trainer.py

# coding=utf-8
# Copyright 2020-present the HuggingFace Inc. team.
#
# 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.
"""
The Trainer class, to easily train a 🤗 Transformers from scratch or finetune it on a new task.
"""

import collections
import inspect
import math
import os
import random
import re
import shutil
import sys
import time
import warnings
from pathlib import Path
from typing import TYPE_CHECKING, Any, Callable, Dict, List, Optional, Tuple, Union

from tqdm.auto import tqdm


# Integrations must be imported before ML frameworks:
from .integrations import (  # isort: split
    default_hp_search_backend,
    get_reporting_integration_callbacks,
    hp_params,
    is_fairscale_available,
    is_optuna_available,
    is_ray_tune_available,
    is_sigopt_available,
    run_hp_search_optuna,
    run_hp_search_ray,
    run_hp_search_sigopt,
)

import numpy as np
import torch
from packaging import version
from torch import nn
from torch.utils.data import DataLoader, Dataset, IterableDataset, RandomSampler, SequentialSampler
from torch.utils.data.distributed import DistributedSampler

from huggingface_hub import Repository

from . import __version__
from .configuration_utils import PretrainedConfig
from .data.data_collator import DataCollator, DataCollatorWithPadding, default_data_collator
from .debug_utils import DebugOption, DebugUnderflowOverflow
from .deepspeed import deepspeed_init, is_deepspeed_zero3_enabled
from .dependency_versions_check import dep_version_check
from .file_utils import (
    CONFIG_NAME,
    WEIGHTS_NAME,
    get_full_repo_name,
    is_apex_available,
    is_datasets_available,
    is_in_notebook,
    is_sagemaker_dp_enabled,
    is_sagemaker_mp_enabled,
    is_torch_tpu_available,
)
from .modelcard import TrainingSummary
from .modeling_utils import PreTrainedModel, unwrap_model
from .models.auto.modeling_auto import MODEL_FOR_QUESTION_ANSWERING_MAPPING_NAMES
from .optimization import Adafactor, AdamW, get_scheduler
from .tokenization_utils_base import PreTrainedTokenizerBase
from .trainer_callback import (
    CallbackHandler,
    DefaultFlowCallback,
    PrinterCallback,
    ProgressCallback,
    TrainerCallback,
    TrainerControl,
    TrainerState,
)
from .trainer_pt_utils import (
    DistributedLengthGroupedSampler,
    DistributedSamplerWithLoop,
    DistributedTensorGatherer,
    IterableDatasetShard,
    LabelSmoother,
    LengthGroupedSampler,
    SequentialDistributedSampler,
    ShardSampler,
    distributed_broadcast_scalars,
    distributed_concat,
    find_batch_size,
    get_parameter_names,
    nested_concat,
    nested_detach,
    nested_numpify,
    nested_truncate,
    nested_xla_mesh_reduce,
    reissue_pt_warnings,
)
from .trainer_utils import (
    PREFIX_CHECKPOINT_DIR,
    BestRun,
    EvalLoopOutput,
    EvalPrediction,
    HPSearchBackend,
    HubStrategy,
    IntervalStrategy,
    PredictionOutput,
    ShardedDDPOption,
    TrainerMemoryTracker,
    TrainOutput,
    default_compute_objective,
    default_hp_space,
    denumpify_detensorize,
    get_last_checkpoint,
    number_of_arguments,
    set_seed,
    speed_metrics,
)
from .training_args import ParallelMode, TrainingArguments
from .utils import logging


_is_torch_generator_available = False
_is_native_amp_available = False

DEFAULT_CALLBACKS = [DefaultFlowCallback]
DEFAULT_PROGRESS_CALLBACK = ProgressCallback

if is_in_notebook():
    from .utils.notebook import NotebookProgressCallback

    DEFAULT_PROGRESS_CALLBACK = NotebookProgressCallback

if is_apex_available():
    from apex import amp

if version.parse(torch.__version__) >= version.parse("1.6"):
    _is_torch_generator_available = True
    _is_native_amp_available = True
    from torch.cuda.amp import autocast

if is_datasets_available():
    import datasets

if is_torch_tpu_available():
    import torch_xla.core.xla_model as xm
    import torch_xla.debug.metrics as met
    import torch_xla.distributed.parallel_loader as pl

if is_fairscale_available():
    dep_version_check("fairscale")
    import fairscale
    from fairscale.nn.data_parallel import FullyShardedDataParallel as FullyShardedDDP
    from fairscale.nn.data_parallel import ShardedDataParallel as ShardedDDP
    from fairscale.nn.wrap import auto_wrap
    from fairscale.optim import OSS
    from fairscale.optim.grad_scaler import ShardedGradScaler

if is_sagemaker_dp_enabled():
    import smdistributed.dataparallel.torch.distributed as dist
    from smdistributed.dataparallel.torch.parallel.distributed import DistributedDataParallel as DDP
else:
    import torch.distributed as dist

if is_sagemaker_mp_enabled():
    import smdistributed.modelparallel.torch as smp

    from .trainer_pt_utils import smp_forward_backward, smp_forward_only, smp_gather, smp_nested_concat


if TYPE_CHECKING:
    import optuna

logger = logging.get_logger(__name__)


# Name of the files used for checkpointing
TRAINING_ARGS_NAME = "training_args.bin"
TRAINER_STATE_NAME = "trainer_state.json"
OPTIMIZER_NAME = "optimizer.pt"
SCHEDULER_NAME = "scheduler.pt"
SCALER_NAME = "scaler.pt"


class Trainer:
    """
    Trainer is a simple but feature-complete training and eval loop for PyTorch, optimized for 🤗 Transformers.

    Args:
        model (:class:`~transformers.PreTrainedModel` or :obj:`torch.nn.Module`, `optional`):
            The model to train, evaluate or use for predictions. If not provided, a ``model_init`` must be passed.

            .. note::

                :class:`~transformers.Trainer` is optimized to work with the :class:`~transformers.PreTrainedModel`
                provided by the library. You can still use your own models defined as :obj:`torch.nn.Module` as long as
                they work the same way as the 🤗 Transformers models.
        args (:class:`~transformers.TrainingArguments`, `optional`):
            The arguments to tweak for training. Will default to a basic instance of
            :class:`~transformers.TrainingArguments` with the ``output_dir`` set to a directory named `tmp_trainer` in
            the current directory if not provided.
        data_collator (:obj:`DataCollator`, `optional`):
            The function to use to form a batch from a list of elements of :obj:`train_dataset` or :obj:`eval_dataset`.
            Will default to :func:`~transformers.default_data_collator` if no ``tokenizer`` is provided, an instance of
            :func:`~transformers.DataCollatorWithPadding` otherwise.
        train_dataset (:obj:`torch.utils.data.Dataset` or :obj:`torch.utils.data.IterableDataset`, `optional`):
            The dataset to use for training. If it is an :obj:`datasets.Dataset`, columns not accepted by the
            ``model.forward()`` method are automatically removed.

            Note that if it's a :obj:`torch.utils.data.IterableDataset` with some randomization and you are training in
            a distributed fashion, your iterable dataset should either use a internal attribute :obj:`generator` that
            is a :obj:`torch.Generator` for the randomization that must be identical on all processes (and the Trainer
            will manually set the seed of this :obj:`generator` at each epoch) or have a :obj:`set_epoch()` method that
            internally sets the seed of the RNGs used.
        eval_dataset (:obj:`torch.utils.data.Dataset`, `optional`):
             The dataset to use for evaluation. If it is an :obj:`datasets.Dataset`, columns not accepted by the
             ``model.forward()`` method are automatically removed.
        tokenizer (:class:`PreTrainedTokenizerBase`, `optional`):
            The tokenizer used to preprocess the data. If provided, will be used to automatically pad the inputs the
            maximum length when batching inputs, and it will be saved along the model to make it easier to rerun an
            interrupted training or reuse the fine-tuned model.
        model_init (:obj:`Callable[[], PreTrainedModel]`, `optional`):
            A function that instantiates the model to be used. If provided, each call to
            :meth:`~transformers.Trainer.train` will start from a new instance of the model as given by this function.

            The function may have zero argument, or a single one containing the optuna/Ray Tune/SigOpt trial object, to
            be able to choose different architectures according to hyper parameters (such as layer count, sizes of
            inner layers, dropout probabilities etc).
        compute_metrics (:obj:`Callable[[EvalPrediction], Dict]`, `optional`):
            The function that will be used to compute metrics at evaluation. Must take a
            :class:`~transformers.EvalPrediction` and return a dictionary string to metric values.
        callbacks (List of :obj:`~transformers.TrainerCallback`, `optional`):
            A list of callbacks to customize the training loop. Will add those to the list of default callbacks
            detailed in :doc:`here <callback>`.

            If you want to remove one of the default callbacks used, use the :meth:`Trainer.remove_callback` method.
        optimizers (:obj:`Tuple[torch.optim.Optimizer, torch.optim.lr_scheduler.LambdaLR]`, `optional`): A tuple
            containing the optimizer and the scheduler to use. Will default to an instance of
            :class:`~transformers.AdamW` on your model and a scheduler given by
            :func:`~transformers.get_linear_schedule_with_warmup` controlled by :obj:`args`.

    Important attributes:

        - **model** -- Always points to the core model. If using a transformers model, it will be a
          :class:`~transformers.PreTrainedModel` subclass.
        - **model_wrapped** -- Always points to the most external model in case one or more other modules wrap the
          original model. This is the model that should be used for the forward pass. For example, under ``DeepSpeed``,
          the inner model is wrapped in ``DeepSpeed`` and then again in ``torch.nn.DistributedDataParallel``. If the
          inner model hasn't been wrapped, then ``self.model_wrapped`` is the same as ``self.model``.
        - **is_model_parallel** -- Whether or not a model has been switched to a model parallel mode (different from
          data parallelism, this means some of the model layers are split on different GPUs).
        - **place_model_on_device** -- Whether or not to automatically place the model on the device - it will be set
          to :obj:`False` if model parallel or deepspeed is used, or if the default
          ``TrainingArguments.place_model_on_device`` is overridden to return :obj:`False` .
        - **is_in_train** -- Whether or not a model is currently running ``train`` (e.g. when ``evaluate`` is called
          while in ``train``)

    """

    from .trainer_pt_utils import _get_learning_rate, log_metrics, metrics_format, save_metrics, save_state

    def __init__(
        self,
        model: Union[PreTrainedModel, nn.Module] = None,
        args: TrainingArguments = None,
        data_collator: Optional[DataCollator] = None,
        train_dataset: Optional[Dataset] = None,
        eval_dataset: Optional[Dataset] = None,
        tokenizer: Optional[PreTrainedTokenizerBase] = None,
        model_init: Callable[[], PreTrainedModel] = None,
        compute_metrics: Optional[Callable[[EvalPrediction], Dict]] = None,
        callbacks: Optional[List[TrainerCallback]] = None,
        optimizers: Tuple[torch.optim.Optimizer, torch.optim.lr_scheduler.LambdaLR] = (None, None),
    ):
        if args is None:
            output_dir = "tmp_trainer"
            logger.info(f"No `TrainingArguments` passed, using `output_dir={output_dir}`.")
            args = TrainingArguments(output_dir=output_dir)
        self.args = args
        # Seed must be set before instantiating the model when using model
        set_seed(self.args.seed)
        self.hp_name = None
        self.deepspeed = None
        self.is_in_train = False

        # memory metrics - must set up as early as possible
        self._memory_tracker = TrainerMemoryTracker(self.args.skip_memory_metrics)
        self._memory_tracker.start()

        # set the correct log level depending on the node
        log_level = args.get_process_log_level()
        logging.set_verbosity(log_level)

        # force device and distributed setup init explicitly
        args._setup_devices

        if model is None:
            if model_init is not None:
                self.model_init = model_init
                model = self.call_model_init()
            else:
                raise RuntimeError("`Trainer` requires either a `model` or `model_init` argument")
        else:
            if model_init is not None:
                warnings.warn(
                    "`Trainer` requires either a `model` or `model_init` argument, but not both. "
                    "`model_init` will overwrite your model when calling the `train` method. This will become a fatal error in the next release.",
                    FutureWarning,
                )
            self.model_init = model_init

        if hasattr(model, "is_parallelizable") and model.is_parallelizable and model.model_parallel:
            self.is_model_parallel = True
        else:
            self.is_model_parallel = False

        # Setup Sharded DDP training
        self.sharded_ddp = None
        if len(args.sharded_ddp) > 0:
            if args.deepspeed:
                raise ValueError(
                    "Using --sharded_ddp xxx together with --deepspeed is not possible, deactivate one of those flags."
                )

            if args.local_rank == -1:
                raise ValueError("Using sharded DDP only works in distributed training.")
            elif not is_fairscale_available():
                raise ImportError("Sharded DDP training requires fairscale: `pip install fairscale`.")
            elif ShardedDDPOption.SIMPLE not in args.sharded_ddp and FullyShardedDDP is None:
                raise ImportError(
                    "Sharded DDP in a mode other than simple training requires fairscale version >= 0.3, found "
                    f"{fairscale.__version__}. Upgrade your fairscale library: `pip install --upgrade fairscale`."
                )
            elif ShardedDDPOption.SIMPLE in args.sharded_ddp:
                self.sharded_ddp = ShardedDDPOption.SIMPLE
            elif ShardedDDPOption.ZERO_DP_2 in args.sharded_ddp:
                self.sharded_ddp = ShardedDDPOption.ZERO_DP_2
            elif ShardedDDPOption.ZERO_DP_3 in args.sharded_ddp:
                self.sharded_ddp = ShardedDDPOption.ZERO_DP_3

        # one place to sort out whether to place the model on device or not
        # postpone switching model to cuda when:
        # 1. MP - since we are trying to fit a much bigger than 1 gpu model
        # 2. fp16-enabled DeepSpeed loads the model in half the size and it doesn't need .to() anyway,
        #    and we only use deepspeed for training at the moment
        # 3. full fp16 eval - since the model needs to be half'ed first
        # 4. Sharded DDP - same as MP
        self.place_model_on_device = args.place_model_on_device
        if (
            self.is_model_parallel
            or args.deepspeed
            or (args.fp16_full_eval and not args.do_train)
            or (self.sharded_ddp in [ShardedDDPOption.ZERO_DP_2, ShardedDDPOption.ZERO_DP_3])
        ):
            self.place_model_on_device = False

        default_collator = default_data_collator if tokenizer is None else DataCollatorWithPadding(tokenizer)
        self.data_collator = data_collator if data_collator is not None else default_collator
        self.train_dataset = train_dataset
        self.eval_dataset = eval_dataset
        self.tokenizer = tokenizer

        if self.place_model_on_device:
            self._move_model_to_device(model, args.device)

        # Force n_gpu to 1 to avoid DataParallel as MP will manage the GPUs
        if self.is_model_parallel:
            self.args._n_gpu = 1

        # later use `self.model is self.model_wrapped` to check if it's wrapped or not
        self.model_wrapped = model
        self.model = model

        self.compute_metrics = compute_metrics
        self.optimizer, self.lr_scheduler = optimizers
        if model_init is not None and (self.optimizer is not None or self.lr_scheduler is not None):
            raise RuntimeError(
                "Passing a `model_init` is incompatible with providing the `optimizers` argument. "
                "You should subclass `Trainer` and override the `create_optimizer_and_scheduler` method."
            )
        default_callbacks = DEFAULT_CALLBACKS + get_reporting_integration_callbacks(self.args.report_to)
        callbacks = default_callbacks if callbacks is None else default_callbacks + callbacks
        self.callback_handler = CallbackHandler(
            callbacks, self.model, self.tokenizer, self.optimizer, self.lr_scheduler
        )
        self.add_callback(PrinterCallback if self.args.disable_tqdm else DEFAULT_PROGRESS_CALLBACK)

        # Will be set to True by `self._setup_loggers()` on first call to `self.log()`.
        self._loggers_initialized = False

        # Create clone of distant repo and output directory if needed
        if self.args.push_to_hub:
            self.init_git_repo()
            # In case of pull, we need to make sure every process has the latest.
            if is_torch_tpu_available():
                xm.rendezvous("init git repo")
            elif args.local_rank != -1:
                dist.barrier()

        if self.args.should_save:
            os.makedirs(self.args.output_dir, exist_ok=True)

        if not callable(self.data_collator) and callable(getattr(self.data_collator, "collate_batch", None)):
            raise ValueError("The `data_collator` should be a simple callable (function, class with `__call__`).")

        if args.max_steps > 0:
            logger.info("max_steps is given, it will override any value given in num_train_epochs")

        if train_dataset is not None and not isinstance(train_dataset, collections.abc.Sized) and args.max_steps <= 0:
            raise ValueError("train_dataset does not implement __len__, max_steps has to be specified")

        self._signature_columns = None

        # Mixed precision setup
        self.use_apex = False
        self.use_amp = False
        self.fp16_backend = None

        if args.fp16:
            if args.fp16_backend == "auto":
                self.fp16_backend = "amp" if _is_native_amp_available else "apex"
            else:
                self.fp16_backend = args.fp16_backend
            logger.info(f"Using {self.fp16_backend} fp16 backend")

        if args.fp16 and not args.deepspeed:  # deepspeed manages its own fp16
            if self.fp16_backend == "amp":
                self.use_amp = True
                if is_sagemaker_mp_enabled():
                    self.scaler = smp.amp.GradScaler()
                elif self.sharded_ddp is not None:
                    self.scaler = ShardedGradScaler()
                else:
                    self.scaler = torch.cuda.amp.GradScaler()
            else:
                if not is_apex_available():
                    raise ImportError(
                        "Using FP16 with APEX but APEX is not installed, please refer to https://www.github.com/nvidia/apex."
                    )
                self.use_apex = True

        # FP16 + model parallelism in SageMaker: gradient clipping does not work for now so we raise a helpful error.
        if is_sagemaker_mp_enabled() and self.use_amp and args.max_grad_norm is not None and args.max_grad_norm > 0:
            raise ValueError(
                "SageMaker Model Parallelism in mixed precision mode does not support gradient clipping yet. Pass "
                "along 'max_grad_norm': 0 in your hyperparameters."
            )

        # Label smoothing
        if self.args.label_smoothing_factor != 0:
            self.label_smoother = LabelSmoother(epsilon=self.args.label_smoothing_factor)
        else:
            self.label_smoother = None

        self.state = TrainerState()
        self.control = TrainerControl()
        # Internal variable to count flos in each process, will be accumulated in `self.state.total_flos` then
        # returned to 0 every time flos need to be logged
        self.current_flos = 0
        self.hp_search_backend = None
        self.use_tune_checkpoints = False
        default_label_names = (
            ["start_positions", "end_positions"]
            if type(self.model).__name__ in MODEL_FOR_QUESTION_ANSWERING_MAPPING_NAMES.values()
            else ["labels"]
        )
        self.label_names = default_label_names if self.args.label_names is None else self.args.label_names
        self.control = self.callback_handler.on_init_end(self.args, self.state, self.control)

        # very last
        self._memory_tracker.stop_and_update_metrics()

    def add_callback(self, callback):
        """
        Add a callback to the current list of :class:`~transformer.TrainerCallback`.

        Args:
           callback (:obj:`type` or :class:`~transformer.TrainerCallback`):
               A :class:`~transformer.TrainerCallback` class or an instance of a :class:`~transformer.TrainerCallback`.
               In the first case, will instantiate a member of that class.
        """
        self.callback_handler.add_callback(callback)

    def pop_callback(self, callback):
        """
        Remove a callback from the current list of :class:`~transformer.TrainerCallback` and returns it.

        If the callback is not found, returns :obj:`None` (and no error is raised).

        Args:
           callback (:obj:`type` or :class:`~transformer.TrainerCallback`):
               A :class:`~transformer.TrainerCallback` class or an instance of a :class:`~transformer.TrainerCallback`.
               In the first case, will pop the first member of that class found in the list of callbacks.

        Returns:
            :class:`~transformer.TrainerCallback`: The callback removed, if found.
        """
        return self.callback_handler.pop_callback(callback)

    def remove_callback(self, callback):
        """
        Remove a callback from the current list of :class:`~transformer.TrainerCallback`.

        Args:
           callback (:obj:`type` or :class:`~transformer.TrainerCallback`):
               A :class:`~transformer.TrainerCallback` class or an instance of a :class:`~transformer.TrainerCallback`.
               In the first case, will remove the first member of that class found in the list of callbacks.
        """
        self.callback_handler.remove_callback(callback)

    def _move_model_to_device(self, model, device):
        model = model.to(device)
        # Moving a model to an XLA device disconnects the tied weights, so we have to retie them.
        if self.args.parallel_mode == ParallelMode.TPU and hasattr(model, "tie_weights"):
            model.tie_weights()

    def _remove_unused_columns(self, dataset: "datasets.Dataset", description: Optional[str] = None):
        if not self.args.remove_unused_columns:
            return dataset
        if self._signature_columns is None:
            # Inspect model forward signature to keep only the arguments it accepts.
            signature = inspect.signature(self.model.forward)
            self._signature_columns = list(signature.parameters.keys())
            # Labels may be named label or label_ids, the default data collator handles that.
            self._signature_columns += ["label", "label_ids"]
        columns = [k for k in self._signature_columns if k in dataset.column_names]
        ignored_columns = list(set(dataset.column_names) - set(self._signature_columns))
        if len(ignored_columns) > 0:
            dset_description = "" if description is None else f"in the {description} set "
            logger.info(
                f"The following columns {dset_description} don't have a corresponding argument in "
                f"`{self.model.__class__.__name__}.forward` and have been ignored: {', '.join(ignored_columns)}."
            )

        if version.parse(datasets.__version__) < version.parse("1.4.0"):
            dataset.set_format(
                type=dataset.format["type"], columns=columns, format_kwargs=dataset.format["format_kwargs"]
            )
            return dataset
        else:
            return dataset.remove_columns(ignored_columns)

    def _get_train_sampler(self) -> Optional[torch.utils.data.Sampler]:
        if not isinstance(self.train_dataset, collections.abc.Sized):
            return None

        generator = None
        if self.args.world_size <= 1 and _is_torch_generator_available:
            generator = torch.Generator()
            generator.manual_seed(int(torch.empty((), dtype=torch.int64).random_().item()))

        # Build the sampler.
        if self.args.group_by_length:
            if is_datasets_available() and isinstance(self.train_dataset, datasets.Dataset):
                lengths = (
                    self.train_dataset[self.args.length_column_name]
                    if self.args.length_column_name in self.train_dataset.column_names
                    else None
                )
            else:
                lengths = None
            model_input_name = self.tokenizer.model_input_names[0] if self.tokenizer is not None else None
            if self.args.world_size <= 1:
                return LengthGroupedSampler(
                    self.args.train_batch_size,
                    dataset=self.train_dataset,
                    lengths=lengths,
                    model_input_name=model_input_name,
                    generator=generator,
                )
            else:
                return DistributedLengthGroupedSampler(
                    self.args.train_batch_size,
                    dataset=self.train_dataset,
                    num_replicas=self.args.world_size,
                    rank=self.args.process_index,
                    lengths=lengths,
                    model_input_name=model_input_name,
                    seed=self.args.seed,
                )

        else:
            if self.args.world_size <= 1:
                if _is_torch_generator_available:
                    return RandomSampler(self.train_dataset, generator=generator)
                return RandomSampler(self.train_dataset)
            elif (
                self.args.parallel_mode in [ParallelMode.TPU, ParallelMode.SAGEMAKER_MODEL_PARALLEL]
                and not self.args.dataloader_drop_last
            ):
                # Use a loop for TPUs when drop_last is False to have all batches have the same size.
                return DistributedSamplerWithLoop(
                    self.train_dataset,
                    batch_size=self.args.per_device_train_batch_size,
                    num_replicas=self.args.world_size,
                    rank=self.args.process_index,
                    seed=self.args.seed,
                )
            else:
                return DistributedSampler(
                    self.train_dataset,
                    num_replicas=self.args.world_size,
                    rank=self.args.process_index,
                    seed=self.args.seed,
                )

    def get_train_dataloader(self) -> DataLoader:
        """
        Returns the training :class:`~torch.utils.data.DataLoader`.

        Will use no sampler if :obj:`self.train_dataset` does not implement :obj:`__len__`, a random sampler (adapted
        to distributed training if necessary) otherwise.

        Subclass and override this method if you want to inject some custom behavior.
        """
        if self.train_dataset is None:
            raise ValueError("Trainer: training requires a train_dataset.")

        train_dataset = self.train_dataset
        if is_datasets_available() and isinstance(train_dataset, datasets.Dataset):
            train_dataset = self._remove_unused_columns(train_dataset, description="training")

        if isinstance(train_dataset, torch.utils.data.IterableDataset):
            if self.args.world_size > 1:
                train_dataset = IterableDatasetShard(
                    train_dataset,
                    batch_size=self.args.train_batch_size,
                    drop_last=self.args.dataloader_drop_last,
                    num_processes=self.args.world_size,
                    process_index=self.args.process_index,
                )

            return DataLoader(
                train_dataset,
                batch_size=self.args.train_batch_size,
                collate_fn=self.data_collator,
                num_workers=self.args.dataloader_num_workers,
                pin_memory=self.args.dataloader_pin_memory,
            )

        train_sampler = self._get_train_sampler()

        return DataLoader(
            train_dataset,
            batch_size=self.args.train_batch_size,
            sampler=train_sampler,
            collate_fn=self.data_collator,
            drop_last=self.args.dataloader_drop_last,
            num_workers=self.args.dataloader_num_workers,
            pin_memory=self.args.dataloader_pin_memory,
        )

    def _get_eval_sampler(self, eval_dataset: Dataset) -> Optional[torch.utils.data.Sampler]:
        # Deprecated code
        if self.args.use_legacy_prediction_loop:
            if is_torch_tpu_available():
                return SequentialDistributedSampler(
                    eval_dataset, num_replicas=xm.xrt_world_size(), rank=xm.get_ordinal()
                )
            elif is_sagemaker_mp_enabled():
                return SequentialDistributedSampler(
                    eval_dataset,
                    num_replicas=smp.dp_size(),
                    rank=smp.dp_rank(),
                    batch_size=self.args.per_device_eval_batch_size,
                )
            elif self.args.local_rank != -1:
                return SequentialDistributedSampler(eval_dataset)
            else:
                return SequentialSampler(eval_dataset)

        if self.args.world_size <= 1:
            return SequentialSampler(eval_dataset)
        else:
            return ShardSampler(
                eval_dataset,
                batch_size=self.args.per_device_eval_batch_size,
                num_processes=self.args.world_size,
                process_index=self.args.process_index,
            )

    def get_eval_dataloader(self, eval_dataset: Optional[Dataset] = None) -> DataLoader:
        """
        Returns the evaluation :class:`~torch.utils.data.DataLoader`.

        Subclass and override this method if you want to inject some custom behavior.

        Args:
            eval_dataset (:obj:`torch.utils.data.Dataset`, `optional`):
                If provided, will override :obj:`self.eval_dataset`. If it is an :obj:`datasets.Dataset`, columns not
                accepted by the ``model.forward()`` method are automatically removed. It must implement :obj:`__len__`.
        """
        if eval_dataset is None and self.eval_dataset is None:
            raise ValueError("Trainer: evaluation requires an eval_dataset.")
        eval_dataset = eval_dataset if eval_dataset is not None else self.eval_dataset

        if is_datasets_available() and isinstance(eval_dataset, datasets.Dataset):
            eval_dataset = self._remove_unused_columns(eval_dataset, description="evaluation")

        if isinstance(eval_dataset, torch.utils.data.IterableDataset):
            if self.args.world_size > 1:
                eval_dataset = IterableDatasetShard(
                    eval_dataset,
                    batch_size=self.args.eval_batch_size,
                    drop_last=self.args.dataloader_drop_last,
                    num_processes=self.args.world_size,
                    process_index=self.args.process_index,
                )
            return DataLoader(
                eval_dataset,
                batch_size=self.args.eval_batch_size,
                collate_fn=self.data_collator,
                num_workers=self.args.dataloader_num_workers,
                pin_memory=self.args.dataloader_pin_memory,
            )

        eval_sampler = self._get_eval_sampler(eval_dataset)

        return DataLoader(
            eval_dataset,
            sampler=eval_sampler,
            batch_size=self.args.eval_batch_size,
            collate_fn=self.data_collator,
            drop_last=self.args.dataloader_drop_last,
            num_workers=self.args.dataloader_num_workers,
            pin_memory=self.args.dataloader_pin_memory,
        )

    def get_test_dataloader(self, test_dataset: Dataset) -> DataLoader:
        """
        Returns the test :class:`~torch.utils.data.DataLoader`.

        Subclass and override this method if you want to inject some custom behavior.

        Args:
            test_dataset (:obj:`torch.utils.data.Dataset`, `optional`):
                The test dataset to use. If it is an :obj:`datasets.Dataset`, columns not accepted by the
                ``model.forward()`` method are automatically removed. It must implement :obj:`__len__`.
        """
        if is_datasets_available() and isinstance(test_dataset, datasets.Dataset):
            test_dataset = self._remove_unused_columns(test_dataset, description="test")

        if isinstance(test_dataset, torch.utils.data.IterableDataset):
            if self.args.world_size > 1:
                test_dataset = IterableDatasetShard(
                    test_dataset,
                    batch_size=self.args.eval_batch_size,
                    drop_last=self.args.dataloader_drop_last,
                    num_processes=self.args.world_size,
                    process_index=self.args.process_index,
                )
            return DataLoader(
                test_dataset,
                batch_size=self.args.eval_batch_size,
                collate_fn=self.data_collator,
                num_workers=self.args.dataloader_num_workers,
                pin_memory=self.args.dataloader_pin_memory,
            )

        test_sampler = self._get_eval_sampler(test_dataset)

        # We use the same batch_size as for eval.
        return DataLoader(
            test_dataset,
            sampler=test_sampler,
            batch_size=self.args.eval_batch_size,
            collate_fn=self.data_collator,
            drop_last=self.args.dataloader_drop_last,
            pin_memory=self.args.dataloader_pin_memory,
        )

    def create_optimizer_and_scheduler(self, num_training_steps: int):
        """
        Setup the optimizer and the learning rate scheduler.

        We provide a reasonable default that works well. If you want to use something else, you can pass a tuple in the
        Trainer's init through :obj:`optimizers`, or subclass and override this method (or :obj:`create_optimizer`
        and/or :obj:`create_scheduler`) in a subclass.
        """
        self.create_optimizer()
        self.create_scheduler(num_training_steps=num_training_steps, optimizer=self.optimizer)

    def create_optimizer(self):
        """
        Setup the optimizer.

        We provide a reasonable default that works well. If you want to use something else, you can pass a tuple in the
        Trainer's init through :obj:`optimizers`, or subclass and override this method in a subclass.
        """
        if self.optimizer is None:
            decay_parameters = get_parameter_names(self.model, [nn.LayerNorm])
            decay_parameters = [name for name in decay_parameters if "bias" not in name]
            optimizer_grouped_parameters = [
                {
                    "params": [p for n, p in self.model.named_parameters() if n in decay_parameters],
                    "weight_decay": self.args.weight_decay,
                },
                {
                    "params": [p for n, p in self.model.named_parameters() if n not in decay_parameters],
                    "weight_decay": 0.0,
                },
            ]
            optimizer_cls = Adafactor if self.args.adafactor else AdamW
            if self.args.adafactor:
                optimizer_cls = Adafactor
                optimizer_kwargs = {"scale_parameter": False, "relative_step": False}
            else:
                optimizer_cls = AdamW
                optimizer_kwargs = {
                    "betas": (self.args.adam_beta1, self.args.adam_beta2),
                    "eps": self.args.adam_epsilon,
                }
            optimizer_kwargs["lr"] = self.args.learning_rate
            if self.sharded_ddp == ShardedDDPOption.SIMPLE:
                self.optimizer = OSS(
                    params=optimizer_grouped_parameters,
                    optim=optimizer_cls,
                    **optimizer_kwargs,
                )
            else:
                self.optimizer = optimizer_cls(optimizer_grouped_parameters, **optimizer_kwargs)

        if is_sagemaker_mp_enabled():
            self.optimizer = smp.DistributedOptimizer(self.optimizer)

        return self.optimizer

    def create_scheduler(self, num_training_steps: int, optimizer: torch.optim.Optimizer = None):
        """
        Setup the scheduler. The optimizer of the trainer must have been set up either before this method is called or
        passed as an argument.

        Args:
            num_training_steps (int): The number of training steps to do.
        """
        if self.lr_scheduler is None:
            self.lr_scheduler = get_scheduler(
                self.args.lr_scheduler_type,
                optimizer=self.optimizer if optimizer is None else optimizer,
                num_warmup_steps=self.args.get_warmup_steps(num_training_steps),
                num_training_steps=num_training_steps,
            )
        return self.lr_scheduler

    def num_examples(self, dataloader: DataLoader) -> int:
        """
        Helper to get number of samples in a :class:`~torch.utils.data.DataLoader` by accessing its dataset.

        Will raise an exception if the underlying dataset does not implement method :obj:`__len__`
        """
        return len(dataloader.dataset)

    def _hp_search_setup(self, trial: Union["optuna.Trial", Dict[str, Any]]):
        """HP search setup code"""
        self._trial = trial

        if self.hp_search_backend is None or trial is None:
            return
        if self.hp_search_backend == HPSearchBackend.OPTUNA:
            params = self.hp_space(trial)
        elif self.hp_search_backend == HPSearchBackend.RAY:
            params = trial
            params.pop("wandb", None)
        elif self.hp_search_backend == HPSearchBackend.SIGOPT:
            params = {k: int(v) if isinstance(v, str) else v for k, v in trial.assignments.items()}

        for key, value in params.items():
            if not hasattr(self.args, key):
                logger.warn(
                    f"Trying to set {key} in the hyperparameter search but there is no corresponding field in `TrainingArguments`."
                )
                continue
            old_attr = getattr(self.args, key, None)
            # Casting value to the proper type
            if old_attr is not None:
                value = type(old_attr)(value)
            setattr(self.args, key, value)
        if self.hp_search_backend == HPSearchBackend.OPTUNA:
            logger.info("Trial:", trial.params)
        if self.hp_search_backend == HPSearchBackend.SIGOPT:
            logger.info(f"SigOpt Assignments: {trial.assignments}")
        if self.args.deepspeed:
            # Rebuild the deepspeed config to reflect the updated training parameters
            from transformers.deepspeed import HfDeepSpeedConfig

            self.args.hf_deepspeed_config = HfDeepSpeedConfig(self.args)

    def _report_to_hp_search(
        self, trial: Union["optuna.Trial", Dict[str, Any]], epoch: int, metrics: Dict[str, float]
    ):
        if self.hp_search_backend is None or trial is None:
            return
        self.objective = self.compute_objective(metrics.copy())
        if self.hp_search_backend == HPSearchBackend.OPTUNA:
            import optuna

            trial.report(self.objective, epoch)
            if trial.should_prune():
                raise optuna.TrialPruned()
        elif self.hp_search_backend == HPSearchBackend.RAY:
            from ray import tune

            if self.control.should_save:
                self._tune_save_checkpoint()
            tune.report(objective=self.objective, **metrics)

    def _tune_save_checkpoint(self):
        from ray import tune

        if not self.use_tune_checkpoints:
            return
        with tune.checkpoint_dir(step=self.state.global_step) as checkpoint_dir:
            output_dir = os.path.join(checkpoint_dir, f"{PREFIX_CHECKPOINT_DIR}-{self.state.global_step}")
            self.save_model(output_dir)
            if self.args.should_save:
                self.state.save_to_json(os.path.join(output_dir, TRAINER_STATE_NAME))
                torch.save(self.optimizer.state_dict(), os.path.join(output_dir, OPTIMIZER_NAME))
                torch.save(self.lr_scheduler.state_dict(), os.path.join(output_dir, SCHEDULER_NAME))

    def call_model_init(self, trial=None):
        model_init_argcount = number_of_arguments(self.model_init)
        if model_init_argcount == 0:
            model = self.model_init()
        elif model_init_argcount == 1:
            model = self.model_init(trial)
        else:
            raise RuntimeError("model_init should have 0 or 1 argument.")

        if model is None:
            raise RuntimeError("model_init should not return None.")

        return model

    def _wrap_model(self, model, training=True):
        if is_sagemaker_mp_enabled():
            # Wrapping the base model twice in a DistributedModel will raise an error.
            if isinstance(self.model_wrapped, smp.model.DistributedModel):
                return self.model_wrapped
            return smp.DistributedModel(model, backward_passes_per_step=self.args.gradient_accumulation_steps)

        # already initialized its own DDP and AMP
        if self.deepspeed:
            return self.deepspeed

        # train/eval could be run multiple-times - if already wrapped, don't re-wrap it again
        if unwrap_model(model) is not model:
            return model

        # Mixed precision training with apex (torch < 1.6)
        if self.use_apex and training:
            model, self.optimizer = amp.initialize(model, self.optimizer, opt_level=self.args.fp16_opt_level)

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

        # Note: in torch.distributed mode, there's no point in wrapping the model
        # inside a DistributedDataParallel as we'll be under `no_grad` anyways.
        if not training:
            return model

        # Distributed training (should be after apex fp16 initialization)
        if self.sharded_ddp is not None:
            # Sharded DDP!
            if self.sharded_ddp == ShardedDDPOption.SIMPLE:
                model = ShardedDDP(model, self.optimizer)
            else:
                mixed_precision = self.args.fp16
                cpu_offload = ShardedDDPOption.OFFLOAD in self.args.sharded_ddp
                zero_3 = self.sharded_ddp == ShardedDDPOption.ZERO_DP_3
                # XXX: Breaking the self.model convention but I see no way around it for now.
                if ShardedDDPOption.AUTO_WRAP in self.args.sharded_ddp:
                    model = auto_wrap(model)
                self.model = model = FullyShardedDDP(
                    model,
                    mixed_precision=mixed_precision,
                    reshard_after_forward=zero_3,
                    cpu_offload=cpu_offload,
                ).to(self.args.device)

        elif is_sagemaker_dp_enabled():
            model = DDP(model, device_ids=[dist.get_local_rank()], broadcast_buffers=False)
        elif self.args.local_rank != -1:
            if self.args.ddp_find_unused_parameters is not None:
                find_unused_parameters = self.args.ddp_find_unused_parameters
            elif isinstance(model, PreTrainedModel):
                # find_unused_parameters breaks checkpointing as per
                # https://github.com/huggingface/transformers/pull/4659#issuecomment-643356021
                find_unused_parameters = not model.is_gradient_checkpointing
            else:
                find_unused_parameters = True
            model = nn.parallel.DistributedDataParallel(
                model,
                device_ids=[self.args.local_rank] if self.args._n_gpu != 0 else None,
                output_device=self.args.local_rank if self.args._n_gpu != 0 else None,
                find_unused_parameters=find_unused_parameters,
            )

        return model

    def train(
        self,
        resume_from_checkpoint: Optional[Union[str, bool]] = None,
        trial: Union["optuna.Trial", Dict[str, Any]] = None,
        ignore_keys_for_eval: Optional[List[str]] = None,
        **kwargs,
    ):
        """
        Main training entry point.

        Args:
            resume_from_checkpoint (:obj:`str` or :obj:`bool`, `optional`):
                If a :obj:`str`, local path to a saved checkpoint as saved by a previous instance of
                :class:`~transformers.Trainer`. If a :obj:`bool` and equals `True`, load the last checkpoint in
                `args.output_dir` as saved by a previous instance of :class:`~transformers.Trainer`. If present,
                training will resume from the model/optimizer/scheduler states loaded here.
            trial (:obj:`optuna.Trial` or :obj:`Dict[str, Any]`, `optional`):
                The trial run or the hyperparameter dictionary for hyperparameter search.
            ignore_keys_for_eval (:obj:`List[str]`, `optional`)
                A list of keys in the output of your model (if it is a dictionary) that should be ignored when
                gathering predictions for evaluation during the training.
            kwargs:
                Additional keyword arguments used to hide deprecated arguments
        """
        resume_from_checkpoint = None if not resume_from_checkpoint else resume_from_checkpoint

        # memory metrics - must set up as early as possible
        self._memory_tracker.start()

        args = self.args

        self.is_in_train = True

        # do_train is not a reliable argument, as it might not be set and .train() still called, so
        # the following is a workaround:
        if args.fp16_full_eval and not args.do_train:
            self._move_model_to_device(self.model, args.device)

        if "model_path" in kwargs:
            resume_from_checkpoint = kwargs.pop("model_path")
            warnings.warn(
                "`model_path` is deprecated and will be removed in a future version. Use `resume_from_checkpoint` "
                "instead.",
                FutureWarning,
            )
        if len(kwargs) > 0:
            raise TypeError(f"train() received got unexpected keyword arguments: {', '.join(list(kwargs.keys()))}.")
        # This might change the seed so needs to run first.
        self._hp_search_setup(trial)

        # Model re-init
        model_reloaded = False
        if self.model_init is not None:
            # Seed must be set before instantiating the model when using model_init.
            set_seed(args.seed)
            self.model = self.call_model_init(trial)
            model_reloaded = True
            # Reinitializes optimizer and scheduler
            self.optimizer, self.lr_scheduler = None, None

        # Load potential model checkpoint
        if isinstance(resume_from_checkpoint, bool) and resume_from_checkpoint:
            resume_from_checkpoint = get_last_checkpoint(args.output_dir)
            if resume_from_checkpoint is None:
                raise ValueError(f"No valid checkpoint found in output directory ({args.output_dir})")

        if resume_from_checkpoint is not None:
            if not os.path.isfile(os.path.join(resume_from_checkpoint, WEIGHTS_NAME)):
                raise ValueError(f"Can't find a valid checkpoint at {resume_from_checkpoint}")

            logger.info(f"Loading model from {resume_from_checkpoint}).")

            if os.path.isfile(os.path.join(resume_from_checkpoint, CONFIG_NAME)):
                config = PretrainedConfig.from_json_file(os.path.join(resume_from_checkpoint, CONFIG_NAME))
                checkpoint_version = config.transformers_version
                if checkpoint_version is not None and checkpoint_version != __version__:
                    logger.warn(
                        f"You are resuming training from a checkpoint trained with {checkpoint_version} of "
                        f"Transformers but your current version is {__version__}. This is not recommended and could "
                        "yield to errors or unwanted behaviors."
                    )

            if args.deepspeed:
                # will be resumed in deepspeed_init
                pass
            else:
                # We load the model state dict on the CPU to avoid an OOM error.
                state_dict = torch.load(os.path.join(resume_from_checkpoint, WEIGHTS_NAME), map_location="cpu")
                # If the model is on the GPU, it still works!
                self._load_state_dict_in_model(state_dict)

                # release memory
                del state_dict

        # If model was re-initialized, put it on the right device and update self.model_wrapped
        if model_reloaded:
            if self.place_model_on_device:
                self._move_model_to_device(self.model, args.device)
            self.model_wrapped = self.model

        # Keeping track whether we can can len() on the dataset or not
        train_dataset_is_sized = isinstance(self.train_dataset, collections.abc.Sized)

        # Data loader and number of training steps
        train_dataloader = self.get_train_dataloader()

        # Setting up training control variables:
        # number of training epochs: num_train_epochs
        # number of training steps per epoch: num_update_steps_per_epoch
        # total number of training steps to execute: max_steps
        total_train_batch_size = args.train_batch_size * args.gradient_accumulation_steps * args.world_size
        if train_dataset_is_sized:
            num_update_steps_per_epoch = len(train_dataloader) // args.gradient_accumulation_steps
            num_update_steps_per_epoch = max(num_update_steps_per_epoch, 1)
            if args.max_steps > 0:
                max_steps = args.max_steps
                num_train_epochs = args.max_steps // num_update_steps_per_epoch + int(
                    args.max_steps % num_update_steps_per_epoch > 0
                )
                # May be slightly incorrect if the last batch in the training datalaoder has a smaller size but it's
                # the best we can do.
                num_train_samples = args.max_steps * total_train_batch_size
            else:
                max_steps = math.ceil(args.num_train_epochs * num_update_steps_per_epoch)
                num_train_epochs = math.ceil(args.num_train_epochs)
                num_train_samples = len(self.train_dataset) * args.num_train_epochs
        else:
            # see __init__. max_steps is set when the dataset has no __len__
            max_steps = args.max_steps
            # Setting a very large number of epochs so we go as many times as necessary over the iterator.
            num_train_epochs = sys.maxsize
            num_update_steps_per_epoch = max_steps
            num_train_samples = args.max_steps * total_train_batch_size

        if DebugOption.UNDERFLOW_OVERFLOW in self.args.debug:
            if self.args.n_gpu > 1:
                # nn.DataParallel(model) replicates the model, creating new variables and module
                # references registered here no longer work on other gpus, breaking the module
                raise ValueError(
                    "Currently --debug underflow_overflow is not supported under DP. Please use DDP (torch.distributed.launch)."
                )
            else:
                debug_overflow = DebugUnderflowOverflow(self.model)  # noqa

        delay_optimizer_creation = self.sharded_ddp is not None and self.sharded_ddp != ShardedDDPOption.SIMPLE
        if args.deepspeed:
            deepspeed_engine, optimizer, lr_scheduler = deepspeed_init(
                self, num_training_steps=max_steps, resume_from_checkpoint=resume_from_checkpoint
            )
            self.model = deepspeed_engine.module
            self.model_wrapped = deepspeed_engine
            self.deepspeed = deepspeed_engine
            self.optimizer = optimizer
            self.lr_scheduler = lr_scheduler
        elif not delay_optimizer_creation:
            self.create_optimizer_and_scheduler(num_training_steps=max_steps)

        self.state = TrainerState()
        self.state.is_hyper_param_search = trial is not None

        # Activate gradient checkpointing if needed
        if args.gradient_checkpointing:
            self.model.gradient_checkpointing_enable()

        model = self._wrap_model(self.model_wrapped)

        # for the rest of this function `model` is the outside model, whether it was wrapped or not
        if model is not self.model:
            self.model_wrapped = model

        if delay_optimizer_creation:
            self.create_optimizer_and_scheduler(num_training_steps=max_steps)

        # Check if saved optimizer or scheduler states exist
        self._load_optimizer_and_scheduler(resume_from_checkpoint)

        # important: at this point:
        # self.model         is the Transformers Model
        # self.model_wrapped is DDP(Transformers Model), Deepspeed(Transformers Model), etc.

        # Train!
        num_examples = (
            self.num_examples(train_dataloader) if train_dataset_is_sized else total_train_batch_size * args.max_steps
        )

        logger.info("***** Running training *****")
        logger.info(f"  Num examples = {num_examples}")
        logger.info(f"  Num Epochs = {num_train_epochs}")
        logger.info(f"  Instantaneous batch size per device = {args.per_device_train_batch_size}")
        logger.info(f"  Total train batch size (w. parallel, distributed & accumulation) = {total_train_batch_size}")
        logger.info(f"  Gradient Accumulation steps = {args.gradient_accumulation_steps}")
        logger.info(f"  Total optimization steps = {max_steps}")

        self.state.epoch = 0
        start_time = time.time()
        epochs_trained = 0
        steps_trained_in_current_epoch = 0
        steps_trained_progress_bar = None

        # Check if continuing training from a checkpoint
        if resume_from_checkpoint is not None and os.path.isfile(
            os.path.join(resume_from_checkpoint, TRAINER_STATE_NAME)
        ):
            self.state = TrainerState.load_from_json(os.path.join(resume_from_checkpoint, TRAINER_STATE_NAME))
            epochs_trained = self.state.global_step // num_update_steps_per_epoch
            if not args.ignore_data_skip:
                steps_trained_in_current_epoch = self.state.global_step % (num_update_steps_per_epoch)
                steps_trained_in_current_epoch *= args.gradient_accumulation_steps
            else:
                steps_trained_in_current_epoch = 0

            logger.info("  Continuing training from checkpoint, will skip to saved global_step")
            logger.info(f"  Continuing training from epoch {epochs_trained}")
            logger.info(f"  Continuing training from global step {self.state.global_step}")
            if not args.ignore_data_skip:
                logger.info(
                    f"  Will skip the first {epochs_trained} epochs then the first {steps_trained_in_current_epoch} "
                    "batches in the first epoch. If this takes a lot of time, you can add the `--ignore_data_skip` "
                    "flag to your launch command, but you will resume the training on data already seen by your model."
                )
                if self.is_local_process_zero() and not args.disable_tqdm:
                    steps_trained_progress_bar = tqdm(total=steps_trained_in_current_epoch)
                    steps_trained_progress_bar.set_description("Skipping the first batches")

        # Update the references
        self.callback_handler.model = self.model
        self.callback_handler.optimizer = self.optimizer
        self.callback_handler.lr_scheduler = self.lr_scheduler
        self.callback_handler.train_dataloader = train_dataloader
        self.state.trial_name = self.hp_name(trial) if self.hp_name is not None else None
        if trial is not None:
            assignments = trial.assignments if self.hp_search_backend == HPSearchBackend.SIGOPT else trial
            self.state.trial_params = hp_params(assignments)
        else:
            self.state.trial_params = None
        # This should be the same if the state has been saved but in case the training arguments changed, it's safer
        # to set this after the load.
        self.state.max_steps = max_steps
        self.state.num_train_epochs = num_train_epochs
        self.state.is_local_process_zero = self.is_local_process_zero()
        self.state.is_world_process_zero = self.is_world_process_zero()

        # tr_loss is a tensor to avoid synchronization of TPUs through .item()
        tr_loss = torch.tensor(0.0).to(args.device)
        # _total_loss_scalar is updated everytime .item() has to be called on tr_loss and stores the sum of all losses
        self._total_loss_scalar = 0.0
        self._globalstep_last_logged = self.state.global_step
        model.zero_grad()

        self.control = self.callback_handler.on_train_begin(args, self.state, self.control)

        # Skip the first epochs_trained epochs to get the random state of the dataloader at the right point.
        if not args.ignore_data_skip:
            for epoch in range(epochs_trained):
                # We just need to begin an iteration to create the randomization of the sampler.
                for _ in train_dataloader:
                    break

        for epoch in range(epochs_trained, num_train_epochs):
            if isinstance(train_dataloader, DataLoader) and isinstance(train_dataloader.sampler, DistributedSampler):
                train_dataloader.sampler.set_epoch(epoch)
            elif isinstance(train_dataloader.dataset, IterableDatasetShard):
                train_dataloader.dataset.set_epoch(epoch)

            if is_torch_tpu_available():
                parallel_loader = pl.ParallelLoader(train_dataloader, [args.device]).per_device_loader(args.device)
                epoch_iterator = parallel_loader
            else:
                epoch_iterator = train_dataloader

            # Reset the past mems state at the beginning of each epoch if necessary.
            if args.past_index >= 0:
                self._past = None

            steps_in_epoch = (
                len(epoch_iterator) if train_dataset_is_sized else args.max_steps * args.gradient_accumulation_steps
            )
            self.control = self.callback_handler.on_epoch_begin(args, self.state, self.control)

            step = -1
            for step, inputs in enumerate(epoch_iterator):

                # Skip past any already trained steps if resuming training
                if steps_trained_in_current_epoch > 0:
                    steps_trained_in_current_epoch -= 1
                    if steps_trained_progress_bar is not None:
                        steps_trained_progress_bar.update(1)
                    if steps_trained_in_current_epoch == 0:
                        self._load_rng_state(resume_from_checkpoint)
                    continue
                elif steps_trained_progress_bar is not None:
                    steps_trained_progress_bar.close()
                    steps_trained_progress_bar = None

                if step % args.gradient_accumulation_steps == 0:
                    self.control = self.callback_handler.on_step_begin(args, self.state, self.control)

                if (
                    ((step + 1) % args.gradient_accumulation_steps != 0)
                    and args.local_rank != -1
                    and args._no_sync_in_gradient_accumulation
                ):
                    # Avoid unnecessary DDP synchronization since there will be no backward pass on this example.
                    with model.no_sync():
                        tr_loss_step = self.training_step(model, inputs)
                else:
                    tr_loss_step = self.training_step(model, inputs)

                if (
                    args.logging_nan_inf_filter
                    and not is_torch_tpu_available()
                    and (torch.isnan(tr_loss_step) or torch.isinf(tr_loss_step))
                ):
                    # if loss is nan or inf simply add the average of previous logged losses
                    tr_loss += tr_loss / (1 + self.state.global_step - self._globalstep_last_logged)
                else:
                    tr_loss += tr_loss_step

                self.current_flos += float(self.floating_point_ops(inputs))

                # Optimizer step for deepspeed must be called on every step regardless of the value of gradient_accumulation_steps
                if self.deepspeed:
                    self.deepspeed.step()

                if (step + 1) % args.gradient_accumulation_steps == 0 or (
                    # last step in epoch but step is always smaller than gradient_accumulation_steps
                    steps_in_epoch <= args.gradient_accumulation_steps
                    and (step + 1) == steps_in_epoch
                ):
                    # Gradient clipping
                    if args.max_grad_norm is not None and args.max_grad_norm > 0 and not self.deepspeed:
                        # deepspeed does its own clipping

                        if self.use_amp:
                            # AMP: gradients need unscaling
                            self.scaler.unscale_(self.optimizer)

                        if hasattr(self.optimizer, "clip_grad_norm"):
                            # Some optimizers (like the sharded optimizer) have a specific way to do gradient clipping
                            self.optimizer.clip_grad_norm(args.max_grad_norm)
                        elif hasattr(model, "clip_grad_norm_"):
                            # Some models (like FullyShardedDDP) have a specific way to do gradient clipping
                            model.clip_grad_norm_(args.max_grad_norm)
                        else:
                            # Revert to normal clipping otherwise, handling Apex or full precision
                            nn.utils.clip_grad_norm_(
                                amp.master_params(self.optimizer) if self.use_apex else model.parameters(),
                                args.max_grad_norm,
                            )

                    # Optimizer step
                    optimizer_was_run = True
                    if self.deepspeed:
                        pass  # called outside the loop
                    elif is_torch_tpu_available():
                        xm.optimizer_step(self.optimizer)
                    elif self.use_amp:
                        scale_before = self.scaler.get_scale()
                        self.scaler.step(self.optimizer)
                        self.scaler.update()
                        scale_after = self.scaler.get_scale()
                        optimizer_was_run = scale_before <= scale_after
                    else:
                        self.optimizer.step()

                    if optimizer_was_run and not self.deepspeed:
                        self.lr_scheduler.step()

                    model.zero_grad()
                    self.state.global_step += 1
                    self.state.epoch = epoch + (step + 1) / steps_in_epoch
                    self.control = self.callback_handler.on_step_end(args, self.state, self.control)

                    self._maybe_log_save_evaluate(tr_loss, model, trial, epoch, ignore_keys_for_eval)
                else:
                    self.control = self.callback_handler.on_substep_end(args, self.state, self.control)

                if self.control.should_epoch_stop or self.control.should_training_stop:
                    break
            if step < 0:
                logger.warning(
                    f"There seems to be not a single sample in your epoch_iterator, stopping training at step"
                    f" {self.state.global_step}! This is expected if you're using an IterableDataset and set"
                    f" num_steps ({max_steps}) higher than the number of available samples."
                )
                self.control.should_training_stop = True

            self.control = self.callback_handler.on_epoch_end(args, self.state, self.control)
            self._maybe_log_save_evaluate(tr_loss, model, trial, epoch, ignore_keys_for_eval)

            if DebugOption.TPU_METRICS_DEBUG in self.args.debug:
                if is_torch_tpu_available():
                    # tpu-comment: Logging debug metrics for PyTorch/XLA (compile, execute times, ops, etc.)
                    xm.master_print(met.metrics_report())
                else:
                    logger.warning(
                        "You enabled PyTorch/XLA debug metrics but you don't have a TPU "
                        "configured. Check your training configuration if this is unexpected."
                    )
            if self.control.should_training_stop:
                break

        if args.past_index and hasattr(self, "_past"):
            # Clean the state at the end of training
            delattr(self, "_past")

        logger.info("\n\nTraining completed. Do not forget to share your model on huggingface.co/models =)\n\n")
        if args.load_best_model_at_end and self.state.best_model_checkpoint is not None:
            # Wait for everyone to get here so we are sur the model has been saved by process 0.
            if is_torch_tpu_available():
                xm.rendezvous("load_best_model_at_end")
            elif args.local_rank != -1:
                dist.barrier()

            logger.info(
                f"Loading best model from {self.state.best_model_checkpoint} (score: {self.state.best_metric})."
            )

            best_model_path = os.path.join(self.state.best_model_checkpoint, WEIGHTS_NAME)
            if os.path.exists(best_model_path):
                # We load the model state dict on the CPU to avoid an OOM error.
                state_dict = torch.load(best_model_path, map_location="cpu")
                # If the model is on the GPU, it still works!
                self._load_state_dict_in_model(state_dict)
            else:
                logger.warn(
                    f"Could not locate the best model at {best_model_path}, if you are running a distributed training "
                    "on multiple nodes, you should activate `--save_on_each_node`."
                )

            if self.deepspeed:
                self.deepspeed.load_checkpoint(
                    self.state.best_model_checkpoint, load_optimizer_states=False, load_lr_scheduler_states=False
                )

        # add remaining tr_loss
        self._total_loss_scalar += tr_loss.item()
        train_loss = self._total_loss_scalar / self.state.global_step

        metrics = speed_metrics("train", start_time, num_samples=num_train_samples, num_steps=self.state.max_steps)
        self.store_flos()
        metrics["total_flos"] = self.state.total_flos
        metrics["train_loss"] = train_loss

        self.is_in_train = False

        self._memory_tracker.stop_and_update_metrics(metrics)

        self.log(metrics)

        self.control = self.callback_handler.on_train_end(args, self.state, self.control)

        return TrainOutput(self.state.global_step, train_loss, metrics)

    def _load_state_dict_in_model(self, state_dict):
        load_result = self.model.load_state_dict(state_dict, strict=False)

        if len(load_result.missing_keys) != 0:
            if self.model._keys_to_ignore_on_save is not None and set(load_result.missing_keys) == set(
                self.model._keys_to_ignore_on_save
            ):
                self.model.tie_weights()
            else:
                logger.warn(f"There were missing keys in the checkpoint model loaded: {load_result.missing_keys}.")
        if len(load_result.unexpected_keys) != 0:
            logger.warn(f"There were unexpected keys in the checkpoint model loaded: {load_result.unexpected_keys}.")

    def _maybe_log_save_evaluate(self, tr_loss, model, trial, epoch, ignore_keys_for_eval):
        if self.control.should_log:
            logs: Dict[str, float] = {}

            # all_gather + mean() to get average loss over all processes
            tr_loss_scalar = self._nested_gather(tr_loss).mean().item()

            # reset tr_loss to zero
            tr_loss -= tr_loss

            logs["loss"] = round(tr_loss_scalar / (self.state.global_step - self._globalstep_last_logged), 4)
            logs["learning_rate"] = self._get_learning_rate()

            self._total_loss_scalar += tr_loss_scalar
            self._globalstep_last_logged = self.state.global_step
            self.store_flos()

            self.log(logs)

        metrics = None
        if self.control.should_evaluate:
            metrics = self.evaluate(ignore_keys=ignore_keys_for_eval)
            self._report_to_hp_search(trial, epoch, metrics)

        if self.control.should_save:
            self._save_checkpoint(model, trial, metrics=metrics)
            self.control = self.callback_handler.on_save(self.args, self.state, self.control)

    def _load_rng_state(self, checkpoint):
        # Load RNG states from `checkpoint`
        if checkpoint is None:
            return

        local_rank = xm.get_local_ordinal() if is_torch_tpu_available() else self.args.local_rank
        if local_rank != -1:
            rng_file = os.path.join(checkpoint, f"rng_state_{local_rank}.pth")
            if not os.path.isfile(os.path.join(checkpoint, rng_file)):
                logger.info(
                    f"Didn't find an RNG file for process {local_rank}, if you are resuming a training that "
                    "wasn't launched in a distributed fashion, reproducibility is not guaranteed."
                )
                return
        else:
            rng_file = os.path.join(checkpoint, "rng_state.pth")
            if not os.path.isfile(rng_file):
                logger.info(
                    "Didn't find an RNG file, if you are resuming a training that was launched in a distributed "
                    "fashion, reproducibility is not guaranteed."
                )
                return

        checkpoint_rng_state = torch.load(rng_file)
        random.setstate(checkpoint_rng_state["python"])
        np.random.set_state(checkpoint_rng_state["numpy"])
        torch.random.set_rng_state(checkpoint_rng_state["cpu"])
        if torch.cuda.is_available():
            if self.args.local_rank != -1:
                torch.cuda.random.set_rng_state(checkpoint_rng_state["cuda"])
            else:
                torch.cuda.random.set_rng_state_all(checkpoint_rng_state["cuda"])
        if is_torch_tpu_available():
            xm.set_rng_state(checkpoint_rng_state["xla"])

    def _save_checkpoint(self, model, trial, metrics=None):
        # In all cases, including ddp/dp/deepspeed, self.model is always a reference to the model we
        # want to save except FullyShardedDDP.
        # assert unwrap_model(model) is self.model, "internal model should be a reference to self.model"

        # Save model checkpoint
        checkpoint_folder = f"{PREFIX_CHECKPOINT_DIR}-{self.state.global_step}"

        if self.hp_search_backend is not None and trial is not None:
            if self.hp_search_backend == HPSearchBackend.OPTUNA:
                run_id = trial.number
            elif self.hp_search_backend == HPSearchBackend.RAY:
                from ray import tune

                run_id = tune.get_trial_id()
            elif self.hp_search_backend == HPSearchBackend.SIGOPT:
                run_id = trial.id
            run_name = self.hp_name(trial) if self.hp_name is not None else f"run-{run_id}"
            run_dir = os.path.join(self.args.output_dir, run_name)
        else:
            run_dir = self.args.output_dir
            self.store_flos()

        output_dir = os.path.join(run_dir, checkpoint_folder)
        self.save_model(output_dir)
        if self.deepspeed:
            # under zero3 model file itself doesn't get saved since it's bogus! Unless deepspeed
            # config `stage3_gather_fp16_weights_on_model_save` is True
            self.deepspeed.save_checkpoint(output_dir)

        # Save optimizer and scheduler
        if self.sharded_ddp == ShardedDDPOption.SIMPLE:
            self.optimizer.consolidate_state_dict()

        if is_torch_tpu_available():
            xm.rendezvous("saving_optimizer_states")
            xm.save(self.optimizer.state_dict(), os.path.join(output_dir, OPTIMIZER_NAME))
            with warnings.catch_warnings(record=True) as caught_warnings:
                xm.save(self.lr_scheduler.state_dict(), os.path.join(output_dir, SCHEDULER_NAME))
                reissue_pt_warnings(caught_warnings)
        elif is_sagemaker_mp_enabled():
            if smp.dp_rank() == 0:
                # Consolidate the state dict on all processed of dp_rank 0
                opt_state_dict = self.optimizer.state_dict()
                # Save it and the scheduler on the main process
                if self.args.should_save:
                    torch.save(opt_state_dict, os.path.join(output_dir, OPTIMIZER_NAME))
                    with warnings.catch_warnings(record=True) as caught_warnings:
                        torch.save(self.lr_scheduler.state_dict(), os.path.join(output_dir, SCHEDULER_NAME))
                    reissue_pt_warnings(caught_warnings)
                    if self.use_amp:
                        torch.save(self.scaler.state_dict(), os.path.join(output_dir, SCALER_NAME))
        elif self.args.should_save and not self.deepspeed:
            # deepspeed.save_checkpoint above saves model/optim/sched
            torch.save(self.optimizer.state_dict(), os.path.join(output_dir, OPTIMIZER_NAME))
            with warnings.catch_warnings(record=True) as caught_warnings:
                torch.save(self.lr_scheduler.state_dict(), os.path.join(output_dir, SCHEDULER_NAME))
            reissue_pt_warnings(caught_warnings)
            if self.use_amp:
                torch.save(self.scaler.state_dict(), os.path.join(output_dir, SCALER_NAME))

        # Determine the new best metric / best model checkpoint
        if metrics is not None and self.args.metric_for_best_model is not None:
            metric_to_check = self.args.metric_for_best_model
            if not metric_to_check.startswith("eval_"):
                metric_to_check = f"eval_{metric_to_check}"
            metric_value = metrics[metric_to_check]

            operator = np.greater if self.args.greater_is_better else np.less
            if (
                self.state.best_metric is None
                or self.state.best_model_checkpoint is None
                or operator(metric_value, self.state.best_metric)
            ):
                self.state.best_metric = metric_value
                self.state.best_model_checkpoint = output_dir

        # Save the Trainer state
        if self.args.should_save:
            self.state.save_to_json(os.path.join(output_dir, TRAINER_STATE_NAME))

        # Save RNG state in non-distributed training
        rng_states = {
            "python": random.getstate(),
            "numpy": np.random.get_state(),
            "cpu": torch.random.get_rng_state(),
        }
        if torch.cuda.is_available():
            if self.args.local_rank == -1:
                # In non distributed, we save the global CUDA RNG state (will take care of DataParallel)
                rng_states["cuda"] = torch.cuda.random.get_rng_state_all()
            else:
                rng_states["cuda"] = torch.cuda.random.get_rng_state()

        if is_torch_tpu_available():
            rng_states["xla"] = xm.get_rng_state()

        # A process can arrive here before the process 0 has a chance to save the model, in which case output_dir may
        # not yet exist.
        os.makedirs(output_dir, exist_ok=True)
        local_rank = xm.get_local_ordinal() if is_torch_tpu_available() else self.args.local_rank
        if local_rank == -1:
            torch.save(rng_states, os.path.join(output_dir, "rng_state.pth"))
        else:
            torch.save(rng_states, os.path.join(output_dir, f"rng_state_{local_rank}.pth"))

        if self.args.push_to_hub:
            self._push_from_checkpoint(output_dir)

        # Maybe delete some older checkpoints.
        if self.args.should_save:
            self._rotate_checkpoints(use_mtime=True, output_dir=run_dir)

    def _load_optimizer_and_scheduler(self, checkpoint):
        """If optimizer and scheduler states exist, load them."""
        if checkpoint is None:
            return

        if self.deepspeed:
            # deepspeed loads optimizer/lr_scheduler together with the model in deepspeed_init
            return

        if os.path.isfile(os.path.join(checkpoint, OPTIMIZER_NAME)) and os.path.isfile(
            os.path.join(checkpoint, SCHEDULER_NAME)
        ):
            # Load in optimizer and scheduler states
            if is_torch_tpu_available():
                # On TPU we have to take some extra precautions to properly load the states on the right device.
                optimizer_state = torch.load(os.path.join(checkpoint, OPTIMIZER_NAME), map_location="cpu")
                with warnings.catch_warnings(record=True) as caught_warnings:
                    lr_scheduler_state = torch.load(os.path.join(checkpoint, SCHEDULER_NAME), map_location="cpu")
                reissue_pt_warnings(caught_warnings)

                xm.send_cpu_data_to_device(optimizer_state, self.args.device)
                xm.send_cpu_data_to_device(lr_scheduler_state, self.args.device)

                self.optimizer.load_state_dict(optimizer_state)
                self.lr_scheduler.load_state_dict(lr_scheduler_state)
            else:
                map_location = "cpu" if is_sagemaker_mp_enabled() else self.args.device
                self.optimizer.load_state_dict(
                    torch.load(os.path.join(checkpoint, OPTIMIZER_NAME), map_location=map_location)
                )
                with warnings.catch_warnings(record=True) as caught_warnings:
                    self.lr_scheduler.load_state_dict(torch.load(os.path.join(checkpoint, SCHEDULER_NAME)))
                reissue_pt_warnings(caught_warnings)
                if self.use_amp and os.path.isfile(os.path.join(checkpoint, SCALER_NAME)):
                    self.scaler.load_state_dict(torch.load(os.path.join(checkpoint, SCALER_NAME)))

    def hyperparameter_search(
        self,
        hp_space: Optional[Callable[["optuna.Trial"], Dict[str, float]]] = None,
        compute_objective: Optional[Callable[[Dict[str, float]], float]] = None,
        n_trials: int = 20,
        direction: str = "minimize",
        backend: Optional[Union["str", HPSearchBackend]] = None,
        hp_name: Optional[Callable[["optuna.Trial"], str]] = None,
        **kwargs,
    ) -> BestRun:
        """
        Launch an hyperparameter search using ``optuna`` or ``Ray Tune`` or ``SigOpt``. The optimized quantity is
        determined by :obj:`compute_objective`, which defaults to a function returning the evaluation loss when no
        metric is provided, the sum of all metrics otherwise.

        .. warning::

            To use this method, you need to have provided a ``model_init`` when initializing your
            :class:`~transformers.Trainer`: we need to reinitialize the model at each new run. This is incompatible
            with the ``optimizers`` argument, so you need to subclass :class:`~transformers.Trainer` and override the
            method :meth:`~transformers.Trainer.create_optimizer_and_scheduler` for custom optimizer/scheduler.

        Args:
            hp_space (:obj:`Callable[["optuna.Trial"], Dict[str, float]]`, `optional`):
                A function that defines the hyperparameter search space. Will default to
                :func:`~transformers.trainer_utils.default_hp_space_optuna` or
                :func:`~transformers.trainer_utils.default_hp_space_ray` or
                :func:`~transformers.trainer_utils.default_hp_space_sigopt` depending on your backend.
            compute_objective (:obj:`Callable[[Dict[str, float]], float]`, `optional`):
                A function computing the objective to minimize or maximize from the metrics returned by the
                :obj:`evaluate` method. Will default to :func:`~transformers.trainer_utils.default_compute_objective`.
            n_trials (:obj:`int`, `optional`, defaults to 100):
                The number of trial runs to test.
            direction(:obj:`str`, `optional`, defaults to :obj:`"minimize"`):
                Whether to optimize greater or lower objects. Can be :obj:`"minimize"` or :obj:`"maximize"`, you should
                pick :obj:`"minimize"` when optimizing the validation loss, :obj:`"maximize"` when optimizing one or
                several metrics.
            backend(:obj:`str` or :class:`~transformers.training_utils.HPSearchBackend`, `optional`):
                The backend to use for hyperparameter search. Will default to optuna or Ray Tune or SigOpt, depending
                on which one is installed. If all are installed, will default to optuna.
            kwargs:
                Additional keyword arguments passed along to :obj:`optuna.create_study` or :obj:`ray.tune.run`. For
                more information see:

                - the documentation of `optuna.create_study
                  <https://optuna.readthedocs.io/en/stable/reference/generated/optuna.study.create_study.html>`__
                - the documentation of `tune.run
                  <https://docs.ray.io/en/latest/tune/api_docs/execution.html#tune-run>`__
                - the documentation of `sigopt <https://app.sigopt.com/docs/endpoints/experiments/create>`__

        Returns:
            :class:`transformers.trainer_utils.BestRun`: All the information about the best run.
        """
        if backend is None:
            backend = default_hp_search_backend()
            if backend is None:
                raise RuntimeError(
                    "At least one of optuna or ray should be installed. "
                    "To install optuna run `pip install optuna`. "
                    "To install ray run `pip install ray[tune]`. "
                    "To install sigopt run `pip install sigopt`."
                )
        backend = HPSearchBackend(backend)
        if backend == HPSearchBackend.OPTUNA and not is_optuna_available():
            raise RuntimeError("You picked the optuna backend, but it is not installed. Use `pip install optuna`.")
        if backend == HPSearchBackend.RAY and not is_ray_tune_available():
            raise RuntimeError(
                "You picked the Ray Tune backend, but it is not installed. Use `pip install 'ray[tune]'`."
            )
        if backend == HPSearchBackend.SIGOPT and not is_sigopt_available():
            raise RuntimeError("You picked the sigopt backend, but it is not installed. Use `pip install sigopt`.")
        self.hp_search_backend = backend
        if self.model_init is None:
            raise RuntimeError(
                "To use hyperparameter search, you need to pass your model through a model_init function."
            )

        self.hp_space = default_hp_space[backend] if hp_space is None else hp_space
        self.hp_name = hp_name
        self.compute_objective = default_compute_objective if compute_objective is None else compute_objective

        backend_dict = {
            HPSearchBackend.OPTUNA: run_hp_search_optuna,
            HPSearchBackend.RAY: run_hp_search_ray,
            HPSearchBackend.SIGOPT: run_hp_search_sigopt,
        }
        best_run = backend_dict[backend](self, n_trials, direction, **kwargs)

        self.hp_search_backend = None
        return best_run

    def log(self, logs: Dict[str, float]) -> None:
        """
        Log :obj:`logs` on the various objects watching training.

        Subclass and override this method to inject custom behavior.

        Args:
            logs (:obj:`Dict[str, float]`):
                The values to log.
        """
        if self.state.epoch is not None:
            logs["epoch"] = round(self.state.epoch, 2)

        output = {**logs, **{"step": self.state.global_step}}
        self.state.log_history.append(output)
        self.control = self.callback_handler.on_log(self.args, self.state, self.control, logs)

    def _prepare_input(self, data: Union[torch.Tensor, Any]) -> Union[torch.Tensor, Any]:
        """
        Prepares one :obj:`data` before feeding it to the model, be it a tensor or a nested list/dictionary of tensors.
        """
        if isinstance(data, dict):
            return type(data)(**{k: self._prepare_input(v) for k, v in data.items()})
        elif isinstance(data, (tuple, list)):
            return type(data)(self._prepare_input(v) for v in data)
        elif isinstance(data, torch.Tensor):
            kwargs = dict(device=self.args.device)
            if self.deepspeed and data.dtype != torch.int64:
                # NLP models inputs are int64 and those get adjusted to the right dtype of the
                # embedding. Other models such as wav2vec2's inputs are already float and thus
                # may need special handling to match the dtypes of the model
                kwargs.update(dict(dtype=self.args.hf_deepspeed_config.dtype()))
            return data.to(**kwargs)
        return data

    def _prepare_inputs(self, inputs: Dict[str, Union[torch.Tensor, Any]]) -> Dict[str, Union[torch.Tensor, Any]]:
        """
        Prepare :obj:`inputs` before feeding them to the model, converting them to tensors if they are not already and
        handling potential state.
        """
        inputs = self._prepare_input(inputs)
        if self.args.past_index >= 0 and self._past is not None:
            inputs["mems"] = self._past

        return inputs

    def training_step(self, model: nn.Module, inputs: Dict[str, Union[torch.Tensor, Any]]) -> torch.Tensor:
        """
        Perform a training step on a batch of inputs.

        Subclass and override to inject custom behavior.

        Args:
            model (:obj:`nn.Module`):
                The model to train.
            inputs (:obj:`Dict[str, Union[torch.Tensor, Any]]`):
                The inputs and targets of the model.

                The dictionary will be unpacked before being fed to the model. Most models expect the targets under the
                argument :obj:`labels`. Check your model's documentation for all accepted arguments.

        Return:
            :obj:`torch.Tensor`: The tensor with training loss on this batch.
        """
        model.train()
        inputs = self._prepare_inputs(inputs)

        if is_sagemaker_mp_enabled():
            scaler = self.scaler if self.use_amp else None
            loss_mb = smp_forward_backward(model, inputs, self.args.gradient_accumulation_steps, scaler=scaler)
            return loss_mb.reduce_mean().detach().to(self.args.device)

        if self.use_amp:
            with autocast():
                loss = self.compute_loss(model, inputs)
        else:
            loss = self.compute_loss(model, inputs)

        if self.args.n_gpu > 1:
            loss = loss.mean()  # mean() to average on multi-gpu parallel training

        if self.args.gradient_accumulation_steps > 1 and not self.deepspeed:
            # deepspeed handles loss scaling by gradient_accumulation_steps in its `backward`
            loss = loss / self.args.gradient_accumulation_steps

        if self.use_amp:
            self.scaler.scale(loss).backward()
        elif self.use_apex:
            with amp.scale_loss(loss, self.optimizer) as scaled_loss:
                scaled_loss.backward()
        elif self.deepspeed:
            # loss gets scaled under gradient_accumulation_steps in deepspeed
            loss = self.deepspeed.backward(loss)
        else:
            loss.backward()

        return loss.detach()

    def compute_loss(self, model, inputs, return_outputs=False):
        """
        How the loss is computed by Trainer. By default, all models return the loss in the first element.

        Subclass and override for custom behavior.
        """
        if self.label_smoother is not None and "labels" in inputs:
            labels = inputs.pop("labels")
        else:
            labels = None
        outputs = model(**inputs)
        # Save past state if it exists
        # TODO: this needs to be fixed and made cleaner later.
        if self.args.past_index >= 0:
            self._past = outputs[self.args.past_index]

        if labels is not None:
            loss = self.label_smoother(outputs, labels)
        else:
            # We don't use .loss here since the model may return tuples instead of ModelOutput.
            loss = outputs["loss"] if isinstance(outputs, dict) else outputs[0]

        return (loss, outputs) if return_outputs else loss

    def is_local_process_zero(self) -> bool:
        """
        Whether or not this process is the local (e.g., on one machine if training in a distributed fashion on several
        machines) main process.
        """
        return self.args.local_process_index == 0

    def is_world_process_zero(self) -> bool:
        """
        Whether or not this process is the global main process (when training in a distributed fashion on several
        machines, this is only going to be :obj:`True` for one process).
        """
        # Special case for SageMaker ModelParallel since there process_index is dp_process_index, not the global
        # process index.
        if is_sagemaker_mp_enabled():
            return smp.rank() == 0
        else:
            return self.args.process_index == 0

    def save_model(self, output_dir: Optional[str] = None):
        """
        Will save the model, so you can reload it using :obj:`from_pretrained()`.

        Will only save from the main process.
        """

        if output_dir is None:
            output_dir = self.args.output_dir

        if is_torch_tpu_available():
            self._save_tpu(output_dir)
        elif is_sagemaker_mp_enabled():
            # Calling the state_dict needs to be done on the wrapped model and on all processes.
            state_dict = self.model_wrapped.state_dict()
            if self.args.should_save:
                self._save(output_dir, state_dict=state_dict)
        elif (
            ShardedDDPOption.ZERO_DP_2 in self.args.sharded_ddp or ShardedDDPOption.ZERO_DP_3 in self.args.sharded_ddp
        ):
            state_dict = self.model.state_dict()

            if self.args.should_save:
                self._save(output_dir, state_dict=state_dict)
        elif self.deepspeed:

            # this takes care of everything as long as we aren't under zero3
            if self.args.should_save:
                self._save(output_dir)

            if is_deepspeed_zero3_enabled():
                # It's too complicated to try to override different places where the weights dump gets
                # saved, so since under zero3 the file is bogus, simply delete it. The user should
                # either user deepspeed checkpoint to resume or to recover full weights use
                # zero_to_fp32.py stored in the checkpoint.
                if self.args.should_save:
                    file = os.path.join(output_dir, WEIGHTS_NAME)
                    if os.path.isfile(file):
                        # logger.info(f"deepspeed zero3: removing {file}, see zero_to_fp32.py to recover weights")
                        os.remove(file)

                # now save the real model if stage3_gather_fp16_weights_on_model_save=True
                # if false it will not be saved.
                # This must be called on all ranks
                self.deepspeed.save_fp16_model(output_dir, WEIGHTS_NAME)

        elif self.args.should_save:
            self._save(output_dir)

    def _save_tpu(self, output_dir: Optional[str] = None):
        output_dir = output_dir if output_dir is not None else self.args.output_dir
        logger.info(f"Saving model checkpoint to {output_dir}")

        if xm.is_master_ordinal():
            os.makedirs(output_dir, exist_ok=True)
            torch.save(self.args, os.path.join(output_dir, TRAINING_ARGS_NAME))

        # Save a trained model and configuration using `save_pretrained()`.
        # They can then be reloaded using `from_pretrained()`
        xm.rendezvous("saving_checkpoint")
        if not isinstance(self.model, PreTrainedModel):
            if isinstance(unwrap_model(self.model), PreTrainedModel):
                unwrap_model(self.model).save_pretrained(
                    output_dir,
                    save_config=self.args.should_save,
                    state_dict=self.model.state_dict(),
                    save_function=xm.save,
                )
            else:
                logger.info("Trainer.model is not a `PreTrainedModel`, only saving its state dict.")
                state_dict = self.model.state_dict()
                xm.save(state_dict, os.path.join(output_dir, WEIGHTS_NAME))
        else:
            self.model.save_pretrained(output_dir, save_config=self.args.should_save, save_function=xm.save)
        if self.tokenizer is not None and self.args.should_save:
            self.tokenizer.save_pretrained(output_dir)

    def _save(self, output_dir: Optional[str] = None, state_dict=None):
        # If we are executing this function, we are the process zero, so we don't check for that.
        output_dir = output_dir if output_dir is not None else self.args.output_dir
        os.makedirs(output_dir, exist_ok=True)
        logger.info(f"Saving model checkpoint to {output_dir}")
        # Save a trained model and configuration using `save_pretrained()`.
        # They can then be reloaded using `from_pretrained()`
        if not isinstance(self.model, PreTrainedModel):
            if isinstance(unwrap_model(self.model), PreTrainedModel):
                if state_dict is None:
                    state_dict = self.model.state_dict()
                unwrap_model(self.model).save_pretrained(output_dir, state_dict=state_dict)
            else:
                logger.info("Trainer.model is not a `PreTrainedModel`, only saving its state dict.")
                if state_dict is None:
                    state_dict = self.model.state_dict()
                torch.save(state_dict, os.path.join(output_dir, WEIGHTS_NAME))
        else:
            self.model.save_pretrained(output_dir, state_dict=state_dict)
        if self.tokenizer is not None:
            self.tokenizer.save_pretrained(output_dir)

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

    def store_flos(self):
        # Storing the number of floating-point operations that went into the model
        if self.args.local_rank != -1:
            self.state.total_flos += (
                distributed_broadcast_scalars([self.current_flos], device=self.args.device).sum().item()
            )
            self.current_flos = 0
        else:
            self.state.total_flos += self.current_flos
            self.current_flos = 0

    def _sorted_checkpoints(
        self, output_dir=None, checkpoint_prefix=PREFIX_CHECKPOINT_DIR, use_mtime=False
    ) -> List[str]:
        ordering_and_checkpoint_path = []

        glob_checkpoints = [str(x) for x in Path(output_dir).glob(f"{checkpoint_prefix}-*")]

        for path in glob_checkpoints:
            if use_mtime:
                ordering_and_checkpoint_path.append((os.path.getmtime(path), path))
            else:
                regex_match = re.match(f".*{checkpoint_prefix}-([0-9]+)", path)
                if regex_match is not None and regex_match.groups() is not None:
                    ordering_and_checkpoint_path.append((int(regex_match.groups()[0]), path))

        checkpoints_sorted = sorted(ordering_and_checkpoint_path)
        checkpoints_sorted = [checkpoint[1] for checkpoint in checkpoints_sorted]
        # Make sure we don't delete the best model.
        if self.state.best_model_checkpoint is not None:
            best_model_index = checkpoints_sorted.index(str(Path(self.state.best_model_checkpoint)))
            for i in range(best_model_index, len(checkpoints_sorted) - 2):
                checkpoints_sorted[i], checkpoints_sorted[i + 1] = checkpoints_sorted[i + 1], checkpoints_sorted[i]
        return checkpoints_sorted

    def _rotate_checkpoints(self, use_mtime=False, output_dir=None) -> None:
        if self.args.save_total_limit is None or self.args.save_total_limit <= 0:
            return

        # Check if we should delete older checkpoint(s)
        checkpoints_sorted = self._sorted_checkpoints(use_mtime=use_mtime, output_dir=output_dir)
        if len(checkpoints_sorted) <= self.args.save_total_limit:
            return

        # If save_total_limit=1 with load_best_model_at_end=True, we could end up deleting the last checkpoint, which
        # we don't do to allow resuming.
        save_total_limit = self.args.save_total_limit
        if (
            self.state.best_model_checkpoint is not None
            and self.args.save_total_limit == 1
            and checkpoints_sorted[-1] != self.state.best_model_checkpoint
        ):
            save_total_limit = 2

        number_of_checkpoints_to_delete = max(0, len(checkpoints_sorted) - save_total_limit)
        checkpoints_to_be_deleted = checkpoints_sorted[:number_of_checkpoints_to_delete]
        for checkpoint in checkpoints_to_be_deleted:
            logger.info(f"Deleting older checkpoint [{checkpoint}] due to args.save_total_limit")
            shutil.rmtree(checkpoint)

    def evaluate(
        self,
        eval_dataset: Optional[Dataset] = None,
        ignore_keys: Optional[List[str]] = None,
        metric_key_prefix: str = "eval",
    ) -> Dict[str, float]:
        """
        Run evaluation and returns metrics.

        The calling script will be responsible for providing a method to compute metrics, as they are task-dependent
        (pass it to the init :obj:`compute_metrics` argument).

        You can also subclass and override this method to inject custom behavior.

        Args:
            eval_dataset (:obj:`Dataset`, `optional`):
                Pass a dataset if you wish to override :obj:`self.eval_dataset`. If it is an :obj:`datasets.Dataset`,
                columns not accepted by the ``model.forward()`` method are automatically removed. It must implement the
                :obj:`__len__` method.
            ignore_keys (:obj:`Lst[str]`, `optional`):
                A list of keys in the output of your model (if it is a dictionary) that should be ignored when
                gathering predictions.
            metric_key_prefix (:obj:`str`, `optional`, defaults to :obj:`"eval"`):
                An optional prefix to be used as the metrics key prefix. For example the metrics "bleu" will be named
                "eval_bleu" if the prefix is "eval" (default)

        Returns:
            A dictionary containing the evaluation loss and the potential metrics computed from the predictions. The
            dictionary also contains the epoch number which comes from the training state.
        """
        # memory metrics - must set up as early as possible
        self._memory_tracker.start()

        eval_dataloader = self.get_eval_dataloader(eval_dataset)
        start_time = time.time()

        eval_loop = self.prediction_loop if self.args.use_legacy_prediction_loop else self.evaluation_loop
        output = eval_loop(
            eval_dataloader,
            description="Evaluation",
            # No point gathering the predictions if there are no metrics, otherwise we defer to
            # self.args.prediction_loss_only
            prediction_loss_only=True if self.compute_metrics is None else None,
            ignore_keys=ignore_keys,
            metric_key_prefix=metric_key_prefix,
        )

        total_batch_size = self.args.eval_batch_size * self.args.world_size
        output.metrics.update(
            speed_metrics(
                metric_key_prefix,
                start_time,
                num_samples=output.num_samples,
                num_steps=math.ceil(output.num_samples / total_batch_size),
            )
        )

        self.log(output.metrics)

        if DebugOption.TPU_METRICS_DEBUG in self.args.debug:
            # tpu-comment: Logging debug metrics for PyTorch/XLA (compile, execute times, ops, etc.)
            xm.master_print(met.metrics_report())

        self.control = self.callback_handler.on_evaluate(self.args, self.state, self.control, output.metrics)

        self._memory_tracker.stop_and_update_metrics(output.metrics)

        return output.metrics

    def predict(
        self, test_dataset: Dataset, ignore_keys: Optional[List[str]] = None, metric_key_prefix: str = "test"
    ) -> PredictionOutput:
        """
        Run prediction and returns predictions and potential metrics.

        Depending on the dataset and your use case, your test dataset may contain labels. In that case, this method
        will also return metrics, like in :obj:`evaluate()`.

        Args:
            test_dataset (:obj:`Dataset`):
                Dataset to run the predictions on. If it is an :obj:`datasets.Dataset`, columns not accepted by the
                ``model.forward()`` method are automatically removed. Has to implement the method :obj:`__len__`
            ignore_keys (:obj:`Lst[str]`, `optional`):
                A list of keys in the output of your model (if it is a dictionary) that should be ignored when
                gathering predictions.
            metric_key_prefix (:obj:`str`, `optional`, defaults to :obj:`"test"`):
                An optional prefix to be used as the metrics key prefix. For example the metrics "bleu" will be named
                "test_bleu" if the prefix is "test" (default)

        .. note::

            If your predictions or labels have different sequence length (for instance because you're doing dynamic
            padding in a token classification task) the predictions will be padded (on the right) to allow for
            concatenation into one array. The padding index is -100.

        Returns: `NamedTuple` A namedtuple with the following keys:

            - predictions (:obj:`np.ndarray`): The predictions on :obj:`test_dataset`.
            - label_ids (:obj:`np.ndarray`, `optional`): The labels (if the dataset contained some).
            - metrics (:obj:`Dict[str, float]`, `optional`): The potential dictionary of metrics (if the dataset
              contained labels).
        """
        # memory metrics - must set up as early as possible
        self._memory_tracker.start()

        test_dataloader = self.get_test_dataloader(test_dataset)
        start_time = time.time()

        eval_loop = self.prediction_loop if self.args.use_legacy_prediction_loop else self.evaluation_loop
        output = eval_loop(
            test_dataloader, description="Prediction", ignore_keys=ignore_keys, metric_key_prefix=metric_key_prefix
        )
        total_batch_size = self.args.eval_batch_size * self.args.world_size
        output.metrics.update(
            speed_metrics(
                metric_key_prefix,
                start_time,
                num_samples=output.num_samples,
                num_steps=math.ceil(output.num_samples / total_batch_size),
            )
        )

        self._memory_tracker.stop_and_update_metrics(output.metrics)

        return PredictionOutput(predictions=output.predictions, label_ids=output.label_ids, metrics=output.metrics)

    def evaluation_loop(
        self,
        dataloader: DataLoader,
        description: str,
        prediction_loss_only: Optional[bool] = None,
        ignore_keys: Optional[List[str]] = None,
        metric_key_prefix: str = "eval",
    ) -> EvalLoopOutput:
        """
        Prediction/evaluation loop, shared by :obj:`Trainer.evaluate()` and :obj:`Trainer.predict()`.

        Works both with or without labels.
        """
        prediction_loss_only = (
            prediction_loss_only if prediction_loss_only is not None else self.args.prediction_loss_only
        )

        # if eval is called w/o train init deepspeed here
        if self.args.deepspeed and not self.deepspeed:

            # XXX: eval doesn't have `resume_from_checkpoint` arg but we should be able to do eval
            # from the checkpoint eventually
            deepspeed_engine, _, _ = deepspeed_init(
                self, num_training_steps=0, resume_from_checkpoint=None, inference=True
            )
            self.model = deepspeed_engine.module
            self.model_wrapped = deepspeed_engine
            self.deepspeed = deepspeed_engine

        model = self._wrap_model(self.model, training=False)

        # if full fp16 is wanted on eval and this ``evaluation`` or ``predict`` isn't called while
        # ``train`` is running, halve it first and then put on device
        if not self.is_in_train and self.args.fp16_full_eval:
            model = model.half().to(self.args.device)

        batch_size = dataloader.batch_size

        logger.info(f"***** Running {description} *****")
        if isinstance(dataloader.dataset, collections.abc.Sized):
            logger.info(f"  Num examples = {self.num_examples(dataloader)}")
        else:
            logger.info("  Num examples: Unknown")
        logger.info(f"  Batch size = {batch_size}")

        model.eval()

        self.callback_handler.eval_dataloader = dataloader
        # Do this before wrapping.
        eval_dataset = dataloader.dataset

        if is_torch_tpu_available():
            dataloader = pl.ParallelLoader(dataloader, [self.args.device]).per_device_loader(self.args.device)

        if self.args.past_index >= 0:
            self._past = None

        # Initialize containers
        # losses/preds/labels on GPU/TPU (accumulated for eval_accumulation_steps)
        losses_host = None
        preds_host = None
        labels_host = None
        # losses/preds/labels on CPU (final containers)
        all_losses = None
        all_preds = None
        all_labels = None
        # Will be useful when we have an iterable dataset so don't know its length.

        observed_num_examples = 0
        # Main evaluation loop
        for step, inputs in enumerate(dataloader):
            # Update the observed num examples
            observed_batch_size = find_batch_size(inputs)
            if observed_batch_size is not None:
                observed_num_examples += observed_batch_size
                # For batch samplers, batch_size is not known by the dataloader in advance.
                if batch_size is None:
                    batch_size = observed_batch_size

            # Prediction step
            loss, logits, labels = self.prediction_step(model, inputs, prediction_loss_only, ignore_keys=ignore_keys)

            # Update containers on host
            if loss is not None:
                losses = self._nested_gather(loss.repeat(batch_size))
                losses_host = losses if losses_host is None else torch.cat((losses_host, losses), dim=0)
            if logits is not None:
                logits = self._pad_across_processes(logits)
                logits = self._nested_gather(logits)
                preds_host = logits if preds_host is None else nested_concat(preds_host, logits, padding_index=-100)
            if labels is not None:
                labels = self._pad_across_processes(labels)
                labels = self._nested_gather(labels)
                labels_host = labels if labels_host is None else nested_concat(labels_host, labels, padding_index=-100)
            self.control = self.callback_handler.on_prediction_step(self.args, self.state, self.control)

            # Gather all tensors and put them back on the CPU if we have done enough accumulation steps.
            if self.args.eval_accumulation_steps is not None and (step + 1) % self.args.eval_accumulation_steps == 0:
                if losses_host is not None:
                    losses = nested_numpify(losses_host)
                    all_losses = losses if all_losses is None else np.concatenate((all_losses, losses), axis=0)
                if preds_host is not None:
                    logits = nested_numpify(preds_host)
                    all_preds = logits if all_preds is None else nested_concat(all_preds, logits, padding_index=-100)
                if labels_host is not None:
                    labels = nested_numpify(labels_host)
                    all_labels = (
                        labels if all_labels is None else nested_concat(all_labels, labels, padding_index=-100)
                    )

                # Set back to None to begin a new accumulation
                losses_host, preds_host, labels_host = None, None, None

        if self.args.past_index and hasattr(self, "_past"):
            # Clean the state at the end of the evaluation loop
            delattr(self, "_past")

        # Gather all remaining tensors and put them back on the CPU
        if losses_host is not None:
            losses = nested_numpify(losses_host)
            all_losses = losses if all_losses is None else np.concatenate((all_losses, losses), axis=0)
        if preds_host is not None:
            logits = nested_numpify(preds_host)
            all_preds = logits if all_preds is None else nested_concat(all_preds, logits, padding_index=-100)
        if labels_host is not None:
            labels = nested_numpify(labels_host)
            all_labels = labels if all_labels is None else nested_concat(all_labels, labels, padding_index=-100)

        # Number of samples
        if not isinstance(eval_dataset, IterableDataset):
            num_samples = len(eval_dataset)
        # The instance check is weird and does not actually check for the type, but whether the dataset has the right
        # methods. Therefore we need to make sure it also has the attribute.
        elif isinstance(eval_dataset, IterableDatasetShard) and hasattr(eval_dataset, "num_examples"):
            num_samples = eval_dataset.num_examples
        else:
            num_samples = observed_num_examples

        # Number of losses has been rounded to a multiple of batch_size and in a distributed training, the number of
        # samplers has been rounded to a multiple of batch_size, so we truncate.
        if all_losses is not None:
            all_losses = all_losses[:num_samples]
        if all_preds is not None:
            all_preds = nested_truncate(all_preds, num_samples)
        if all_labels is not None:
            all_labels = nested_truncate(all_labels, num_samples)

        # Metrics!
        if self.compute_metrics is not None and all_preds is not None and all_labels is not None:
            metrics = self.compute_metrics(EvalPrediction(predictions=all_preds, label_ids=all_labels))
        else:
            metrics = {}

        # To be JSON-serializable, we need to remove numpy types or zero-d tensors
        metrics = denumpify_detensorize(metrics)

        if all_losses is not None:
            metrics[f"{metric_key_prefix}_loss"] = all_losses.mean().item()

        # Prefix all keys with metric_key_prefix + '_'
        for key in list(metrics.keys()):
            if not key.startswith(f"{metric_key_prefix}_"):
                metrics[f"{metric_key_prefix}_{key}"] = metrics.pop(key)

        return EvalLoopOutput(predictions=all_preds, label_ids=all_labels, metrics=metrics, num_samples=num_samples)

    def _nested_gather(self, tensors, name=None):
        """
        Gather value of `tensors` (tensor or list/tuple of nested tensors) and convert them to numpy before
        concatenating them to `gathered`
        """
        if tensors is None:
            return
        if is_torch_tpu_available():
            if name is None:
                name = "nested_gather"
            tensors = nested_xla_mesh_reduce(tensors, name)
        elif is_sagemaker_mp_enabled():
            tensors = smp_gather(tensors)
        elif self.args.local_rank != -1:
            tensors = distributed_concat(tensors)
        return tensors

    # Copied from Accelerate.
    def _pad_across_processes(self, tensor, pad_index=-100):
        """
        Recursively pad the tensors in a nested list/tuple/dictionary of tensors from all devices to the same size so
        they can safely be gathered.
        """
        if isinstance(tensor, (list, tuple)):
            return type(tensor)(self._pad_across_processes(t, pad_index=pad_index) for t in tensor)
        elif isinstance(tensor, dict):
            return type(tensor)({k: self._pad_across_processes(v, pad_index=pad_index) for k, v in tensor.items()})
        elif not isinstance(tensor, torch.Tensor):
            raise TypeError(
                f"Can't pad the values of type {type(tensor)}, only of nested list/tuple/dicts of tensors."
            )

        if len(tensor.shape) < 2:
            return tensor
        # Gather all sizes
        size = torch.tensor(tensor.shape, device=tensor.device)[None]
        sizes = self._nested_gather(size).cpu()

        max_size = max(s[1] for s in sizes)
        if tensor.shape[1] == max_size:
            return tensor

        # Then pad to the maximum size
        old_size = tensor.shape
        new_size = list(old_size)
        new_size[1] = max_size
        new_tensor = tensor.new_zeros(tuple(new_size)) + pad_index
        new_tensor[:, : old_size[1]] = tensor
        return new_tensor

    def prediction_step(
        self,
        model: nn.Module,
        inputs: Dict[str, Union[torch.Tensor, Any]],
        prediction_loss_only: bool,
        ignore_keys: Optional[List[str]] = None,
    ) -> Tuple[Optional[torch.Tensor], Optional[torch.Tensor], Optional[torch.Tensor]]:
        """
        Perform an evaluation step on :obj:`model` using obj:`inputs`.

        Subclass and override to inject custom behavior.

        Args:
            model (:obj:`nn.Module`):
                The model to evaluate.
            inputs (:obj:`Dict[str, Union[torch.Tensor, Any]]`):
                The inputs and targets of the model.

                The dictionary will be unpacked before being fed to the model. Most models expect the targets under the
                argument :obj:`labels`. Check your model's documentation for all accepted arguments.
            prediction_loss_only (:obj:`bool`):
                Whether or not to return the loss only.
            ignore_keys (:obj:`Lst[str]`, `optional`):
                A list of keys in the output of your model (if it is a dictionary) that should be ignored when
                gathering predictions.

        Return:
            Tuple[Optional[torch.Tensor], Optional[torch.Tensor], Optional[torch.Tensor]]: A tuple with the loss,
            logits and labels (each being optional).
        """
        has_labels = all(inputs.get(k) is not None for k in self.label_names)
        inputs = self._prepare_inputs(inputs)
        if ignore_keys is None:
            if hasattr(self.model, "config"):
                ignore_keys = getattr(self.model.config, "keys_to_ignore_at_inference", [])
            else:
                ignore_keys = []

        # labels may be popped when computing the loss (label smoothing for instance) so we grab them first.
        if has_labels:
            labels = nested_detach(tuple(inputs.get(name) for name in self.label_names))
            if len(labels) == 1:
                labels = labels[0]
        else:
            labels = None

        with torch.no_grad():
            if is_sagemaker_mp_enabled():
                raw_outputs = smp_forward_only(model, inputs)
                if has_labels:
                    if isinstance(raw_outputs, dict):
                        loss_mb = raw_outputs["loss"]
                        logits_mb = tuple(v for k, v in raw_outputs.items() if k not in ignore_keys + ["loss"])
                    else:
                        loss_mb = raw_outputs[0]
                        logits_mb = raw_outputs[1:]

                    loss = loss_mb.reduce_mean().detach().cpu()
                    logits = smp_nested_concat(logits_mb)
                else:
                    loss = None
                    if isinstance(raw_outputs, dict):
                        logits_mb = tuple(v for k, v in raw_outputs.items() if k not in ignore_keys)
                    else:
                        logits_mb = raw_outputs
                    logits = smp_nested_concat(logits_mb)
            else:
                if has_labels:
                    if self.use_amp:
                        with autocast():
                            loss, outputs = self.compute_loss(model, inputs, return_outputs=True)
                    else:
                        loss, outputs = self.compute_loss(model, inputs, return_outputs=True)
                    loss = loss.mean().detach()
                    if isinstance(outputs, dict):
                        logits = tuple(v for k, v in outputs.items() if k not in ignore_keys + ["loss"])
                    else:
                        logits = outputs[1:]
                else:
                    loss = None
                    if self.use_amp:
                        with autocast():
                            outputs = model(**inputs)
                    else:
                        outputs = model(**inputs)
                    if isinstance(outputs, dict):
                        logits = tuple(v for k, v in outputs.items() if k not in ignore_keys)
                    else:
                        logits = outputs
                    # TODO: this needs to be fixed and made cleaner later.
                    if self.args.past_index >= 0:
                        self._past = outputs[self.args.past_index - 1]

        if prediction_loss_only:
            return (loss, None, None)

        logits = nested_detach(logits)
        if len(logits) == 1:
            logits = logits[0]

        return (loss, logits, labels)

    def floating_point_ops(self, inputs: Dict[str, Union[torch.Tensor, Any]]):
        """
        For models that inherit from :class:`~transformers.PreTrainedModel`, uses that method to compute the number of
        floating point operations for every backward + forward pass. If using another model, either implement such a
        method in the model or subclass and override this method.

        Args:
            inputs (:obj:`Dict[str, Union[torch.Tensor, Any]]`):
                The inputs and targets of the model.

        Returns:
            :obj:`int`: The number of floating-point operations.
        """
        if hasattr(self.model, "floating_point_ops"):
            return self.model.floating_point_ops(inputs)
        else:
            return 0

    def init_git_repo(self):
        """
        Initializes a git repo in :obj:`self.args.hub_model_id`.
        """
        if not self.is_world_process_zero():
            return
        use_auth_token = True if self.args.hub_token is None else self.args.hub_token
        if self.args.hub_model_id is None:
            repo_name = Path(self.args.output_dir).absolute().name
        else:
            repo_name = self.args.hub_model_id
        if "/" not in repo_name:
            repo_name = get_full_repo_name(repo_name, token=self.args.hub_token)

        try:
            self.repo = Repository(
                self.args.output_dir,
                clone_from=repo_name,
                use_auth_token=use_auth_token,
            )
        except EnvironmentError:
            if self.args.overwrite_output_dir:
                # Try again after wiping output_dir
                shutil.rmtree(self.args.output_dir)
                self.repo = Repository(
                    self.args.output_dir,
                    clone_from=repo_name,
                    use_auth_token=use_auth_token,
                )
            else:
                raise

        self.repo.git_pull()

        # By default, ignore the checkpoint folders
        if (
            not os.path.exists(os.path.join(self.args.output_dir, ".gitignore"))
            and self.args.hub_strategy != HubStrategy.ALL_CHECKPOINTS
        ):
            with open(os.path.join(self.args.output_dir, ".gitignore"), "w", encoding="utf-8") as writer:
                writer.writelines(["checkpoint-*/"])

        self.push_in_progress = None

    def create_model_card(
        self,
        language: Optional[str] = None,
        license: Optional[str] = None,
        tags: Optional[str] = None,
        model_name: Optional[str] = None,
        finetuned_from: Optional[str] = None,
        tasks: Optional[str] = None,
        dataset_tags: Optional[Union[str, List[str]]] = None,
        dataset: Optional[Union[str, List[str]]] = None,
        dataset_args: Optional[Union[str, List[str]]] = None,
    ):
        training_summary = TrainingSummary.from_trainer(
            self,
            language=language,
            license=license,
            tags=tags,
            model_name=model_name,
            finetuned_from=finetuned_from,
            tasks=tasks,
            dataset_tags=dataset_tags,
            dataset=dataset,
            dataset_args=dataset_args,
        )
        model_card = training_summary.to_model_card()
        with open(os.path.join(self.args.output_dir, "README.md"), "w") as f:
            f.write(model_card)

    def _push_from_checkpoint(self, checkpoint_folder):
        # Only push from one node.
        if not self.is_world_process_zero() or self.args.hub_strategy == HubStrategy.END:
            return
        # If we haven't finished the last push, we don't do this one.
        if self.push_in_progress is not None and not self.push_in_progress.is_done:
            return

        output_dir = self.args.output_dir
        # To avoid a new synchronization of all model weights, we just copy the file from the checkpoint folder
        modeling_files = [CONFIG_NAME, WEIGHTS_NAME]
        for modeling_file in modeling_files:
            if os.path.isfile(os.path.join(checkpoint_folder, modeling_file)):
                shutil.copy(os.path.join(checkpoint_folder, modeling_file), os.path.join(output_dir, modeling_file))
        # Saving the tokenizer is fast and we don't know how many files it may have spawned, so we resave it to be sure.
        if self.tokenizer is not None:
            self.tokenizer.save_pretrained(output_dir)
        # Same for the training arguments
        torch.save(self.args, os.path.join(output_dir, TRAINING_ARGS_NAME))

        try:
            if self.args.hub_strategy == HubStrategy.CHECKPOINT:
                # Temporarily move the checkpoint just saved for the push
                tmp_checkpoint = os.path.join(output_dir, "last-checkpoint")
                # We have to remove the "last-checkpoint" dir if it exists, otherwise the checkpoint is moved as a
                # subfolder.
                if os.path.isdir(tmp_checkpoint):
                    shutil.rmtree(tmp_checkpoint)
                shutil.move(checkpoint_folder, tmp_checkpoint)

            if self.args.save_strategy == IntervalStrategy.STEPS:
                commit_message = f"Training in progress, step {self.state.global_step}"
            else:
                commit_message = f"Training in progress, epoch {int(self.state.epoch)}"
            _, self.push_in_progress = self.repo.push_to_hub(
                commit_message=commit_message, blocking=False, auto_lfs_prune=True
            )
        finally:
            if self.args.hub_strategy == HubStrategy.CHECKPOINT:
                # Move back the checkpoint to its place
                shutil.move(tmp_checkpoint, checkpoint_folder)

    def push_to_hub(self, commit_message: Optional[str] = "End of training", blocking: bool = True, **kwargs) -> str:
        """
        Upload `self.model` and `self.tokenizer` to the 🤗 model hub on the repo `self.args.hub_model_id`.

        Parameters:
            commit_message (:obj:`str`, `optional`, defaults to :obj:`"End of training"`):
                Message to commit while pushing.
            blocking (:obj:`bool`, `optional`, defaults to :obj:`True`):
                Whether the function should return only when the :obj:`git push` has finished.
            kwargs:
                Additional keyword arguments passed along to :meth:`~transformers.Trainer.create_model_card`.

        Returns:
            The url of the commit of your model in the given repository if :obj:`blocking=False`, a tuple with the url
            of the commit and an object to track the progress of the commit if :obj:`blocking=True`
        """

        if self.args.should_save:
            if self.args.hub_model_id is None:
                model_name = Path(self.args.output_dir).name
            else:
                model_name = self.args.hub_model_id.split("/")[-1]
        # Needs to be executed on all processes for TPU training, but will only save on the processed determined by
        # self.args.should_save.
        self.save_model()

        # Only push from one node.
        if not self.is_world_process_zero():
            return

        git_head_commit_url = self.repo.push_to_hub(
            commit_message=commit_message, blocking=blocking, auto_lfs_prune=True
        )
        # push separately the model card to be independant from the rest of the model
        if self.args.should_save:
            self.create_model_card(model_name=model_name, **kwargs)
            try:
                self.repo.push_to_hub(
                    commit_message="update model card README.md", blocking=blocking, auto_lfs_prune=True
                )
            except EnvironmentError as exc:
                logger.error(f"Error pushing update to the model card. Please read logs and retry.\n${exc}")

        return git_head_commit_url

    #
    # Deprecated code
    #

    def prediction_loop(
        self,
        dataloader: DataLoader,
        description: str,
        prediction_loss_only: Optional[bool] = None,
        ignore_keys: Optional[List[str]] = None,
        metric_key_prefix: str = "eval",
    ) -> PredictionOutput:
        """
        Prediction/evaluation loop, shared by :obj:`Trainer.evaluate()` and :obj:`Trainer.predict()`.

        Works both with or without labels.
        """
        if not isinstance(dataloader.dataset, collections.abc.Sized):
            raise ValueError("dataset must implement __len__")
        prediction_loss_only = (
            prediction_loss_only if prediction_loss_only is not None else self.args.prediction_loss_only
        )

        # if eval is called w/o train init deepspeed here
        if self.args.deepspeed and not self.deepspeed:

            # XXX: eval doesn't have `resume_from_checkpoint` arg but we should be able to do eval
            # from the checkpoint eventually
            deepspeed_engine, _, _ = deepspeed_init(self, num_training_steps=0, resume_from_checkpoint=None)
            self.model = deepspeed_engine.module
            self.model_wrapped = deepspeed_engine
            self.deepspeed = deepspeed_engine
            # XXX: we don't need optim/sched for inference, but this needs to be sorted out, since
            # for example the Z3-optimizer is a must for zero3 to work even for inference - what we
            # don't need is the deepspeed basic optimizer which is self.optimizer.optimizer
            deepspeed_engine.optimizer.optimizer = None
            deepspeed_engine.lr_scheduler = None

        model = self._wrap_model(self.model, training=False)

        # if full fp16 is wanted on eval and this ``evaluation`` or ``predict`` isn't called while
        # ``train`` is running, halve it first and then put on device
        if not self.is_in_train and self.args.fp16_full_eval:
            model = model.half().to(self.args.device)

        batch_size = dataloader.batch_size
        num_examples = self.num_examples(dataloader)
        logger.info(f"***** Running {description} *****")
        logger.info(f"  Num examples = {num_examples}")
        logger.info(f"  Batch size = {batch_size}")
        losses_host: torch.Tensor = None
        preds_host: Union[torch.Tensor, List[torch.Tensor]] = None
        labels_host: Union[torch.Tensor, List[torch.Tensor]] = None

        world_size = max(1, self.args.world_size)

        eval_losses_gatherer = DistributedTensorGatherer(world_size, num_examples, make_multiple_of=batch_size)
        if not prediction_loss_only:
            # The actual number of eval_sample can be greater than num_examples in distributed settings (when we pass
            # a batch size to the sampler)
            make_multiple_of = None
            if hasattr(dataloader, "sampler") and isinstance(dataloader.sampler, SequentialDistributedSampler):
                make_multiple_of = dataloader.sampler.batch_size
            preds_gatherer = DistributedTensorGatherer(world_size, num_examples, make_multiple_of=make_multiple_of)
            labels_gatherer = DistributedTensorGatherer(world_size, num_examples, make_multiple_of=make_multiple_of)

        model.eval()

        if is_torch_tpu_available():
            dataloader = pl.ParallelLoader(dataloader, [self.args.device]).per_device_loader(self.args.device)

        if self.args.past_index >= 0:
            self._past = None

        self.callback_handler.eval_dataloader = dataloader

        for step, inputs in enumerate(dataloader):
            loss, logits, labels = self.prediction_step(model, inputs, prediction_loss_only, ignore_keys=ignore_keys)
            if loss is not None:
                losses = loss.repeat(batch_size)
                losses_host = losses if losses_host is None else torch.cat((losses_host, losses), dim=0)
            if logits is not None:
                preds_host = logits if preds_host is None else nested_concat(preds_host, logits, padding_index=-100)
            if labels is not None:
                labels_host = labels if labels_host is None else nested_concat(labels_host, labels, padding_index=-100)
            self.control = self.callback_handler.on_prediction_step(self.args, self.state, self.control)

            # Gather all tensors and put them back on the CPU if we have done enough accumulation steps.
            if self.args.eval_accumulation_steps is not None and (step + 1) % self.args.eval_accumulation_steps == 0:
                eval_losses_gatherer.add_arrays(self._gather_and_numpify(losses_host, "eval_losses"))
                if not prediction_loss_only:
                    preds_gatherer.add_arrays(self._gather_and_numpify(preds_host, "eval_preds"))
                    labels_gatherer.add_arrays(self._gather_and_numpify(labels_host, "eval_label_ids"))

                # Set back to None to begin a new accumulation
                losses_host, preds_host, labels_host = None, None, None

        if self.args.past_index and hasattr(self, "_past"):
            # Clean the state at the end of the evaluation loop
            delattr(self, "_past")

        # Gather all remaining tensors and put them back on the CPU
        eval_losses_gatherer.add_arrays(self._gather_and_numpify(losses_host, "eval_losses"))
        if not prediction_loss_only:
            preds_gatherer.add_arrays(self._gather_and_numpify(preds_host, "eval_preds"))
            labels_gatherer.add_arrays(self._gather_and_numpify(labels_host, "eval_label_ids"))

        eval_loss = eval_losses_gatherer.finalize()
        preds = preds_gatherer.finalize() if not prediction_loss_only else None
        label_ids = labels_gatherer.finalize() if not prediction_loss_only else None

        if self.compute_metrics is not None and preds is not None and label_ids is not None:
            metrics = self.compute_metrics(EvalPrediction(predictions=preds, label_ids=label_ids))
        else:
            metrics = {}

        # To be JSON-serializable, we need to remove numpy types or zero-d tensors
        metrics = denumpify_detensorize(metrics)

        if eval_loss is not None:
            metrics[f"{metric_key_prefix}_loss"] = eval_loss.mean().item()

        # Prefix all keys with metric_key_prefix + '_'
        for key in list(metrics.keys()):
            if not key.startswith(f"{metric_key_prefix}_"):
                metrics[f"{metric_key_prefix}_{key}"] = metrics.pop(key)

        return PredictionOutput(predictions=preds, label_ids=label_ids, metrics=metrics)

    def _gather_and_numpify(self, tensors, name):
        """
        Gather value of `tensors` (tensor or list/tuple of nested tensors) and convert them to numpy before
        concatenating them to `gathered`
        """
        if tensors is None:
            return
        if is_torch_tpu_available():
            tensors = nested_xla_mesh_reduce(tensors, name)
        elif is_sagemaker_mp_enabled():
            tensors = smp_gather(tensors)
        elif self.args.local_rank != -1:
            tensors = distributed_concat(tensors)

        return nested_numpify(tensors)