        # stuff so we will just skip the example.
                        #
                        # Note that this means for training mode, every example is NOT
                        # guaranteed to be preserved.
                        actual_text = " ".join(doc_tokens[start_position:(end_position + 1)])
                        cleaned_answer_text = " ".join(
                            whitespace_tokenize(orig_answer_text))
                        if actual_text.find(cleaned_answer_text) == -1:
                            logger.warning("Could not find answer: '%s' vs. '%s'",
                                           actual_text, cleaned_answer_text)
                            continue
                    else:
                        start_position = -1
                        end_position = -1
                        orig_answer_text = ""

                example = SquadExample(
                    qas_id=qas_id,
                    question_text=question_text,
                    doc_tokens=doc_tokens,
                    orig_answer_text=orig_answer_text,
                    start_position=start_position,
                    end_position=end_position,
                    is_impossible=is_impossible,
                    que_heads=dqtag["head_que"],
                    que_types=dqtag["type_que"],
                    que_span=dqtag["span_que"],
                    doc_heads=new_head_doc,
                    doc_types=new_type_doc,
                    doc_span=new_span_doc,
                    token_doc=dqtag["token_doc"],
                    token_que=dqtag["token_que"]
                )
                examples.append(example)
    return examples


def convert_examples_to_features(examples, tokenizer, max_seq_length,
                                 doc_stride, max_query_length, is_training):
    """Loads a data file into a list of `InputBatch`s."""

    unique_id = 1000000000
    features = []
    flag = True
    for (example_index, example) in enumerate(tqdm(examples,ncols=50)):
        que_span = example.que_span
        org_que_token = example.token_que
        org_doc_token = example.token_doc
        all_doc_span = example.doc_span

        query_tokens = tokenizer.tokenize(example.question_text)

        que_tokens = []
        prev_is_whitespace = True
        for c in example.question_text:
            if is_whitespace(c):
                prev_is_whitespace = True
            else:
                if prev_is_whitespace:
                    que_tokens.append(c)
                else:
                    que_tokens[-1] += c
                prev_is_whitespace = False

        sub_que_span = []

        que_org_to_split_map = {}
        pre_tok_len = 0
        for idx, que_token in enumerate(que_tokens):
            sub_que_tok = tokenizer.tokenize(que_token)
            que_org_to_split_map[idx] = (pre_tok_len, len(sub_que_tok) + pre_tok_len - 1)
            pre_tok_len += len(sub_que_tok)

        for idx, (start_ix, end_ix) in enumerate(que_span):
            head_start, head_end = que_org_to_split_map[idx]

            # sub_start_idx and sub_end_idx of children of head node
            head_spans = [(que_org_to_split_map[start_ix-1][0], que_org_to_split_map[end_ix-1][1])]
            # all other head sub_tok point to first head sub_tok
            if head_start != head_end:
                head_spans.append((head_start+1, head_end))
                sub_que_span.append(head_spans)

                for i in range(head_start+1, head_end+1):
                    sub_que_span.append([(i, i)])
            else:
                sub_que_span.append(head_spans)

        assert len(sub_que_span) == len(query_tokens)

        tok_to_orig_index = []
        orig_to_tok_index = []
        all_doc_tokens = []

        for (i, token) in enumerate(example.doc_tokens):
            orig_to_tok_index.append(len(all_doc_tokens))
            sub_tokens = tokenizer.tokenize(token)
            for sub_token in sub_tokens:
                tok_to_orig_index.append(i)
                all_doc_tokens.append(sub_token)

        doc_org_to_split_map = {}
        pre_tok_len = 0
        for idx, doc_token in enumerate(example.doc_tokens):
            sub_doc_tok = tokenizer.tokenize(doc_token)
            doc_org_to_split_map[idx] = (pre_tok_len, len(sub_doc_tok) + pre_tok_len - 1)
            pre_tok_len += len(sub_doc_tok)

        cnt_span = 0
        for sen_idx, sen_span in enumerate(all_doc_span):
            for idx, (start_ix, end_ix) in enumerate(sen_span):
                assert (start_ix <= len(sen_span) and end_ix <= len(sen_span))
                cnt_span += 1

        assert cnt_span == len(example.doc_tokens)

        sub_doc_span = []
        pre_sen_len = 0
        for sen_idx, sen_span in enumerate(all_doc_span):
            sen_offset = pre_sen_len
            pre_sen_len += len(sen_span)
            for idx, (start_ix, end_ix) in enumerate(sen_span):
                head_start, head_end = doc_org_to_split_map[sen_offset+idx]
                # sub_start_idx and sub_end_idx of children of head node
                head_spans = [(doc_org_to_split_map[sen_offset+start_ix-1][0], doc_org_to_split_map[sen_offset+end_ix-1][1])]
                # all other head sub_tok point to first head sub_tok
                if head_start != head_end:
                    head_spans.append((head_start + 1, head_end))
                    sub_doc_span.append(head_spans)

                    for i in range(head_start + 1, head_end + 1):
                        sub_doc_span.append([(i, i)])
                else:
                    sub_doc_span.append(head_spans)

        assert len(sub_doc_span) == len(all_doc_tokens)

        # making masks
        que_span_mask = np.zeros((len(sub_que_span), len(sub_que_span)))
        for idx, span_list in enumerate(sub_que_span):
            for (start_ix, end_ix) in span_list:
                if start_ix != end_ix:
                    que_span_mask[start_ix:end_ix + 1, idx] = 1

        doc_span_mask = np.zeros((len(sub_doc_span), len(sub_doc_span)))

        for idx, span_list in enumerate(sub_doc_span):
            for (start_ix, end_ix) in span_list:
                if start_ix != end_ix:
                   doc_span_mask[start_ix:end_ix + 1, idx] = 1

        if len(query_tokens) > max_query_length:
            query_tokens = query_tokens[0:max_query_length]
            que_span_mask = que_span_mask[:max_query_length,:max_query_length]


        tok_start_position = None
        tok_end_position = None
        if is_training and example.is_impossible:
            tok_start_position = -1
            tok_end_position = -1
        if is_training and not example.is_impossible:
            tok_start_position = orig_to_tok_index[example.start_position]
            if example.end_position < len(example.doc_tokens) - 1:
                tok_end_position = orig_to_tok_index[example.end_position + 1] - 1
            else:
                tok_end_position = len(all_doc_tokens) - 1
            (tok_start_position, tok_end_position) = _improve_answer_span(
                all_doc_tokens, tok_start_position, tok_end_position, tokenizer,
                example.orig_answer_text)

        # The -3 accounts for [CLS], [SEP] and [SEP]
        max_tokens_for_doc = max_seq_length - len(query_tokens) - 3

        # We can have documents that are longer than the maximum sequence length.
        # To deal with this we do a sliding window approach, where we take chunks
        # of the up to our max length with a stride of `doc_stride`.
        _DocSpan = collections.namedtuple(  # pylint: disable=invalid-name
            "DocSpan", ["start", "length"])
        doc_spans = []
        start_offset = 0
        while start_offset < len(all_doc_tokens):
            length = len(all_doc_tokens) - start_offset
            if length > max_tokens_for_doc:
                length = max_tokens_for_doc
            doc_spans.append(_DocSpan(start=start_offset, length=length))
            if start_offset + length == len(all_doc_tokens):
                break
            start_offset += min(length, doc_stride)

        for (doc_span_index, doc_span) in enumerate(doc_spans):
            tokens = []
            token_to_orig_map = {}
            token_is_max_context = {}
            # use this idx list to select from doc span mask
            head_select_idx = []
            segment_ids = []
            tokens.append("[CLS]")
            segment_ids.append(0)

            for i in range(doc_span.length):
                split_token_index = doc_span.start + i
                token_to_orig_map[len(tokens)] = tok_to_orig_index[split_token_index]

                is_max_context = _check_is_max_context(doc_spans, doc_span_index,
                                                       split_token_index)
                token_is_max_context[len(tokens)] = is_max_context
                tokens.append(all_doc_tokens[split_token_index])
                head_select_idx.append(split_token_index)

                segment_ids.append(0)

            tokens.append("[SEP]")
            segment_ids.append(0)
            start_doc_ix = head_select_idx[0]
            end_doc_ix = head_select_idx[-1]
            select_doc_len = end_doc_ix - start_doc_ix + 1
            select_que_len = len(query_tokens)
            assert len(head_select_idx) == select_doc_len

            input_span_mask = np.zeros((max_seq_length, max_seq_length))
            # 0 count for [CLS] and select_doc_len+1 count for [SEP]
            input_span_mask[1:select_doc_len + 1, 1:select_doc_len + 1] = doc_span_mask[start_doc_ix:end_doc_ix + 1,
                                                                          start_doc_ix:end_doc_ix + 1]
            input_span_mask[select_doc_len + 2:select_doc_len + select_que_len + 2,
            select_doc_len + 2:select_doc_len + select_que_len + 2] = que_span_mask
            record_mask = []
            for i in range(max_seq_length):
                i_mask = []
                for j in range(max_seq_length):
                    if input_span_mask[i, j] == 1:
                        i_mask.append(j)
                record_mask.append(i_mask)

            for idx, token in enumerate(query_tokens):
                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
            assert len(input_span_mask) == max_seq_length

            start_position = None
            end_position = None
            is_impossible = example.is_impossible

            if is_training and not example.is_impossible:
                # For training, if our document chunk does not contain an annotation
                # we throw it out, since there is nothing to predict.
                doc_start = doc_span.start
                doc_end = doc_span.start + doc_span.length - 1
                out_of_span = False
                if (example.start_position < doc_start or
                        example.end_position < doc_start or
                        example.start_position > doc_end or example.end_position > doc_end):
                    out_of_span = True
                if out_of_span:
                    start_position = 0
                    end_position = 0
                    is_impossible = True

                else:
                    #doc_offset = len(query_tokens) + 2
                    doc_offset = 1 # [CLS]
                    start_position = tok_start_position - doc_start + doc_offset
                    end_position = tok_end_position - doc_start + doc_offset
                    is_impossible = False

            if is_training and example.is_impossible:
                start_position = 0
                end_position = 0
                is_impossible =True

            features.append(
                InputFeatures(
                    unique_id=unique_id,
                    example_index=example_index,
                    doc_span_index=doc_span_index,
                    tokens=tokens,
                    token_to_orig_map=token_to_orig_map,
                    token_is_max_context=token_is_max_context,
                    input_ids=input_ids,
                    input_mask=input_mask,
                    segment_ids=segment_ids,
                    start_position=start_position,
                    end_position=end_position,
                    is_impossible=is_impossible,
                    input_span_mask=record_mask
                ))
            unique_id += 1

    return features


def _improve_answer_span(doc_tokens, input_start, input_end, tokenizer,
                         orig_answer_text):
    """Returns tokenized answer spans that better match the annotated answer."""

    # The SQuAD annotations are character based. We first project them to
    # whitespace-tokenized words. But then after WordPiece tokenization, we can
    # often find a "better match". For example:
    #
    #   Question: What year was John Smith born?
    #   Context: The leader was John Smith (1895-1943).
    #   Answer: 1895
    #
    # The original whitespace-tokenized answer will be "(1895-1943).". However
    # after tokenization, our tokens will be "( 1895 - 1943 ) .". So we can match
    # the exact answer, 1895.
    #
    # However, this is not always possible. Consider the following:
    #
    #   Question: What country is the top exporter of electornics?
    #   Context: The Japanese electronics industry is the lagest in the world.
    #   Answer: Japan
    #
    # In this case, the annotator chose "Japan" as a character sub-span of
    # the word "Japanese". Since our WordPiece tokenizer does not split
    # "Japanese", we just use "Japanese" as the annotation. This is fairly rare
    # in SQuAD, but does happen.
    tok_answer_text = " ".join(tokenizer.tokenize(orig_answer_text))

    for new_start in range(input_start, input_end + 1):
        for new_end in range(input_end, new_start - 1, -1):
            text_span = " ".join(doc_tokens[new_start:(new_end + 1)])
            if text_span == tok_answer_text:
                return (new_start, new_end)

    return (input_start, input_end)


def _check_is_max_context(doc_spans, cur_span_index, position):
    """Check if this is the 'max context' doc span for the token."""

    # Because of the sliding window approach taken to scoring documents, a single
    # token can appear in multiple documents. E.g.
    #  Doc: the man went to the store and bought a gallon of milk
    #  Span A: the man went to the
    #  Span B: to the store and bought
    #  Span C: and bought a gallon of
    #  ...
    #
    # Now the word 'bought' will have two scores from spans B and C. We only
    # want to consider the score with "maximum context", which we define as
    # the *minimum* of its left and right context (the *sum* of left and
    # right context will always be the same, of course).
    #
    # In the example the maximum context for 'bought' would be span C since
    # it has 1 left context and 3 right context, while span B has 4 left context
    # and 0 right context.
    best_score = None
    best_span_index = None
    for (span_index, doc_span) in enumerate(doc_spans):
        end = doc_span.start + doc_span.length - 1
        if position < doc_span.start:
            continue
        if position > end:
            continue
        num_left_context = position - doc_span.start
        num_right_context = end - position
        score = min(num_left_context, num_right_context) + 0.01 * doc_span.length
        if best_score is None or score > best_score:
            best_score = score
            best_span_index = span_index

    return cur_span_index == best_span_index


RawResult = collections.namedtuple("RawResult",
                                   ["unique_id", "start_logits", "end_logits", "choice_logits"])


def write_predictions(all_examples, all_features, all_results, n_best_size,
                      max_answer_length, do_lower_case, output_prediction_file,
                      output_nbest_file, output_null_log_odds_file, verbose_logging,
                      version_2_with_negative, null_score_diff_threshold):
    """Write final predictions to the json file and log-odds of null if needed."""
    logger.info("Writing predictions to: %s" % (output_prediction_file))
    logger.info("Writing nbest to: %s" % (output_nbest_file))

    example_index_to_features = collections.defaultdict(list)
    for feature in all_features:
        example_index_to_features[feature.example_index].append(feature)

    unique_id_to_result = {}
    for result in all_results:
        unique_id_to_result[result.unique_id] = result

    _PrelimPrediction = collections.namedtuple(  # pylint: disable=invalid-name
        "PrelimPrediction",
        ["feature_index", "start_index", "end_index", "start_logit", "end_logit"])

    all_predictions = collections.OrderedDict()
    all_nbest_json = collections.OrderedDict()
    scores_diff_json = collections.OrderedDict()

    for (example_index, example) in enumerate(all_examples):
        features = example_index_to_features[example_index]

        prelim_predictions = []
        # keep track of the minimum score of null start+end of position 0
        score_null = 1000000  # large and positive
        min_null_feature_index = 0  # the paragraph slice with min null score
        null_start_logit = 0  # the start logit at the slice with min null score
        null_end_logit = 0  # the end logit at the slice with min null score
        for (feature_index, feature) in enumerate(features):
            result = unique_id_to_result[feature.unique_id]
            start_indexes = _get_best_indexes(result.start_logits, n_best_size)
            end_indexes = _get_best_indexes(result.end_logits, n_best_size)
            # if we could have irrelevant answers, get the min score of irrelevant
            if version_2_with_negative:
                feature_null_score = result.start_logits[0] + result.end_logits[0]
                if feature_null_score < score_null:
                    score_null = feature_null_score
                    min_null_feature_index = feature_index
                    null_start_logit = result.start_logits[0]
                    null_end_logit = result.end_logits[0]
            for start_index in start_indexes:
                for end_index in end_indexes:
                    # We could hypothetically create invalid predictions, e.g., predict
                    # that the start of the span is in the question. We throw out all
                    # invalid predictions.
                    if start_index >= len(feature.tokens):
                        continue
                    if end_index >= len(feature.tokens):
                        continue
                    if start_index not in feature.token_to_orig_map:
                        continue
                    if end_index not in feature.token_to_orig_map:
                        continue
                    if not feature.token_is_max_context.get(start_index, False):
                        continue
                    if end_index < start_index:
                        continue
                    length = end_index - start_index + 1
                    if length > max_answer_length:
                        continue
                    prelim_predictions.append(
                        _PrelimPrediction(
                            feature_index=feature_index,
                            start_index=start_index,
                            end_index=end_index,
                            start_logit=result.start_logits[start_index],
                            end_logit=result.end_logits[end_index]))
        if version_2_with_negative:
            prelim_predictions.append(
                _PrelimPrediction(
                    feature_index=min_null_feature_index,
                    start_index=0,
                    end_index=0,
                    start_logit=null_start_logit,
                    end_logit=null_end_logit))
        prelim_predictions = sorted(
            prelim_predictions,
            key=lambda x: (x.start_logit + x.end_logit),
            reverse=True)

        _NbestPrediction = collections.namedtuple(  # pylint: disable=invalid-name
            "NbestPrediction", ["text", "start_logit", "end_logit"])

        seen_predictions = {}
        nbest = []
        for pred in prelim_predictions:
            if len(nbest) >= n_best_size:
                break
            feature = features[pred.feature_index]
            if pred.start_index > 0:  # this is a non-null prediction
                tok_tokens = feature.tokens[pred.start_index:(pred.end_index + 1)]
                orig_doc_start = feature.token_to_orig_map[pred.start_index]
                orig_doc_end = feature.token_to_orig_map[pred.end_index]
                orig_tokens = example.doc_tokens[orig_doc_start:(orig_doc_end + 1)]
                tok_text = " ".join(tok_tokens)

                # De-tokenize WordPieces that have been split off.
                tok_text = tok_text.replace(" ##", "")
                tok_text = tok_text.replace("##", "")

                # Clean whitespace
                tok_text = tok_text.strip()
                tok_text = " ".join(tok_text.split())
                orig_text = " ".join(orig_tokens)

                final_text = get_final_text(tok_text, orig_text, do_lower_case, verbose_logging)
                if final_text in seen_predictions:
                    continue

                seen_predictions[final_text] = True
            else:
                final_text = ""
                seen_predictions[final_text] = True

            nbest.append(
                _NbestPrediction(
                    text=final_text,
                    start_logit=pred.start_logit,
                    end_logit=pred.end_logit))
        # if we didn't include the empty option in the n-best, include it
        if version_2_with_negative:
            if "" not in seen_predictions:
                nbest.append(
                    _NbestPrediction(
                        text="",
                        start_logit=null_start_logit,
                        end_logit=null_end_logit))

            # In very rare edge cases we could only have single null prediction.
            # So we just create a nonce prediction in this case to avoid failure.
            if len(nbest) == 1:
                nbest.insert(0,
                             _NbestPrediction(text="empty", start_logit=0.0, end_logit=0.0))

        # In very rare edge cases we could have no valid predictions. So we
        # just create a nonce prediction in this case to avoid failure.
        if not nbest:
            nbest.append(
                _NbestPrediction(text="empty", start_logit=0.0, end_logit=0.0))

        assert len(nbest) >= 1

        total_scores = []
        best_non_null_entry = None
        for entry in nbest:
            total_scores.append(entry.start_logit + entry.end_logit)
            if not best_non_null_entry:
                if entry.text:
                    best_non_null_entry = entry

        probs = _compute_softmax(total_scores)

        nbest_json = []
        for (i, entry) in enumerate(nbest):
            output = collections.OrderedDict()
            output["text"] = entry.text
            output["probability"] = probs[i]
            output["start_logit"] = entry.start_logit
            output["end_logit"] = entry.end_logit
            nbest_json.append(output)

        assert len(nbest_json) >= 1

        if not version_2_with_negative:
            all_predictions[example.qas_id] = nbest_json[0]["text"]
        else:
            # predict "" iff the null score - the score of best non-null > threshold
            score_diff = score_null - best_non_null_entry.start_logit - (
                best_non_null_entry.end_logit)
            scores_diff_json[example.qas_id] = score_diff
            if score_diff > null_score_diff_threshold:
                all_predictions[example.qas_id] = ""
            else:
                all_predictions[example.qas_id] = best_non_null_entry.text
            all_nbest_json[example.qas_id] = nbest_json

    with open(output_prediction_file, "w") as writer:
        writer.write(json.dumps(all_predictions, indent=4) + "\n")

    with open(output_nbest_file, "w") as writer:
        writer.write(json.dumps(all_nbest_json, indent=4) + "\n")

    if version_2_with_negative:
        with open(output_null_log_odds_file, "w") as writer:
            writer.write(json.dumps(scores_diff_json, indent=4) + "\n")


def get_final_text(pred_text, orig_text, do_lower_case, verbose_logging=False):
    """Project the tokenized prediction back to the original text."""

    # When we created the data, we kept track of the alignment between original
    # (whitespace tokenized) tokens and our WordPiece tokenized tokens. So
    # now `orig_text` contains the span of our original text corresponding to the
    # span that we predicted.
    #
    # However, `orig_text` may contain extra characters that we don't want in
    # our prediction.
    #
    # For example, let's say:
    #   pred_text = steve smith
    #   orig_text = Steve Smith's
    #
    # We don't want to return `orig_text` because it contains the extra "'s".
    #
    # We don't want to return `pred_text` because it's already been normalized
    # (the SQuAD eval script also does punctuation stripping/lower casing but
    # our tokenizer does additional normalization like stripping accent
    # characters).
    #
    # What we really want to return is "Steve Smith".
    #
    # Therefore, we have to apply a semi-complicated alignment heruistic between
    # `pred_text` and `orig_text` to get a character-to-charcter alignment. This
    # can fail in certain cases in which case we just return `orig_text`.

    def _strip_spaces(text):
        ns_chars = []
        ns_to_s_map = collections.OrderedDict()
        for (i, c) in enumerate(text):
            if c == " ":
                continue
            ns_to_s_map[len(ns_chars)] = i
            ns_chars.append(c)
        ns_text = "".join(ns_chars)
        return (ns_text, ns_to_s_map)

    # We first tokenize `orig_text`, strip whitespace from the result
    # and `pred_text`, and check if they are the same length. If they are
    # NOT the same length, the heuristic has failed. If they are the same
    # length, we assume the characters are one-to-one aligned.
    tokenizer = BasicTokenizer(do_lower_case=do_lower_case)

    tok_text = " ".join(tokenizer.tokenize(orig_text))

    start_position = tok_text.find(pred_text)
    if start_position == -1:
        if verbose_logging:
            logger.info(
                "Unable to find text: '%s' in '%s'" % (pred_text, orig_text))
        return orig_text
    end_position = start_position + len(pred_text) - 1

    (orig_ns_text, orig_ns_to_s_map) = _strip_spaces(orig_text)
    (tok_ns_text, tok_ns_to_s_map) = _strip_spaces(tok_text)

    if len(orig_ns_text) != len(tok_ns_text):
        if verbose_logging:
            logger.info("Length not equal after stripping spaces: '%s' vs '%s'",
                        orig_ns_text, tok_ns_text)
        return orig_text

    # We then project the characters in `pred_text` back to `orig_text` using
    # the character-to-character alignment.
    tok_s_to_ns_map = {}
    for (i, tok_index) in tok_ns_to_s_map.items():
        tok_s_to_ns_map[tok_index] = i

    orig_start_position = None
    if start_position in tok_s_to_ns_map:
        ns_start_position = tok_s_to_ns_map[start_position]
        if ns_start_position in orig_ns_to_s_map:
            orig_start_position = orig_ns_to_s_map[ns_start_position]

    if orig_start_position is None:
        if verbose_logging:
            logger.info("Couldn't map start position")
        return orig_text

    orig_end_position = None
    if end_position in tok_s_to_ns_map:
        ns_end_position = tok_s_to_ns_map[end_position]
        if ns_end_position in orig_ns_to_s_map:
            orig_end_position = orig_ns_to_s_map[ns_end_position]

    if orig_end_position is None:
        if verbose_logging:
            logger.info("Couldn't map end position")
        return orig_text

    output_text = orig_text[orig_start_position:(orig_end_position + 1)]
    return output_text


def _get_best_indexes(logits, n_best_size):
    """Get the n-best logits from a list."""
    index_and_score = sorted(enumerate(logits), key=lambda x: x[1], reverse=True)

    best_indexes = []
    for i in range(len(index_and_score)):
        if i >= n_best_size:
            break
        best_indexes.append(index_and_score[i][0])
    return best_indexes


def _compute_softmax(scores):
    """Compute softmax probability over raw logits."""
    if not scores:
        return []

    max_score = None
    for score in scores:
        if max_score is None or score > max_score:
            max_score = score

    exp_scores = []
    total_sum = 0.0
    for score in scores:
        x = math.exp(score - max_score)
        exp_scores.append(x)
        total_sum += x

    probs = []
    for score in exp_scores:
        probs.append(score / total_sum)
    return probs


def warmup_linear(x, warmup=0.002):
    if x < warmup:
        return x / warmup
    return 1.0 - x


def main():
    parser = argparse.ArgumentParser()

    ## Required parameters
    parser.add_argument("--bert_model", default="bert-base-cased", type=str,
                        help="Bert pre-trained model selected in the list: bert-base-uncased, "
                             "bert-large-uncased, bert-base-cased, bert-base-multilingual, bert-base-chinese.")
    parser.add_argument("--output_dir", default="PN", type=str,
                        help="The output directory where the model checkpoints and predictions will be written.")

    ## Other parameters
    parser.add_argument("--train_file", default="data/squad/squad_sample.json",
                        type=str, help="SQuAD json for training. E.g., train-v1.1.json")
    parser.add_argument("--predict_file", default="data/squad/squad_sample.json",
                        type=str,
                        help="SQuAD json for predictions. E.g., dev-v1.1.json or test-v1.1.json")
    parser.add_argument("--train_tag_file", default="data/squad/squad_span_sample.json", type=str)
    parser.add_argument("--predict_tag_file", default="data/squad/squad_span_sample.json", type=str)

    parser.add_argument("--max_seq_length", default=384, type=int,
                        help="The maximum total input sequence length after WordPiece tokenization. Sequences "
                             "longer than this will be truncated, and sequences shorter than this will be padded.")
    parser.add_argument("--doc_stride", default=128, type=int,
                        help="When splitting up a long document into chunks, how much stride to take between chunks.")
    parser.add_argument("--max_query_length", default=64, type=int,
                        help="The maximum number of tokens for the question. Questions longer than this will "
                             "be truncated to this length.")
    parser.add_argument("--do_train", default=False, action='store_true', help="Whether to run training.")
    parser.add_argument("--do_predict", default=False, action='store_true', help="Whether to run eval on the dev set.")
    parser.add_argument("--train_batch_size", default=4, type=int, help="Total batch size for training.")
    parser.add_argument("--predict_batch_size", default=8, type=int, help="Total batch size for predictions.")
    parser.add_argument("--learning_rate", default=5e-5, type=float, help="The initial learning rate for Adam.")
    parser.add_argument("--num_train_epochs", default=3.0, type=float,
                        help="Total number of training epochs to perform.")
    parser.add_argument("--warmup_proportion", default=0.1, type=float,
                        help="Proportion of training to perform linear learning rate warmup for. E.g., 0.1 = 10% "
                             "of training.")
    parser.add_argument("--n_best_size", default=20, type=int,
                        help="The total number of n-best predictions to generate in the nbest_predictions.json "
                             "output file.")
    parser.add_argument("--max_answer_length", default=30, type=int,
                        help="The maximum length of an answer that can be generated. This is needed because the start "
                             "and end predictions are not conditioned on one another.")
    parser.add_argument("--verbose_logging", default=False, action='store_true',
                        help="If true, all of the warnings related to data processing will be printed. "
                             "A number of warnings are expected for a normal SQuAD evaluation.")
    parser.add_argument("--no_cuda",
                        default=False,
                        action='store_true',
                        help="Whether not to use CUDA when available")
    parser.add_argument('--seed',
                        type=int,
                        default=42,
                        help="random seed for initialization")
    parser.add_argument('--gradient_accumulation_steps',
                        type=int,
                        default=1,
                        help="Number of updates steps to accumulate before performing a backward/update pass.")
    parser.add_argument("--do_lower_case",
                        action='store_true',
                        help="Whether to lower case the input text. True for uncased models, False for cased models.")
    parser.add_argument("--local_rank",
                        type=int,
                        default=-1,
                        help="local_rank for distributed training on gpus")
    parser.add_argument('--fp16',
                        default=False,
                        action='store_true',
                        help="Whether to use 16-bit float precision instead of 32-bit")
    parser.add_argument('--loss_scale',
                        type=float, default=0,
                        help="Loss scaling to improve fp16 numeric stability. Only used when fp16 set to True.\n"
                             "0 (default value): dynamic loss scaling.\n"
                             "Positive power of 2: static loss scaling value.\n")
    parser.add_argument('--null_score_diff_threshold',
                        type=float, default=0.0,
                        help="If null_score - best_non_null is greater than the threshold predict null.")

    args = parser.parse_args()

    if args.local_rank == -1 or args.no_cuda:
        device = torch.device("cuda" if torch.cuda.is_available() and not args.no_cuda else "cpu")
        n_gpu = torch.cuda.device_count()
    else:
        torch.cuda.set_device(args.local_rank)
        device = torch.device("cuda", args.local_rank)
        n_gpu = 1
        # Initializes the distributed backend which will take care of sychronizing nodes/GPUs
        torch.distributed.init_process_group(backend='nccl')
    logger.info("device: {} n_gpu: {}, distributed training: {}, 16-bits training: {}".format(
        device, n_gpu, bool(args.local_rank != -1), args.fp16))

    if args.gradient_accumulation_steps < 1:
        raise ValueError("Invalid gradient_accumulation_steps parameter: {}, should be >= 1".format(
            args.gradient_accumulation_steps))

    args.train_batch_size = int(args.train_batch_size / args.gradient_accumulation_steps)

    random.seed(args.seed)
    np.random.seed(args.seed)
    torch.manual_seed(args.seed)
    if n_gpu > 0:
        torch.cuda.manual_seed_all(args.seed)

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

    if args.do_train:
        if not args.train_file:
            raise ValueError(
                "If `do_train` is True, then `train_file` must be specified.")
    if args.do_predict:
        if not args.predict_file:
            raise ValueError(
                "If `do_predict` is True, then `predict_file` must be specified.")

    os.makedirs(args.output_dir, exist_ok=True)

    tokenizer = BertTokenizer.from_pretrained(args.bert_model, do_lower_case=args.do_lower_case)

    train_examples = None
    num_train_steps = None

    if args.do_train:
        print("reading train...")
        train_examples = read_squad_examples(
            input_file=args.train_file, input_tag_file=args.train_tag_file, is_training=True)
        num_train_steps = int(
            len(train_examples) / args.train_batch_size / args.gradient_accumulation_steps * args.num_train_epochs)

    # Prepare model
    model = BertForQuestionAnsweringSpanMask.from_pretrained(args.bert_model,
                                                           cache_dir=PYTORCH_PRETRAINED_BERT_CACHE / 'distributed_{}'.format(
                                                           args.local_rank))

    if args.fp16:
        model.half()
    model.to(device)
    if args.local_rank != -1:
        try:
            from apex.parallel import DistributedDataParallel as DDP
        except ImportError:
            raise ImportError(
                "Please install apex from https://www.github.com/nvidia/apex to use distributed and fp16 training.")

        model = DDP(model)
    elif n_gpu > 1:
        model = torch.nn.DataParallel(model)

    # Prepare optimizer
    param_optimizer = list(model.named_parameters())

    # hack to remove pooler, which is not used
    # thus it produce None grad that break apex
    param_optimizer = [n for n in param_optimizer if 'pooler' not in n[0]]

    no_decay = ['bias', 'LayerNorm.bias', 'LayerNorm.weight']
    optimizer_grouped_parameters = [
        {'params': [p for n, p in param_optimizer if not any(nd in n for nd in no_decay)], 'weight_decay': 0.01},
        {'params': [p for n, p in param_optimizer if any(nd in n for nd in no_decay)], 'weight_decay': 0.0}
    ]

    t_total = num_train_steps
    if args.local_rank != -1:
        t_total = t_total // torch.distributed.get_world_size()
    if args.fp16:
        try:
            from apex.optimizers import FP16_Optimizer
            from apex.optimizers import FusedAdam
        except ImportError:
            raise ImportError(
                "Please install apex from https://www.github.com/nvidia/apex to use distributed and fp16 training.")

        optimizer = FusedAdam(optimizer_grouped_parameters,
                              lr=args.learning_rate,
                              bias_correction=False,
                              max_grad_norm=1.0)
        if args.loss_scale == 0:
            optimizer = FP16_Optimizer(optimizer, dynamic_loss_scale=True)
        else:
            optimizer = FP16_Optimizer(optimizer, static_loss_scale=args.loss_scale)
    else:
        optimizer = BertAdam(optimizer_grouped_parameters,
                             lr=args.learning_rate,
                             warmup=args.warmup_proportion,
                             t_total=t_total)

    global_step = 0
    TrainLoss = []
    if args.do_train:
        cached_train_features_file = "train_features.pkl"

        try:
            with open(cached_train_features_file, "rb") as reader:
                train_features = pickle.load(reader)
        except:
            train_features = convert_examples_to_features(
                examples=train_examples,
                tokenizer=tokenizer,
                max_seq_length=args.max_seq_length,
                doc_stride=args.doc_stride,
                max_query_length=args.max_query_length,
                is_training=True)

            with open(cached_train_features_file, "wb") as writer:
                pickle.dump(train_features, writer)

        logger.info("***** Running training *****")
        logger.info("  Num orig examples = %d", len(train_examples))
        logger.info("  Num split examples = %d", len(train_features))
        logger.info("  Batch size = %d", args.train_batch_size)
        logger.info("  Num steps = %d", num_train_steps)
        all_input_ids = torch.tensor([f.input_ids for f in train_features], dtype=torch.long)
        all_input_mask = torch.tensor([f.input_mask for f in train_features], dtype=torch.long)
        all_segment_ids = torch.tensor([f.segment_ids for f in train_features], dtype=torch.long)
        all_start_positions = torch.tensor([f.start_position for f in train_features], dtype=torch.long)
        all_end_positions = torch.tensor([f.end_position for f in train_features], dtype=torch.long)
        all_is_impossibles = torch.tensor([int(f.is_impossible) for f in train_features], dtype=torch.long)
        all_example_index = torch.arange(all_input_ids.size(0), dtype=torch.long)

        train_data = TensorDataset(all_input_ids, all_input_mask, all_segment_ids,
                                   all_start_positions, all_end_positions, all_is_impossibles, all_example_index)

        if args.local_rank == -1:
            train_sampler = RandomSampler(train_data)
        else:
            train_sampler = DistributedSampler(train_data)
        train_dataloader = DataLoader(train_data, sampler=train_sampler, batch_size=args.train_batch_size)
        step = 0
        for epoch in trange(int(args.num_train_epochs), desc="Epoch"):
            model.train()

            train_loss = 0
            for step, batch in enumerate(tqdm(train_dataloader, desc="Iteration", ncols=50)):
                if n_gpu == 1:
                    batch = tuple(t.to(device) for t in batch)  # multi-gpu does scattering it-self
                input_ids, input_mask, segment_ids, start_positions, end_positions, is_impossibles, example_indices = batch

                input_span_mask = np.zeros((input_ids.size(0), input_ids.size(1), input_ids.size(1)))
                for batch_idx, ex_idx in enumerate(example_indices):
                    train_feature = train_features[ex_idx.item()]
                    train_span_mask = train_feature.input_span_mask
                    for idx, i_mask in enumerate(train_span_mask):
                        for j in i_mask:
                            input_span_mask[batch_idx, idx, j] = 1
                input_span_mask = torch.tensor(input_span_mask, dtype=torch.long)
                input_span_mask = input_span_mask.to(device)

                loss = model(input_ids, segment_ids, input_mask, input_span_mask, start_positions,
                                                     end_positions, is_impossibles)
                if n_gpu > 1:
                    loss = loss.mean()  # mean() to average on multi-gpu.
                if args.gradient_accumulation_steps > 1:
                    loss = loss / args.gradient_accumulation_steps

                if args.fp16:
                    optimizer.backward(loss)
                else:
                    loss.backward()
                if (step + 1) % args.gradient_accumulation_steps == 0:
                    # modify learning rate with special warm up BERT uses
                    lr_this_step = args.learning_rate * warmup_linear(global_step / t_total, args.warmup_proportion)
                    for param_group in optimizer.param_groups:
                        param_group['lr'] = lr_this_step
                    optimizer.step()
                    optimizer.zero_grad()
                    global_step += 1
                train_loss += loss.item()
            train_loss = train_loss / (step + 1)
            TrainLoss.append(train_loss)

            # Save a trained model
            model_to_save = model.module if hasattr(model, 'module') else model  # Only save the model it-self
            output_model_file = os.path.join(args.output_dir, "epoch_" + str(epoch) + "_pytorch_model.bin")
            torch.save(model_to_save.state_dict(), output_model_file)

        with open(os.path.join(args.output_dir, "train_loss.pkl"), 'wb') as f:
            pickle.dump(TrainLoss, f)

    if args.do_predict and (args.local_rank == -1 or torch.distributed.get_rank() == 0):
        print("reading dev...")
        cached_eval_features_file = "eval_features.pkl"

        eval_examples = read_squad_examples(
            input_file=args.predict_file, input_tag_file=args.predict_tag_file, is_training=False)
        try:
            with open(cached_eval_features_file, "rb") as reader:
                eval_features = pickle.load(reader)
        except:
            eval_features = convert_examples_to_features(
                examples=eval_examples,
                tokenizer=tokenizer,
                max_seq_length=args.max_seq_length,
                doc_stride=args.doc_stride,
                max_query_length=args.max_query_length,
                is_training=False)
            with open(cached_eval_features_file, "wb") as writer:
                pickle.dump(eval_features, writer)

        all_input_ids = torch.tensor([f.input_ids for f in eval_features], dtype=torch.long)
        all_input_mask = torch.tensor([f.input_mask for f in eval_features], dtype=torch.long)
        all_segment_ids = torch.tensor([f.segment_ids for f in eval_features], dtype=torch.long)
        all_example_index = torch.arange(all_input_ids.size(0), dtype=torch.long)

        eval_data = TensorDataset(all_input_ids, all_input_mask, all_segment_ids, all_example_index)

        # Run prediction for full data
        eval_sampler = SequentialSampler(eval_data)
        eval_dataloader = DataLoader(eval_data, sampler=eval_sampler, batch_size=args.predict_batch_size)

        with open(os.path.join(args.output_dir, "train_loss.pkl"), 'rb') as f:
            TrainLoss = pickle.load(f)

        all_results = []

        for epoch in trange(int(args.num_train_epochs), desc="Epoch"):
            train_loss = TrainLoss[epoch]
            output_model_file = os.path.join(args.output_dir, "epoch_" + str(epoch) + "_pytorch_model.bin")
            model_state_dict = torch.load(output_model_file)
            model = BertForQuestionAnsweringSpanMask.from_pretrained(args.bert_model, state_dict=model_state_dict)
            model.to(device)
            model.eval()
            model.half()

            logger.info("Start evaluating")
            for input_ids, input_mask, segment_ids, example_indices in tqdm(eval_dataloader, desc="Evaluating",
                                                                            ncols=50):
                if len(all_results) % 1000 == 0:
                    logger.info("Processing example: %d" % (len(all_results)))
                input_ids = input_ids.to(device)
                input_mask = input_mask.to(device)
                segment_ids = segment_ids.to(device)
                input_span_mask = np.zeros((input_ids.size(0), input_ids.size(1), input_ids.size(1)))
                for batch_idx, ex_idx in enumerate(example_indices):
                    eval_feature = eval_features[ex_idx.item()]
                    eval_span_mask = eval_feature.input_span_mask
                    for idx, i_mask in enumerate(eval_span_mask):
                        for j in i_mask:
                            input_span_mask[batch_idx, idx, j] = 1
                input_span_mask = torch.tensor(input_span_mask, dtype=torch.long)
                input_span_mask = input_span_mask.to(device)

                with torch.no_grad():
                    batch_start_logits, batch_end_logits = model(input_ids, segment_ids, input_mask, input_span_mask)
                for i, example_index in enumerate(example_indices):
                    start_logits = batch_start_logits[i].detach().cpu().tolist()
                    end_logits = batch_end_logits[i].detach().cpu().tolist()
                    eval_feature = eval_features[example_index.item()]
                    unique_id = int(eval_feature.unique_id)
                    all_results.append(RawResult(unique_id=unique_id,
                                                 start_logits=start_logits,
                                                 end_logits=end_logits,
                                                 choice_logits=end_logits))
            output_prediction_file = os.path.join(args.output_dir, "epoch_" + str(epoch) + "_predictions.json")
            output_nbest_file = os.path.join(args.output_dir, "epoch_" + str(epoch) + "_nbest_predictions.json")
            output_null_log_odds_file = os.path.join(args.output_dir, "epoch_" + str(epoch) + "_null_odds.json")

            write_predictions(eval_examples, eval_features, all_results,
                              args.n_best_size, args.max_answer_length,
                              args.do_lower_case, output_prediction_file,
                              output_nbest_file, output_null_log_odds_file, args.verbose_logging,
                              True, args.null_score_diff_threshold)

            result = {'train_loss': train_loss}

            with open(os.path.join(args.output_dir, "result.txt"), "a") as writer:
                writer.write("\t\n***** Eval results Epoch %d  %s *****\t\n" % (epoch, time.strftime(
                    '%Y-%m-%d %H:%M:%S', time.localtime(time.time()))))
                for key in sorted(result.keys()):
                    logger.info("  %s = %s", key, str(result[key]))
                    writer.write("%s = %s\t" % (key, str(result[key])))
                    writer.write("\t\n")


if __name__ == "__main__":
    main()