Transformers documentation

Trainer

You are viewing v4.37.0 version. A newer version v4.49.0 is available.
Hugging Face's logo
Join the Hugging Face community

and get access to the augmented documentation experience

to get started

Trainer

Trainer クラスは、ほずんどの暙準的なナヌスケヌスに察しお、PyTorch で機胜を完党にトレヌニングするための API を提䟛したす。これは、サンプル スクリプト のほずんどで䜿甚されおいたす。

Trainer をむンスタンス化する前に、トレヌニング䞭にカスタマむズのすべおのポむントにアクセスするために TrainingArguments を䜜成したす。

この API は、耇数の GPU/TPU での分散トレヌニング、NVIDIA Apex および PyTorch のネむティブ AMP による混合粟床をサポヌトしたす。

Trainer には、䞊蚘の機胜をサポヌトする基本的なトレヌニング ルヌプが含たれおいたす。カスタム動䜜を挿入するには、それらをサブクラス化し、次のメ゜ッドをオヌバヌラむドしたす。

  • get_train_dataloader — トレヌニング デヌタロヌダヌを䜜成したす。
  • get_eval_dataloader — 評䟡甚デヌタロヌダヌを䜜成したす。
  • get_test_dataloader — テスト デヌタロヌダヌを䜜成したす。
  • log — トレヌニングを監芖しおいるさたざたなオブゞェクトに関する情報をログに蚘録したす。
  • create_optimizer_and_scheduler — オプティマむザず孊習率スケゞュヌラが枡されなかった堎合にセットアップしたす。 初期化。 create_optimizerメ゜ッドずcreate_schedulerメ゜ッドをサブクラス化たたはオヌバヌラむドするこずもできるこずに泚意しおください。 別々に。
  • create_optimizer — init で枡されなかった堎合にオプティマむザヌをセットアップしたす。
  • create_scheduler — init で枡されなかった堎合、孊習率スケゞュヌラを蚭定したす。
  • compute_loss - トレヌニング入力のバッチの損倱を蚈算したす。
  • training_step — トレヌニング ステップを実行したす。
  • prediction_step — 評䟡/テスト ステップを実行したす。
  • evaluate — 評䟡ルヌプを実行し、メトリクスを返したす。
  • predict — テスト セットの予枬 (ラベルが䜿甚可胜な堎合はメトリクスも含む) を返したす。

Trainer クラスは 🀗 Transformers モデル甚に最適化されおおり、驚くべき動䜜をする可胜性がありたす 他の機皮で䜿甚する堎合。独自のモデルで䜿甚する堎合は、次の点を確認しおください。

  • モデルは垞に ModelOutput のタプルたたはサブクラスを返したす。
  • labels 匕数が指定され、その損倱が最初の倀ずしお返される堎合、モデルは損倱を蚈算できたす。 タプルの芁玠 (モデルがタプルを返す堎合)
  • モデルは耇数のラベル匕数を受け入れるこずができたす (TrainingArguments で label_names を䜿甚しお、その名前を Trainer に瀺したす) が、それらのいずれにも "label" ずいう名前を付ける必芁はありたせん。

以䞋は、加重損倱を䜿甚するように Trainer をカスタマむズする方法の䟋です (䞍均衡なトレヌニング セットがある堎合に圹立ちたす)。

from torch import nn
from transformers import Trainer


class CustomTrainer(Trainer):
    def compute_loss(self, model, inputs, return_outputs=False):
        labels = inputs.pop("labels")
        # forward pass
        outputs = model(**inputs)
        logits = outputs.get("logits")
        # compute custom loss (suppose one has 3 labels with different weights)
        loss_fct = nn.CrossEntropyLoss(weight=torch.tensor([1.0, 2.0, 3.0], device=model.device))
        loss = loss_fct(logits.view(-1, self.model.config.num_labels), labels.view(-1))
        return (loss, outputs) if return_outputs else loss

PyTorch Trainer のトレヌニング ルヌプの動䜜をカスタマむズするもう 1 ぀の方法は、トレヌニング ルヌプの状態を怜査できる callbacks を䜿甚するこずです (進行状況レポヌト、TensorBoard たたは他の ML プラットフォヌムでのログ蚘録など)。決定早期停止など。

Trainer

class transformers.Trainer

< >

( model: Union = Noneargs: TrainingArguments = Nonedata_collator: Optional = Nonetrain_dataset: Optional = Noneeval_dataset: Union = Nonetokenizer: Optional = Nonemodel_init: Optional = Nonecompute_metrics: Optional = Nonecallbacks: Optional = Noneoptimizers: Tuple = (None, None)preprocess_logits_for_metrics: Optional = None )

Parameters

  • model (PreTrainedModel or torch.nn.Module, optional) — The model to train, evaluate or use for predictions. If not provided, a model_init must be passed.

    Trainer is optimized to work with the PreTrainedModel provided by the library. You can still use your own models defined as torch.nn.Module as long as they work the same way as the 🀗 Transformers models.

  • args (TrainingArguments, optional) — The arguments to tweak for training. Will default to a basic instance of TrainingArguments with the output_dir set to a directory named tmp_trainer in the current directory if not provided.
  • data_collator (DataCollator, optional) — The function to use to form a batch from a list of elements of train_dataset or eval_dataset. Will default to default_data_collator() if no tokenizer is provided, an instance of DataCollatorWithPadding otherwise.
  • train_dataset (torch.utils.data.Dataset or torch.utils.data.IterableDataset, optional) — The dataset to use for training. If it is a Dataset, columns not accepted by the model.forward() method are automatically removed.

    Note that if it’s a torch.utils.data.IterableDataset with some randomization and you are training in a distributed fashion, your iterable dataset should either use a internal attribute generator that is a torch.Generator for the randomization that must be identical on all processes (and the Trainer will manually set the seed of this generator at each epoch) or have a set_epoch() method that internally sets the seed of the RNGs used.

  • eval_dataset (Union[torch.utils.data.Dataset, Dict[str, torch.utils.data.Dataset]), optional) — The dataset to use for evaluation. If it is a Dataset, columns not accepted by the model.forward() method are automatically removed. If it is a dictionary, it will evaluate on each dataset prepending the dictionary key to the metric name.
  • tokenizer (PreTrainedTokenizerBase, optional) — The tokenizer used to preprocess the data. If provided, will be used to automatically pad the inputs to 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 (Callable[[], PreTrainedModel], optional) — A function that instantiates the model to be used. If provided, each call to 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 (Callable[[EvalPrediction], Dict], optional) — The function that will be used to compute metrics at evaluation. Must take a EvalPrediction and return a dictionary string to metric values.
  • callbacks (List of TrainerCallback, optional) — A list of callbacks to customize the training loop. Will add those to the list of default callbacks detailed in here.

    If you want to remove one of the default callbacks used, use the Trainer.remove_callback() method.

  • optimizers (Tuple[torch.optim.Optimizer, torch.optim.lr_scheduler.LambdaLR], optional, defaults to (None, None)) — A tuple containing the optimizer and the scheduler to use. Will default to an instance of AdamW on your model and a scheduler given by get_linear_schedule_with_warmup() controlled by args.
  • preprocess_logits_for_metrics (Callable[[torch.Tensor, torch.Tensor], torch.Tensor], optional) — A function that preprocess the logits right before caching them at each evaluation step. Must take two tensors, the logits and the labels, and return the logits once processed as desired. The modifications made by this function will be reflected in the predictions received by compute_metrics.

    Note that the labels (second parameter) will be None if the dataset does not have them.

Trainer is a simple but feature-complete training and eval loop for PyTorch, optimized for 🀗 Transformers.

Important attributes:

  • model — Always points to the core model. If using a transformers model, it will be a 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 False if model parallel or deepspeed is used, or if the default TrainingArguments.place_model_on_device is overridden to return False .
  • is_in_train — Whether or not a model is currently running train (e.g. when evaluate is called while in train)

add_callback

< >

( callback )

Parameters

Add a callback to the current list of TrainerCallback.

autocast_smart_context_manager

< >

( cache_enabled: Optional = True )

A helper wrapper that creates an appropriate context manager for autocast while feeding it the desired arguments, depending on the situation.

compute_loss

< >

( modelinputsreturn_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.

compute_loss_context_manager

< >

( )

A helper wrapper to group together context managers.

create_model_card

< >

( language: Optional = Nonelicense: Optional = Nonetags: Union = Nonemodel_name: Optional = Nonefinetuned_from: Optional = Nonetasks: Union = Nonedataset_tags: Union = Nonedataset: Union = Nonedataset_args: Union = None )

Parameters

  • language (str, optional) — The language of the model (if applicable)
  • license (str, optional) — The license of the model. Will default to the license of the pretrained model used, if the original model given to the Trainer comes from a repo on the Hub.
  • tags (str or List[str], optional) — Some tags to be included in the metadata of the model card.
  • model_name (str, optional) — The name of the model.
  • finetuned_from (str, optional) — The name of the model used to fine-tune this one (if applicable). Will default to the name of the repo of the original model given to the Trainer (if it comes from the Hub).
  • tasks (str or List[str], optional) — One or several task identifiers, to be included in the metadata of the model card.
  • dataset_tags (str or List[str], optional) — One or several dataset tags, to be included in the metadata of the model card.
  • dataset (str or List[str], optional) — One or several dataset identifiers, to be included in the metadata of the model card.
  • dataset_args (str or List[str], optional) — One or several dataset arguments, to be included in the metadata of the model card.

Creates a draft of a model card using the information available to the Trainer.

create_optimizer

< >

( )

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 optimizers, or subclass and override this method in a subclass.

create_optimizer_and_scheduler

< >

( 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 optimizers, or subclass and override this method (or create_optimizer and/or create_scheduler) in a subclass.

create_scheduler

< >

( num_training_steps: intoptimizer: Optimizer = None )

Parameters

  • num_training_steps (int) — The number of training steps to do.

Setup the scheduler. The optimizer of the trainer must have been set up either before this method is called or passed as an argument.

evaluate

< >

( eval_dataset: Union = Noneignore_keys: Optional = Nonemetric_key_prefix: str = 'eval' )

Parameters

  • eval_dataset (Union[Dataset, Dict[str, Dataset]), optional) — Pass a dataset if you wish to override self.eval_dataset. If it is a Dataset, columns not accepted by the model.forward() method are automatically removed. If it is a dictionary, it will evaluate on each dataset, prepending the dictionary key to the metric name. Datasets must implement the __len__ method.

    If you pass a dictionary with names of datasets as keys and datasets as values, evaluate will run separate evaluations on each dataset. This can be useful to monitor how training affects other datasets or simply to get a more fine-grained evaluation. When used with load_best_model_at_end, make sure metric_for_best_model references exactly one of the datasets. If you, for example, pass in {"data1": data1, "data2": data2} for two datasets data1 and data2, you could specify metric_for_best_model="eval_data1_loss" for using the loss on data1 and metric_for_best_model="eval_data1_loss" for the loss on data2.

  • 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)

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.

evaluation_loop

< >

( dataloader: DataLoaderdescription: strprediction_loss_only: Optional = Noneignore_keys: Optional = Nonemetric_key_prefix: str = 'eval' )

Prediction/evaluation loop, shared by Trainer.evaluate() and Trainer.predict().

Works both with or without labels.

floating_point_ops

< >

( inputs: Dict ) → int

Parameters

  • inputs (Dict[str, Union[torch.Tensor, Any]]) — The inputs and targets of the model.

Returns

int

The number of floating-point operations.

For models that inherit from 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.

get_decay_parameter_names

< >

( model )

Get all parameter names that weight decay will be applied to

Note that some models implement their own layernorm instead of calling nn.LayerNorm, weight decay could still apply to those modules since this function only filter out instance of nn.LayerNorm

get_eval_dataloader

< >

( eval_dataset: Optional = None )

Parameters

  • eval_dataset (torch.utils.data.Dataset, optional) — If provided, will override self.eval_dataset. If it is a Dataset, columns not accepted by the model.forward() method are automatically removed. It must implement __len__.

Returns the evaluation ~torch.utils.data.DataLoader.

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

get_optimizer_cls_and_kwargs

< >

( args: TrainingArguments )

Parameters

  • args (transformers.training_args.TrainingArguments) — The training arguments for the training session.

Returns the optimizer class and optimizer parameters based on the training arguments.

get_test_dataloader

< >

( test_dataset: Dataset )

Parameters

  • test_dataset (torch.utils.data.Dataset, optional) — The test dataset to use. If it is a Dataset, columns not accepted by the model.forward() method are automatically removed. It must implement __len__.

Returns the test ~torch.utils.data.DataLoader.

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

get_train_dataloader

< >

( )

Returns the training ~torch.utils.data.DataLoader.

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

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

hyperparameter_search

< >

( hp_space: Optional = Nonecompute_objective: Optional = Nonen_trials: int = 20direction: Union = 'minimize'backend: Union = Nonehp_name: Optional = None**kwargs ) → [trainer_utils.BestRun or List[trainer_utils.BestRun]]

Parameters

  • hp_space (Callable[["optuna.Trial"], Dict[str, float]], optional) — A function that defines the hyperparameter search space. Will default to default_hp_space_optuna() or default_hp_space_ray() or default_hp_space_sigopt() depending on your backend.
  • compute_objective (Callable[[Dict[str, float]], float], optional) — A function computing the objective to minimize or maximize from the metrics returned by the evaluate method. Will default to default_compute_objective().
  • n_trials (int, optional, defaults to 100) — The number of trial runs to test.
  • direction (str or List[str], optional, defaults to "minimize") — If it’s single objective optimization, direction is str, can be "minimize" or "maximize", you should pick "minimize" when optimizing the validation loss, "maximize" when optimizing one or several metrics. If it’s multi objectives optimization, direction is List[str], can be List of "minimize" and "maximize", you should pick "minimize" when optimizing the validation loss, "maximize" when optimizing one or several metrics.
  • backend (str or ~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.
  • hp_name (Callable[["optuna.Trial"], str]], optional) — A function that defines the trial/run name. Will default to None.
  • kwargs (Dict[str, Any], optional) — Additional keyword arguments passed along to optuna.create_study or ray.tune.run. For more information see:

Returns

[trainer_utils.BestRun or List[trainer_utils.BestRun]]

All the information about the best run or best runs for multi-objective optimization. Experiment summary can be found in run_summary attribute for Ray backend.

Launch an hyperparameter search using optuna or Ray Tune or SigOpt. The optimized quantity is determined by compute_objective, which defaults to a function returning the evaluation loss when no metric is provided, the sum of all metrics otherwise.

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

init_hf_repo

< >

( )

Initializes a git repo in self.args.hub_model_id.

is_local_process_zero

< >

( )

Whether or not this process is the local (e.g., on one machine if training in a distributed fashion on several machines) main process.

is_world_process_zero

< >

( )

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 True for one process).

log

< >

( logs: Dict )

Parameters

  • logs (Dict[str, float]) — The values to log.

Log logs on the various objects watching training.

Subclass and override this method to inject custom behavior.

log_metrics

< >

( splitmetrics )

Parameters

  • split (str) — Mode/split name: one of train, eval, test
  • metrics (Dict[str, float]) — The metrics returned from train/evaluate/predictmetrics: metrics dict

Log metrics in a specially formatted way

Under distributed environment this is done only for a process with rank 0.

Notes on memory reports:

In order to get memory usage report you need to install psutil. You can do that with pip install psutil.

Now when this method is run, you will see a report that will include: :

init_mem_cpu_alloc_delta   =     1301MB
init_mem_cpu_peaked_delta  =      154MB
init_mem_gpu_alloc_delta   =      230MB
init_mem_gpu_peaked_delta  =        0MB
train_mem_cpu_alloc_delta  =     1345MB
train_mem_cpu_peaked_delta =        0MB
train_mem_gpu_alloc_delta  =      693MB
train_mem_gpu_peaked_delta =        7MB

Understanding the reports:

  • the first segment, e.g., train__, tells you which stage the metrics are for. Reports starting with init_ will be added to the first stage that gets run. So that if only evaluation is run, the memory usage for the __init__ will be reported along with the eval_ metrics.
  • the third segment, is either cpu or gpu, tells you whether it’s the general RAM or the gpu0 memory metric.
  • *_alloc_delta - is the difference in the used/allocated memory counter between the end and the start of the stage - it can be negative if a function released more memory than it allocated.
  • *_peaked_delta - is any extra memory that was consumed and then freed - relative to the current allocated memory counter - it is never negative. When you look at the metrics of any stage you add up alloc_delta + peaked_delta and you know how much memory was needed to complete that stage.

The reporting happens only for process of rank 0 and gpu 0 (if there is a gpu). Typically this is enough since the main process does the bulk of work, but it could be not quite so if model parallel is used and then other GPUs may use a different amount of gpu memory. This is also not the same under DataParallel where gpu0 may require much more memory than the rest since it stores the gradient and optimizer states for all participating GPUS. Perhaps in the future these reports will evolve to measure those too.

The CPU RAM metric measures RSS (Resident Set Size) includes both the memory which is unique to the process and the memory shared with other processes. It is important to note that it does not include swapped out memory, so the reports could be imprecise.

The CPU peak memory is measured using a sampling thread. Due to python’s GIL it may miss some of the peak memory if that thread didn’t get a chance to run when the highest memory was used. Therefore this report can be less than reality. Using tracemalloc would have reported the exact peak memory, but it doesn’t report memory allocations outside of python. So if some C++ CUDA extension allocated its own memory it won’t be reported. And therefore it was dropped in favor of the memory sampling approach, which reads the current process memory usage.

The GPU allocated and peak memory reporting is done with torch.cuda.memory_allocated() and torch.cuda.max_memory_allocated(). This metric reports only “deltas” for pytorch-specific allocations, as torch.cuda memory management system doesn’t track any memory allocated outside of pytorch. For example, the very first cuda call typically loads CUDA kernels, which may take from 0.5 to 2GB of GPU memory.

Note that this tracker doesn’t account for memory allocations outside of Trainer’s __init__, train, evaluate and predict calls.

Because evaluation calls may happen during train, we can’t handle nested invocations because torch.cuda.max_memory_allocated is a single counter, so if it gets reset by a nested eval call, train’s tracker will report incorrect info. If this pytorch issue gets resolved it will be possible to change this class to be re-entrant. Until then we will only track the outer level of train, evaluate and predict methods. Which means that if eval is called during train, it’s the latter that will account for its memory usage and that of the former.

This also means that if any other tool that is used along the Trainer calls torch.cuda.reset_peak_memory_stats, the gpu peak memory stats could be invalid. And the Trainer will disrupt the normal behavior of any such tools that rely on calling torch.cuda.reset_peak_memory_stats themselves.

For best performance you may want to consider turning the memory profiling off for production runs.

metrics_format

< >

( metrics: Dict ) → metrics (Dict[str, float])

Parameters

  • metrics (Dict[str, float]) — The metrics returned from train/evaluate/predict

Returns

metrics (Dict[str, float])

The reformatted metrics

Reformat Trainer metrics values to a human-readable format

num_examples

< >

( dataloader: DataLoader )

Helper to get number of samples in a ~torch.utils.data.DataLoader by accessing its dataset. When dataloader.dataset does not exist or has no length, estimates as best it can

num_tokens

< >

( train_dl: DataLoadermax_steps: Optional = None )

Helper to get number of tokens in a ~torch.utils.data.DataLoader by enumerating dataloader.

pop_callback

< >

( callback ) → TrainerCallback

Parameters

Returns

TrainerCallback

The callback removed, if found.

Remove a callback from the current list of TrainerCallback and returns it.

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

predict

< >

( test_dataset: Datasetignore_keys: Optional = Nonemetric_key_prefix: str = 'test' )

Parameters

  • test_dataset (Dataset) — Dataset to run the predictions on. If it is an datasets.Dataset, columns not accepted by the model.forward() method are automatically removed. Has to implement the method __len__
  • 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 "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)

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 evaluate().

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 (np.ndarray): The predictions on test_dataset.
  • label_ids (np.ndarray, optional): The labels (if the dataset contained some).
  • metrics (Dict[str, float], optional): The potential dictionary of metrics (if the dataset contained labels).

prediction_loop

< >

( dataloader: DataLoaderdescription: strprediction_loss_only: Optional = Noneignore_keys: Optional = Nonemetric_key_prefix: str = 'eval' )

Prediction/evaluation loop, shared by Trainer.evaluate() and Trainer.predict().

Works both with or without labels.

prediction_step

< >

( model: Moduleinputs: Dictprediction_loss_only: boolignore_keys: Optional = None ) → Tuple[Optional[torch.Tensor], Optional[torch.Tensor], Optional[torch.Tensor]]

Parameters

  • 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.

Returns

Tuple[Optional[torch.Tensor], Optional[torch.Tensor], Optional[torch.Tensor]]

A tuple with the loss, logits and labels (each being optional).

Perform an evaluation step on model using inputs.

Subclass and override to inject custom behavior.

propagate_args_to_deepspeed

< >

( auto_find_batch_size = False )

Sets values in the deepspeed plugin based on the Trainer args

push_to_hub

< >

( commit_message: Optional = 'End of training'blocking: bool = True**kwargs )

Parameters

  • commit_message (str, optional, defaults to "End of training") — Message to commit while pushing.
  • blocking (bool, optional, defaults to True) — Whether the function should return only when the git push has finished.
  • kwargs (Dict[str, Any], optional) — Additional keyword arguments passed along to create_model_card().

Upload self.model and self.tokenizer to the 🀗 model hub on the repo self.args.hub_model_id.

remove_callback

< >

( callback )

Parameters

Remove a callback from the current list of TrainerCallback.

save_metrics

< >

( splitmetricscombined = True )

Parameters

  • split (str) — Mode/split name: one of train, eval, test, all
  • metrics (Dict[str, float]) — The metrics returned from train/evaluate/predict
  • combined (bool, optional, defaults to True) — Creates combined metrics by updating all_results.json with metrics of this call

Save metrics into a json file for that split, e.g. train_results.json.

Under distributed environment this is done only for a process with rank 0.

To understand the metrics please read the docstring of log_metrics(). The only difference is that raw unformatted numbers are saved in the current method.

save_model

< >

( output_dir: Optional = None_internal_call: bool = False )

Will save the model, so you can reload it using from_pretrained().

Will only save from the main process.

save_state

< >

( )

Saves the Trainer state, since Trainer.save_model saves only the tokenizer with the model

Under distributed environment this is done only for a process with rank 0.

train

< >

( resume_from_checkpoint: Union = Nonetrial: Union = Noneignore_keys_for_eval: Optional = None**kwargs )

Parameters

  • resume_from_checkpoint (str or bool, optional) — If a str, local path to a saved checkpoint as saved by a previous instance of Trainer. If a bool and equals True, load the last checkpoint in args.output_dir as saved by a previous instance of Trainer. If present, training will resume from the model/optimizer/scheduler states loaded here.
  • trial (optuna.Trial or Dict[str, Any], optional) — The trial run or the hyperparameter dictionary for hyperparameter search.
  • ignore_keys_for_eval (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 (Dict[str, Any], optional) — Additional keyword arguments used to hide deprecated arguments

Main training entry point.

training_step

< >

( model: Moduleinputs: Dict ) → torch.Tensor

Parameters

  • 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.

Returns

torch.Tensor

The tensor with training loss on this batch.

Perform a training step on a batch of inputs.

Subclass and override to inject custom behavior.

Seq2SeqTrainer

class transformers.Seq2SeqTrainer

< >

( model: Union = Noneargs: TrainingArguments = Nonedata_collator: Optional = Nonetrain_dataset: Optional = Noneeval_dataset: Union = Nonetokenizer: Optional = Nonemodel_init: Optional = Nonecompute_metrics: Optional = Nonecallbacks: Optional = Noneoptimizers: Tuple = (None, None)preprocess_logits_for_metrics: Optional = None )

evaluate

< >

( eval_dataset: Optional = Noneignore_keys: Optional = Nonemetric_key_prefix: str = 'eval'**gen_kwargs )

Parameters

  • eval_dataset (Dataset, optional) — Pass a dataset if you wish to override self.eval_dataset. If it is an 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)
  • max_length (int, optional) — The maximum target length to use when predicting with the generate method.
  • num_beams (int, optional) — Number of beams for beam search that will be used when predicting with the generate method. 1 means no beam search. gen_kwargs — Additional generate specific kwargs.

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.

predict

< >

( test_dataset: Datasetignore_keys: Optional = Nonemetric_key_prefix: str = 'test'**gen_kwargs )

Parameters

  • test_dataset (Dataset) — Dataset to run the predictions on. If it is a Dataset, columns not accepted by the model.forward() method are automatically removed. Has to implement the method __len__
  • 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)
  • max_length (int, optional) — The maximum target length to use when predicting with the generate method.
  • num_beams (int, optional) — Number of beams for beam search that will be used when predicting with the generate method. 1 means no beam search. gen_kwargs — Additional generate specific kwargs.

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 evaluate().

If your predictions or labels have different sequence lengths (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 (np.ndarray): The predictions on test_dataset.
  • label_ids (np.ndarray, optional): The labels (if the dataset contained some).
  • metrics (Dict[str, float], optional): The potential dictionary of metrics (if the dataset contained labels).

TrainingArguments

class transformers.TrainingArguments

< >

( output_dir: stroverwrite_output_dir: bool = Falsedo_train: bool = Falsedo_eval: bool = Falsedo_predict: bool = Falseevaluation_strategy: Union = 'no'prediction_loss_only: bool = Falseper_device_train_batch_size: int = 8per_device_eval_batch_size: int = 8per_gpu_train_batch_size: Optional = Noneper_gpu_eval_batch_size: Optional = Nonegradient_accumulation_steps: int = 1eval_accumulation_steps: Optional = Noneeval_delay: Optional = 0learning_rate: float = 5e-05weight_decay: float = 0.0adam_beta1: float = 0.9adam_beta2: float = 0.999adam_epsilon: float = 1e-08max_grad_norm: float = 1.0num_train_epochs: float = 3.0max_steps: int = -1lr_scheduler_type: Union = 'linear'lr_scheduler_kwargs: Optional = <factory>warmup_ratio: float = 0.0warmup_steps: int = 0log_level: Optional = 'passive'log_level_replica: Optional = 'warning'log_on_each_node: bool = Truelogging_dir: Optional = Nonelogging_strategy: Union = 'steps'logging_first_step: bool = Falselogging_steps: float = 500logging_nan_inf_filter: bool = Truesave_strategy: Union = 'steps'save_steps: float = 500save_total_limit: Optional = Nonesave_safetensors: Optional = Truesave_on_each_node: bool = Falsesave_only_model: bool = Falseno_cuda: bool = Falseuse_cpu: bool = Falseuse_mps_device: bool = Falseseed: int = 42data_seed: Optional = Nonejit_mode_eval: bool = Falseuse_ipex: bool = Falsebf16: bool = Falsefp16: bool = Falsefp16_opt_level: str = 'O1'half_precision_backend: str = 'auto'bf16_full_eval: bool = Falsefp16_full_eval: bool = Falsetf32: Optional = Nonelocal_rank: int = -1ddp_backend: Optional = Nonetpu_num_cores: Optional = Nonetpu_metrics_debug: bool = Falsedebug: Union = ''dataloader_drop_last: bool = Falseeval_steps: Optional = Nonedataloader_num_workers: int = 0past_index: int = -1run_name: Optional = Nonedisable_tqdm: Optional = Noneremove_unused_columns: Optional = Truelabel_names: Optional = Noneload_best_model_at_end: Optional = Falsemetric_for_best_model: Optional = Nonegreater_is_better: Optional = Noneignore_data_skip: bool = Falsefsdp: Union = ''fsdp_min_num_params: int = 0fsdp_config: Optional = Nonefsdp_transformer_layer_cls_to_wrap: Optional = Nonedeepspeed: Optional = Nonelabel_smoothing_factor: float = 0.0optim: Union = 'adamw_torch'optim_args: Optional = Noneadafactor: bool = Falsegroup_by_length: bool = Falselength_column_name: Optional = 'length'report_to: Optional = Noneddp_find_unused_parameters: Optional = Noneddp_bucket_cap_mb: Optional = Noneddp_broadcast_buffers: Optional = Nonedataloader_pin_memory: bool = Truedataloader_persistent_workers: bool = Falseskip_memory_metrics: bool = Trueuse_legacy_prediction_loop: bool = Falsepush_to_hub: bool = Falseresume_from_checkpoint: Optional = Nonehub_model_id: Optional = Nonehub_strategy: Union = 'every_save'hub_token: Optional = Nonehub_private_repo: bool = Falsehub_always_push: bool = Falsegradient_checkpointing: bool = Falsegradient_checkpointing_kwargs: Optional = Noneinclude_inputs_for_metrics: bool = Falsefp16_backend: str = 'auto'push_to_hub_model_id: Optional = Nonepush_to_hub_organization: Optional = Nonepush_to_hub_token: Optional = Nonemp_parameters: str = ''auto_find_batch_size: bool = Falsefull_determinism: bool = Falsetorchdynamo: Optional = Noneray_scope: Optional = 'last'ddp_timeout: Optional = 1800torch_compile: bool = Falsetorch_compile_backend: Optional = Nonetorch_compile_mode: Optional = Nonedispatch_batches: Optional = Nonesplit_batches: Optional = Falseinclude_tokens_per_second: Optional = Falseinclude_num_input_tokens_seen: Optional = Falseneftune_noise_alpha: float = None )

Parameters

  • output_dir (str) — The output directory where the model predictions and checkpoints will be written.
  • overwrite_output_dir (bool, optional, defaults to False) — If True, overwrite the content of the output directory. Use this to continue training if output_dir points to a checkpoint directory.
  • do_train (bool, optional, defaults to False) — Whether to run training or not. This argument is not directly used by Trainer, it’s intended to be used by your training/evaluation scripts instead. See the example scripts for more details.
  • do_eval (bool, optional) — Whether to run evaluation on the validation set or not. Will be set to True if evaluation_strategy is different from "no". This argument is not directly used by Trainer, it’s intended to be used by your training/evaluation scripts instead. See the example scripts for more details.
  • do_predict (bool, optional, defaults to False) — Whether to run predictions on the test set or not. This argument is not directly used by Trainer, it’s intended to be used by your training/evaluation scripts instead. See the example scripts for more details.
  • evaluation_strategy (str or IntervalStrategy, optional, defaults to "no") — The evaluation strategy to adopt during training. Possible values are:

    • "no": No evaluation is done during training.
    • "steps": Evaluation is done (and logged) every eval_steps.
    • "epoch": Evaluation is done at the end of each epoch.
  • prediction_loss_only (bool, optional, defaults to False) — When performing evaluation and generating predictions, only returns the loss.
  • per_device_train_batch_size (int, optional, defaults to 8) — The batch size per GPU/XPU/TPU/MPS/NPU core/CPU for training.
  • per_device_eval_batch_size (int, optional, defaults to 8) — The batch size per GPU/XPU/TPU/MPS/NPU core/CPU for evaluation.
  • gradient_accumulation_steps (int, optional, defaults to 1) — Number of updates steps to accumulate the gradients for, before performing a backward/update pass.

    When using gradient accumulation, one step is counted as one step with backward pass. Therefore, logging, evaluation, save will be conducted every gradient_accumulation_steps * xxx_step training examples.

  • eval_accumulation_steps (int, optional) — Number of predictions steps to accumulate the output tensors for, before moving the results to the CPU. If left unset, the whole predictions are accumulated on GPU/NPU/TPU before being moved to the CPU (faster but requires more memory).
  • eval_delay (float, optional) — Number of epochs or steps to wait for before the first evaluation can be performed, depending on the evaluation_strategy.
  • learning_rate (float, optional, defaults to 5e-5) — The initial learning rate for AdamW optimizer.
  • weight_decay (float, optional, defaults to 0) — The weight decay to apply (if not zero) to all layers except all bias and LayerNorm weights in AdamW optimizer.
  • adam_beta1 (float, optional, defaults to 0.9) — The beta1 hyperparameter for the AdamW optimizer.
  • adam_beta2 (float, optional, defaults to 0.999) — The beta2 hyperparameter for the AdamW optimizer.
  • adam_epsilon (float, optional, defaults to 1e-8) — The epsilon hyperparameter for the AdamW optimizer.
  • max_grad_norm (float, optional, defaults to 1.0) — Maximum gradient norm (for gradient clipping).
  • num_train_epochs(float, optional, defaults to 3.0) — Total number of training epochs to perform (if not an integer, will perform the decimal part percents of the last epoch before stopping training).
  • max_steps (int, optional, defaults to -1) — If set to a positive number, the total number of training steps to perform. Overrides num_train_epochs. For a finite dataset, training is reiterated through the dataset (if all data is exhausted) until max_steps is reached.
  • lr_scheduler_type (str or SchedulerType, optional, defaults to "linear") — The scheduler type to use. See the documentation of SchedulerType for all possible values.
  • lr_scheduler_kwargs (‘dict’, optional, defaults to {}) — The extra arguments for the lr_scheduler. See the documentation of each scheduler for possible values.
  • warmup_ratio (float, optional, defaults to 0.0) — Ratio of total training steps used for a linear warmup from 0 to learning_rate.
  • warmup_steps (int, optional, defaults to 0) — Number of steps used for a linear warmup from 0 to learning_rate. Overrides any effect of warmup_ratio.
  • log_level (str, optional, defaults to passive) — Logger log level to use on the main process. Possible choices are the log levels as strings: ‘debug’, ‘info’, ‘warning’, ‘error’ and ‘critical’, plus a ‘passive’ level which doesn’t set anything and keeps the current log level for the Transformers library (which will be "warning" by default).
  • log_level_replica (str, optional, defaults to "warning") — Logger log level to use on replicas. Same choices as log_level”
  • log_on_each_node (bool, optional, defaults to True) — In multinode distributed training, whether to log using log_level once per node, or only on the main node.
  • logging_dir (str, optional) — TensorBoard log directory. Will default to *output_dir/runs/CURRENT_DATETIME_HOSTNAME*.
  • logging_strategy (str or IntervalStrategy, optional, defaults to "steps") — The logging strategy to adopt during training. Possible values are:

    • "no": No logging is done during training.
    • "epoch": Logging is done at the end of each epoch.
    • "steps": Logging is done every logging_steps.
  • logging_first_step (bool, optional, defaults to False) — Whether to log and evaluate the first global_step or not.
  • logging_steps (int or float, optional, defaults to 500) — Number of update steps between two logs if logging_strategy="steps". Should be an integer or a float in range [0,1). If smaller than 1, will be interpreted as ratio of total training steps.
  • logging_nan_inf_filter (bool, optional, defaults to True) — Whether to filter nan and inf losses for logging. If set to True the loss of every step that is nan or inf is filtered and the average loss of the current logging window is taken instead.

    logging_nan_inf_filter only influences the logging of loss values, it does not change the behavior the gradient is computed or applied to the model.

  • save_strategy (str or IntervalStrategy, optional, defaults to "steps") — The checkpoint save strategy to adopt during training. Possible values are:

    • "no": No save is done during training.
    • "epoch": Save is done at the end of each epoch.
    • "steps": Save is done every save_steps.
  • save_steps (int or float, optional, defaults to 500) — Number of updates steps before two checkpoint saves if save_strategy="steps". Should be an integer or a float in range [0,1). If smaller than 1, will be interpreted as ratio of total training steps.
  • save_total_limit (int, optional) — If a value is passed, will limit the total amount of checkpoints. Deletes the older checkpoints in output_dir. When load_best_model_at_end is enabled, the “best” checkpoint according to metric_for_best_model will always be retained in addition to the most recent ones. For example, for save_total_limit=5 and load_best_model_at_end, the four last checkpoints will always be retained alongside the best model. When save_total_limit=1 and load_best_model_at_end, it is possible that two checkpoints are saved: the last one and the best one (if they are different).
  • save_safetensors (bool, optional, defaults to True) — Use safetensors saving and loading for state dicts instead of default torch.load and torch.save.
  • save_on_each_node (bool, optional, defaults to False) — When doing multi-node distributed training, whether to save models and checkpoints on each node, or only on the main one.

    This should not be activated when the different nodes use the same storage as the files will be saved with the same names for each node.

  • save_only_model (bool, optional, defaults to False) — When checkpointing, whether to only save the model, or also the optimizer, scheduler & rng state. Note that when this is true, you won’t be able to resume training from checkpoint. This enables you to save storage by not storing the optimizer, scheduler & rng state. You can only load the model using from_pretrained with this option set to True.
  • use_cpu (bool, optional, defaults to False) — Whether or not to use cpu. If set to False, we will use cuda or mps device if available.
  • seed (int, optional, defaults to 42) — Random seed that will be set at the beginning of training. To ensure reproducibility across runs, use the ~Trainer.model_init function to instantiate the model if it has some randomly initialized parameters.
  • data_seed (int, optional) — Random seed to be used with data samplers. If not set, random generators for data sampling will use the same seed as seed. This can be used to ensure reproducibility of data sampling, independent of the model seed.
  • jit_mode_eval (bool, optional, defaults to False) — Whether or not to use PyTorch jit trace for inference.
  • use_ipex (bool, optional, defaults to False) — Use Intel extension for PyTorch when it is available. IPEX installation.
  • bf16 (bool, optional, defaults to False) — Whether to use bf16 16-bit (mixed) precision training instead of 32-bit training. Requires Ampere or higher NVIDIA architecture or using CPU (use_cpu) or Ascend NPU. This is an experimental API and it may change.
  • fp16 (bool, optional, defaults to False) — Whether to use fp16 16-bit (mixed) precision training instead of 32-bit training.
  • fp16_opt_level (str, optional, defaults to ‘O1’) — For fp16 training, Apex AMP optimization level selected in [‘O0’, ‘O1’, ‘O2’, and ‘O3’]. See details on the Apex documentation.
  • fp16_backend (str, optional, defaults to "auto") — This argument is deprecated. Use half_precision_backend instead.
  • half_precision_backend (str, optional, defaults to "auto") — The backend to use for mixed precision training. Must be one of "auto", "apex", "cpu_amp". "auto" will use CPU/CUDA AMP or APEX depending on the PyTorch version detected, while the other choices will force the requested backend.
  • bf16_full_eval (bool, optional, defaults to False) — Whether to use full bfloat16 evaluation instead of 32-bit. This will be faster and save memory but can harm metric values. This is an experimental API and it may change.
  • fp16_full_eval (bool, optional, defaults to False) — Whether to use full float16 evaluation instead of 32-bit. This will be faster and save memory but can harm metric values.
  • tf32 (bool, optional) — Whether to enable the TF32 mode, available in Ampere and newer GPU architectures. The default value depends on PyTorch’s version default of torch.backends.cuda.matmul.allow_tf32. For more details please refer to the TF32 documentation. This is an experimental API and it may change.
  • local_rank (int, optional, defaults to -1) — Rank of the process during distributed training.
  • ddp_backend (str, optional) — The backend to use for distributed training. Must be one of "nccl", "mpi", "ccl", "gloo", "hccl".
  • tpu_num_cores (int, optional) — When training on TPU, the number of TPU cores (automatically passed by launcher script).
  • dataloader_drop_last (bool, optional, defaults to False) — Whether to drop the last incomplete batch (if the length of the dataset is not divisible by the batch size) or not.
  • eval_steps (int or float, optional) — Number of update steps between two evaluations if evaluation_strategy="steps". Will default to the same value as logging_steps if not set. Should be an integer or a float in range [0,1). If smaller than 1, will be interpreted as ratio of total training steps.
  • dataloader_num_workers (int, optional, defaults to 0) — Number of subprocesses to use for data loading (PyTorch only). 0 means that the data will be loaded in the main process.
  • past_index (int, optional, defaults to -1) — Some models like TransformerXL or XLNet can make use of the past hidden states for their predictions. If this argument is set to a positive int, the Trainer will use the corresponding output (usually index 2) as the past state and feed it to the model at the next training step under the keyword argument mems.
  • run_name (str, optional) — A descriptor for the run. Typically used for wandb and mlflow logging.
  • disable_tqdm (bool, optional) — Whether or not to disable the tqdm progress bars and table of metrics produced by ~notebook.NotebookTrainingTracker in Jupyter Notebooks. Will default to True if the logging level is set to warn or lower (default), False otherwise.
  • remove_unused_columns (bool, optional, defaults to True) — Whether or not to automatically remove the columns unused by the model forward method.
  • label_names (List[str], optional) — The list of keys in your dictionary of inputs that correspond to the labels.

    Will eventually default to the list of argument names accepted by the model that contain the word “label”, except if the model used is one of the XxxForQuestionAnswering in which case it will also include the ["start_positions", "end_positions"] keys.

  • load_best_model_at_end (bool, optional, defaults to False) — Whether or not to load the best model found during training at the end of training. When this option is enabled, the best checkpoint will always be saved. See save_total_limit for more.

    When set to True, the parameters save_strategy needs to be the same as evaluation_strategy, and in the case it is “steps”, save_steps must be a round multiple of eval_steps.

  • metric_for_best_model (str, optional) — Use in conjunction with load_best_model_at_end to specify the metric to use to compare two different models. Must be the name of a metric returned by the evaluation with or without the prefix "eval_". Will default to "loss" if unspecified and load_best_model_at_end=True (to use the evaluation loss).

    If you set this value, greater_is_better will default to True. Don’t forget to set it to False if your metric is better when lower.

  • greater_is_better (bool, optional) — Use in conjunction with load_best_model_at_end and metric_for_best_model to specify if better models should have a greater metric or not. Will default to:

    • True if metric_for_best_model is set to a value that isn’t "loss" or "eval_loss".
    • False if metric_for_best_model is not set, or set to "loss" or "eval_loss".
  • ignore_data_skip (bool, optional, defaults to False) — When resuming training, whether or not to skip the epochs and batches to get the data loading at the same stage as in the previous training. If set to True, the training will begin faster (as that skipping step can take a long time) but will not yield the same results as the interrupted training would have.
  • fsdp (bool, str or list of FSDPOption, optional, defaults to '') — Use PyTorch Distributed Parallel Training (in distributed training only).

    A list of options along the following:

    • "full_shard": Shard parameters, gradients and optimizer states.
    • "shard_grad_op": Shard optimizer states and gradients.
    • "hybrid_shard": Apply FULL_SHARD within a node, and replicate parameters across nodes.
    • "hybrid_shard_zero2": Apply SHARD_GRAD_OP within a node, and replicate parameters across nodes.
    • "offload": Offload parameters and gradients to CPUs (only compatible with "full_shard" and "shard_grad_op").
    • "auto_wrap": Automatically recursively wrap layers with FSDP using default_auto_wrap_policy.
  • fsdp_config (str or dict, optional) — Config to be used with fsdp (Pytorch Distributed Parallel Training). The value is either a location of fsdp json config file (e.g., fsdp_config.json) or an already loaded json file as dict.

    A List of config and its options:

    • min_num_params (int, optional, defaults to 0): FSDP’s minimum number of parameters for Default Auto Wrapping. (useful only when fsdp field is passed).

    • transformer_layer_cls_to_wrap (List[str], optional): List of transformer layer class names (case-sensitive) to wrap, e.g, BertLayer, GPTJBlock, T5Block 
 (useful only when fsdp flag is passed).

    • backward_prefetch (str, optional) FSDP’s backward prefetch mode. Controls when to prefetch next set of parameters (useful only when fsdp field is passed).

      A list of options along the following:

      • "backward_pre" : Prefetches the next set of parameters before the current set of parameter’s gradient computation.
      • "backward_post" : This prefetches the next set of parameters after the current set of parameter’s gradient computation.
    • forward_prefetch (bool, optional, defaults to False) FSDP’s forward prefetch mode (useful only when fsdp field is passed). If "True", then FSDP explicitly prefetches the next upcoming all-gather while executing in the forward pass.

    • limit_all_gathers (bool, optional, defaults to False) FSDP’s limit_all_gathers (useful only when fsdp field is passed). If "True", FSDP explicitly synchronizes the CPU thread to prevent too many in-flight all-gathers.

    • use_orig_params (bool, optional, defaults to True) If "True", allows non-uniform requires_grad during init, which means support for interspersed frozen and trainable paramteres. Useful in cases such as parameter-efficient fine-tuning. Please refer this [blog](https://dev-discuss.pytorch.org/t/rethinking-pytorch-fully-sharded-data-parallel-fsdp-from-first-principles/1019

    • sync_module_states (bool, optional, defaults to True) If "True", each individually wrapped FSDP unit will broadcast module parameters from rank 0 to ensure they are the same across all ranks after initialization

    • activation_checkpointing (bool, optional, defaults to False): If "True", activation checkpointing is a technique to reduce memory usage by clearing activations of certain layers and recomputing them during a backward pass. Effectively, this trades extra computation time for reduced memory usage.

    • xla (bool, optional, defaults to False): Whether to use PyTorch/XLA Fully Sharded Data Parallel Training. This is an experimental feature and its API may evolve in the future.

    • xla_fsdp_settings (dict, optional) The value is a dictionary which stores the XLA FSDP wrapping parameters.

      For a complete list of options, please see here.

    • xla_fsdp_grad_ckpt (bool, optional, defaults to False): Will use gradient checkpointing over each nested XLA FSDP wrapped layer. This setting can only be used when the xla flag is set to true, and an auto wrapping policy is specified through fsdp_min_num_params or fsdp_transformer_layer_cls_to_wrap.

  • deepspeed (str or dict, optional) — Use Deepspeed. This is an experimental feature and its API may evolve in the future. The value is either the location of DeepSpeed json config file (e.g., ds_config.json) or an already loaded json file as a dict”
  • label_smoothing_factor (float, optional, defaults to 0.0) — The label smoothing factor to use. Zero means no label smoothing, otherwise the underlying onehot-encoded labels are changed from 0s and 1s to label_smoothing_factor/num_labels and 1 - label_smoothing_factor + label_smoothing_factor/num_labels respectively.
  • debug (str or list of DebugOption, optional, defaults to "") — Enable one or more debug features. This is an experimental feature.

    Possible options are:

    • "underflow_overflow": detects overflow in model’s input/outputs and reports the last frames that led to the event
    • "tpu_metrics_debug": print debug metrics on TPU

    The options should be separated by whitespaces.

  • optim (str or training_args.OptimizerNames, optional, defaults to "adamw_torch") — The optimizer to use: adamw_hf, adamw_torch, adamw_torch_fused, adamw_apex_fused, adamw_anyprecision or adafactor.
  • optim_args (str, optional) — Optional arguments that are supplied to AnyPrecisionAdamW.
  • group_by_length (bool, optional, defaults to False) — Whether or not to group together samples of roughly the same length in the training dataset (to minimize padding applied and be more efficient). Only useful if applying dynamic padding.
  • length_column_name (str, optional, defaults to "length") — Column name for precomputed lengths. If the column exists, grouping by length will use these values rather than computing them on train startup. Ignored unless group_by_length is True and the dataset is an instance of Dataset.
  • report_to (str or List[str], optional, defaults to "all") — The list of integrations to report the results and logs to. Supported platforms are "azure_ml", "clearml", "codecarbon", "comet_ml", "dagshub", "dvclive", "flyte", "mlflow", "neptune", "tensorboard", and "wandb". Use "all" to report to all integrations installed, "none" for no integrations.
  • ddp_find_unused_parameters (bool, optional) — When using distributed training, the value of the flag find_unused_parameters passed to DistributedDataParallel. Will default to False if gradient checkpointing is used, True otherwise.
  • ddp_bucket_cap_mb (int, optional) — When using distributed training, the value of the flag bucket_cap_mb passed to DistributedDataParallel.
  • ddp_broadcast_buffers (bool, optional) — When using distributed training, the value of the flag broadcast_buffers passed to DistributedDataParallel. Will default to False if gradient checkpointing is used, True otherwise.
  • dataloader_pin_memory (bool, optional, defaults to True) — Whether you want to pin memory in data loaders or not. Will default to True.
  • dataloader_persistent_workers (bool, optional, defaults to False) — If True, the data loader will not shut down the worker processes after a dataset has been consumed once. This allows to maintain the workers Dataset instances alive. Can potentially speed up training, but will increase RAM usage. Will default to False.
  • skip_memory_metrics (bool, optional, defaults to True) — Whether to skip adding of memory profiler reports to metrics. This is skipped by default because it slows down the training and evaluation speed.
  • push_to_hub (bool, optional, defaults to False) — Whether or not to push the model to the Hub every time the model is saved. If this is activated, output_dir will begin a git directory synced with the repo (determined by hub_model_id) and the content will be pushed each time a save is triggered (depending on your save_strategy). Calling save_model() will also trigger a push.

    If output_dir exists, it needs to be a local clone of the repository to which the Trainer will be pushed.

  • resume_from_checkpoint (str, optional) — The path to a folder with a valid checkpoint for your model. This argument is not directly used by Trainer, it’s intended to be used by your training/evaluation scripts instead. See the example scripts for more details.
  • hub_model_id (str, optional) — The name of the repository to keep in sync with the local output_dir. It can be a simple model ID in which case the model will be pushed in your namespace. Otherwise it should be the whole repository name, for instance "user_name/model", which allows you to push to an organization you are a member of with "organization_name/model". Will default to user_name/output_dir_name with output_dir_name being the name of output_dir.

    Will default to the name of output_dir.

  • hub_strategy (str or HubStrategy, optional, defaults to "every_save") — Defines the scope of what is pushed to the Hub and when. Possible values are:

    • "end": push the model, its configuration, the tokenizer (if passed along to the Trainer) and a draft of a model card when the save_model() method is called.
    • "every_save": push the model, its configuration, the tokenizer (if passed along to the Trainer) and a draft of a model card each time there is a model save. The pushes are asynchronous to not block training, and in case the save are very frequent, a new push is only attempted if the previous one is finished. A last push is made with the final model at the end of training.
    • "checkpoint": like "every_save" but the latest checkpoint is also pushed in a subfolder named last-checkpoint, allowing you to resume training easily with trainer.train(resume_from_checkpoint="last-checkpoint").
    • "all_checkpoints": like "checkpoint" but all checkpoints are pushed like they appear in the output folder (so you will get one checkpoint folder per folder in your final repository)
  • hub_token (str, optional) — The token to use to push the model to the Hub. Will default to the token in the cache folder obtained with huggingface-cli login.
  • hub_private_repo (bool, optional, defaults to False) — If True, the Hub repo will be set to private.
  • hub_always_push (bool, optional, defaults to False) — Unless this is True, the Trainer will skip pushing a checkpoint when the previous push is not finished.
  • gradient_checkpointing (bool, optional, defaults to False) — If True, use gradient checkpointing to save memory at the expense of slower backward pass.
  • gradient_checkpointing_kwargs (dict, optional, defaults to None) — Key word arguments to be passed to the gradient_checkpointing_enable method.
  • include_inputs_for_metrics (bool, optional, defaults to False) — Whether or not the inputs will be passed to the compute_metrics function. This is intended for metrics that need inputs, predictions and references for scoring calculation in Metric class.
  • auto_find_batch_size (bool, optional, defaults to False) — Whether to find a batch size that will fit into memory automatically through exponential decay, avoiding CUDA Out-of-Memory errors. Requires accelerate to be installed (pip install accelerate)
  • full_determinism (bool, optional, defaults to False) — If True, enable_full_determinism() is called instead of set_seed() to ensure reproducible results in distributed training. Important: this will negatively impact the performance, so only use it for debugging.
  • torchdynamo (str, optional) — If set, the backend compiler for TorchDynamo. Possible choices are "eager", "aot_eager", "inductor", "nvfuser", "aot_nvfuser", "aot_cudagraphs", "ofi", "fx2trt", "onnxrt" and "ipex".
  • ray_scope (str, optional, defaults to "last") — The scope to use when doing hyperparameter search with Ray. By default, "last" will be used. Ray will then use the last checkpoint of all trials, compare those, and select the best one. However, other options are also available. See the Ray documentation for more options.
  • ddp_timeout (int, optional, defaults to 1800) — The timeout for torch.distributed.init_process_group calls, used to avoid GPU socket timeouts when performing slow operations in distributed runnings. Please refer the [PyTorch documentation] (https://pytorch.org/docs/stable/distributed.html#torch.distributed.init_process_group) for more information.
  • use_mps_device (bool, optional, defaults to False) — This argument is deprecated.mps device will be used if it is available similar to cuda device.
  • torch_compile (bool, optional, defaults to False) — Whether or not to compile the model using PyTorch 2.0 torch.compile.

    This will use the best defaults for the torch.compile API. You can customize the defaults with the argument torch_compile_backend and torch_compile_mode but we don’t guarantee any of them will work as the support is progressively rolled in in PyTorch.

    This flag and the whole compile API is experimental and subject to change in future releases.

  • torch_compile_backend (str, optional) — The backend to use in torch.compile. If set to any value, torch_compile will be set to True.

    Refer to the PyTorch doc for possible values and note that they may change across PyTorch versions.

    This flag is experimental and subject to change in future releases.

  • torch_compile_mode (str, optional) — The mode to use in torch.compile. If set to any value, torch_compile will be set to True.

    Refer to the PyTorch doc for possible values and note that they may change across PyTorch versions.

    This flag is experimental and subject to change in future releases.

  • split_batches (bool, optional) — Whether or not the accelerator should split the batches yielded by the dataloaders across the devices during distributed training. If

    set to True, the actual batch size used will be the same on any kind of distributed processes, but it must be a

    round multiple of the number of processes you are using (such as GPUs).

  • include_tokens_per_second (bool, optional) — Whether or not to compute the number of tokens per second per device for training speed metrics.

    This will iterate over the entire training dataloader once beforehand,

    and will slow down the entire process.

  • include_num_input_tokens_seen (bool, optional) — Whether or not to track the number of input tokens seen throughout training.

    May be slower in distributed training as gather operations must be called.

  • neftune_noise_alpha (Optional[float]) — If not None, this will activate NEFTune noise embeddings. This can drastically improve model performance for instruction fine-tuning. Check out the original paper and the original code. Support transformers PreTrainedModel and also PeftModel from peft.

TrainingArguments is the subset of the arguments we use in our example scripts which relate to the training loop itself.

Using HfArgumentParser we can turn this class into argparse arguments that can be specified on the command line.

get_process_log_level

< >

( )

Returns the log level to be used depending on whether this process is the main process of node 0, main process of node non-0, or a non-main process.

For the main process the log level defaults to the logging level set (logging.WARNING if you didn’t do anything) unless overridden by log_level argument.

For the replica processes the log level defaults to logging.WARNING unless overridden by log_level_replica argument.

The choice between the main and replica process settings is made according to the return value of should_log.

get_warmup_steps

< >

( num_training_steps: int )

Get number of steps used for a linear warmup.

main_process_first

< >

( local = Truedesc = 'work' )

Parameters

  • local (bool, optional, defaults to True) — if True first means process of rank 0 of each node if False first means process of rank 0 of node rank 0 In multi-node environment with a shared filesystem you most likely will want to use local=False so that only the main process of the first node will do the processing. If however, the filesystem is not shared, then the main process of each node will need to do the processing, which is the default behavior.
  • desc (str, optional, defaults to "work") — a work description to be used in debug logs

A context manager for torch distributed environment where on needs to do something on the main process, while blocking replicas, and when it’s finished releasing the replicas.

One such use is for datasets’s map feature which to be efficient should be run once on the main process, which upon completion saves a cached version of results and which then automatically gets loaded by the replicas.

set_dataloader

< >

( train_batch_size: int = 8eval_batch_size: int = 8drop_last: bool = Falsenum_workers: int = 0pin_memory: bool = Truepersistent_workers: bool = Falseauto_find_batch_size: bool = Falseignore_data_skip: bool = Falsesampler_seed: Optional = None )

Parameters

  • drop_last (bool, optional, defaults to False) — Whether to drop the last incomplete batch (if the length of the dataset is not divisible by the batch size) or not.
  • num_workers (int, optional, defaults to 0) — Number of subprocesses to use for data loading (PyTorch only). 0 means that the data will be loaded in the main process.
  • pin_memory (bool, optional, defaults to True) — Whether you want to pin memory in data loaders or not. Will default to True.
  • persistent_workers (bool, optional, defaults to False) — If True, the data loader will not shut down the worker processes after a dataset has been consumed once. This allows to maintain the workers Dataset instances alive. Can potentially speed up training, but will increase RAM usage. Will default to False.
  • auto_find_batch_size (bool, optional, defaults to False) — Whether to find a batch size that will fit into memory automatically through exponential decay, avoiding CUDA Out-of-Memory errors. Requires accelerate to be installed (pip install accelerate)
  • ignore_data_skip (bool, optional, defaults to False) — When resuming training, whether or not to skip the epochs and batches to get the data loading at the same stage as in the previous training. If set to True, the training will begin faster (as that skipping step can take a long time) but will not yield the same results as the interrupted training would have.
  • sampler_seed (int, optional) — Random seed to be used with data samplers. If not set, random generators for data sampling will use the same seed as self.seed. This can be used to ensure reproducibility of data sampling, independent of the model seed.

A method that regroups all arguments linked to the dataloaders creation.

Example:

>>> from transformers import TrainingArguments

>>> args = TrainingArguments("working_dir")
>>> args = args.set_dataloader(train_batch_size=16, eval_batch_size=64)
>>> args.per_device_train_batch_size
16

set_evaluate

< >

( strategy: Union = 'no'steps: int = 500batch_size: int = 8accumulation_steps: Optional = Nonedelay: Optional = Noneloss_only: bool = Falsejit_mode: bool = False )

Parameters

  • strategy (str or IntervalStrategy, optional, defaults to "no") — The evaluation strategy to adopt during training. Possible values are:

    • "no": No evaluation is done during training.
    • "steps": Evaluation is done (and logged) every steps.
    • "epoch": Evaluation is done at the end of each epoch.

    Setting a strategy different from "no" will set self.do_eval to True.

  • steps (int, optional, defaults to 500) — Number of update steps between two evaluations if strategy="steps".
  • batch_size (int optional, defaults to 8) — The batch size per device (GPU/TPU core/CPU
) used for evaluation.
  • accumulation_steps (int, optional) — Number of predictions steps to accumulate the output tensors for, before moving the results to the CPU. If left unset, the whole predictions are accumulated on GPU/TPU before being moved to the CPU (faster but requires more memory).
  • delay (float, optional) — Number of epochs or steps to wait for before the first evaluation can be performed, depending on the evaluation_strategy.
  • loss_only (bool, optional, defaults to False) — Ignores all outputs except the loss.
  • jit_mode (bool, optional) — Whether or not to use PyTorch jit trace for inference.

A method that regroups all arguments linked to evaluation.

Example:

>>> from transformers import TrainingArguments

>>> args = TrainingArguments("working_dir")
>>> args = args.set_evaluate(strategy="steps", steps=100)
>>> args.eval_steps
100

set_logging

< >

( strategy: Union = 'steps'steps: int = 500report_to: Union = 'none'level: str = 'passive'first_step: bool = Falsenan_inf_filter: bool = Falseon_each_node: bool = Falsereplica_level: str = 'passive' )

Parameters

  • strategy (str or IntervalStrategy, optional, defaults to "steps") — The logging strategy to adopt during training. Possible values are:

    • "no": No save is done during training.
    • "epoch": Save is done at the end of each epoch.
    • "steps": Save is done every save_steps.
  • steps (int, optional, defaults to 500) — Number of update steps between two logs if strategy="steps".
  • level (str, optional, defaults to "passive") — Logger log level to use on the main process. Possible choices are the log levels as strings: "debug", "info", "warning", "error" and "critical", plus a "passive" level which doesn’t set anything and lets the application set the level.
  • report_to (str or List[str], optional, defaults to "all") — The list of integrations to report the results and logs to. Supported platforms are "azure_ml", "clearml", "codecarbon", "comet_ml", "dagshub", "dvclive", "flyte", "mlflow", "neptune", "tensorboard", and "wandb". Use "all" to report to all integrations installed, "none" for no integrations.
  • first_step (bool, optional, defaults to False) — Whether to log and evaluate the first global_step or not.
  • nan_inf_filter (bool, optional, defaults to True) — Whether to filter nan and inf losses for logging. If set to True the loss of every step that is nan or inf is filtered and the average loss of the current logging window is taken instead.

    nan_inf_filter only influences the logging of loss values, it does not change the behavior the gradient is computed or applied to the model.

  • on_each_node (bool, optional, defaults to True) — In multinode distributed training, whether to log using log_level once per node, or only on the main node.
  • replica_level (str, optional, defaults to "passive") — Logger log level to use on replicas. Same choices as log_level

A method that regroups all arguments linked to logging.

Example:

>>> from transformers import TrainingArguments

>>> args = TrainingArguments("working_dir")
>>> args = args.set_logging(strategy="steps", steps=100)
>>> args.logging_steps
100

set_lr_scheduler

< >

( name: Union = 'linear'num_epochs: float = 3.0max_steps: int = -1warmup_ratio: float = 0warmup_steps: int = 0 )

Parameters

  • name (str or SchedulerType, optional, defaults to "linear") — The scheduler type to use. See the documentation of SchedulerType for all possible values.
  • num_epochs(float, optional, defaults to 3.0) — Total number of training epochs to perform (if not an integer, will perform the decimal part percents of the last epoch before stopping training).
  • max_steps (int, optional, defaults to -1) — If set to a positive number, the total number of training steps to perform. Overrides num_train_epochs. For a finite dataset, training is reiterated through the dataset (if all data is exhausted) until max_steps is reached.
  • warmup_ratio (float, optional, defaults to 0.0) — Ratio of total training steps used for a linear warmup from 0 to learning_rate.
  • warmup_steps (int, optional, defaults to 0) — Number of steps used for a linear warmup from 0 to learning_rate. Overrides any effect of warmup_ratio.

A method that regroups all arguments linked to the learning rate scheduler and its hyperparameters.

Example:

>>> from transformers import TrainingArguments

>>> args = TrainingArguments("working_dir")
>>> args = args.set_lr_scheduler(name="cosine", warmup_ratio=0.05)
>>> args.warmup_ratio
0.05

set_optimizer

< >

( name: Union = 'adamw_torch'learning_rate: float = 5e-05weight_decay: float = 0beta1: float = 0.9beta2: float = 0.999epsilon: float = 1e-08args: Optional = None )

Parameters

  • name (str or training_args.OptimizerNames, optional, defaults to "adamw_torch") — The optimizer to use: "adamw_hf", "adamw_torch", "adamw_torch_fused", "adamw_apex_fused", "adamw_anyprecision" or "adafactor".
  • learning_rate (float, optional, defaults to 5e-5) — The initial learning rate.
  • weight_decay (float, optional, defaults to 0) — The weight decay to apply (if not zero) to all layers except all bias and LayerNorm weights.
  • beta1 (float, optional, defaults to 0.9) — The beta1 hyperparameter for the adam optimizer or its variants.
  • beta2 (float, optional, defaults to 0.999) — The beta2 hyperparameter for the adam optimizer or its variants.
  • epsilon (float, optional, defaults to 1e-8) — The epsilon hyperparameter for the adam optimizer or its variants.
  • args (str, optional) — Optional arguments that are supplied to AnyPrecisionAdamW (only useful when optim="adamw_anyprecision").

A method that regroups all arguments linked to the optimizer and its hyperparameters.

Example:

>>> from transformers import TrainingArguments

>>> args = TrainingArguments("working_dir")
>>> args = args.set_optimizer(name="adamw_torch", beta1=0.8)
>>> args.optim
'adamw_torch'

set_push_to_hub

< >

( model_id: strstrategy: Union = 'every_save'token: Optional = Noneprivate_repo: bool = Falsealways_push: bool = False )

Parameters

  • model_id (str) — The name of the repository to keep in sync with the local output_dir. It can be a simple model ID in which case the model will be pushed in your namespace. Otherwise it should be the whole repository name, for instance "user_name/model", which allows you to push to an organization you are a member of with "organization_name/model".
  • strategy (str or HubStrategy, optional, defaults to "every_save") — Defines the scope of what is pushed to the Hub and when. Possible values are:

    • "end": push the model, its configuration, the tokenizer (if passed along to the Trainer) and a draft of a model card when the save_model() method is called.
    • "every_save": push the model, its configuration, the tokenizer (if passed along to the Trainer) and a draft of a model card each time there is a model save. The pushes are asynchronous to not block training, and in case the save are very frequent, a new push is only attempted if the previous one is finished. A last push is made with the final model at the end of training.
    • "checkpoint": like "every_save" but the latest checkpoint is also pushed in a subfolder named last-checkpoint, allowing you to resume training easily with trainer.train(resume_from_checkpoint="last-checkpoint").
    • "all_checkpoints": like "checkpoint" but all checkpoints are pushed like they appear in the output folder (so you will get one checkpoint folder per folder in your final repository)
  • token (str, optional) — The token to use to push the model to the Hub. Will default to the token in the cache folder obtained with huggingface-cli login.
  • private_repo (bool, optional, defaults to False) — If True, the Hub repo will be set to private.
  • always_push (bool, optional, defaults to False) — Unless this is True, the Trainer will skip pushing a checkpoint when the previous push is not finished.

A method that regroups all arguments linked to synchronizing checkpoints with the Hub.

Calling this method will set self.push_to_hub to True, which means the output_dir will begin a git directory synced with the repo (determined by model_id) and the content will be pushed each time a save is triggered (depending onself.save_strategy). Calling save_model() will also trigger a push.

Example:

>>> from transformers import TrainingArguments

>>> args = TrainingArguments("working_dir")
>>> args = args.set_push_to_hub("me/awesome-model")
>>> args.hub_model_id
'me/awesome-model'

set_save

< >

( strategy: Union = 'steps'steps: int = 500total_limit: Optional = Noneon_each_node: bool = False )

Parameters

  • strategy (str or IntervalStrategy, optional, defaults to "steps") — The checkpoint save strategy to adopt during training. Possible values are:

    • "no": No save is done during training.
    • "epoch": Save is done at the end of each epoch.
    • "steps": Save is done every save_steps.
  • steps (int, optional, defaults to 500) — Number of updates steps before two checkpoint saves if strategy="steps".
  • total_limit (int, optional) — If a value is passed, will limit the total amount of checkpoints. Deletes the older checkpoints in output_dir.
  • on_each_node (bool, optional, defaults to False) — When doing multi-node distributed training, whether to save models and checkpoints on each node, or only on the main one.

    This should not be activated when the different nodes use the same storage as the files will be saved with the same names for each node.

A method that regroups all arguments linked to checkpoint saving.

Example:

>>> from transformers import TrainingArguments

>>> args = TrainingArguments("working_dir")
>>> args = args.set_save(strategy="steps", steps=100)
>>> args.save_steps
100

set_testing

< >

( batch_size: int = 8loss_only: bool = Falsejit_mode: bool = False )

Parameters

  • batch_size (int optional, defaults to 8) — The batch size per device (GPU/TPU core/CPU
) used for testing.
  • loss_only (bool, optional, defaults to False) — Ignores all outputs except the loss.
  • jit_mode (bool, optional) — Whether or not to use PyTorch jit trace for inference.

A method that regroups all basic arguments linked to testing on a held-out dataset.

Calling this method will automatically set self.do_predict to True.

Example:

>>> from transformers import TrainingArguments

>>> args = TrainingArguments("working_dir")
>>> args = args.set_testing(batch_size=32)
>>> args.per_device_eval_batch_size
32

set_training

< >

( learning_rate: float = 5e-05batch_size: int = 8weight_decay: float = 0num_epochs: float = 3max_steps: int = -1gradient_accumulation_steps: int = 1seed: int = 42gradient_checkpointing: bool = False )

Parameters

  • learning_rate (float, optional, defaults to 5e-5) — The initial learning rate for the optimizer.
  • batch_size (int optional, defaults to 8) — The batch size per device (GPU/TPU core/CPU
) used for training.
  • weight_decay (float, optional, defaults to 0) — The weight decay to apply (if not zero) to all layers except all bias and LayerNorm weights in the optimizer.
  • num_train_epochs(float, optional, defaults to 3.0) — Total number of training epochs to perform (if not an integer, will perform the decimal part percents of the last epoch before stopping training).
  • max_steps (int, optional, defaults to -1) — If set to a positive number, the total number of training steps to perform. Overrides num_train_epochs. For a finite dataset, training is reiterated through the dataset (if all data is exhausted) until max_steps is reached.
  • gradient_accumulation_steps (int, optional, defaults to 1) — Number of updates steps to accumulate the gradients for, before performing a backward/update pass.

    When using gradient accumulation, one step is counted as one step with backward pass. Therefore, logging, evaluation, save will be conducted every gradient_accumulation_steps * xxx_step training examples.

  • seed (int, optional, defaults to 42) — Random seed that will be set at the beginning of training. To ensure reproducibility across runs, use the ~Trainer.model_init function to instantiate the model if it has some randomly initialized parameters.
  • gradient_checkpointing (bool, optional, defaults to False) — If True, use gradient checkpointing to save memory at the expense of slower backward pass.

A method that regroups all basic arguments linked to the training.

Calling this method will automatically set self.do_train to True.

Example:

>>> from transformers import TrainingArguments

>>> args = TrainingArguments("working_dir")
>>> args = args.set_training(learning_rate=1e-4, batch_size=32)
>>> args.learning_rate
1e-4

to_dict

< >

( )

Serializes this instance while replace Enum by their values (for JSON serialization support). It obfuscates the token values by removing their value.

to_json_string

< >

( )

Serializes this instance to a JSON string.

to_sanitized_dict

< >

( )

Sanitized serialization to use with TensorBoard’s hparams

Seq2SeqTrainingArguments

class transformers.Seq2SeqTrainingArguments

< >

( output_dir: stroverwrite_output_dir: bool = Falsedo_train: bool = Falsedo_eval: bool = Falsedo_predict: bool = Falseevaluation_strategy: Union = 'no'prediction_loss_only: bool = Falseper_device_train_batch_size: int = 8per_device_eval_batch_size: int = 8per_gpu_train_batch_size: Optional = Noneper_gpu_eval_batch_size: Optional = Nonegradient_accumulation_steps: int = 1eval_accumulation_steps: Optional = Noneeval_delay: Optional = 0learning_rate: float = 5e-05weight_decay: float = 0.0adam_beta1: float = 0.9adam_beta2: float = 0.999adam_epsilon: float = 1e-08max_grad_norm: float = 1.0num_train_epochs: float = 3.0max_steps: int = -1lr_scheduler_type: Union = 'linear'lr_scheduler_kwargs: Optional = <factory>warmup_ratio: float = 0.0warmup_steps: int = 0log_level: Optional = 'passive'log_level_replica: Optional = 'warning'log_on_each_node: bool = Truelogging_dir: Optional = Nonelogging_strategy: Union = 'steps'logging_first_step: bool = Falselogging_steps: float = 500logging_nan_inf_filter: bool = Truesave_strategy: Union = 'steps'save_steps: float = 500save_total_limit: Optional = Nonesave_safetensors: Optional = Truesave_on_each_node: bool = Falsesave_only_model: bool = Falseno_cuda: bool = Falseuse_cpu: bool = Falseuse_mps_device: bool = Falseseed: int = 42data_seed: Optional = Nonejit_mode_eval: bool = Falseuse_ipex: bool = Falsebf16: bool = Falsefp16: bool = Falsefp16_opt_level: str = 'O1'half_precision_backend: str = 'auto'bf16_full_eval: bool = Falsefp16_full_eval: bool = Falsetf32: Optional = Nonelocal_rank: int = -1ddp_backend: Optional = Nonetpu_num_cores: Optional = Nonetpu_metrics_debug: bool = Falsedebug: Union = ''dataloader_drop_last: bool = Falseeval_steps: Optional = Nonedataloader_num_workers: int = 0past_index: int = -1run_name: Optional = Nonedisable_tqdm: Optional = Noneremove_unused_columns: Optional = Truelabel_names: Optional = Noneload_best_model_at_end: Optional = Falsemetric_for_best_model: Optional = Nonegreater_is_better: Optional = Noneignore_data_skip: bool = Falsefsdp: Union = ''fsdp_min_num_params: int = 0fsdp_config: Optional = Nonefsdp_transformer_layer_cls_to_wrap: Optional = Nonedeepspeed: Optional = Nonelabel_smoothing_factor: float = 0.0optim: Union = 'adamw_torch'optim_args: Optional = Noneadafactor: bool = Falsegroup_by_length: bool = Falselength_column_name: Optional = 'length'report_to: Optional = Noneddp_find_unused_parameters: Optional = Noneddp_bucket_cap_mb: Optional = Noneddp_broadcast_buffers: Optional = Nonedataloader_pin_memory: bool = Truedataloader_persistent_workers: bool = Falseskip_memory_metrics: bool = Trueuse_legacy_prediction_loop: bool = Falsepush_to_hub: bool = Falseresume_from_checkpoint: Optional = Nonehub_model_id: Optional = Nonehub_strategy: Union = 'every_save'hub_token: Optional = Nonehub_private_repo: bool = Falsehub_always_push: bool = Falsegradient_checkpointing: bool = Falsegradient_checkpointing_kwargs: Optional = Noneinclude_inputs_for_metrics: bool = Falsefp16_backend: str = 'auto'push_to_hub_model_id: Optional = Nonepush_to_hub_organization: Optional = Nonepush_to_hub_token: Optional = Nonemp_parameters: str = ''auto_find_batch_size: bool = Falsefull_determinism: bool = Falsetorchdynamo: Optional = Noneray_scope: Optional = 'last'ddp_timeout: Optional = 1800torch_compile: bool = Falsetorch_compile_backend: Optional = Nonetorch_compile_mode: Optional = Nonedispatch_batches: Optional = Nonesplit_batches: Optional = Falseinclude_tokens_per_second: Optional = Falseinclude_num_input_tokens_seen: Optional = Falseneftune_noise_alpha: float = Nonesortish_sampler: bool = Falsepredict_with_generate: bool = Falsegeneration_max_length: Optional = Nonegeneration_num_beams: Optional = Nonegeneration_config: Union = None )

Parameters

  • output_dir (str) — The output directory where the model predictions and checkpoints will be written.
  • overwrite_output_dir (bool, optional, defaults to False) — If True, overwrite the content of the output directory. Use this to continue training if output_dir points to a checkpoint directory.
  • do_train (bool, optional, defaults to False) — Whether to run training or not. This argument is not directly used by Trainer, it’s intended to be used by your training/evaluation scripts instead. See the example scripts for more details.
  • do_eval (bool, optional) — Whether to run evaluation on the validation set or not. Will be set to True if evaluation_strategy is different from "no". This argument is not directly used by Trainer, it’s intended to be used by your training/evaluation scripts instead. See the example scripts for more details.
  • do_predict (bool, optional, defaults to False) — Whether to run predictions on the test set or not. This argument is not directly used by Trainer, it’s intended to be used by your training/evaluation scripts instead. See the example scripts for more details.
  • evaluation_strategy (str or IntervalStrategy, optional, defaults to "no") — The evaluation strategy to adopt during training. Possible values are:

    • "no": No evaluation is done during training.
    • "steps": Evaluation is done (and logged) every eval_steps.
    • "epoch": Evaluation is done at the end of each epoch.
  • prediction_loss_only (bool, optional, defaults to False) — When performing evaluation and generating predictions, only returns the loss.
  • per_device_train_batch_size (int, optional, defaults to 8) — The batch size per GPU/XPU/TPU/MPS/NPU core/CPU for training.
  • per_device_eval_batch_size (int, optional, defaults to 8) — The batch size per GPU/XPU/TPU/MPS/NPU core/CPU for evaluation.
  • gradient_accumulation_steps (int, optional, defaults to 1) — Number of updates steps to accumulate the gradients for, before performing a backward/update pass.

    When using gradient accumulation, one step is counted as one step with backward pass. Therefore, logging, evaluation, save will be conducted every gradient_accumulation_steps * xxx_step training examples.

  • eval_accumulation_steps (int, optional) — Number of predictions steps to accumulate the output tensors for, before moving the results to the CPU. If left unset, the whole predictions are accumulated on GPU/NPU/TPU before being moved to the CPU (faster but requires more memory).
  • eval_delay (float, optional) — Number of epochs or steps to wait for before the first evaluation can be performed, depending on the evaluation_strategy.
  • learning_rate (float, optional, defaults to 5e-5) — The initial learning rate for AdamW optimizer.
  • weight_decay (float, optional, defaults to 0) — The weight decay to apply (if not zero) to all layers except all bias and LayerNorm weights in AdamW optimizer.
  • adam_beta1 (float, optional, defaults to 0.9) — The beta1 hyperparameter for the AdamW optimizer.
  • adam_beta2 (float, optional, defaults to 0.999) — The beta2 hyperparameter for the AdamW optimizer.
  • adam_epsilon (float, optional, defaults to 1e-8) — The epsilon hyperparameter for the AdamW optimizer.
  • max_grad_norm (float, optional, defaults to 1.0) — Maximum gradient norm (for gradient clipping).
  • num_train_epochs(float, optional, defaults to 3.0) — Total number of training epochs to perform (if not an integer, will perform the decimal part percents of the last epoch before stopping training).
  • max_steps (int, optional, defaults to -1) — If set to a positive number, the total number of training steps to perform. Overrides num_train_epochs. For a finite dataset, training is reiterated through the dataset (if all data is exhausted) until max_steps is reached.
  • lr_scheduler_type (str or SchedulerType, optional, defaults to "linear") — The scheduler type to use. See the documentation of SchedulerType for all possible values.
  • lr_scheduler_kwargs (‘dict’, optional, defaults to {}) — The extra arguments for the lr_scheduler. See the documentation of each scheduler for possible values.
  • warmup_ratio (float, optional, defaults to 0.0) — Ratio of total training steps used for a linear warmup from 0 to learning_rate.
  • warmup_steps (int, optional, defaults to 0) — Number of steps used for a linear warmup from 0 to learning_rate. Overrides any effect of warmup_ratio.
  • log_level (str, optional, defaults to passive) — Logger log level to use on the main process. Possible choices are the log levels as strings: ‘debug’, ‘info’, ‘warning’, ‘error’ and ‘critical’, plus a ‘passive’ level which doesn’t set anything and keeps the current log level for the Transformers library (which will be "warning" by default).
  • log_level_replica (str, optional, defaults to "warning") — Logger log level to use on replicas. Same choices as log_level”
  • log_on_each_node (bool, optional, defaults to True) — In multinode distributed training, whether to log using log_level once per node, or only on the main node.
  • logging_dir (str, optional) — TensorBoard log directory. Will default to *output_dir/runs/CURRENT_DATETIME_HOSTNAME*.
  • logging_strategy (str or IntervalStrategy, optional, defaults to "steps") — The logging strategy to adopt during training. Possible values are:

    • "no": No logging is done during training.
    • "epoch": Logging is done at the end of each epoch.
    • "steps": Logging is done every logging_steps.
  • logging_first_step (bool, optional, defaults to False) — Whether to log and evaluate the first global_step or not.
  • logging_steps (int or float, optional, defaults to 500) — Number of update steps between two logs if logging_strategy="steps". Should be an integer or a float in range [0,1). If smaller than 1, will be interpreted as ratio of total training steps.
  • logging_nan_inf_filter (bool, optional, defaults to True) — Whether to filter nan and inf losses for logging. If set to True the loss of every step that is nan or inf is filtered and the average loss of the current logging window is taken instead.

    logging_nan_inf_filter only influences the logging of loss values, it does not change the behavior the gradient is computed or applied to the model.

  • save_strategy (str or IntervalStrategy, optional, defaults to "steps") — The checkpoint save strategy to adopt during training. Possible values are:

    • "no": No save is done during training.
    • "epoch": Save is done at the end of each epoch.
    • "steps": Save is done every save_steps.
  • save_steps (int or float, optional, defaults to 500) — Number of updates steps before two checkpoint saves if save_strategy="steps". Should be an integer or a float in range [0,1). If smaller than 1, will be interpreted as ratio of total training steps.
  • save_total_limit (int, optional) — If a value is passed, will limit the total amount of checkpoints. Deletes the older checkpoints in output_dir. When load_best_model_at_end is enabled, the “best” checkpoint according to metric_for_best_model will always be retained in addition to the most recent ones. For example, for save_total_limit=5 and load_best_model_at_end, the four last checkpoints will always be retained alongside the best model. When save_total_limit=1 and load_best_model_at_end, it is possible that two checkpoints are saved: the last one and the best one (if they are different).
  • save_safetensors (bool, optional, defaults to True) — Use safetensors saving and loading for state dicts instead of default torch.load and torch.save.
  • save_on_each_node (bool, optional, defaults to False) — When doing multi-node distributed training, whether to save models and checkpoints on each node, or only on the main one.

    This should not be activated when the different nodes use the same storage as the files will be saved with the same names for each node.

  • save_only_model (bool, optional, defaults to False) — When checkpointing, whether to only save the model, or also the optimizer, scheduler & rng state. Note that when this is true, you won’t be able to resume training from checkpoint. This enables you to save storage by not storing the optimizer, scheduler & rng state. You can only load the model using from_pretrained with this option set to True.
  • use_cpu (bool, optional, defaults to False) — Whether or not to use cpu. If set to False, we will use cuda or mps device if available.
  • seed (int, optional, defaults to 42) — Random seed that will be set at the beginning of training. To ensure reproducibility across runs, use the ~Trainer.model_init function to instantiate the model if it has some randomly initialized parameters.
  • data_seed (int, optional) — Random seed to be used with data samplers. If not set, random generators for data sampling will use the same seed as seed. This can be used to ensure reproducibility of data sampling, independent of the model seed.
  • jit_mode_eval (bool, optional, defaults to False) — Whether or not to use PyTorch jit trace for inference.
  • use_ipex (bool, optional, defaults to False) — Use Intel extension for PyTorch when it is available. IPEX installation.
  • bf16 (bool, optional, defaults to False) — Whether to use bf16 16-bit (mixed) precision training instead of 32-bit training. Requires Ampere or higher NVIDIA architecture or using CPU (use_cpu) or Ascend NPU. This is an experimental API and it may change.
  • fp16 (bool, optional, defaults to False) — Whether to use fp16 16-bit (mixed) precision training instead of 32-bit training.
  • fp16_opt_level (str, optional, defaults to ‘O1’) — For fp16 training, Apex AMP optimization level selected in [‘O0’, ‘O1’, ‘O2’, and ‘O3’]. See details on the Apex documentation.
  • fp16_backend (str, optional, defaults to "auto") — This argument is deprecated. Use half_precision_backend instead.
  • half_precision_backend (str, optional, defaults to "auto") — The backend to use for mixed precision training. Must be one of "auto", "apex", "cpu_amp". "auto" will use CPU/CUDA AMP or APEX depending on the PyTorch version detected, while the other choices will force the requested backend.
  • bf16_full_eval (bool, optional, defaults to False) — Whether to use full bfloat16 evaluation instead of 32-bit. This will be faster and save memory but can harm metric values. This is an experimental API and it may change.
  • fp16_full_eval (bool, optional, defaults to False) — Whether to use full float16 evaluation instead of 32-bit. This will be faster and save memory but can harm metric values.
  • tf32 (bool, optional) — Whether to enable the TF32 mode, available in Ampere and newer GPU architectures. The default value depends on PyTorch’s version default of torch.backends.cuda.matmul.allow_tf32. For more details please refer to the TF32 documentation. This is an experimental API and it may change.
  • local_rank (int, optional, defaults to -1) — Rank of the process during distributed training.
  • ddp_backend (str, optional) — The backend to use for distributed training. Must be one of "nccl", "mpi", "ccl", "gloo", "hccl".
  • tpu_num_cores (int, optional) — When training on TPU, the number of TPU cores (automatically passed by launcher script).
  • dataloader_drop_last (bool, optional, defaults to False) — Whether to drop the last incomplete batch (if the length of the dataset is not divisible by the batch size) or not.
  • eval_steps (int or float, optional) — Number of update steps between two evaluations if evaluation_strategy="steps". Will default to the same value as logging_steps if not set. Should be an integer or a float in range [0,1). If smaller than 1, will be interpreted as ratio of total training steps.
  • dataloader_num_workers (int, optional, defaults to 0) — Number of subprocesses to use for data loading (PyTorch only). 0 means that the data will be loaded in the main process.
  • past_index (int, optional, defaults to -1) — Some models like TransformerXL or XLNet can make use of the past hidden states for their predictions. If this argument is set to a positive int, the Trainer will use the corresponding output (usually index 2) as the past state and feed it to the model at the next training step under the keyword argument mems.
  • run_name (str, optional) — A descriptor for the run. Typically used for wandb and mlflow logging.
  • disable_tqdm (bool, optional) — Whether or not to disable the tqdm progress bars and table of metrics produced by ~notebook.NotebookTrainingTracker in Jupyter Notebooks. Will default to True if the logging level is set to warn or lower (default), False otherwise.
  • remove_unused_columns (bool, optional, defaults to True) — Whether or not to automatically remove the columns unused by the model forward method.
  • label_names (List[str], optional) — The list of keys in your dictionary of inputs that correspond to the labels.

    Will eventually default to the list of argument names accepted by the model that contain the word “label”, except if the model used is one of the XxxForQuestionAnswering in which case it will also include the ["start_positions", "end_positions"] keys.

  • load_best_model_at_end (bool, optional, defaults to False) — Whether or not to load the best model found during training at the end of training. When this option is enabled, the best checkpoint will always be saved. See save_total_limit for more.

    When set to True, the parameters save_strategy needs to be the same as evaluation_strategy, and in the case it is “steps”, save_steps must be a round multiple of eval_steps.

  • metric_for_best_model (str, optional) — Use in conjunction with load_best_model_at_end to specify the metric to use to compare two different models. Must be the name of a metric returned by the evaluation with or without the prefix "eval_". Will default to "loss" if unspecified and load_best_model_at_end=True (to use the evaluation loss).

    If you set this value, greater_is_better will default to True. Don’t forget to set it to False if your metric is better when lower.

  • greater_is_better (bool, optional) — Use in conjunction with load_best_model_at_end and metric_for_best_model to specify if better models should have a greater metric or not. Will default to:

    • True if metric_for_best_model is set to a value that isn’t "loss" or "eval_loss".
    • False if metric_for_best_model is not set, or set to "loss" or "eval_loss".
  • ignore_data_skip (bool, optional, defaults to False) — When resuming training, whether or not to skip the epochs and batches to get the data loading at the same stage as in the previous training. If set to True, the training will begin faster (as that skipping step can take a long time) but will not yield the same results as the interrupted training would have.
  • fsdp (bool, str or list of FSDPOption, optional, defaults to '') — Use PyTorch Distributed Parallel Training (in distributed training only).

    A list of options along the following:

    • "full_shard": Shard parameters, gradients and optimizer states.
    • "shard_grad_op": Shard optimizer states and gradients.
    • "hybrid_shard": Apply FULL_SHARD within a node, and replicate parameters across nodes.
    • "hybrid_shard_zero2": Apply SHARD_GRAD_OP within a node, and replicate parameters across nodes.
    • "offload": Offload parameters and gradients to CPUs (only compatible with "full_shard" and "shard_grad_op").
    • "auto_wrap": Automatically recursively wrap layers with FSDP using default_auto_wrap_policy.
  • fsdp_config (str or dict, optional) — Config to be used with fsdp (Pytorch Distributed Parallel Training). The value is either a location of fsdp json config file (e.g., fsdp_config.json) or an already loaded json file as dict.

    A List of config and its options:

    • min_num_params (int, optional, defaults to 0): FSDP’s minimum number of parameters for Default Auto Wrapping. (useful only when fsdp field is passed).

    • transformer_layer_cls_to_wrap (List[str], optional): List of transformer layer class names (case-sensitive) to wrap, e.g, BertLayer, GPTJBlock, T5Block 
 (useful only when fsdp flag is passed).

    • backward_prefetch (str, optional) FSDP’s backward prefetch mode. Controls when to prefetch next set of parameters (useful only when fsdp field is passed).

      A list of options along the following:

      • "backward_pre" : Prefetches the next set of parameters before the current set of parameter’s gradient computation.
      • "backward_post" : This prefetches the next set of parameters after the current set of parameter’s gradient computation.
    • forward_prefetch (bool, optional, defaults to False) FSDP’s forward prefetch mode (useful only when fsdp field is passed). If "True", then FSDP explicitly prefetches the next upcoming all-gather while executing in the forward pass.

    • limit_all_gathers (bool, optional, defaults to False) FSDP’s limit_all_gathers (useful only when fsdp field is passed). If "True", FSDP explicitly synchronizes the CPU thread to prevent too many in-flight all-gathers.

    • use_orig_params (bool, optional, defaults to True) If "True", allows non-uniform requires_grad during init, which means support for interspersed frozen and trainable paramteres. Useful in cases such as parameter-efficient fine-tuning. Please refer this [blog](https://dev-discuss.pytorch.org/t/rethinking-pytorch-fully-sharded-data-parallel-fsdp-from-first-principles/1019

    • sync_module_states (bool, optional, defaults to True) If "True", each individually wrapped FSDP unit will broadcast module parameters from rank 0 to ensure they are the same across all ranks after initialization

    • activation_checkpointing (bool, optional, defaults to False): If "True", activation checkpointing is a technique to reduce memory usage by clearing activations of certain layers and recomputing them during a backward pass. Effectively, this trades extra computation time for reduced memory usage.

    • xla (bool, optional, defaults to False): Whether to use PyTorch/XLA Fully Sharded Data Parallel Training. This is an experimental feature and its API may evolve in the future.

    • xla_fsdp_settings (dict, optional) The value is a dictionary which stores the XLA FSDP wrapping parameters.

      For a complete list of options, please see here.

    • xla_fsdp_grad_ckpt (bool, optional, defaults to False): Will use gradient checkpointing over each nested XLA FSDP wrapped layer. This setting can only be used when the xla flag is set to true, and an auto wrapping policy is specified through fsdp_min_num_params or fsdp_transformer_layer_cls_to_wrap.

  • deepspeed (str or dict, optional) — Use Deepspeed. This is an experimental feature and its API may evolve in the future. The value is either the location of DeepSpeed json config file (e.g., ds_config.json) or an already loaded json file as a dict”
  • label_smoothing_factor (float, optional, defaults to 0.0) — The label smoothing factor to use. Zero means no label smoothing, otherwise the underlying onehot-encoded labels are changed from 0s and 1s to label_smoothing_factor/num_labels and 1 - label_smoothing_factor + label_smoothing_factor/num_labels respectively.
  • debug (str or list of DebugOption, optional, defaults to "") — Enable one or more debug features. This is an experimental feature.

    Possible options are:

    • "underflow_overflow": detects overflow in model’s input/outputs and reports the last frames that led to the event
    • "tpu_metrics_debug": print debug metrics on TPU

    The options should be separated by whitespaces.

  • optim (str or training_args.OptimizerNames, optional, defaults to "adamw_torch") — The optimizer to use: adamw_hf, adamw_torch, adamw_torch_fused, adamw_apex_fused, adamw_anyprecision or adafactor.
  • optim_args (str, optional) — Optional arguments that are supplied to AnyPrecisionAdamW.
  • group_by_length (bool, optional, defaults to False) — Whether or not to group together samples of roughly the same length in the training dataset (to minimize padding applied and be more efficient). Only useful if applying dynamic padding.
  • length_column_name (str, optional, defaults to "length") — Column name for precomputed lengths. If the column exists, grouping by length will use these values rather than computing them on train startup. Ignored unless group_by_length is True and the dataset is an instance of Dataset.
  • report_to (str or List[str], optional, defaults to "all") — The list of integrations to report the results and logs to. Supported platforms are "azure_ml", "clearml", "codecarbon", "comet_ml", "dagshub", "dvclive", "flyte", "mlflow", "neptune", "tensorboard", and "wandb". Use "all" to report to all integrations installed, "none" for no integrations.
  • ddp_find_unused_parameters (bool, optional) — When using distributed training, the value of the flag find_unused_parameters passed to DistributedDataParallel. Will default to False if gradient checkpointing is used, True otherwise.
  • ddp_bucket_cap_mb (int, optional) — When using distributed training, the value of the flag bucket_cap_mb passed to DistributedDataParallel.
  • ddp_broadcast_buffers (bool, optional) — When using distributed training, the value of the flag broadcast_buffers passed to DistributedDataParallel. Will default to False if gradient checkpointing is used, True otherwise.
  • dataloader_pin_memory (bool, optional, defaults to True) — Whether you want to pin memory in data loaders or not. Will default to True.
  • dataloader_persistent_workers (bool, optional, defaults to False) — If True, the data loader will not shut down the worker processes after a dataset has been consumed once. This allows to maintain the workers Dataset instances alive. Can potentially speed up training, but will increase RAM usage. Will default to False.
  • skip_memory_metrics (bool, optional, defaults to True) — Whether to skip adding of memory profiler reports to metrics. This is skipped by default because it slows down the training and evaluation speed.
  • push_to_hub (bool, optional, defaults to False) — Whether or not to push the model to the Hub every time the model is saved. If this is activated, output_dir will begin a git directory synced with the repo (determined by hub_model_id) and the content will be pushed each time a save is triggered (depending on your save_strategy). Calling save_model() will also trigger a push.

    If output_dir exists, it needs to be a local clone of the repository to which the Trainer will be pushed.

  • resume_from_checkpoint (str, optional) — The path to a folder with a valid checkpoint for your model. This argument is not directly used by Trainer, it’s intended to be used by your training/evaluation scripts instead. See the example scripts for more details.
  • hub_model_id (str, optional) — The name of the repository to keep in sync with the local output_dir. It can be a simple model ID in which case the model will be pushed in your namespace. Otherwise it should be the whole repository name, for instance "user_name/model", which allows you to push to an organization you are a member of with "organization_name/model". Will default to user_name/output_dir_name with output_dir_name being the name of output_dir.

    Will default to the name of output_dir.

  • hub_strategy (str or HubStrategy, optional, defaults to "every_save") — Defines the scope of what is pushed to the Hub and when. Possible values are:

    • "end": push the model, its configuration, the tokenizer (if passed along to the Trainer) and a draft of a model card when the save_model() method is called.
    • "every_save": push the model, its configuration, the tokenizer (if passed along to the Trainer) and a draft of a model card each time there is a model save. The pushes are asynchronous to not block training, and in case the save are very frequent, a new push is only attempted if the previous one is finished. A last push is made with the final model at the end of training.
    • "checkpoint": like "every_save" but the latest checkpoint is also pushed in a subfolder named last-checkpoint, allowing you to resume training easily with trainer.train(resume_from_checkpoint="last-checkpoint").
    • "all_checkpoints": like "checkpoint" but all checkpoints are pushed like they appear in the output folder (so you will get one checkpoint folder per folder in your final repository)
  • hub_token (str, optional) — The token to use to push the model to the Hub. Will default to the token in the cache folder obtained with huggingface-cli login.
  • hub_private_repo (bool, optional, defaults to False) — If True, the Hub repo will be set to private.
  • hub_always_push (bool, optional, defaults to False) — Unless this is True, the Trainer will skip pushing a checkpoint when the previous push is not finished.
  • gradient_checkpointing (bool, optional, defaults to False) — If True, use gradient checkpointing to save memory at the expense of slower backward pass.
  • gradient_checkpointing_kwargs (dict, optional, defaults to None) — Key word arguments to be passed to the gradient_checkpointing_enable method.
  • include_inputs_for_metrics (bool, optional, defaults to False) — Whether or not the inputs will be passed to the compute_metrics function. This is intended for metrics that need inputs, predictions and references for scoring calculation in Metric class.
  • auto_find_batch_size (bool, optional, defaults to False) — Whether to find a batch size that will fit into memory automatically through exponential decay, avoiding CUDA Out-of-Memory errors. Requires accelerate to be installed (pip install accelerate)
  • full_determinism (bool, optional, defaults to False) — If True, enable_full_determinism() is called instead of set_seed() to ensure reproducible results in distributed training. Important: this will negatively impact the performance, so only use it for debugging.
  • torchdynamo (str, optional) — If set, the backend compiler for TorchDynamo. Possible choices are "eager", "aot_eager", "inductor", "nvfuser", "aot_nvfuser", "aot_cudagraphs", "ofi", "fx2trt", "onnxrt" and "ipex".
  • ray_scope (str, optional, defaults to "last") — The scope to use when doing hyperparameter search with Ray. By default, "last" will be used. Ray will then use the last checkpoint of all trials, compare those, and select the best one. However, other options are also available. See the Ray documentation for more options.
  • ddp_timeout (int, optional, defaults to 1800) — The timeout for torch.distributed.init_process_group calls, used to avoid GPU socket timeouts when performing slow operations in distributed runnings. Please refer the [PyTorch documentation] (https://pytorch.org/docs/stable/distributed.html#torch.distributed.init_process_group) for more information.
  • use_mps_device (bool, optional, defaults to False) — This argument is deprecated.mps device will be used if it is available similar to cuda device.
  • torch_compile (bool, optional, defaults to False) — Whether or not to compile the model using PyTorch 2.0 torch.compile.

    This will use the best defaults for the torch.compile API. You can customize the defaults with the argument torch_compile_backend and torch_compile_mode but we don’t guarantee any of them will work as the support is progressively rolled in in PyTorch.

    This flag and the whole compile API is experimental and subject to change in future releases.

  • torch_compile_backend (str, optional) — The backend to use in torch.compile. If set to any value, torch_compile will be set to True.

    Refer to the PyTorch doc for possible values and note that they may change across PyTorch versions.

    This flag is experimental and subject to change in future releases.

  • torch_compile_mode (str, optional) — The mode to use in torch.compile. If set to any value, torch_compile will be set to True.

    Refer to the PyTorch doc for possible values and note that they may change across PyTorch versions.

    This flag is experimental and subject to change in future releases.

  • split_batches (bool, optional) — Whether or not the accelerator should split the batches yielded by the dataloaders across the devices during distributed training. If

    set to True, the actual batch size used will be the same on any kind of distributed processes, but it must be a

    round multiple of the number of processes you are using (such as GPUs).

  • include_tokens_per_second (bool, optional) — Whether or not to compute the number of tokens per second per device for training speed metrics.

    This will iterate over the entire training dataloader once beforehand,

    and will slow down the entire process.

  • include_num_input_tokens_seen (bool, optional) — Whether or not to track the number of input tokens seen throughout training.

    May be slower in distributed training as gather operations must be called.

  • neftune_noise_alpha (Optional[float]) — If not None, this will activate NEFTune noise embeddings. This can drastically improve model performance for instruction fine-tuning. Check out the original paper and the original code. Support transformers PreTrainedModel and also PeftModel from peft.
  • sortish_sampler (bool, optional, defaults to False) — Whether to use a sortish sampler or not. Only possible if the underlying datasets are Seq2SeqDataset for now but will become generally available in the near future.

    It sorts the inputs according to lengths in order to minimize the padding size, with a bit of randomness for the training set.

  • predict_with_generate (bool, optional, defaults to False) — Whether to use generate to calculate generative metrics (ROUGE, BLEU).
  • generation_max_length (int, optional) — The max_length to use on each evaluation loop when predict_with_generate=True. Will default to the max_length value of the model configuration.
  • generation_num_beams (int, optional) — The num_beams to use on each evaluation loop when predict_with_generate=True. Will default to the num_beams value of the model configuration.
  • generation_config (str or Path or GenerationConfig, optional) — Allows to load a GenerationConfig from the from_pretrained method. This can be either:

    • a string, the model id of a pretrained model configuration hosted inside a model repo on huggingface.co. Valid model ids can be located at the root-level, like bert-base-uncased, or namespaced under a user or organization name, like dbmdz/bert-base-german-cased.
    • a path to a directory containing a configuration file saved using the save_pretrained() method, e.g., ./my_model_directory/.
    • a GenerationConfig object.

TrainingArguments is the subset of the arguments we use in our example scripts which relate to the training loop itself.

Using HfArgumentParser we can turn this class into argparse arguments that can be specified on the command line.

to_dict

< >

( )

Serializes this instance while replace Enum by their values and GenerationConfig by dictionaries (for JSON serialization support). It obfuscates the token values by removing their value.

Checkpoints

デフォルトでは、Trainer はすべおのチェックポむントを、 TrainingArguments を䜿甚しおいたす。これらは、xxx を含むcheckpoint-xxxずいう名前のサブフォルダヌに保存されたす。 それはトレヌニングの段階でした。

チェックポむントからトレヌニングを再開するには、次のいずれかを䜿甚しお Trainer.train() を呌び出したす。

  • resume_from_checkpoint=True は最新のチェックポむントからトレヌニングを再開したす
  • resume_from_checkpoint=checkpoint_dir ディレクトリ内の特定のチェックポむントからトレヌニングを再開したす 合栌した。

さらに、push_to_hub=True を䜿甚するず、モデル ハブにチェックポむントを簡単に保存できたす。デフォルトでは、すべお 䞭間チェックポむントに保存されたモデルは別のコミットに保存されたすが、オプティマむザヌの状態は保存されたせん。適応できたす TrainingArguments の hub-strategy 倀を次のいずれかにしたす。

  • "checkpoint": 最新のチェックポむントも last-checkpoint ずいう名前のサブフォルダヌにプッシュされたす。 trainer.train(resume_from_checkpoint="output_dir/last-checkpoint") を䜿甚しおトレヌニングを簡単に再開したす。
  • "all_checkpoints": すべおのチェックポむントは、出力フォルダヌに衚瀺されるようにプッシュされたす (したがっお、1 ぀のチェックポむントが埗られたす) 最終リポゞトリ内のフォルダヌごずのチェックポむント フォルダヌ)

Logging

デフォルトでは、Trainer はメむンプロセスに logging.INFO を䜿甚し、レプリカがある堎合には logging.WARNING を䜿甚したす。

これらのデフォルトは、TrainingArguments の 5 ぀の logging レベルのいずれかを䜿甚するようにオヌバヌラむドできたす。 匕数:

  • log_level - メむンプロセス甚
  • log_level_replica - レプリカ甚

さらに、TrainingArguments の log_on_each_node が False に蚭定されおいる堎合、メむン ノヌドのみが メむン プロセスのログ レベル蚭定を䜿甚するず、他のすべおのノヌドはレプリカのログ レベル蚭定を䜿甚したす。

Trainer は、transformers のログ レベルをノヌドごずに個別に蚭定するこずに泚意しおください。 Trainer.__init__()。したがっお、他の機胜を利甚する堎合は、これをより早く蚭定するこずをお勧めしたす (次の䟋を参照)。 Trainer オブゞェクトを䜜成する前の transformers 機胜。

これをアプリケヌションで䜿甚する方法の䟋を次に瀺したす。

[...]
logger = logging.getLogger(__name__)

# Setup logging
logging.basicConfig(
    format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
    datefmt="%m/%d/%Y %H:%M:%S",
    handlers=[logging.StreamHandler(sys.stdout)],
)

# set the main code and the modules it uses to the same log-level according to the node
log_level = training_args.get_process_log_level()
logger.setLevel(log_level)
datasets.utils.logging.set_verbosity(log_level)
transformers.utils.logging.set_verbosity(log_level)

trainer = Trainer(...)

そしお、メむン ノヌドず他のすべおのノヌドで重耇する可胜性が高いものを出力しないように譊告するだけを衚瀺したい堎合は、 è­Šå‘Š: 次のように実行できたす。

my_app.py ... --log_level warning --log_level_replica error

マルチノヌド環境で、各ノヌドのメむンプロセスのログを繰り返したくない堎合は、次のようにしたす。 䞊蚘を次のように倉曎したす。

my_app.py ... --log_level warning --log_level_replica error --log_on_each_node 0

その埌、最初のノヌドのメむン プロセスのみが「譊告」レベルでログに蚘録され、メむン ノヌド䞊の他のすべおのプロセスはログに蚘録されたす。 ノヌドず他のノヌド䞊のすべおのプロセスは「゚ラヌ」レベルでログに蚘録されたす。

アプリケヌションをできるだけ静かにする必芁がある堎合は、次のようにしたす。

my_app.py ... --log_level error --log_level_replica error --log_on_each_node 0

(マルチノヌド環境の堎合は --log_on_each_node 0 を远加したす)

Randomness

Trainer によっお生成されたチェックポむントから再開する堎合、すべおの努力がその状態を埩元するために行われたす。 python、numpy、および pytorch の RNG 状態は、そのチェックポむントを保存した時点ず同じ状態になりたす。 これにより、「停止しお再開」ずいうスタむルのトレヌニングが、ノンストップトレヌニングに可胜な限り近づけられるはずです。

ただし、さたざたなデフォルトの非決定的な pytorch 蚭定により、これは完党に機胜しない可胜性がありたす。フルをご垌望の堎合は 決定論に぀いおは、ランダム性の゜ヌスの制埡 を参照しおください。ドキュメントで説明されおいるように、これらの蚭定の䞀郚は 物事を決定論的にするもの (䟋: torch.backends.cudnn.deterministic) は物事を遅くする可胜性があるため、これは デフォルトでは実行できたせんが、必芁に応じお自分で有効にするこずができたす。

Specific GPUs Selection

どの GPU をどのような順序で䜿甚するかをプログラムに指瀺する方法に぀いお説明したす。

DistributedDataParallel を䜿甚しお GPU のサブセットのみを䜿甚する堎合、䜿甚する GPU の数を指定するだけです。 。たずえば、GPU が 4 ぀あるが、最初の 2 ぀を䜿甚したい堎合は、次のようにしたす。

torchrun --nproc_per_node=2  trainer-program.py ...

accelerate たたは deepspeed がむンストヌルされおいる堎合は、次を䜿甚しお同じこずを達成するこずもできたす。の䞀぀

accelerate launch --num_processes 2 trainer-program.py ...
deepspeed --num_gpus 2 trainer-program.py ...

これらのランチャヌを䜿甚するために、Accelerate たたは Deepspeed 統合 機胜を䜿甚する必芁はありたせん。

これたでは、プログラムに䜿甚する GPU の数を指瀺できたした。次に、特定の GPU を遞択し、その順序を制埡する方法に぀いお説明したす。

次の環境倉数は、䜿甚する GPU ずその順序を制埡するのに圹立ちたす。

CUDA_VISIBLE_DEVICES

耇数の GPU があり、そのうちの 1 ぀たたはいく぀かの GPU だけを䜿甚したい堎合は、環境倉数 CUDA_VISIBLE_DEVICES を䜿甚する GPU のリストに蚭定したす。

たずえば、4 ぀の GPU (0、1、2、3) があるずしたす。物理 GPU 0 ず 2 のみで実行するには、次のようにしたす。

CUDA_VISIBLE_DEVICES=0,2 torchrun trainer-program.py ...

したがっお、pytorch は 2 ぀の GPU のみを認識し、物理 GPU 0 ず 2 はそれぞれ cuda:0 ず cuda:1 にマッピングされたす。

順序を倉曎するこずもできたす。

CUDA_VISIBLE_DEVICES=2,0 torchrun trainer-program.py ...

ここでは、物理 GPU 0 ず 2 がそれぞれcuda:1ずcuda:0にマッピングされおいたす。

䞊蚘の䟋はすべお DistributedDataParallel 䜿甚パタヌンのものですが、同じ方法が DataParallel でも機胜したす。

CUDA_VISIBLE_DEVICES=2,0 python trainer-program.py ...

GPU のない環境を゚ミュレヌトするには、次のようにこの環境倉数を空の倀に蚭定するだけです。

CUDA_VISIBLE_DEVICES= python trainer-program.py ...

他の環境倉数ず同様に、これらをコマンド ラむンに远加する代わりに、次のように゚クスポヌトするこずもできたす。

export CUDA_VISIBLE_DEVICES=0,2
torchrun trainer-program.py ...

ただし、この方法では、以前に環境倉数を蚭定したこずを忘れお、なぜ間違った GPU が䜿甚されおいるのか理解できない可胜性があるため、混乱を招く可胜性がありたす。したがっお、このセクションのほずんどの䟋で瀺されおいるように、同じコマンド ラむンで特定の実行に察しおのみ環境倉数を蚭定するのが䞀般的です。

CUDA_DEVICE_ORDER

物理デバむスの順序を制埡する远加の環境倉数 CUDA_DEVICE_ORDER がありたす。遞択肢は次の 2 ぀です。

  1. PCIe バス ID 順 (nvidia-smi の順序ず䞀臎) - これがデフォルトです。
export CUDA_DEVICE_ORDER=PCI_BUS_ID
  1. GPU コンピュヌティング胜力順に䞊べる
export CUDA_DEVICE_ORDER=FASTEST_FIRST

ほずんどの堎合、この環境倉数を気にする必芁はありたせんが、叀い GPU ず新しい GPU が物理的に挿入されおいるため、遅い叀いカヌドが遅くなっおいるように芋えるような偏ったセットアップを行っおいる堎合には、非垞に圹立ちたす。初め。これを解決する 1 ぀の方法は、カヌドを亀換するこずです。ただし、カヌドを亀換できない堎合 (デバむスの冷华が圱響を受けた堎合など)、CUDA_DEVICE_ORDER=FASTEST_FIRSTを蚭定するず、垞に新しい高速カヌドが最初に配眮されたす。ただし、nvidia-smiは䟝然ずしお PCIe の順序でレポヌトするため、倚少混乱するでしょう。

順序を入れ替えるもう 1 ぀の解決策は、以䞋を䜿甚するこずです。

export CUDA_VISIBLE_DEVICES=1,0

この䟋では 2 ぀の GPU だけを䜿甚しおいたすが、もちろん、コンピュヌタヌに搭茉されおいる数の GPU にも同じこずが圓おはたりたす。

たた、この環境倉数を蚭定する堎合は、~/.bashrc ファむルたたはその他の起動蚭定ファむルに蚭定しお、忘れるのが最善です。

Trainer Integrations

Trainer は、トレヌニングを劇的に改善する可胜性のあるラむブラリをサポヌトするように拡匵されたした。 時間ずはるかに倧きなモデルに適合したす。

珟圚、サヌドパヌティの゜リュヌション DeepSpeed および PyTorch FSDP をサポヌトしおいたす。論文 [ZeRO: メモリの最適化] 兆パラメヌタ モデルのトレヌニングに向けお、Samyam Rajbhandari、Jeff Rasley、Olatunji Ruwase、Yuxiong He 著](https://arxiv.org/abs/1910.02054)。

この提䟛されるサポヌトは、この蚘事の執筆時点では新しくお実隓的なものです。 DeepSpeed ず PyTorch FSDP のサポヌトはアクティブであり、それに関する問題は歓迎したすが、FairScale 統合は PyTorch メむンに統合されおいるため、もうサポヌトしおいたせん (PyTorch FSDP 統合)

CUDA Extension Installation Notes

この蚘事の執筆時点では、Deepspeed を䜿甚するには、CUDA C++ コヌドをコンパむルする必芁がありたす。

すべおのむンストヌルの問題は、Deepspeed の察応する GitHub の問題を通じお察凊する必芁がありたすが、ビルド䞭に発生する可胜性のある䞀般的な問題がいく぀かありたす。 CUDA 拡匵機胜を構築する必芁がある PyTorch 拡匵機胜。

したがっお、次の操䜜を実行䞭に CUDA 関連のビルドの問題が発生した堎合は、次のずおりです。

pip install deepspeed

たず次の泚意事項をお読みください。

これらのノヌトでは、pytorch が CUDA 10.2 でビルドされた堎合に䜕をすべきかの䟋を瀺したす。あなたの状況が次のような堎合 異なる堎合は、バヌゞョン番号を目的のバヌゞョンに調敎するこずを忘れないでください。

Possible problem #1

Pytorch には独自の CUDA ツヌルキットが付属しおいたすが、これら 2 ぀のプロゞェクトをビルドするには、同䞀バヌゞョンの CUDA が必芁です。 システム党䜓にむンストヌルされたす。

たずえば、Python 環境に cudatoolkit==10.2 を指定しお pytorch をむンストヌルした堎合は、次のものも必芁です。 CUDA 10.2 がシステム党䜓にむンストヌルされたした。

正確な堎所はシステムによっお異なる堎合がありたすが、倚くのシステムでは/usr/local/cuda-10.2が最も䞀般的な堎所です。 Unix システム。 CUDA が正しく蚭定され、PATH環境倉数に远加されるず、 次のようにしおむンストヌル堎所を指定したす。

which nvcc

CUDA がシステム党䜓にむンストヌルされおいない堎合は、最初にむンストヌルしおください。お気に入りを䜿甚しお手順を芋぀けるこずができたす 怜玢゚ンゞン。たずえば、Ubuntu を䜿甚しおいる堎合は、ubuntu cuda 10.2 install を怜玢するずよいでしょう。

Possible problem #2

もう 1 ぀の考えられる䞀般的な問題は、システム党䜓に耇数の CUDA ツヌルキットがむンストヌルされおいる可胜性があるこずです。たずえばあなた がある可胜性があり

/usr/local/cuda-10.2
/usr/local/cuda-11.0

この状況では、PATH および LD_LIBRARY_PATH 環境倉数に以䞋が含たれおいるこずを確認する必芁がありたす。 目的の CUDA バヌゞョンぞの正しいパス。通垞、パッケヌゞ むンストヌラヌは、これらに、 最埌のバヌゞョンがむンストヌルされたした。適切なパッケヌゞが芋぀からないためにパッケヌゞのビルドが倱敗するずいう問題が発生した堎合は、 CUDA バヌゞョンがシステム党䜓にむンストヌルされおいるにもかかわらず、前述の 2 ぀を調敎する必芁があるこずを意味したす 環境倉数。

たず、その内容を芋おみたしょう。

echo $PATH
echo $LD_LIBRARY_PATH

それで、䞭に䜕が入っおいるかがわかりたす。

LD_LIBRARY_PATH が空である可胜性がありたす。

PATH は実行可胜ファむルが存圚する堎所をリストし、LD_LIBRARY_PATH は共有ラむブラリの堎所を瀺したす。 探すこずです。どちらの堎合も、前の゚ントリが埌の゚ントリより優先されたす。 : は耇数を区切るために䜿甚されたす ゚ントリ。

ここで、ビルド プログラムに特定の CUDA ツヌルキットの堎所を指瀺するには、最初にリストされる垌望のパスを挿入したす。 やっおいるこず

export PATH=/usr/local/cuda-10.2/bin:$PATH
export LD_LIBRARY_PATH=/usr/local/cuda-10.2/lib64:$LD_LIBRARY_PATH

既存の倀を䞊曞きするのではなく、先頭に远加するこずに泚意しおください。

もちろん、必芁に応じおバヌゞョン番号やフルパスを調敎したす。割り圓おたディレクトリが実際に機胜するこずを確認しおください 存圚する。 lib64 サブディレクトリは、libcudart.so などのさたざたな CUDA .so オブゞェクトが存圚する堎所です。 システムでは別の名前が付けられたすが、珟実を反映するように調敎しおください。

Possible problem #3

䞀郚の叀い CUDA バヌゞョンは、新しいコンパむラでのビルドを拒吊する堎合がありたす。たずえば、あなたはgcc-9を持っおいたすが、それが必芁です gcc-7。

それにはさたざたな方法がありたす。

最新の CUDA ツヌルキットをむンストヌルできる堎合は、通垞、新しいコンパむラがサポヌトされおいるはずです。

あるいは、既に所有しおいるコンパむラに加えお、䞋䜍バヌゞョンのコンパむラをむンストヌルするこずもできたす。 すでに存圚したすが、デフォルトではないため、ビルドシステムはそれを認識できたせん。 「gcc-7」がむンストヌルされおいるが、 ビルドシステムが芋぀からないずいうメッセヌゞを衚瀺する堎合は、次の方法で解決できる可胜性がありたす。

sudo ln -s /usr/bin/gcc-7  /usr/local/cuda-10.2/bin/gcc
sudo ln -s /usr/bin/g++-7  /usr/local/cuda-10.2/bin/g++

ここでは、/usr/local/cuda-10.2/bin/gcc から gcc-7 ぞのシンボリックリンクを䜜成しおいたす。 /usr/local/cuda-10.2/bin/ は PATH 環境倉数内にある必芁がありたす (前の問題の解決策を参照)。 gcc-7 (および g++7) が芋぀かるはずで、ビルドは成功したす。

い぀ものように、状況に合わせお䟋のパスを線集しおください。

PyTorch Fully Sharded Data parallel

より倧きなバッチ サむズで巚倧なモデルのトレヌニングを高速化するには、完党にシャヌド化されたデヌタ䞊列モデルを䜿甚できたす。 このタむプのデヌタ䞊列パラダむムでは、オプティマむザヌの状態、募配、パラメヌタヌをシャヌディングするこずで、より倚くのデヌタず倧芏暡なモデルをフィッティングできたす。 この機胜ずその利点の詳现に぀いおは、完党シャヌディング デヌタ䞊列ブログ をご芧ください。 最新の PyTorch の Fully Sharded Data Parallel (FSDP) トレヌニング機胜を統合したした。 必芁なのは、蚭定を通じお有効にするこずだけです。

FSDP サポヌトに必芁な PyTorch バヌゞョン: PyTorch Nightly (リリヌス埌にこれを読んだ堎合は 1.12.0) FSDP を有効にしたモデルの保存は、最近の修正でのみ利甚できるためです。

䜿甚法

  • 配垃されたランチャヌが远加されおいるこずを確認しおください ただ䜿甚しおいない堎合は、-m torch.distributed.launch --nproc_per_node=NUMBER_OF_GPUS_YOU_HAVEを䜿甚したす。

  • シャヌディング戊略:

    • FULL_SHARD : デヌタ䞊列ワヌカヌ/GPU にわたるシャヌド オプティマむザヌの状態 + 募配 + モデル パラメヌタヌ。 このためには、コマンドラむン匕数に--fsdp full_shardを远加したす。
    • SHARD_GRAD_OP : シャヌド オプティマむザヌの状態 + デヌタ䞊列ワヌカヌ/GPU 党䜓の募配。 このためには、コマンドラむン匕数に--fsdp shard_grad_opを远加したす。
    • NO_SHARD : シャヌディングなし。このためには、コマンドラむン匕数に--fsdp no_shardを远加したす。
  • パラメヌタず募配を CPU にオフロヌドするには、 コマンドラむン匕数に--fsdp "full_shard offload"たたは--fsdp "shard_grad_op offload"を远加したす。

  • default_auto_wrap_policy を䜿甚しお FSDP でレむダヌを自動的に再垰的にラップするには、 コマンドラむン匕数に--fsdp "full_shard auto_wrap"たたは--fsdp "shard_grad_op auto_wrap"を远加したす。

  • CPU オフロヌドず自動ラッピングの䞡方を有効にするには、 コマンドラむン匕数に--fsdp "full_shard offload auto_wrap"たたは--fsdp "shard_grad_op offload auto_wrap"を远加したす。

  • 残りの FSDP 構成は、--fsdp_config <path_to_fsdp_config.json>を介しお枡されたす。それは、次のいずれかの堎所です。 FSDP json 構成ファむル (䟋: fsdp_config.json)、たたはすでにロヌドされおいる json ファむルを dict ずしお䜿甚したす。

    • 自動ラッピングが有効な堎合は、トランスベヌスの自動ラップ ポリシヌたたはサむズ ベヌスの自動ラップ ポリシヌを䜿甚できたす。
      • トランスフォヌマヌベヌスの自動ラップポリシヌの堎合、構成ファむルで fsdp_transformer_layer_cls_to_wrap を指定するこずをお勧めしたす。指定しない堎合、䜿甚可胜な堎合、デフォルト倀は model._no_split_modules になりたす。 これは、ラップするトランスフォヌマヌ局クラス名のリスト (倧文字ず小文字を区別) を指定したす (䟋: BertLayer、GPTJBlock、T5Block 
)。 重みを共有するサブモゞュヌル (埋め蟌み局など) が異なる FSDP ラップされたナニットにならないようにする必芁があるため、これは重芁です。 このポリシヌを䜿甚するず、マルチヘッド アテンションずそれに続くいく぀かの MLP レむダヌを含むブロックごずにラッピングが発生したす。 共有埋め蟌みを含む残りの局は、同じ最も倖偎の FSDP ナニットにラップされるのが䟿利です。 したがっお、トランスベヌスのモデルにはこれを䜿甚しおください。
      • サむズベヌスの自動ラップポリシヌの堎合は、蚭定ファむルにfsdp_min_num_paramsを远加しおください。 自動ラッピングのための FSDP のパラメヌタの最小数を指定したす。
    • 蚭定ファむルで fsdp_backward_prefetch を指定できるようになりたした。次のパラメヌタのセットをい぀プリフェッチするかを制埡したす。 backward_pre ず backward_pos が利甚可胜なオプションです。 詳现に぀いおは、torch.distributed.fsdp.full_sharded_data_Parallel.BackwardPrefetchを参照しおください。
    • 蚭定ファむルで fsdp_forward_prefetch を指定できるようになりたした。次のパラメヌタのセットをい぀プリフェッチするかを制埡したす。 Trueの堎合、FSDP はフォワヌド パスでの実行䞭に、次に来るオヌルギャザヌを明瀺的にプリフェッチしたす。
    • 蚭定ファむルで limit_all_gathers を指定できるようになりたした。 Trueの堎合、FSDP は CPU スレッドを明瀺的に同期しお、実行䞭のオヌルギャザが倚すぎるのを防ぎたす。
    • activation_checkpointingを蚭定ファむルで指定できるようになりたした。 Trueの堎合、FSDP アクティベヌション チェックポむントは、FSDP のアクティベヌションをクリアするこずでメモリ䜿甚量を削枛する手法です。 特定のレむダヌを凊理し、バックワヌド パス䞭にそれらを再蚈算したす。事実䞊、これは䜙分な蚈算時間を犠牲にしたす メモリ䜿甚量を削枛したす。

泚意すべき泚意点がいく぀かありたす

  • これは generate ず互換性がないため、 --predict_with_generate ずも互換性がありたせん すべおの seq2seq/clm スクリプト (翻蚳/芁玄/clm など)。 問題 #21667 を参照しおください。

PyTorch/XLA Fully Sharded Data parallel

TPU ナヌザヌの皆様に朗報です。 PyTorch/XLA は FSDP をサポヌトするようになりたした。 最新の Fully Sharded Data Parallel (FSDP) トレヌニングがすべおサポヌトされおいたす。 詳现に぀いおは、FSDP を䜿甚した Cloud TPU での PyTorch モデルのスケヌリング および PyTorch/XLA 実装 を参照しおください。 FSDP の 必芁なのは、蚭定を通じお有効にするこずだけです。

FSDP サポヌトに必芁な PyTorch/XLA バヌゞョン: >=2.0

䜿甚法

--fsdp "full shard" を、--fsdp_config <path_to_fsdp_config.json> に加えられる次の倉曎ずずもに枡したす。

  • PyTorch/XLA FSDP を有効にするには、xlaをTrueに蚭定する必芁がありたす。
  • xla_fsdp_settings 倀は、XLA FSDP ラッピング パラメヌタを栌玍する蟞曞です。 オプションの完党なリストに぀いおは、こちら。
  • xla_fsdp_grad_ckpt。 Trueの堎合、ネストされた XLA FSDP でラップされた各レむダヌ䞊で募配チェックポむントを䜿甚したす。 この蚭定は、xla フラグが true に蚭定されおおり、自動ラッピング ポリシヌが指定されおいる堎合にのみ䜿甚できたす。 fsdp_min_num_params たたは fsdp_transformer_layer_cls_to_wrap。
  • トランスフォヌマヌ ベヌスの自動ラップ ポリシヌたたはサむズ ベヌスの自動ラップ ポリシヌのいずれかを䜿甚できたす。
    • トランスフォヌマヌベヌスの自動ラップポリシヌの堎合、構成ファむルで fsdp_transformer_layer_cls_to_wrap を指定するこずをお勧めしたす。指定しない堎合、䜿甚可胜な堎合、デフォルト倀は model._no_split_modules になりたす。 これは、ラップするトランスフォヌマヌ局クラス名のリスト (倧文字ず小文字を区別) を指定したす (䟋: BertLayer、GPTJBlock、T5Block 
)。 重みを共有するサブモゞュヌル (埋め蟌み局など) が異なる FSDP ラップされたナニットにならないようにする必芁があるため、これは重芁です。 このポリシヌを䜿甚するず、マルチヘッド アテンションずそれに続くいく぀かの MLP レむダヌを含むブロックごずにラッピングが発生したす。 共有埋め蟌みを含む残りの局は、同じ最も倖偎の FSDP ナニットにラップされるのが䟿利です。 したがっお、トランスベヌスのモデルにはこれを䜿甚しおください。
    • サむズベヌスの自動ラップポリシヌの堎合は、蚭定ファむルにfsdp_min_num_paramsを远加しおください。 自動ラッピングのための FSDP のパラメヌタの最小数を指定したす。

Using Trainer for accelerated PyTorch Training on Mac

PyTorch v1.12 リリヌスにより、開発者ず研究者は Apple シリコン GPU を利甚しおモデル トレヌニングを倧幅に高速化できたす。 これにより、プロトタむピングや埮調敎などの機械孊習ワヌクフロヌを Mac 䞊でロヌカルで実行できるようになりたす。 PyTorch のバック゚ンドずしおの Apple の Metal Performance Shaders (MPS) はこれを可胜にし、新しい "mps" デバむス経由で䜿甚できたす。 これにより、蚈算グラフずプリミティブが MPS Graph フレヌムワヌクず MPS によっお提䟛される調敎されたカヌネルにマッピングされたす。 詳现に぀いおは、公匏ドキュメント Mac での Accelerated PyTorch Training の玹介 を参照しおください。 および MPS バック゚ンド。

MacOS マシンに PyTorch >= 1.13 (執筆時点ではナむトリヌ バヌゞョン) をむンストヌルするこずを匷くお勧めしたす。 トランスベヌスのモデルのモデルの正確性ずパフォヌマンスの向䞊に関連する䞻芁な修正が行われおいたす。 詳现に぀いおは、https://github.com/pytorch/pytorch/issues/82707 を参照しおください。

Apple Silicon チップを䜿甚したトレヌニングず掚論の利点

  1. ナヌザヌがロヌカルで倧芏暡なネットワヌクやバッチ サむズをトレヌニングできるようにしたす
  2. ナニファむド メモリ アヌキテクチャにより、デヌタ取埗の遅延が短瞮され、GPU がメモリ ストア党䜓に盎接アクセスできるようになりたす。 したがっお、゚ンドツヌ゚ンドのパフォヌマンスが向䞊したす。
  3. クラりドベヌスの開発に関連するコストや远加のロヌカル GPU の必芁性を削枛したす。

前提条件: mps サポヌトを備えたトヌチをむンストヌルするには、 この玠晎らしいメディア蚘事 GPU アクセラレヌションが M1 Mac の PyTorch に登堎 に埓っおください。 。

䜿甚法 mps デバむスは、cuda デバむスが䜿甚される方法ず同様に利甚可胜な堎合、デフォルトで䜿甚されたす。 したがっお、ナヌザヌによるアクションは必芁ありたせん。 たずえば、以䞋のコマンドを䜿甚しお、Apple Silicon GPU を䜿甚しお公匏の Glue テキスト分類タスクを (ルヌト フォルダヌから) 実行できたす。

export TASK_NAME=mrpc

python examples/pytorch/text-classification/run_glue.py \
  --model_name_or_path bert-base-cased \
  --task_name $TASK_NAME \
  --do_train \
  --do_eval \
  --max_seq_length 128 \
  --per_device_train_batch_size 32 \
  --learning_rate 2e-5 \
  --num_train_epochs 3 \
  --output_dir /tmp/$TASK_NAME/ \
  --overwrite_output_dir

泚意すべきいく぀かの泚意事項

  1. 䞀郚の PyTorch 操䜜は mps に実装されおいないため、゚ラヌがスロヌされたす。 これを回避する 1 ぀の方法は、環境倉数 PYTORCH_ENABLE_MPS_FALLBACK=1 を蚭定するこずです。 これらの操䜜では CPU にフォヌルバックしたす。ただし、それでも UserWarning がスロヌされたす。
  2. 分散セットアップglooおよびncclは、mpsデバむスでは動䜜したせん。 これは、珟圚「mps」デバむス タむプの単䞀 GPU のみを䜿甚できるこずを意味したす。

最埌に、芚えおおいおください。 🀗 Trainer は MPS バック゚ンドのみを統合するため、 MPS バック゚ンドの䜿甚に関しお問題や質問がある堎合は、 PyTorch GitHub に問題を提出しおください。

Using Accelerate Launcher with Trainer

加速しおトレヌナヌにパワヌを䞎えたしょう。ナヌザヌが期埅するこずに関しおは、次のずおりです。

  • トレヌナヌ匕数に察しお FSDP、DeepSpeed などのトレヌナヌ むンテレヌションを倉曎せずに䜿甚し続けるこずができたす。
  • トレヌナヌで Accelerate Launcher を䜿甚できるようになりたした (掚奚)。

トレヌナヌで Accelerate Launcher を䜿甚する手順:

  1. 🀗 Accelerate がむンストヌルされおいるこずを確認しおください。Accelerate がないず Trainer を䜿甚するこずはできたせん。そうでない堎合は、pip install accelerateしおください。 Accelerate のバヌゞョンを曎新する必芁がある堎合もありたす: pip install activate --upgrade

  2. accelerate configを実行し、アンケヌトに蚘入したす。以䞋は加速蚭定の䟋です。  DDP マルチノヌド マルチ GPU 構成:

    compute_environment: LOCAL_MACHINE                                                                                             
    distributed_type: MULTI_GPU                                                                                                    
    downcast_bf16: 'no'
    gpu_ids: all
    machine_rank: 0 #change rank as per the node
    main_process_ip: 192.168.20.1
    main_process_port: 9898
    main_training_function: main
    mixed_precision: fp16
    num_machines: 2
    num_processes: 8
    rdzv_backend: static
    same_network: true
    tpu_env: []
    tpu_use_cluster: false
    tpu_use_sudo: false
    use_cpu: false

    b. FSDP config:

    compute_environment: LOCAL_MACHINE
    distributed_type: FSDP
    downcast_bf16: 'no'
    fsdp_config:
      fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
      fsdp_backward_prefetch_policy: BACKWARD_PRE
      fsdp_forward_prefetch: true
      fsdp_offload_params: false
      fsdp_sharding_strategy: 1
      fsdp_state_dict_type: FULL_STATE_DICT
      fsdp_sync_module_states: true
      fsdp_transformer_layer_cls_to_wrap: BertLayer
      fsdp_use_orig_params: true
    machine_rank: 0
    main_training_function: main
    mixed_precision: bf16
    num_machines: 1
    num_processes: 2
    rdzv_backend: static
    same_network: true
    tpu_env: []
    tpu_use_cluster: false
    tpu_use_sudo: false
    use_cpu: false

    c.ファむルを指す DeepSpeed 構成:

    compute_environment: LOCAL_MACHINE
    deepspeed_config:
      deepspeed_config_file: /home/user/configs/ds_zero3_config.json
      zero3_init_flag: true
    distributed_type: DEEPSPEED
    downcast_bf16: 'no'
    machine_rank: 0
    main_training_function: main
    num_machines: 1
    num_processes: 4
    rdzv_backend: static
    same_network: true
    tpu_env: []
    tpu_use_cluster: false
    tpu_use_sudo: false
    use_cpu: false

    d.加速プラグむンを䜿甚した DeepSpeed 構成:

    compute_environment: LOCAL_MACHINE                                                                                             
    deepspeed_config:                                                                                                              
      gradient_accumulation_steps: 1
      gradient_clipping: 0.7
      offload_optimizer_device: cpu
      offload_param_device: cpu
      zero3_init_flag: true
      zero_stage: 2
    distributed_type: DEEPSPEED
    downcast_bf16: 'no'
    machine_rank: 0
    main_training_function: main
    mixed_precision: bf16
    num_machines: 1
    num_processes: 4
    rdzv_backend: static
    same_network: true
    tpu_env: []
    tpu_use_cluster: false
    tpu_use_sudo: false
    use_cpu: false
  3. 加速蚭定たたはランチャヌ匕数によっお䞊蚘で凊理された匕数以倖の匕数を䜿甚しお、トレヌナヌ スクリプトを実行したす。 以䞋は、䞊蚘の FSDP 構成でaccelerate launcherを䜿甚しおrun_glue.pyを実行する䟋です。

cd transformers

accelerate launch \
./examples/pytorch/text-classification/run_glue.py \
--model_name_or_path bert-base-cased \
--task_name $TASK_NAME \
--do_train \
--do_eval \
--max_seq_length 128 \
--per_device_train_batch_size 16 \
--learning_rate 5e-5 \
--num_train_epochs 3 \
--output_dir /tmp/$TASK_NAME/ \
--overwrite_output_dir
  1. accelerate launchするための cmd 匕数を盎接䜿甚するこずもできたす。䞊の䟋は次のようにマッピングされたす。
cd transformers

accelerate launch --num_processes=2 \
--use_fsdp \
--mixed_precision=bf16 \
--fsdp_auto_wrap_policy=TRANSFORMER_BASED_WRAP  \
--fsdp_transformer_layer_cls_to_wrap="BertLayer" \
--fsdp_sharding_strategy=1 \
--fsdp_state_dict_type=FULL_STATE_DICT \
./examples/pytorch/text-classification/run_glue.py
--model_name_or_path bert-base-cased \
--task_name $TASK_NAME \
--do_train \
--do_eval \
--max_seq_length 128 \
--per_device_train_batch_size 16 \
--learning_rate 5e-5 \
--num_train_epochs 3 \
--output_dir /tmp/$TASK_NAME/ \
--overwrite_output_dir

詳现に぀いおは、🀗 Accelerate CLI ガむドを参照しおください: 🀗 Accelerate スクリプトの起動。

移動されたセクション:

[ DeepSpeed | Installation | Deployment with multiple GPUs | Deployment with one GPU | Deployment in Notebooks | Configuration | Passing Configuration | Shared Configuration | ZeRO | ZeRO-2 Config | ZeRO-3 Config | NVMe Support | ZeRO-2 vs ZeRO-3 Performance | ZeRO-2 Example | ZeRO-3 Example | Optimizer | Scheduler | fp32 Precision | Automatic Mixed Precision | Batch Size | Gradient Accumulation | Gradient Clipping | Getting The Model Weights Out ]