trainer.py

import numpy as np
import torch
import torch.nn as nn
import torch.distributed as dist
from torch.utils.data import DataLoader, Dataset, RandomSampler, SequentialSampler
import math
import time
import os
import sys
import peft
import shutil
from packaging import version
from typing import Dict, Union, Any, Optional, Callable, List, Tuple
from transformers import Trainer, DataCollatorForSeq2Seq
from transformers.utils import (
    is_sagemaker_mp_enabled,
    is_torch_tpu_available,
    is_accelerate_available,
    is_apex_available,
    logging,
)
from transformers.training_args import ParallelMode, TrainingArguments
from transformers.tokenization_utils_base import PreTrainedTokenizerBase
from transformers.trainer_utils import (
    has_length,
    HPSearchBackend,
    TrainOutput,
    speed_metrics,
    EvalPrediction,
    EvalLoopOutput,
    denumpify_detensorize,
    ShardedDDPOption,
)
from transformers.debug_utils import (
    DebugOption,
    DebugUnderflowOverflow,
)
from transformers.integrations import (
    hp_params,
)
from transformers.integrations.deepspeed import (
    deepspeed_init, 
    deepspeed_load_checkpoint,
    is_deepspeed_available,
)
from transformers.trainer_callback import (
    TrainerState,
    TrainerCallback,
)
from transformers.trainer_pt_utils import (
    get_model_param_count,
    get_dataloader_sampler,
    find_batch_size,
    nested_concat,
    nested_numpify,
    IterableDatasetShard,
    nested_detach,
)
from transformers.trainer import TRAINER_STATE_NAME
from transformers.pytorch_utils import is_torch_less_than_1_11
from transformers.modeling_utils import PreTrainedModel
from transformers.data.data_collator import DataCollator


if is_apex_available():
    from apex import amp


if is_torch_tpu_available(check_device=False):
    import torch_xla.core.xla_model as xm
    import torch_xla.debug.metrics as met


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

    IS_SAGEMAKER_MP_POST_1_10 = version.parse(SMP_VERSION) >= version.parse("1.10")

    from transformers.trainer_pt_utils import smp_forward_backward, smp_forward_only, smp_gather, smp_nested_concat
else:
    IS_SAGEMAKER_MP_POST_1_10 = False


if is_accelerate_available():
    from accelerate import Accelerator, skip_first_batches
    from accelerate import __version__ as accelerate_version
    from accelerate.utils import DistributedDataParallelKwargs, GradientAccumulationPlugin

    if version.parse(accelerate_version) > version.parse("0.20.3"):
        from accelerate.utils import (
            load_fsdp_model,
            load_fsdp_optimizer,
            save_fsdp_model,
            save_fsdp_optimizer,
        )
    DATA_SAMPLERS = [RandomSampler]
    if version.parse(accelerate_version) > version.parse("0.23.0"):
        from accelerate.data_loader import SeedableRandomSampler

        DATA_SAMPLERS += [SeedableRandomSampler]

    if is_deepspeed_available():
        from accelerate.utils import DeepSpeedSchedulerWrapper


logger = logging.get_logger(__name__)


class CustomTrainer(Trainer):
    def _inner_training_loop(
        self, batch_size=None, args=None, resume_from_checkpoint=None, trial=None, ignore_keys_for_eval=None
    ):
        self.accelerator.free_memory()
        self._train_batch_size = batch_size
        logger.debug(f"Currently training with a batch size of: {self._train_batch_size}")
        # 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 = self._train_batch_size * args.gradient_accumulation_steps * args.world_size

        len_dataloader = None
        num_train_tokens = None
        if has_length(train_dataloader):
            len_dataloader = len(train_dataloader)
            num_update_steps_per_epoch = len_dataloader // args.gradient_accumulation_steps
            num_update_steps_per_epoch = max(num_update_steps_per_epoch, 1)
            num_examples = self.num_examples(train_dataloader)
            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 dataloader has a smaller size but it's
                # the best we can do.
                num_train_samples = args.max_steps * total_train_batch_size
                if args.include_tokens_per_second:
                    num_train_tokens = (
                        self.num_tokens(train_dataloader, args.max_steps) * args.gradient_accumulation_steps
                    )
            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 = self.num_examples(train_dataloader) * args.num_train_epochs
                if args.include_tokens_per_second:
                    num_train_tokens = self.num_tokens(train_dataloader) * args.num_train_epochs
        elif args.max_steps > 0:  # Rely on max_steps when dataloader does not have a working size
            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_examples = total_train_batch_size * args.max_steps
            num_train_samples = args.max_steps * total_train_batch_size
            if args.include_tokens_per_second:
                num_train_tokens = self.num_tokens(train_dataloader, args.max_steps) * args.gradient_accumulation_steps
        else:
            raise ValueError(
                "args.max_steps must be set to a positive value if dataloader does not have a length, was"
                f" {args.max_steps}"
            )

        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
            or is_sagemaker_mp_enabled()
            or self.fsdp is not None
            or self.is_fsdp_enabled
        )

        # We need to reset the scheduler, as its parameters may be different on subsequent calls
        if self._created_lr_scheduler:
            self.lr_scheduler = None
            self._created_lr_scheduler = False

        if self.is_deepspeed_enabled:
            self.optimizer, self.lr_scheduler = deepspeed_init(self, num_training_steps=max_steps)

        if 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

        # Compute absolute values for logging, eval, and save if given as ratio
        if args.logging_steps is not None:
            if args.logging_steps < 1:
                self.state.logging_steps = math.ceil(max_steps * args.logging_steps)
            else:
                self.state.logging_steps = args.logging_steps
        if args.eval_steps is not None:
            if args.eval_steps < 1:
                self.state.eval_steps = math.ceil(max_steps * args.eval_steps)
            else:
                self.state.eval_steps = args.eval_steps
        if args.save_steps is not None:
            if args.save_steps < 1:
                self.state.save_steps = math.ceil(max_steps * args.save_steps)
            else:
                self.state.save_steps = args.save_steps

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

        model = self._wrap_model(self.model_wrapped)

        # as the model is wrapped, don't use `accelerator.prepare`
        # this is for unhandled cases such as
        # FSDP-XLA, SageMaker MP/DP, DataParallel, IPEX
        use_accelerator_prepare = True if model is self.model else False

        if delay_optimizer_creation:
            if use_accelerator_prepare:
                self.model = self.accelerator.prepare(self.model)
            self.create_optimizer_and_scheduler(num_training_steps=max_steps)

        # prepare using `accelerator` prepare
        if use_accelerator_prepare:
            self.model.train()
            if hasattr(self.lr_scheduler, "step"):
                if self.use_apex:
                    model = self.accelerator.prepare(self.model)
                else:
                    model, self.optimizer = self.accelerator.prepare(self.model, self.optimizer)
            else:
                # to handle cases wherein we pass "DummyScheduler" such as when it is specified in DeepSpeed config.
                model, self.optimizer, self.lr_scheduler = self.accelerator.prepare(
                    self.model, self.optimizer, self.lr_scheduler
                )

        if self.is_fsdp_enabled:
            self.model = self.model_wrapped = model

        # 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

        # backward compatibility
        if self.is_deepspeed_enabled:
            self.deepspeed = self.model_wrapped

        # ckpt loading
        if resume_from_checkpoint is not None:
            if self.is_deepspeed_enabled:
                deepspeed_load_checkpoint(self.model_wrapped, resume_from_checkpoint)
            elif is_sagemaker_mp_enabled() or self.is_fsdp_enabled:
                self._load_from_checkpoint(resume_from_checkpoint, self.model_wrapped)

        # 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),
        # FSDP(Transformers Model), Dynamo Optimized Module(Transformers Model) etc.

        # Train!
        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 = {self.args.per_device_train_batch_size:,}")
        if self.args.per_device_train_batch_size != self._train_batch_size:
            logger.info(f"  Training with DataParallel so batch size has been adjusted to: {self._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:,}")
        logger.info(f"  Number of trainable parameters = {get_model_param_count(model, trainable_only=True):,}")

        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"
                    f" {steps_trained_in_current_epoch} batches in the first epoch."
                )

        # 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
        if self.hp_name is not None and self._trial is not None:
            # use self._trial because the SigOpt/Optuna hpo only call `_hp_search_setup(trial)` instead of passing trial
            # parameter to Train when using DDP.
            self.state.trial_name = self.hp_name(self._trial)
        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)
        tr_core_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._total_core_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):
                sampler = get_dataloader_sampler(train_dataloader)
                is_random_sampler = isinstance(sampler, RandomSampler)
                if is_torch_less_than_1_11 or not is_random_sampler:
                    # We just need to begin an iteration to create the randomization of the sampler.
                    for _ in train_dataloader:
                        break
                else:
                    # Otherwise we need to call the whooooole sampler cause there is some random operation added
                    # AT THE VERY END!
                    sampler = sampler if sampler is not None else []
                    _ = list(sampler)

        total_batched_samples = 0
        for epoch in range(epochs_trained, num_train_epochs):
            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 len_dataloader is not None
                else args.max_steps * args.gradient_accumulation_steps
            )
            self.control = self.callback_handler.on_epoch_begin(args, self.state, self.control)

            if epoch == epochs_trained and resume_from_checkpoint is not None and steps_trained_in_current_epoch == 0:
                self._load_rng_state(resume_from_checkpoint)

            rng_to_sync = False
            steps_skipped = 0
            if steps_trained_in_current_epoch > 0:
                epoch_iterator = skip_first_batches(epoch_iterator, steps_trained_in_current_epoch)
                steps_skipped = steps_trained_in_current_epoch
                steps_trained_in_current_epoch = 0
                rng_to_sync = True

            step = -1
            for step, inputs in enumerate(epoch_iterator):
                total_batched_samples += 1
                if rng_to_sync:
                    self._load_rng_state(resume_from_checkpoint)
                    rng_to_sync = False

                # 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)

                with self.accelerator.accumulate(model):
                    tr_loss_step, tr_core_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

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

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

                is_last_step_and_steps_less_than_grad_acc = (
                    steps_in_epoch <= args.gradient_accumulation_steps and (step + 1) == steps_in_epoch
                )

                if (
                    total_batched_samples % args.gradient_accumulation_steps == 0
                    or
                    # last step in epoch but step is always smaller than gradient_accumulation_steps
                    is_last_step_and_steps_less_than_grad_acc
                ):
                    # the `or` condition of `is_last_step_and_steps_less_than_grad_acc` is not covered
                    # in accelerate. So, explicitly enable sync gradients to True in that case.
                    if is_last_step_and_steps_less_than_grad_acc or (
                        version.parse(accelerate_version) <= version.parse("0.20.3")
                    ):
                        self.accelerator.gradient_state._set_sync_gradients(True)

                    # Gradient clipping
                    if args.max_grad_norm is not None and args.max_grad_norm > 0:
                        # deepspeed does its own clipping

                        if self.do_grad_scaling:
                            # Reduce gradients first for XLA
                            if is_torch_tpu_available():
                                gradients = xm._fetch_gradients(self.optimizer)
                                xm.all_reduce("sum", gradients, scale=1.0 / xm.xrt_world_size())
                            # AMP: gradients need unscaling
                            self.scaler.unscale_(self.optimizer)

                        if is_sagemaker_mp_enabled() and args.fp16:
                            self.optimizer.clip_master_grads(args.max_grad_norm)
                        elif 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)
                        elif self.use_apex:
                            # Revert to normal clipping otherwise, handling Apex or full precision
                            nn.utils.clip_grad_norm_(
                                amp.master_params(self.optimizer),
                                args.max_grad_norm,
                            )
                        else:
                            self.accelerator.clip_grad_norm_(
                                model.parameters(),
                                args.max_grad_norm,
                            )

                    # Optimizer step
                    optimizer_was_run = True
                    if is_torch_tpu_available():
                        if self.do_grad_scaling:
                            self.scaler.step(self.optimizer)
                            self.scaler.update()
                        else:
                            # tpu-comment: accelerate wrapped optimizers call xm.optimizer_step
                            self.optimizer.step()
                    elif self.do_grad_scaling:
                        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()
                        optimizer_was_run = not self.accelerator.optimizer_step_was_skipped

                    if optimizer_was_run:
                        # Delay optimizer scheduling until metrics are generated
                        if not isinstance(self.lr_scheduler, torch.optim.lr_scheduler.ReduceLROnPlateau):
                            self.lr_scheduler.step()

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

                    self._maybe_log_save_evaluate(tr_loss, tr_core_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(
                    "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, tr_core_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 sure the model has been saved by process 0.
            if is_torch_tpu_available():
                xm.rendezvous("load_best_model_at_end")
            elif args.parallel_mode == ParallelMode.DISTRIBUTED:
                dist.barrier()
            elif is_sagemaker_mp_enabled():
                smp.barrier()

            self._load_best_model()

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

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

        self.is_in_train = False

        self._memory_tracker.stop_and_update_metrics(metrics)

        self.log(metrics)

        run_dir = self._get_output_dir(trial)
        checkpoints_sorted = self._sorted_checkpoints(use_mtime=False, output_dir=run_dir)

        # Delete the last checkpoint when save_total_limit=1 if it's different from the best checkpoint and process allowed to save.
        if self.args.should_save and self.state.best_model_checkpoint is not None and self.args.save_total_limit == 1:
            for checkpoint in checkpoints_sorted:
                if not os.path.samefile(checkpoint, self.state.best_model_checkpoint):
                    logger.info(f"Deleting older checkpoint [{checkpoint}] due to args.save_total_limit")
                    shutil.rmtree(checkpoint)

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

        # Wait for the checkpoint to be uploaded.
        self._finish_current_push()

        return TrainOutput(self.state.global_step, train_loss, metrics)
    
    def _maybe_log_save_evaluate(self, tr_loss, tr_core_loss, model, trial, epoch, ignore_keys_for_eval):
        if self.control.should_log:
            if is_torch_tpu_available():
                xm.mark_step()

            logs: Dict[str, float] = {}

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

            # reset tr_loss to zero
            tr_loss -= tr_loss
            tr_core_loss -= tr_core_loss

            logs["loss"] = round(tr_loss_scalar / (self.state.global_step - self._globalstep_last_logged), 4)
            logs["core_loss"] = round(tr_core_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._total_core_loss_scalar += tr_core_loss_scalar
            self._globalstep_last_logged = self.state.global_step
            self.store_flos()

            self.log(logs)
        
        metrics = None
        if self.control.should_evaluate:
            if isinstance(self.eval_dataset, dict):
                metrics = {}
                for eval_dataset_name, eval_dataset in self.eval_dataset.items():
                    dataset_metrics = self.evaluate(
                        eval_dataset=eval_dataset,
                        ignore_keys=ignore_keys_for_eval,
                        metric_key_prefix=f"eval_{eval_dataset_name}",
                    )
                    metrics.update(dataset_metrics)
            else:
                metrics = self.evaluate(ignore_keys=ignore_keys_for_eval)
            self._report_to_hp_search(trial, self.state.global_step, metrics)

            # Run delayed LR scheduler now that metrics are populated
            if isinstance(self.lr_scheduler, torch.optim.lr_scheduler.ReduceLROnPlateau):
                metric_to_check = self.args.metric_for_best_model
                if not metric_to_check.startswith("eval_"):
                    metric_to_check = f"eval_{metric_to_check}"
                self.lr_scheduler.step(metrics[metric_to_check])

        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 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 (`nn.Module`):
                The model to train.
            inputs (`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 `labels`. Check your model's documentation for all accepted arguments.

        Return:
            `torch.Tensor`: The tensor with training loss on this batch.
        """

        model.train()
        inputs = self._prepare_inputs(inputs)

        if is_sagemaker_mp_enabled():
            loss_mb = smp_forward_backward(model, inputs, self.args.gradient_accumulation_steps)
            return loss_mb.reduce_mean().detach().to(self.args.device)

        with self.compute_loss_context_manager():
            loss, core_loss = self.compute_loss(model, inputs)

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

        if self.do_grad_scaling:
            self.scaler.scale(loss).backward()
        elif self.use_apex:
            raise NotImplementedError
            # with amp.scale_loss(loss, self.optimizer) as scaled_loss:
            #     scaled_loss.backward()
        else:
            self.accelerator.backward(loss)

        # https://stackoverflow.com/questions/54924582/is-it-possible-to-freeze-only-certain-embedding-weights-in-the-embedding-layer-i
        # Tested on peft 0.6.2 and 0.7.0
        if self.args.num_added_tokens > 0:
            if isinstance(self.model.model.model.embed_tokens, peft.utils.other.ModulesToSaveWrapper):
                try:
                    param = self.model.model.model.embed_tokens.modules_to_save['default'].weight
                except KeyError:
                    print(self.model.model.model.embed_tokens.modules_to_save)
                    raise
                
                
            elif isinstance(self.model.model.model.embed_tokens, nn.Embedding):
                param = self.model.model.model.embed_tokens.weight
            else:
                raise ValueError(type(self.model.model.model.embed_tokens))
            assert param.requires_grad
            num_all = param.shape[0]
            num_normal_tokens = num_all - self.args.num_added_tokens
            param.grad[0: num_normal_tokens] = 0.0

            assert isinstance(self.model.model.lm_head, peft.utils.other.ModulesToSaveWrapper)
            # try:
            #     lm_head = self.model.model.lm_head
            #     self.model.model.lm_head.weight.grad[0: num_normal_tokens] = 0.0
            #     if lm_head.bias is not None:
            #         self.model.model.lm_head.bias.grad[0: num_normal_tokens] = 0.0
            # except AttributeError:
            #     print(type(lm_head))
            #     # print(lm_head)
            #     print(type(self.model.model))
            #     # print(self.model.model)
            #     raise

        return loss.detach() / self.args.gradient_accumulation_steps, core_loss.detach() / self.args.gradient_accumulation_steps

    def compute_loss(self, model, inputs, return_outputs=False):
        labels = inputs.pop("labels")
        try:
            core_mask = inputs.pop("core_mask")
        except KeyError:
            # print(inputs)
            raise
        assert labels is not None
        # forward pass
        outputs = model(**inputs)
        logits = outputs.get("logits")
        shift_logits = logits[..., :-1, :].contiguous()
        shift_labels = labels[..., 1:].contiguous()
        core_mask = core_mask[..., 1:].contiguous()
        num_core_tokens = core_mask.sum()
        # compute custom loss (suppose one has 3 labels with different weights)
        loss_fct = nn.CrossEntropyLoss(reduction='none')
        shift_logits = shift_logits.view(-1, shift_logits.shape[-1])
        shift_labels = shift_labels.view(-1)
        core_mask = core_mask.view(-1)
        # Enable model parallelism
        shift_labels = shift_labels.to(shift_logits.device)
        loss = loss_fct(shift_logits, shift_labels)

        num = len(shift_labels)
        whole_loss = loss.sum() / num

        core_mask = core_mask.to(loss)
        if num_core_tokens == 0:
            num_core_tokens = 1  # To avoid zero devide
        
        core_loss = (loss * core_mask).sum() / num_core_tokens

        loss = whole_loss

        return (loss, core_loss, outputs) if return_outputs else (loss, core_loss)
    
    def _set_signature_columns_if_needed(self):
        super()._set_signature_columns_if_needed()
        self._signature_columns.append('core_mask')

    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 `compute_metrics` argument).

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

        Args:
            eval_dataset (`Dataset`, *optional*):
                Pass a dataset if you wish to override `self.eval_dataset`. If it is a [`~datasets.Dataset`], columns
                not accepted by the `model.forward()` method are automatically removed. It must implement the `__len__`
                method.
            ignore_keys (`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.
            metric_key_prefix (`str`, *optional*, defaults to `"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()

        if self.args.use_legacy_prediction_loop:
            raise NotImplementedError
        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
        if f"{metric_key_prefix}_jit_compilation_time" in output.metrics:
            start_time += output.metrics[f"{metric_key_prefix}_jit_compilation_time"]
        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 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 `Trainer.evaluate()` and `Trainer.predict()`.

        Works both with or without labels.
        """
        args = self.args

        prediction_loss_only = prediction_loss_only if prediction_loss_only is not None else args.prediction_loss_only

        # if eval is called w/o train, handle model prep here
        if self.is_deepspeed_enabled and self.deepspeed is None:
            _, _ = deepspeed_init(self, num_training_steps=0, inference=True)

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

        if len(self.accelerator._models) == 0 and model is self.model:
            model = (
                self.accelerator.prepare(model)
                if self.is_deepspeed_enabled
                else self.accelerator.prepare_model(model, evaluation_mode=True)
            )

            if self.is_fsdp_enabled:
                self.model = model

            # 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

            # backward compatibility
            if self.is_deepspeed_enabled:
                self.deepspeed = self.model_wrapped

        # if full fp16 or bf16 eval is wanted and this ``evaluation`` or ``predict`` isn't called
        # while ``train`` is running, cast it to the right dtype first and then put on device
        if not self.is_in_train:
            if args.fp16_full_eval:
                model = model.to(dtype=torch.float16, device=args.device)
            elif args.bf16_full_eval:
                model = model.to(dtype=torch.bfloat16, device=args.device)

        batch_size = self.args.eval_batch_size

        logger.info(f"***** Running {description} *****")
        if has_length(dataloader):
            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 = getattr(dataloader, "dataset", None)

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

        # Initialize containers
        # losses/preds/labels on GPU/TPU (accumulated for eval_accumulation_steps)
        losses_host = None
        core_losses_host = None
        preds_host = None
        labels_host = None
        inputs_host = None

        # losses/preds/labels on CPU (final containers)
        all_losses = None
        all_core_losses = None
        all_preds = None
        all_labels = None
        all_inputs = 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, core_loss, logits, labels = self.prediction_step(model, inputs, prediction_loss_only, ignore_keys=ignore_keys)
            main_input_name = getattr(self.model, "main_input_name", "input_ids")
            inputs_decode = self._prepare_input(inputs[main_input_name]) if args.include_inputs_for_metrics else None

            if is_torch_tpu_available():
                xm.mark_step()

            # Update containers on host
            if loss is not None:
                losses = self.accelerator.gather_for_metrics((loss.repeat(batch_size)))
                losses_host = losses if losses_host is None else nested_concat(losses_host, losses, padding_index=-100)
            if core_loss is not None:
                core_losses = self.accelerator.gather_for_metrics((core_loss.repeat(batch_size)))
                core_losses_host = core_losses if core_losses_host is None else nested_concat(core_losses_host, core_losses, padding_index=-100)
            if labels is not None:
                labels = self.accelerator.pad_across_processes(labels, dim=1, pad_index=-100)
            if inputs_decode is not None:
                inputs_decode = self.accelerator.pad_across_processes(inputs_decode, dim=1, pad_index=-100)
                inputs_decode = self.accelerator.gather_for_metrics((inputs_decode))
                inputs_host = (
                    inputs_decode
                    if inputs_host is None
                    else nested_concat(inputs_host, inputs_decode, padding_index=-100)
                )
            if logits is not None:
                logits = self.accelerator.pad_across_processes(logits, dim=1, pad_index=-100)
                if self.preprocess_logits_for_metrics is not None:
                    logits = self.preprocess_logits_for_metrics(logits, labels)
                logits = self.accelerator.gather_for_metrics((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.accelerator.gather_for_metrics((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(args, self.state, self.control)

            # Gather all tensors and put them back on the CPU if we have done enough accumulation steps.
            if (
                args.eval_accumulation_steps is not None
                and (step + 1) % args.eval_accumulation_steps == 0
                and (self.accelerator.sync_gradients or version.parse(accelerate_version) > version.parse("0.20.3"))
            ):
                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 core_losses_host is not None:
                    core_losses = nested_numpify(core_losses_host)
                    all_core_losses = core_losses if all_core_losses is None else np.concatenate((all_core_losses, core_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 inputs_host is not None:
                    inputs_decode = nested_numpify(inputs_host)
                    all_inputs = (
                        inputs_decode
                        if all_inputs is None
                        else nested_concat(all_inputs, inputs_decode, 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, inputs_host, labels_host = None, None, None, None

        if 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 core_losses_host is not None:
            core_losses = nested_numpify(core_losses_host)
            all_core_losses = core_losses if all_core_losses is None else np.concatenate((all_core_losses, core_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 inputs_host is not None:
            inputs_decode = nested_numpify(inputs_host)
            all_inputs = (
                inputs_decode if all_inputs is None else nested_concat(all_inputs, inputs_decode, 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 has_length(eval_dataset):
            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 getattr(eval_dataset, "num_examples", 0) > 0:
            num_samples = eval_dataset.num_examples
        else:
            if has_length(dataloader):
                num_samples = self.num_examples(dataloader)
            else:  # both len(dataloader.dataset) and len(dataloader) fail
                num_samples = observed_num_examples
        if num_samples == 0 and observed_num_examples > 0:
            num_samples = observed_num_examples

        # Metrics!
        if self.compute_metrics is not None and all_preds is not None and all_labels is not None:
            if args.include_inputs_for_metrics:
                metrics = self.compute_metrics(
                    EvalPrediction(predictions=all_preds, label_ids=all_labels, inputs=all_inputs)
                )
            else:
                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()
        if all_core_losses is not None:
            metrics[f"{metric_key_prefix}_core_loss"] = all_core_losses.mean().item()
        if hasattr(self, "jit_compilation_time"):
            metrics[f"{metric_key_prefix}_jit_compilation_time"] = self.jit_compilation_time

        # 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 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 `model` using `inputs`.

        Subclass and override to inject custom behavior.

        Args:
            model (`nn.Module`):
                The model to evaluate.
            inputs (`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 `labels`. Check your model's documentation for all accepted arguments.
            prediction_loss_only (`bool`):
                Whether or not to return the loss only.
            ignore_keys (`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.

        Return:
            Tuple[Optional[torch.Tensor], Optional[torch.Tensor], Optional[torch.Tensor]]: A tuple with the loss,
            logits and labels (each being optional).
        """
        has_labels = False if len(self.label_names) == 0 else all(inputs.get(k) is not None for k in self.label_names)
        # For CLIP-like models capable of returning loss values.
        # If `return_loss` is not specified or being `None` in `inputs`, we check if the default value of `return_loss`
        # is `True` in `model.forward`.
        return_loss = inputs.get("return_loss", None)
        if return_loss is None:
            return_loss = self.can_return_loss
        loss_without_labels = True if len(self.label_names) == 0 and return_loss else False

        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 or loss_without_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 or loss_without_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 or loss_without_labels:
                    with self.compute_loss_context_manager():
                        loss, core_loss, outputs = self.compute_loss(model, inputs, return_outputs=True)
                    loss = loss.mean().detach()
                    core_loss = core_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
                    core_loss = None
                    with self.compute_loss_context_manager():
                        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, core_loss, None, None)

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

        return (loss, core_loss, logits, labels)


class CustomDataCollator(DataCollatorForSeq2Seq):
    def __call__(self, features, return_tensors=None):
        if return_tensors is None:
            return_tensors = self.return_tensors

        labels = [feature["core_mask"] for feature in features] if "core_mask" in features[0].keys() else None
        # We have to pad the labels before calling `tokenizer.pad` as this method won't pad them and needs them of the
        # same length to return tensors.
        # batch_core_mask = None
        if labels is not None:
            max_label_length = max(len(l) for l in labels)
            if self.pad_to_multiple_of is not None:
                max_label_length = (
                    (max_label_length + self.pad_to_multiple_of - 1)
                    // self.pad_to_multiple_of
                    * self.pad_to_multiple_of
                )

            padding_side = self.tokenizer.padding_side
            for feature in features:
                remainder = [0] * (max_label_length - len(feature["core_mask"]))
                if isinstance(feature["core_mask"], list):
                    feature["core_mask"] = (
                        feature["core_mask"] + remainder if padding_side == "right" else remainder + feature["core_mask"]
                    )
                elif padding_side == "right":
                    feature["core_mask"] = np.concatenate([feature["core_mask"], remainder]).astype(np.int64)
                else:
                    feature["core_mask"] = np.concatenate([remainder, feature["core_mask"]]).astype(np.int64)

            # batch_core_mask = torch.LongTensor([feature['core_mask'] for feature in features])

        labels = [feature["labels"] for feature in features] if "labels" in features[0].keys() else None
        # We have to pad the labels before calling `tokenizer.pad` as this method won't pad them and needs them of the
        # same length to return tensors.
        if labels is not None:
            max_label_length = max(len(l) for l in labels)
            if self.pad_to_multiple_of is not None:
                max_label_length = (
                    (max_label_length + self.pad_to_multiple_of - 1)
                    // self.pad_to_multiple_of
                    * self.pad_to_multiple_of
                )

            padding_side = self.tokenizer.padding_side
            for feature in features:
                remainder = [self.label_pad_token_id] * (max_label_length - len(feature["labels"]))
                if isinstance(feature["labels"], list):
                    feature["labels"] = (
                        feature["labels"] + remainder if padding_side == "right" else remainder + feature["labels"]
                    )
                elif padding_side == "right":
                    feature["labels"] = np.concatenate([feature["labels"], remainder]).astype(np.int64)
                else:
                    feature["labels"] = np.concatenate([remainder, feature["labels"]]).astype(np.int64)


        features = self.tokenizer.pad(
            features,
            padding=self.padding,
            max_length=self.max_length,
            pad_to_multiple_of=self.pad_to_multiple_of,
            return_tensors=return_tensors,
        )

        # if batch_core_mask is not None:
        #     assert features['labels'].shape == features['core_mask'].shape
        #     features['core_mask'] = batch_core_mask

        # prepare decoder_input_ids
        if (
            labels is not None
            and self.model is not None
            and hasattr(self.model, "prepare_decoder_input_ids_from_labels")
        ):
            decoder_input_ids = self.model.prepare_decoder_input_ids_from_labels(labels=features["labels"])
            features["decoder_input_ids"] = decoder_input_ids

        return features