run_qa.py

import collections
import logging
import sys
import shutil
import math
import time
from typing import Optional, Tuple
from dataclasses import dataclass, field

import transformers
import numpy as np
from tqdm.auto import tqdm
from datasets import load_dataset
from trainer_qa import QuestionAnsweringTrainer
from transformers import (
    TrainerCallback,
    AutoModelForQuestionAnswering,
    AutoTokenizer,
    DataCollatorWithPadding,
    EvalPrediction,
    HfArgumentParser,
    PreTrainedTokenizerFast,
    TrainingArguments,
    default_data_collator,
    set_seed,
)
from transformers.trainer_utils import get_last_checkpoint, is_main_process
from transformers.utils import check_min_version
from datasets import set_caching_enabled

from utils import *


# Will error if the minimal version of Transformers is not installed. Remove at your own risks.
check_min_version("4.5.0")

logger = logging.getLogger(__name__)


@dataclass
class ModelArguments:
    """
    Arguments pertaining to which model/config/tokenizer we are going to fine-tune from.
    """

    model_name_or_path: str = field(
        metadata={"help": "Path to pretrained model or model identifier from huggingface.co/models"}
    )
    tokenizer_name: Optional[str] = field(
        default=None, metadata={"help": "Pretrained tokenizer name or path if not the same as model_name"}
    )
    eval_per_epoch: Optional[int] = field(
        default=5,
        metadata={"help": "How many times it evaluates on dev set per epoch."},
    )

@dataclass
class DataTrainingArguments:
    """
    Arguments pertaining to what data we are going to input our model for training and eval.
    """

    train_jsonl: Optional[str] = field(
        default=None, metadata={"help": "Path for train dataset."}
    )
    eval_jsonl: Optional[str] = field(
        default=None, metadata={"help": "Path for eval dataset."}
    )
    eval_idx_lst: Optional[str] = field(
        default=None, metadata={"help": "Path for idx lst to subsample the eval dataset."}
    )
    output_pred_path: Optional[str] = field(
        default=None, metadata={"help": "Path for output predictions."}
    )
    output_metrics_only: bool = field(
        default=False, metadata={"help": "Whether to only output metrics when output_pred_path is set."}
    )
    overwrite_cache: bool = field(
        default=False, metadata={"help": "Overwrite the cached training and evaluation sets"}
    )
    preprocessing_num_workers: Optional[int] = field(
        default=None,
        metadata={"help": "The number of processes to use for the preprocessing."},
    )
    max_seq_length: int = field(
        default=384,
        metadata={
            "help": "The maximum total input sequence length after tokenization. Sequences longer "
            "than this will be truncated, sequences shorter will be padded."
        },
    )
    pad_to_max_length: bool = field(
        default=False,
        metadata={
            "help": "Whether to pad all samples to `max_seq_length`. "
            "If False, will pad the samples dynamically when batching to the maximum length in the batch (which can "
            "be faster on GPU but will be slower on TPU)."
        },
    )
    max_train_samples: Optional[int] = field(
        default=None,
        metadata={
            "help": "For debugging purposes or quicker training, truncate the number of training examples to this "
            "value if set."
        },
    )
    max_val_samples: Optional[int] = field(
        default=None,
        metadata={
            "help": "For debugging purposes or quicker training, truncate the number of validation examples to this "
            "value if set."
        },
    )
    max_test_samples: Optional[int] = field(
        default=None,
        metadata={
            "help": "For debugging purposes or quicker training, truncate the number of test examples to this "
            "value if set."
        },
    )
    version_2_with_negative: bool = field(
        default=False, metadata={"help": "If true, some of the examples do not have an answer."}
    )
    null_score_diff_threshold: float = field(
        default=0.0,
        metadata={
            "help": "The threshold used to select the null answer: if the best answer has a score that is less than "
            "the score of the null answer minus this threshold, the null answer is selected for this example. "
            "Only useful when `version_2_with_negative=True`."
        },
    )
    doc_stride: int = field(
        default=128,
        metadata={"help": "When splitting up a long document into chunks, how much stride to take between chunks."},
    )
    n_best_size: int = field(
        default=20,
        metadata={"help": "The total number of n-best predictions to generate when looking for an answer."},
    )
    max_answer_length: int = field(
        default=30,
        metadata={
            "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."
        },
    )
    disable_caching: bool = field(
        default=False,
        metadata={
            "help": "Disable caching."
        },
    )


def main():
    # See all possible arguments in src/transformers/training_args.py
    # or by passing the --help flag to this script.
    # We now keep distinct sets of args, for a cleaner separation of concerns.

    parser = HfArgumentParser((ModelArguments, DataTrainingArguments, TrainingArguments))
    model_args, data_args, training_args = read_args(parser, sys.argv)

    # Setup logging
    logging.basicConfig(
        format="%(asctime)s - %(levelname)s - %(name)s -   %(message)s",
        datefmt="%m/%d/%Y %H:%M:%S",
        handlers=[logging.StreamHandler(sys.stdout)],
    )
    logger.setLevel(logging.INFO if is_main_process(training_args.local_rank) else logging.WARN)

    if data_args.disable_caching:
        set_caching_enabled(False)
        logger.info(f"Caching disabled.")
    # Detecting last checkpoint.
    last_checkpoint = None
    if os.path.isdir(training_args.output_dir) and training_args.do_train and not training_args.overwrite_output_dir:
        last_checkpoint = get_last_checkpoint(training_args.output_dir)
        if last_checkpoint is None and len(os.listdir(training_args.output_dir)) > 0:
            raise ValueError(
                f"Output directory ({training_args.output_dir}) already exists and is not empty. "
                "Use --overwrite_output_dir to overcome."
            )
        elif last_checkpoint is not None:
            logger.info(
                f"Checkpoint detected, resuming training at {last_checkpoint}. To avoid this behavior, change "
                "the `--output_dir` or add `--overwrite_output_dir` to train from scratch."
            )

    if 'spanbert-large-cased' in model_args.model_name_or_path:
        training_args.fp16 = False
    else:
        training_args.fp16 = True

    # Log on each process the small summary:
    logger.warning(
        f"Process rank: {training_args.local_rank}, device: {training_args.device}, n_gpu: {training_args.n_gpu}"
        + f"distributed training: {bool(training_args.local_rank != -1)}, 16-bits training: {training_args.fp16}"
    )
    # Set the verbosity to info of the Transformers logger (on main process only):
    if is_main_process(training_args.local_rank):
        transformers.utils.logging.set_verbosity_info()
        transformers.utils.logging.enable_default_handler()
        transformers.utils.logging.enable_explicit_format()
    logger.info(f"Training/evaluation parameters {training_args}")

    # Set seed before initializing model.
    set_seed(training_args.seed)

    # Get the datasets: you can either provide your own CSV/JSON/TXT training and evaluation files (see below)
    # or just provide the name of one of the public datasets available on the hub at https://huggingface.co/datasets/
    # (the dataset will be downloaded automatically from the datasets Hub).
    #
    # For CSV/JSON files, this script will use the column called 'text' or the first column if no column called
    # 'text' is found. You can easily tweak this behavior (see below).
    #
    # In distributed training, the load_dataset function guarantee that only one local process can concurrently
    # download the dataset.
    data_files = {}
    if training_args.do_train:
        data_files['train'] = data_args.train_jsonl
    if training_args.do_eval:
        data_files['eval'] = data_args.eval_jsonl
        if not training_args.do_train:
            if os.path.exists(data_args.output_pred_path):
                logger.info(f'already exists: {data_args.output_pred_path}')
                exit(0)
    if data_args.disable_caching:
        tmp_folder = f'/tmp/{time.time()}'
        logger.info(f'Temporarily caching at {tmp_folder}.')
        datasets = load_dataset('json', data_files=data_files, cache_dir=tmp_folder)
    else:
        datasets = load_dataset('json', data_files=data_files)
    # See more about loading any type of standard or custom dataset (from files, python dict, pandas DataFrame, etc) at
    # https://huggingface.co/docs/datasets/loading_datasets.html.

    # Load pretrained model and tokenizer
    #
    # Distributed training:
    # The .from_pretrained methods guarantee that only one local process can concurrently
    # download model & vocab.
    if 'spanbert' in model_args.model_name_or_path:
        if model_args.tokenizer_name is not None and model_args.tokenizer_name != "none":
            tokenizer = AutoTokenizer.from_pretrained(model_args.tokenizer_name, do_lower_case=False)
        else:
            tokenizer = AutoTokenizer.from_pretrained(model_args.model_name_or_path, do_lower_case=False)
        logger.info("***We manually correct the tokenizer for spanbert to be case-sensitive.***")
    else:
        if model_args.tokenizer_name is not None and model_args.tokenizer_name != "none":
            tokenizer = AutoTokenizer.from_pretrained(model_args.tokenizer_name)
        else:
            tokenizer = AutoTokenizer.from_pretrained(model_args.model_name_or_path)
    model = AutoModelForQuestionAnswering.from_pretrained(model_args.model_name_or_path, output_hidden_states=False)
    model.resize_token_embeddings(len(tokenizer))

    # Add [INK] separator in MRQA datasets
    special_tokens_dict = {'additional_special_tokens': ['[INK]']}
    num_added_toks = tokenizer.add_special_tokens(special_tokens_dict)
    logger.info(f"{num_added_toks} additional tokens added.")
    model.resize_token_embeddings(len(tokenizer))

    # Tokenizer check: this script requires a fast tokenizer.
    if not isinstance(tokenizer, PreTrainedTokenizerFast):
        raise ValueError(
            "This example script only works for models that have a fast tokenizer. Checkout the big table of models "
            "at https://huggingface.co/transformers/index.html#bigtable to find the model types that meet this "
            "requirement"
        )

    # Preprocessing the datasets.
    # Preprocessing is slighlty different for training and evaluation.
    if training_args.do_train:
        column_names = datasets["train"].column_names
    elif training_args.do_eval:
        column_names = datasets["eval"].column_names
    else:
        raise ValueError()
    question_column_name = "question"
    context_column_name = "context"
    train_answer_column_name = "detected_answers"
    eval_answer_column_name = "answers"

    # Padding side determines if we do (question|context) or (context|question).
    pad_on_right = tokenizer.padding_side == "right"

    if data_args.max_seq_length > tokenizer.model_max_length:
        logger.warning(
            f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the"
            f"model ({tokenizer.model_max_length}). Using max_seq_length={tokenizer.model_max_length}."
        )
    max_seq_length = min(data_args.max_seq_length, tokenizer.model_max_length)

    # Training preprocessing
    def prepare_train_features(examples):
        # Tokenize our examples with truncation and maybe padding, but keep the overflows using a stride. This results
        # in one example possible giving several features when a context is long, each of those features having a
        # context that overlaps a bit the context of the previous feature.
        tokenized_examples = tokenizer(
            examples[question_column_name if pad_on_right else context_column_name],
            examples[context_column_name if pad_on_right else question_column_name],
            truncation="only_second" if pad_on_right else "only_first",
            max_length=max_seq_length,
            stride=data_args.doc_stride,
            return_overflowing_tokens=True,
            return_offsets_mapping=True,
            padding="max_length" if data_args.pad_to_max_length else False,
        )

        # Since one example might give us several features if it has a long context, we need a map from a feature to
        # its corresponding example. This key gives us just that.
        sample_mapping = tokenized_examples.pop("overflow_to_sample_mapping")
        # The offset mappings will give us a map from token to character position in the original context. This will
        # help us compute the start_positions and end_positions.
        offset_mapping = tokenized_examples.pop("offset_mapping")

        # Let's label those examples!
        tokenized_examples["start_positions"] = []
        tokenized_examples["end_positions"] = []

        for i, offsets in enumerate(offset_mapping):
            # We will label impossible answers with the index of the CLS token.
            input_ids = tokenized_examples["input_ids"][i]
            cls_index = input_ids.index(tokenizer.cls_token_id)

            # Grab the sequence corresponding to that example (to know what is the context and what is the question).
            sequence_ids = tokenized_examples.sequence_ids(i)

            # One example can give several spans, this is the index of the example containing this span of text.
            sample_index = sample_mapping[i]
            answers = examples[train_answer_column_name][sample_index]
            # If no answers are given, set the cls_index as answer.
            if len(answers["answer_start"]) == 0:
                tokenized_examples["start_positions"].append(cls_index)
                tokenized_examples["end_positions"].append(cls_index)
            else:
                # <del>If there are multiple annotations for the answer span, choose the first one</del> We discard examples with multiple answer spans during preprocessing
                # Start/end character index of the answer in the text.
                start_char = answers["answer_start"][0]
                end_char = start_char + len(answers["text"][0])

                # Start token index of the current span in the text.
                token_start_index = 0
                while sequence_ids[token_start_index] != (1 if pad_on_right else 0):
                    token_start_index += 1

                # End token index of the current span in the text.
                token_end_index = len(input_ids) - 1
                while sequence_ids[token_end_index] != (1 if pad_on_right else 0):
                    token_end_index -= 1

                # Detect if the answer is out of the span (in which case this feature is labeled with the CLS index).
                if not (offsets[token_start_index][0] <= start_char and offsets[token_end_index][1] >= end_char):
                    tokenized_examples["start_positions"].append(cls_index)
                    tokenized_examples["end_positions"].append(cls_index)
                else:
                    # Otherwise move the token_start_index and token_end_index to the two ends of the answer.
                    # Note: we could go after the last offset if the answer is the last word (edge case).
                    while token_start_index < len(offsets) and offsets[token_start_index][0] <= start_char:
                        token_start_index += 1
                    tokenized_examples["start_positions"].append(token_start_index - 1)
                    while offsets[token_end_index][1] >= end_char:
                        token_end_index -= 1
                    tokenized_examples["end_positions"].append(token_end_index + 1)

        return tokenized_examples

    if training_args.do_train:
        train_dataset = datasets["train"]
        if data_args.max_train_samples is not None:
            # We will select sample from whole data if agument is specified
            train_dataset = train_dataset.select(range(data_args.max_train_samples))
        # Create train feature from dataset
        train_dataset = train_dataset.map(
            prepare_train_features,
            batched=True,
            num_proc=data_args.preprocessing_num_workers,
            remove_columns=column_names,
            load_from_cache_file=not data_args.overwrite_cache,
        )
        if data_args.max_train_samples is not None:
            # Number of samples might increase during Feature Creation, We select only specified max samples
            train_dataset = train_dataset.select(range(data_args.max_train_samples))
        num_batch_per_epoch = math.ceil(len(train_dataset) / training_args.per_device_train_batch_size)
        eval_step = max(1, num_batch_per_epoch // model_args.eval_per_epoch)
        training_args.logging_steps = eval_step
        training_args.eval_steps = eval_step
        training_args.save_steps = eval_step

    # Validation preprocessing
    def prepare_validation_features(examples):
        # Tokenize our examples with truncation and maybe padding, but keep the overflows using a stride. This results
        # in one example possible giving several features when a context is long, each of those features having a
        # context that overlaps a bit the context of the previous feature.
        tokenized_examples = tokenizer(
            examples[question_column_name if pad_on_right else context_column_name],
            examples[context_column_name if pad_on_right else question_column_name],
            truncation="only_second" if pad_on_right else "only_first",
            max_length=max_seq_length,
            stride=data_args.doc_stride,
            return_overflowing_tokens=True,
            return_offsets_mapping=True,
            padding="max_length" if data_args.pad_to_max_length else False,
        )

        # Since one example might give us several features if it has a long context, we need a map from a feature to
        # its corresponding example. This key gives us just that.
        sample_mapping = tokenized_examples.pop("overflow_to_sample_mapping")

        # For evaluation, we will need to convert our predictions to substrings of the context, so we keep the
        # corresponding example_id and we will store the offset mappings.
        tokenized_examples["example_id"] = []

        for i in range(len(tokenized_examples["input_ids"])):
            # Grab the sequence corresponding to that example (to know what is the context and what is the question).
            sequence_ids = tokenized_examples.sequence_ids(i)
            context_index = 1 if pad_on_right else 0

            # One example can give several spans, this is the index of the example containing this span of text.
            sample_index = sample_mapping[i]
            tokenized_examples["example_id"].append(examples["id"][sample_index])

            # Set to None the offset_mapping that are not part of the context so it's easy to determine if a token
            # position is part of the context or not.
            tokenized_examples["offset_mapping"][i] = [
                (o if sequence_ids[k] == context_index else None)
                for k, o in enumerate(tokenized_examples["offset_mapping"][i])
            ]

        return tokenized_examples

    if training_args.do_eval:
        eval_examples = datasets["eval"]
        if data_args.max_val_samples is not None:
            # We will select sample from whole data
            eval_examples = eval_examples.select(range(data_args.max_val_samples))
        # Validation Feature Creation
        eval_dataset = eval_examples.map(
            prepare_validation_features,
            batched=True,
            num_proc=data_args.preprocessing_num_workers,
            remove_columns=column_names,
            load_from_cache_file=not data_args.overwrite_cache,
        )
        if data_args.max_val_samples is not None:
            # During Feature creation dataset samples might increase, we will select required samples again
            eval_dataset = eval_dataset.select(range(data_args.max_val_samples))

    # Data collator
    # We have already padded to max length if the corresponding flag is True, otherwise we need to pad in the data
    # collator.
    data_collator = (
        default_data_collator
        if data_args.pad_to_max_length
        else DataCollatorWithPadding(tokenizer, pad_to_multiple_of=8 if training_args.fp16 else None)
    )

    # Post-processing:
    def post_processing_function(examples, features, predictions, stage="eval"):
        # Post-processing: we match the start logits and end logits to answers in the original context.
        predictions = postprocess_qa_predictions(
            examples=examples,
            features=features,
            predictions=predictions,
            version_2_with_negative=data_args.version_2_with_negative,
            n_best_size=data_args.n_best_size,
            max_answer_length=data_args.max_answer_length,
            null_score_diff_threshold=data_args.null_score_diff_threshold,
            output_dir=training_args.output_dir,
            is_world_process_zero=trainer.is_world_process_zero(),
            prefix=stage,
        )
        # Format the result to the format the metric expects.
        if data_args.version_2_with_negative:
            formatted_predictions = [
                {"id": k, "prediction_text": v, "no_answer_probability": 0.0} for k, v in predictions.items()
            ]
        else:
            formatted_predictions = [{"id": k, "prediction_text": v} for k, v in predictions.items()]

        references = [{"id": ex["id"], "answers": ex[eval_answer_column_name]} for ex in examples]
        return EvalPrediction(predictions=formatted_predictions, label_ids=references)

    # metric = load_metric("squad_v2" if data_args.version_2_with_negative else "squad")
    #
    # def compute_metrics(p: EvalPrediction):
    #     return metric.compute(predictions=p.predictions, references=p.label_ids)

    class SaveLimitCallback(TrainerCallback):
        def on_train_end(self, args, state, control, **kwargs):
            if state.is_local_process_zero and args.save_total_limit == 0:
                checkpoints_to_be_deleted = [str(x) for x in Path(args.output_dir).glob("checkpoint-*")]
                if state.best_model_checkpoint is not None:
                    best_checkpoint_path_obj = Path(state.best_model_checkpoint)
                    os.remove(str(best_checkpoint_path_obj / 'optimizer.pt'))
                    best_checkpoint_name = best_checkpoint_path_obj.name
                    link_src = best_checkpoint_name
                    link_dst = str(Path(args.output_dir) / 'best')
                    try:
                        os.symlink(link_src, link_dst)
                    except FileExistsError:
                        os.remove(link_dst)
                        os.symlink(link_src, link_dst)
                    logger.info(f"Symbol link: {link_src} -> {link_dst}")
                    best_checkpoint = str(best_checkpoint_path_obj)
                    checkpoints_to_be_deleted = (set(checkpoints_to_be_deleted) - {best_checkpoint})
                for checkpoint in checkpoints_to_be_deleted:
                    logger.info(f"Deleting non-optimal checkpoint [{checkpoint}] due to args.save_total_limit==0")
                    shutil.rmtree(checkpoint)
                try:
                    checkpoint = str(Path(args.output_dir) / 'pytorch_model.bin')
                    logger.info(f"Deleting the last checkpoint [{checkpoint}] due to args.save_total_limit==0")
                    os.remove(checkpoint)
                except:
                    print(f"Error while deleting file {checkpoint}")

    # Initialize our Trainer
    trainer = QuestionAnsweringTrainer(
        model=model,
        args=training_args,
        train_dataset=train_dataset if training_args.do_train else None,
        eval_dataset=eval_dataset if training_args.do_eval else None,
        eval_examples=eval_examples if training_args.do_eval else None,
        tokenizer=tokenizer,
        data_collator=data_collator,
        post_process_function=post_processing_function,
        # compute_metrics=compute_metrics,
        output_pred_path=data_args.output_pred_path,
        output_metrics_only=data_args.output_metrics_only,
        callbacks=[SaveLimitCallback]
    )

    # Training
    if training_args.do_train:
        if training_args.overwrite_output_dir:  # if want to resume training, don't set overwrite_output_dir
            train_result = trainer.train()
        else:
            if last_checkpoint is not None:
                checkpoint = last_checkpoint
            elif os.path.isdir(model_args.model_name_or_path):
                checkpoint = model_args.model_name_or_path
            else:
                checkpoint = None
            logger.info(f'Resuming training from {checkpoint}...')
            train_result = trainer.train(resume_from_checkpoint=checkpoint)
        trainer.save_model()  # Saves the tokenizer too for easy upload

        metrics = train_result.metrics
        max_train_samples = (
            data_args.max_train_samples if data_args.max_train_samples is not None else len(train_dataset)
        )
        metrics["train_samples"] = min(max_train_samples, len(train_dataset))

        trainer.log_metrics("train", metrics)
        trainer.save_metrics("train", metrics)
        trainer.save_state()

    # Evaluation
    if training_args.do_eval:
        logger.info("*** Evaluate ***")
        metrics = trainer.evaluate()
        logger.info(metrics)

    if data_args.disable_caching:
        shutil.rmtree(tmp_folder)
        logger.info('Cache removed.')


def postprocess_qa_predictions(
    examples,
    features,
    predictions: Tuple[np.ndarray, np.ndarray],
    version_2_with_negative: bool = False,
    n_best_size: int = 20,
    max_answer_length: int = 30,
    null_score_diff_threshold: float = 0.0,
    output_dir: Optional[str] = None,
    prefix: Optional[str] = None,
    is_world_process_zero: bool = True,
):
    """
    Post-processes the predictions of a question-answering model to convert them to answers that are substrings of the
    original contexts. This is the base postprocessing functions for models that only return start and end logits.

    Args:
        examples: The non-preprocessed dataset (see the main script for more information).
        features: The processed dataset (see the main script for more information).
        predictions (:obj:`Tuple[np.ndarray, np.ndarray]`):
            The predictions of the model: two arrays containing the start logits and the end logits respectively. Its
            first dimension must match the number of elements of :obj:`features`.
        version_2_with_negative (:obj:`bool`, `optional`, defaults to :obj:`False`):
            Whether or not the underlying dataset contains examples with no answers.
        n_best_size (:obj:`int`, `optional`, defaults to 20):
            The total number of n-best predictions to generate when looking for an answer.
        max_answer_length (:obj:`int`, `optional`, defaults to 30):
            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.
        null_score_diff_threshold (:obj:`float`, `optional`, defaults to 0):
            The threshold used to select the null answer: if the best answer has a score that is less than the score of
            the null answer minus this threshold, the null answer is selected for this example (note that the score of
            the null answer for an example giving several features is the minimum of the scores for the null answer on
            each feature: all features must be aligned on the fact they `want` to predict a null answer).

            Only useful when :obj:`version_2_with_negative` is :obj:`True`.
        output_dir (:obj:`str`, `optional`):
            If provided, the dictionaries of predictions, n_best predictions (with their scores and logits) and, if
            :obj:`version_2_with_negative=True`, the dictionary of the scores differences between best and null
            answers, are saved in `output_dir`.
        prefix (:obj:`str`, `optional`):
            If provided, the dictionaries mentioned above are saved with `prefix` added to their names.
        is_world_process_zero (:obj:`bool`, `optional`, defaults to :obj:`True`):
            Whether this process is the main process or not (used to determine if logging/saves should be done).
    """
    assert len(predictions) == 2, "`predictions` should be a tuple with two elements (start_logits, end_logits)."
    all_start_logits, all_end_logits = predictions

    assert len(predictions[0]) == len(features), f"Got {len(predictions[0])} predictions and {len(features)} features."

    # Build a map example to its corresponding features.
    example_id_to_index = {k: i for i, k in enumerate(examples["id"])}
    features_per_example = collections.defaultdict(list)
    for i, feature in enumerate(features):
        features_per_example[example_id_to_index[feature["example_id"]]].append(i)

    # The dictionaries we have to fill.
    all_predictions = collections.OrderedDict()
    all_nbest_json = collections.OrderedDict()
    if version_2_with_negative:
        scores_diff_json = collections.OrderedDict()

    # Logging.
    logger.setLevel(logging.INFO if is_world_process_zero else logging.WARN)
    logger.info(f"Post-processing {len(examples)} example predictions split into {len(features)} features.")

    # Let's loop over all the examples!
    for example_index, example in enumerate(tqdm(examples, desc='post')):
        # Those are the indices of the features associated to the current example.
        feature_indices = features_per_example[example_index]

        min_null_prediction = None
        prelim_predictions = []

        # Looping through all the features associated to the current example.
        for feature_index in feature_indices:
            # We grab the predictions of the model for this feature.
            start_logits = all_start_logits[feature_index]
            end_logits = all_end_logits[feature_index]
            # This is what will allow us to map some the positions in our logits to span of texts in the original
            # context.
            offset_mapping = features[feature_index]["offset_mapping"]
            # Optional `token_is_max_context`, if provided we will remove answers that do not have the maximum context
            # available in the current feature.
            token_is_max_context = features[feature_index].get("token_is_max_context", None)

            # Update minimum null prediction.
            feature_null_score = start_logits[0] + end_logits[0]
            if min_null_prediction is None or min_null_prediction["score"] > feature_null_score:
                min_null_prediction = {
                    "offsets": (0, 0),
                    "score": feature_null_score,
                    "start_logit": start_logits[0],
                    "end_logit": end_logits[0],
                }

            # Go through all possibilities for the `n_best_size` greater start and end logits.
            start_indexes = np.argsort(start_logits)[-1 : -n_best_size - 1 : -1].tolist()
            end_indexes = np.argsort(end_logits)[-1 : -n_best_size - 1 : -1].tolist()
            for start_index in start_indexes:
                for end_index in end_indexes:
                    # Don't consider out-of-scope answers, either because the indices are out of bounds or correspond
                    # to part of the input_ids that are not in the context.
                    if (
                        start_index >= len(offset_mapping)
                        or end_index >= len(offset_mapping)
                        or offset_mapping[start_index] is None
                        or offset_mapping[end_index] is None
                    ):
                        continue
                    # Don't consider answers with a length that is either < 0 or > max_answer_length.
                    if end_index < start_index or end_index - start_index + 1 > max_answer_length:
                        continue
                    # Don't consider answer that don't have the maximum context available (if such information is
                    # provided).
                    if token_is_max_context is not None and not token_is_max_context.get(str(start_index), False):
                        continue
                    prelim_predictions.append(
                        {
                            "offsets": (offset_mapping[start_index][0], offset_mapping[end_index][1]),
                            "score": start_logits[start_index] + end_logits[end_index],
                            "start_logit": start_logits[start_index],
                            "end_logit": end_logits[end_index],
                        }
                    )
        if version_2_with_negative:
            # Add the minimum null prediction
            prelim_predictions.append(min_null_prediction)
            null_score = min_null_prediction["score"]

        # Only keep the best `n_best_size` predictions.
        predictions = sorted(prelim_predictions, key=lambda x: x["score"], reverse=True)[:n_best_size]

        # Add back the minimum null prediction if it was removed because of its low score.
        if version_2_with_negative and not any(p["offsets"] == (0, 0) for p in predictions):
            predictions.append(min_null_prediction)

        # Use the offsets to gather the answer text in the original context.
        context = example["context"]
        for pred in predictions:
            offsets = pred.pop("offsets")
            pred["text"] = context[offsets[0] : offsets[1]]

        # In the very rare edge case we have not a single non-null prediction, we create a fake prediction to avoid
        # failure.
        if len(predictions) == 0 or (len(predictions) == 1 and predictions[0]["text"] == ""):
            predictions.insert(0, {"text": "empty", "start_logit": 0.0, "end_logit": 0.0, "score": 0.0})

        # Compute the softmax of all scores (we do it with numpy to stay independent from torch/tf in this file, using
        # the LogSumExp trick).
        scores = np.array([pred.pop("score") for pred in predictions])
        exp_scores = np.exp(scores - np.max(scores))
        probs = exp_scores / exp_scores.sum()

        # Include the probabilities in our predictions.
        for prob, pred in zip(probs, predictions):
            pred["probability"] = prob

        # Pick the best prediction. If the null answer is not possible, this is easy.
        if not version_2_with_negative:
            all_predictions[example["id"]] = predictions[0]["text"]
        else:
            # Otherwise we first need to find the best non-empty prediction.
            i = 0
            while predictions[i]["text"] == "":
                i += 1
            best_non_null_pred = predictions[i]

            # Then we compare to the null prediction using the threshold.
            score_diff = null_score - best_non_null_pred["start_logit"] - best_non_null_pred["end_logit"]
            scores_diff_json[example["id"]] = float(score_diff)  # To be JSON-serializable.
            if score_diff > null_score_diff_threshold:
                all_predictions[example["id"]] = ""
            else:
                all_predictions[example["id"]] = best_non_null_pred["text"]

        # Make `predictions` JSON-serializable by casting np.float back to float.
        all_nbest_json[example["id"]] = [
            {k: (float(v) if isinstance(v, (np.float16, np.float32, np.float64)) else v) for k, v in pred.items()}
            for pred in predictions
        ]

    # If we have an output_dir, let's save all those dicts.
    # if output_dir is not None and output_dir.lower() != 'none':
    #     assert os.path.isdir(output_dir), f"{output_dir} is not a directory."
    #
    #     prediction_file = os.path.join(
    #         output_dir, "predictions.json" if prefix is None else f"{prefix}_predictions.json"
    #     )
    #     nbest_file = os.path.join(
    #         output_dir, "nbest_predictions.json" if prefix is None else f"{prefix}_nbest_predictions.json"
    #     )
    #     if version_2_with_negative:
    #         null_odds_file = os.path.join(
    #             output_dir, "null_odds.json" if prefix is None else f"{prefix}_null_odds.json"
    #         )
    #
    #     logger.info(f"Saving predictions to {prediction_file}.")
    #     with open(prediction_file, "w") as writer:
    #         writer.write(json.dumps(all_predictions, indent=4) + "\n")
    #     logger.info(f"Saving nbest_preds to {nbest_file}.")
    #     with open(nbest_file, "w") as writer:
    #         writer.write(json.dumps(all_nbest_json, indent=4) + "\n")
    #     if version_2_with_negative:
    #         logger.info(f"Saving null_odds to {null_odds_file}.")
    #         with open(null_odds_file, "w") as writer:
    #             writer.write(json.dumps(scores_diff_json, indent=4) + "\n")

    return all_predictions


def _mp_fn(index):
    # For xla_spawn (TPUs)
    main()


if __name__ == "__main__":
    main()