run_swag.py

# coding=utf-8
# Copyright 2018 The Google AI Language Team Authors and The HuggingFace 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.
"""BERT finetuning runner.
   Finetuning the library models for multiple choice on SWAG (Bert).
"""


import argparse
import csv
import glob
import logging
import os
import random

import numpy as np
import torch
from torch.utils.data import DataLoader, RandomSampler, SequentialSampler, TensorDataset
from torch.utils.data.distributed import DistributedSampler
from tqdm import tqdm, trange

from transformers import WEIGHTS_NAME, AdamW, AutoConfig, AutoTokenizer, get_linear_schedule_with_warmup
from transformers.modeling_auto import AutoModelForMultipleChoice


try:
    from torch.utils.tensorboard import SummaryWriter
except ImportError:
    from tensorboardX import SummaryWriter


logger = logging.getLogger(__name__)


class SwagExample(object):
    """A single training/test example for the SWAG dataset."""

    def __init__(self, swag_id, context_sentence, start_ending, ending_0, ending_1, ending_2, ending_3, label=None):
        self.swag_id = swag_id
        self.context_sentence = context_sentence
        self.start_ending = start_ending
        self.endings = [
            ending_0,
            ending_1,
            ending_2,
            ending_3,
        ]
        self.label = label

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

    def __repr__(self):
        attributes = [
            "swag_id: {}".format(self.swag_id),
            "context_sentence: {}".format(self.context_sentence),
            "start_ending: {}".format(self.start_ending),
            "ending_0: {}".format(self.endings[0]),
            "ending_1: {}".format(self.endings[1]),
            "ending_2: {}".format(self.endings[2]),
            "ending_3: {}".format(self.endings[3]),
        ]

        if self.label is not None:
            attributes.append("label: {}".format(self.label))

        return ", ".join(attributes)


class InputFeatures(object):
    def __init__(self, example_id, choices_features, label):
        self.example_id = example_id
        self.choices_features = [
            {"input_ids": input_ids, "input_mask": input_mask, "segment_ids": segment_ids}
            for _, input_ids, input_mask, segment_ids in choices_features
        ]
        self.label = label


def read_swag_examples(input_file, is_training=True):
    with open(input_file, "r", encoding="utf-8") as f:
        lines = list(csv.reader(f))

    if is_training and lines[0][-1] != "label":
        raise ValueError("For training, the input file must contain a label column.")

    examples = [
        SwagExample(
            swag_id=line[2],
            context_sentence=line[4],
            start_ending=line[5],  # in the swag dataset, the
            # common beginning of each
            # choice is stored in "sent2".
            ending_0=line[7],
            ending_1=line[8],
            ending_2=line[9],
            ending_3=line[10],
            label=int(line[11]) if is_training else None,
        )
        for line in lines[1:]  # we skip the line with the column names
    ]

    return examples


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

    # Swag is a multiple choice task. To perform this task using Bert,
    # we will use the formatting proposed in "Improving Language
    # Understanding by Generative Pre-Training" and suggested by
    # @jacobdevlin-google in this issue
    # https://github.com/google-research/bert/issues/38.
    #
    # Each choice will correspond to a sample on which we run the
    # inference. For a given Swag example, we will create the 4
    # following inputs:
    # - [CLS] context [SEP] choice_1 [SEP]
    # - [CLS] context [SEP] choice_2 [SEP]
    # - [CLS] context [SEP] choice_3 [SEP]
    # - [CLS] context [SEP] choice_4 [SEP]
    # The model will output a single value for each input. To get the
    # final decision of the model, we will run a softmax over these 4
    # outputs.
    features = []
    for example_index, example in tqdm(enumerate(examples)):
        context_tokens = tokenizer.tokenize(example.context_sentence)
        start_ending_tokens = tokenizer.tokenize(example.start_ending)

        choices_features = []
        for ending_index, ending in enumerate(example.endings):
            # We create a copy of the context tokens in order to be
            # able to shrink it according to ending_tokens
            context_tokens_choice = context_tokens[:]
            ending_tokens = start_ending_tokens + tokenizer.tokenize(ending)
            # Modifies `context_tokens_choice` and `ending_tokens` in
            # place so that the total length is less than the
            # specified length.  Account for [CLS], [SEP], [SEP] with
            # "- 3"
            _truncate_seq_pair(context_tokens_choice, ending_tokens, max_seq_length - 3)

            tokens = ["[CLS]"] + context_tokens_choice + ["[SEP]"] + ending_tokens + ["[SEP]"]
            segment_ids = [0] * (len(context_tokens_choice) + 2) + [1] * (len(ending_tokens) + 1)

            input_ids = tokenizer.convert_tokens_to_ids(tokens)
            input_mask = [1] * len(input_ids)

            # Zero-pad up to the sequence length.
            padding = [0] * (max_seq_length - len(input_ids))
            input_ids += padding
            input_mask += padding
            segment_ids += padding

            assert len(input_ids) == max_seq_length
            assert len(input_mask) == max_seq_length
            assert len(segment_ids) == max_seq_length

            choices_features.append((tokens, input_ids, input_mask, segment_ids))

        label = example.label
        if example_index < 5:
            logger.info("*** Example ***")
            logger.info("swag_id: {}".format(example.swag_id))
            for choice_idx, (tokens, input_ids, input_mask, segment_ids) in enumerate(choices_features):
                logger.info("choice: {}".format(choice_idx))
                logger.info("tokens: {}".format(" ".join(tokens)))
                logger.info("input_ids: {}".format(" ".join(map(str, input_ids))))
                logger.info("input_mask: {}".format(" ".join(map(str, input_mask))))
                logger.info("segment_ids: {}".format(" ".join(map(str, segment_ids))))
            if is_training:
                logger.info("label: {}".format(label))

        features.append(InputFeatures(example_id=example.swag_id, choices_features=choices_features, label=label))

    return features


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

    # This is a simple heuristic which will always truncate the longer sequence
    # one token at a time. This makes more sense than truncating an equal percent
    # of tokens from each, since if one sequence is very short then each token
    # that's truncated likely contains more information than a longer sequence.
    while True:
        total_length = len(tokens_a) + len(tokens_b)
        if total_length <= max_length:
            break
        if len(tokens_a) > len(tokens_b):
            tokens_a.pop()
        else:
            tokens_b.pop()


def accuracy(out, labels):
    outputs = np.argmax(out, axis=1)
    return np.sum(outputs == labels)


def select_field(features, field):
    return [[choice[field] for choice in feature.choices_features] for feature in features]


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


def load_and_cache_examples(args, tokenizer, evaluate=False, output_examples=False):
    if args.local_rank not in [-1, 0]:
        torch.distributed.barrier()  # Make sure only the first process in distributed training process the dataset, and the others will use the cache

    # Load data features from cache or dataset file
    input_file = args.predict_file if evaluate else args.train_file
    cached_features_file = os.path.join(
        os.path.dirname(input_file),
        "cached_{}_{}_{}".format(
            "dev" if evaluate else "train",
            list(filter(None, args.model_name_or_path.split("/"))).pop(),
            str(args.max_seq_length),
        ),
    )
    if os.path.exists(cached_features_file) and not args.overwrite_cache and not output_examples:
        logger.info("Loading features from cached file %s", cached_features_file)
        features = torch.load(cached_features_file)
    else:
        logger.info("Creating features from dataset file at %s", input_file)
        examples = read_swag_examples(input_file)
        features = convert_examples_to_features(examples, tokenizer, args.max_seq_length, not evaluate)

        if args.local_rank in [-1, 0]:
            logger.info("Saving features into cached file %s", cached_features_file)
            torch.save(features, cached_features_file)

    if args.local_rank == 0:
        torch.distributed.barrier()  # Make sure only the first process in distributed training process the dataset, and the others will use the cache

    # Convert to Tensors and build dataset
    all_input_ids = torch.tensor(select_field(features, "input_ids"), dtype=torch.long)
    all_input_mask = torch.tensor(select_field(features, "input_mask"), dtype=torch.long)
    all_segment_ids = torch.tensor(select_field(features, "segment_ids"), dtype=torch.long)
    all_label = torch.tensor([f.label for f in features], dtype=torch.long)

    if evaluate:
        dataset = TensorDataset(all_input_ids, all_input_mask, all_segment_ids, all_label)
    else:
        dataset = TensorDataset(all_input_ids, all_input_mask, all_segment_ids, all_label)

    if output_examples:
        return dataset, examples, features
    return dataset


def train(args, train_dataset, model, tokenizer):
    """ Train the model """
    if args.local_rank in [-1, 0]:
        tb_writer = SummaryWriter()

    args.train_batch_size = args.per_gpu_train_batch_size * max(1, args.n_gpu)
    train_sampler = RandomSampler(train_dataset) if args.local_rank == -1 else DistributedSampler(train_dataset)
    train_dataloader = DataLoader(train_dataset, sampler=train_sampler, batch_size=args.train_batch_size)

    if args.max_steps > 0:
        t_total = args.max_steps
        args.num_train_epochs = args.max_steps // (len(train_dataloader) // args.gradient_accumulation_steps) + 1
    else:
        t_total = len(train_dataloader) // args.gradient_accumulation_steps * args.num_train_epochs

    # Prepare optimizer and schedule (linear warmup and decay)
    no_decay = ["bias", "LayerNorm.weight"]
    optimizer_grouped_parameters = [
        {
            "params": [p for n, p in model.named_parameters() if not any(nd in n for nd in no_decay)],
            "weight_decay": args.weight_decay,
        },
        {"params": [p for n, p in model.named_parameters() if any(nd in n for nd in no_decay)], "weight_decay": 0.0},
    ]
    optimizer = AdamW(optimizer_grouped_parameters, lr=args.learning_rate, eps=args.adam_epsilon)
    scheduler = get_linear_schedule_with_warmup(
        optimizer, num_warmup_steps=args.warmup_steps, num_training_steps=t_total
    )
    if args.fp16:
        try:
            from apex import amp
        except ImportError:
            raise ImportError("Please install apex from https://www.github.com/nvidia/apex to use fp16 training.")
        model, optimizer = amp.initialize(model, optimizer, opt_level=args.fp16_opt_level)

    # multi-gpu training (should be after apex fp16 initialization)
    if args.n_gpu > 1:
        model = torch.nn.DataParallel(model)

    # Distributed training (should be after apex fp16 initialization)
    if args.local_rank != -1:
        model = torch.nn.parallel.DistributedDataParallel(
            model, device_ids=[args.local_rank], output_device=args.local_rank, find_unused_parameters=True
        )

    # Train!
    logger.info("***** Running training *****")
    logger.info("  Num examples = %d", len(train_dataset))
    logger.info("  Num Epochs = %d", args.num_train_epochs)
    logger.info("  Instantaneous batch size per GPU = %d", args.per_gpu_train_batch_size)
    logger.info(
        "  Total train batch size (w. parallel, distributed & accumulation) = %d",
        args.train_batch_size
        * args.gradient_accumulation_steps
        * (torch.distributed.get_world_size() if args.local_rank != -1 else 1),
    )
    logger.info("  Gradient Accumulation steps = %d", args.gradient_accumulation_steps)
    logger.info("  Total optimization steps = %d", t_total)

    global_step = 0
    tr_loss, logging_loss = 0.0, 0.0
    model.zero_grad()
    train_iterator = trange(int(args.num_train_epochs), desc="Epoch", disable=args.local_rank not in [-1, 0])
    set_seed(args)  # Added here for reproductibility
    for _ in train_iterator:
        epoch_iterator = tqdm(train_dataloader, desc="Iteration", disable=args.local_rank not in [-1, 0])
        for step, batch in enumerate(epoch_iterator):
            model.train()
            batch = tuple(t.to(args.device) for t in batch)
            inputs = {
                "input_ids": batch[0],
                "attention_mask": batch[1],
                # 'token_type_ids':  None if args.model_type == 'xlm' else batch[2],
                "token_type_ids": batch[2],
                "labels": batch[3],
            }
            # if args.model_type in ['xlnet', 'xlm']:
            #     inputs.update({'cls_index': batch[5],
            #                    'p_mask':       batch[6]})
            outputs = model(**inputs)
            loss = outputs[0]  # model outputs are always tuple in transformers (see doc)

            if args.n_gpu > 1:
                loss = loss.mean()  # mean() to average on multi-gpu parallel (not distributed) training
            if args.gradient_accumulation_steps > 1:
                loss = loss / args.gradient_accumulation_steps

            if args.fp16:
                with amp.scale_loss(loss, optimizer) as scaled_loss:
                    scaled_loss.backward()
                torch.nn.utils.clip_grad_norm_(amp.master_params(optimizer), args.max_grad_norm)
            else:
                loss.backward()
                torch.nn.utils.clip_grad_norm_(model.parameters(), args.max_grad_norm)

            tr_loss += loss.item()
            if (step + 1) % args.gradient_accumulation_steps == 0:
                optimizer.step()
                scheduler.step()  # Update learning rate schedule
                model.zero_grad()
                global_step += 1

                if args.local_rank in [-1, 0] and args.logging_steps > 0 and global_step % args.logging_steps == 0:
                    # Log metrics
                    if (
                        args.local_rank == -1 and args.evaluate_during_training
                    ):  # Only evaluate when single GPU otherwise metrics may not average well
                        results = evaluate(args, model, tokenizer)
                        for key, value in results.items():
                            tb_writer.add_scalar("eval_{}".format(key), value, global_step)
                    tb_writer.add_scalar("lr", scheduler.get_lr()[0], global_step)
                    tb_writer.add_scalar("loss", (tr_loss - logging_loss) / args.logging_steps, global_step)
                    logging_loss = tr_loss

                if args.local_rank in [-1, 0] and args.save_steps > 0 and global_step % args.save_steps == 0:
                    # Save model checkpoint
                    output_dir = os.path.join(args.output_dir, "checkpoint-{}".format(global_step))
                    model_to_save = (
                        model.module if hasattr(model, "module") else model
                    )  # Take care of distributed/parallel training
                    model_to_save.save_pretrained(output_dir)
                    tokenizer.save_vocabulary(output_dir)
                    torch.save(args, os.path.join(output_dir, "training_args.bin"))
                    logger.info("Saving model checkpoint to %s", output_dir)

            if args.max_steps > 0 and global_step > args.max_steps:
                epoch_iterator.close()
                break
        if args.max_steps > 0 and global_step > args.max_steps:
            train_iterator.close()
            break

    if args.local_rank in [-1, 0]:
        tb_writer.close()

    return global_step, tr_loss / global_step


def evaluate(args, model, tokenizer, prefix=""):
    dataset, examples, features = load_and_cache_examples(args, tokenizer, evaluate=True, output_examples=True)

    if not os.path.exists(args.output_dir) and args.local_rank in [-1, 0]:
        os.makedirs(args.output_dir)

    args.eval_batch_size = args.per_gpu_eval_batch_size * max(1, args.n_gpu)
    # Note that DistributedSampler samples randomly
    eval_sampler = SequentialSampler(dataset) if args.local_rank == -1 else DistributedSampler(dataset)
    eval_dataloader = DataLoader(dataset, sampler=eval_sampler, batch_size=args.eval_batch_size)

    # Eval!
    logger.info("***** Running evaluation {} *****".format(prefix))
    logger.info("  Num examples = %d", len(dataset))
    logger.info("  Batch size = %d", args.eval_batch_size)

    eval_loss, eval_accuracy = 0, 0
    nb_eval_steps, nb_eval_examples = 0, 0

    for batch in tqdm(eval_dataloader, desc="Evaluating"):
        model.eval()
        batch = tuple(t.to(args.device) for t in batch)
        with torch.no_grad():
            inputs = {
                "input_ids": batch[0],
                "attention_mask": batch[1],
                # 'token_type_ids': None if args.model_type == 'xlm' else batch[2]  # XLM don't use segment_ids
                "token_type_ids": batch[2],
                "labels": batch[3],
            }

            # if args.model_type in ['xlnet', 'xlm']:
            #     inputs.update({'cls_index': batch[4],
            #                    'p_mask':    batch[5]})
            outputs = model(**inputs)
            tmp_eval_loss, logits = outputs[:2]
            eval_loss += tmp_eval_loss.mean().item()

        logits = logits.detach().cpu().numpy()
        label_ids = inputs["labels"].to("cpu").numpy()
        tmp_eval_accuracy = accuracy(logits, label_ids)
        eval_accuracy += tmp_eval_accuracy

        nb_eval_steps += 1
        nb_eval_examples += inputs["input_ids"].size(0)

    eval_loss = eval_loss / nb_eval_steps
    eval_accuracy = eval_accuracy / nb_eval_examples
    result = {"eval_loss": eval_loss, "eval_accuracy": eval_accuracy}

    output_eval_file = os.path.join(args.output_dir, "eval_results.txt")
    with open(output_eval_file, "w") as writer:
        logger.info("***** Eval results *****")
        for key in sorted(result.keys()):
            logger.info("%s = %s", key, str(result[key]))
            writer.write("%s = %s\n" % (key, str(result[key])))

    return result


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

    # Required parameters
    parser.add_argument(
        "--train_file", default=None, type=str, required=True, help="SWAG csv for training. E.g., train.csv"
    )
    parser.add_argument(
        "--predict_file",
        default=None,
        type=str,
        required=True,
        help="SWAG csv for predictions. E.g., val.csv or test.csv",
    )
    parser.add_argument(
        "--model_name_or_path",
        default=None,
        type=str,
        required=True,
        help="Path to pretrained model or model identifier from huggingface.co/models",
    )
    parser.add_argument(
        "--output_dir",
        default=None,
        type=str,
        required=True,
        help="The output directory where the model checkpoints and predictions will be written.",
    )

    # Other parameters
    parser.add_argument(
        "--config_name", default="", type=str, help="Pretrained config name or path if not the same as model_name"
    )
    parser.add_argument(
        "--tokenizer_name",
        default="",
        type=str,
        help="Pretrained tokenizer name or path if not the same as model_name",
    )
    parser.add_argument(
        "--max_seq_length",
        default=384,
        type=int,
        help="The maximum total input sequence length after tokenization. Sequences "
        "longer than this will be truncated, and sequences shorter than this will be padded.",
    )
    parser.add_argument("--do_train", action="store_true", help="Whether to run training.")
    parser.add_argument("--do_eval", action="store_true", help="Whether to run eval on the dev set.")
    parser.add_argument(
        "--evaluate_during_training", action="store_true", help="Rul evaluation during training at each logging step."
    )

    parser.add_argument("--per_gpu_train_batch_size", default=8, type=int, help="Batch size per GPU/CPU for training.")
    parser.add_argument(
        "--per_gpu_eval_batch_size", default=8, type=int, help="Batch size per GPU/CPU for evaluation."
    )
    parser.add_argument("--learning_rate", default=5e-5, type=float, help="The initial learning rate for Adam.")
    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("--weight_decay", default=0.0, type=float, help="Weight deay if we apply some.")
    parser.add_argument("--adam_epsilon", default=1e-8, type=float, help="Epsilon for Adam optimizer.")
    parser.add_argument("--max_grad_norm", default=1.0, type=float, help="Max gradient norm.")
    parser.add_argument(
        "--num_train_epochs", default=3.0, type=float, help="Total number of training epochs to perform."
    )
    parser.add_argument(
        "--max_steps",
        default=-1,
        type=int,
        help="If > 0: set total number of training steps to perform. Override num_train_epochs.",
    )
    parser.add_argument("--warmup_steps", default=0, type=int, help="Linear warmup over warmup_steps.")

    parser.add_argument("--logging_steps", type=int, default=50, help="Log every X updates steps.")
    parser.add_argument("--save_steps", type=int, default=50, help="Save checkpoint every X updates steps.")
    parser.add_argument(
        "--eval_all_checkpoints",
        action="store_true",
        help="Evaluate all checkpoints starting with the same prefix as model_name ending and ending with step number",
    )
    parser.add_argument("--no_cuda", action="store_true", help="Whether not to use CUDA when available")
    parser.add_argument(
        "--overwrite_output_dir", action="store_true", help="Overwrite the content of the output directory"
    )
    parser.add_argument(
        "--overwrite_cache", action="store_true", help="Overwrite the cached training and evaluation sets"
    )
    parser.add_argument("--seed", type=int, default=42, help="random seed for initialization")

    parser.add_argument("--local_rank", type=int, default=-1, help="local_rank for distributed training on gpus")
    parser.add_argument(
        "--fp16",
        action="store_true",
        help="Whether to use 16-bit (mixed) precision (through NVIDIA apex) instead of 32-bit",
    )
    parser.add_argument(
        "--fp16_opt_level",
        type=str,
        default="O1",
        help="For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']."
        "See details at https://nvidia.github.io/apex/amp.html",
    )
    parser.add_argument("--server_ip", type=str, default="", help="Can be used for distant debugging.")
    parser.add_argument("--server_port", type=str, default="", help="Can be used for distant debugging.")
    args = parser.parse_args()

    if (
        os.path.exists(args.output_dir)
        and os.listdir(args.output_dir)
        and args.do_train
        and not args.overwrite_output_dir
    ):
        raise ValueError(
            "Output directory ({}) already exists and is not empty. Use --overwrite_output_dir to overcome.".format(
                args.output_dir
            )
        )

    # Setup distant debugging if needed
    if args.server_ip and args.server_port:
        # Distant debugging - see https://code.visualstudio.com/docs/python/debugging#_attach-to-a-local-script
        import ptvsd

        print("Waiting for debugger attach")
        ptvsd.enable_attach(address=(args.server_ip, args.server_port), redirect_output=True)
        ptvsd.wait_for_attach()

    # Setup CUDA, GPU & distributed training
    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")
        args.n_gpu = 0 if args.no_cuda else torch.cuda.device_count()
    else:  # Initializes the distributed backend which will take care of sychronizing nodes/GPUs
        torch.cuda.set_device(args.local_rank)
        device = torch.device("cuda", args.local_rank)
        torch.distributed.init_process_group(backend="nccl")
        args.n_gpu = 1
    args.device = device

    # Setup logging
    logging.basicConfig(
        format="%(asctime)s - %(levelname)s - %(name)s -   %(message)s",
        datefmt="%m/%d/%Y %H:%M:%S",
        level=logging.INFO if args.local_rank in [-1, 0] else logging.WARN,
    )
    logger.warning(
        "Process rank: %s, device: %s, n_gpu: %s, distributed training: %s, 16-bits training: %s",
        args.local_rank,
        device,
        args.n_gpu,
        bool(args.local_rank != -1),
        args.fp16,
    )

    # Set seed
    set_seed(args)

    # Load pretrained model and tokenizer
    if args.local_rank not in [-1, 0]:
        torch.distributed.barrier()  # Make sure only the first process in distributed training will download model & vocab

    config = AutoConfig.from_pretrained(args.config_name if args.config_name else args.model_name_or_path)
    tokenizer = AutoTokenizer.from_pretrained(args.tokenizer_name if args.tokenizer_name else args.model_name_or_path,)
    model = AutoModelForMultipleChoice.from_pretrained(
        args.model_name_or_path, from_tf=bool(".ckpt" in args.model_name_or_path), config=config
    )

    if args.local_rank == 0:
        torch.distributed.barrier()  # Make sure only the first process in distributed training will download model & vocab

    model.to(args.device)

    logger.info("Training/evaluation parameters %s", args)

    # Training
    if args.do_train:
        train_dataset = load_and_cache_examples(args, tokenizer, evaluate=False, output_examples=False)
        global_step, tr_loss = train(args, train_dataset, model, tokenizer)
        logger.info(" global_step = %s, average loss = %s", global_step, tr_loss)

    # Save the trained model and the tokenizer
    if args.local_rank == -1 or torch.distributed.get_rank() == 0:
        logger.info("Saving model checkpoint to %s", args.output_dir)
        # Save a trained model, configuration and tokenizer using `save_pretrained()`.
        # They can then be reloaded using `from_pretrained()`
        model_to_save = (
            model.module if hasattr(model, "module") else model
        )  # Take care of distributed/parallel training
        model_to_save.save_pretrained(args.output_dir)
        tokenizer.save_pretrained(args.output_dir)

        # Good practice: save your training arguments together with the trained model
        torch.save(args, os.path.join(args.output_dir, "training_args.bin"))

        # Load a trained model and vocabulary that you have fine-tuned
        model = AutoModelForMultipleChoice.from_pretrained(args.output_dir)
        tokenizer = AutoTokenizer.from_pretrained(args.output_dir)
        model.to(args.device)

    # Evaluation - we can ask to evaluate all the checkpoints (sub-directories) in a directory
    results = {}
    if args.do_eval and args.local_rank in [-1, 0]:
        if args.do_train:
            checkpoints = [args.output_dir]
        else:
            # if do_train is False and do_eval is true, load model directly from pretrained.
            checkpoints = [args.model_name_or_path]

        if args.eval_all_checkpoints:
            checkpoints = list(
                os.path.dirname(c) for c in sorted(glob.glob(args.output_dir + "/**/" + WEIGHTS_NAME, recursive=True))
            )
            logging.getLogger("transformers.modeling_utils").setLevel(logging.WARN)  # Reduce model loading logs

        logger.info("Evaluate the following checkpoints: %s", checkpoints)

        for checkpoint in checkpoints:
            # Reload the model
            global_step = checkpoint.split("-")[-1] if len(checkpoints) > 1 else ""
            model = AutoModelForMultipleChoice.from_pretrained(checkpoint)
            tokenizer = AutoTokenizer.from_pretrained(checkpoint)
            model.to(args.device)

            # Evaluate
            result = evaluate(args, model, tokenizer, prefix=global_step)

            result = dict((k + ("_{}".format(global_step) if global_step else ""), v) for k, v in result.items())
            results.update(result)

    logger.info("Results: {}".format(results))

    return results


if __name__ == "__main__":
    main()