Data Collator

Data collators are objects that will form a batch by using a list of dataset elements as input. These elements are of the same type as the elements of train_dataset or eval_dataset.

To be able to build batches, data collators may apply some processing (like padding). Some of them (like DataCollatorForLanguageModeling) also apply some random data augmentation (like random masking) on the formed batch.

Examples of use can be found in the example scripts or example notebooks.

Default data collator

transformers.data.data_collator.default_data_collator(features: List[NewType.<locals>.new_type], return_tensors='pt') → Dict[str, Any]

Very simple data collator that simply collates batches of dict-like objects and performs special handling for potential keys named:

• label: handles a single value (int or float) per object

• label_ids: handles a list of values per object

Does not do any additional preprocessing: property names of the input object will be used as corresponding inputs to the model. See glue and ner for example of how it’s useful.

DataCollatorWithPadding

class transformers.data.data_collator.DataCollatorWithPadding(tokenizer: transformers.tokenization_utils_base.PreTrainedTokenizerBase, padding: Union[bool, str, transformers.file_utils.PaddingStrategy] = True, max_length: Optional[int] = None, pad_to_multiple_of: Optional[int] = None, return_tensors: str = 'pt')

Data collator that will dynamically pad the inputs received.

Parameters

• tokenizer (PreTrainedTokenizer or PreTrainedTokenizerFast) – The tokenizer used for encoding the data.

• padding (bool, str or PaddingStrategy, optional, defaults to True) –

  Select a strategy to pad the returned sequences (according to the model’s padding side and padding index) among:

  • True or 'longest': Pad to the longest sequence in the batch (or no padding if only a single sequence if provided).

  • 'max_length': Pad to a maximum length specified with the argument max_length or to the maximum acceptable input length for the model if that argument is not provided.

  • False or 'do_not_pad' (default): No padding (i.e., can output a batch with sequences of different lengths).

• max_length (int, optional) – Maximum length of the returned list and optionally padding length (see above).

• pad_to_multiple_of (int, optional) –

  If set will pad the sequence to a multiple of the provided value.

  This is especially useful to enable the use of Tensor Cores on NVIDIA hardware with compute capability >= 7.5 (Volta).

DataCollatorForTokenClassification

class transformers.data.data_collator.DataCollatorForTokenClassification(tokenizer: transformers.tokenization_utils_base.PreTrainedTokenizerBase, padding: Union[bool, str, transformers.file_utils.PaddingStrategy] = True, max_length: Optional[int] = None, pad_to_multiple_of: Optional[int] = None, label_pad_token_id: int = - 100, return_tensors: str = 'pt')

Data collator that will dynamically pad the inputs received, as well as the labels.

Parameters

• tokenizer (PreTrainedTokenizer or PreTrainedTokenizerFast) – The tokenizer used for encoding the data.

• padding (bool, str or PaddingStrategy, optional, defaults to True) –

  Select a strategy to pad the returned sequences (according to the model’s padding side and padding index) among:

  • True or 'longest': Pad to the longest sequence in the batch (or no padding if only a single sequence if provided).

  • 'max_length': Pad to a maximum length specified with the argument max_length or to the maximum acceptable input length for the model if that argument is not provided.

  • False or 'do_not_pad' (default): No padding (i.e., can output a batch with sequences of different lengths).

• max_length (int, optional) – Maximum length of the returned list and optionally padding length (see above).

• pad_to_multiple_of (int, optional) –

  If set will pad the sequence to a multiple of the provided value.

  This is especially useful to enable the use of Tensor Cores on NVIDIA hardware with compute capability >= 7.5 (Volta).

• label_pad_token_id (int, optional, defaults to -100) – The id to use when padding the labels (-100 will be automatically ignore by PyTorch loss functions).

DataCollatorForSeq2Seq

class transformers.data.data_collator.DataCollatorForSeq2Seq(tokenizer: transformers.tokenization_utils_base.PreTrainedTokenizerBase, model: Optional[Any] = None, padding: Union[bool, str, transformers.file_utils.PaddingStrategy] = True, max_length: Optional[int] = None, pad_to_multiple_of: Optional[int] = None, label_pad_token_id: int = - 100, return_tensors: str = 'pt')

Data collator that will dynamically pad the inputs received, as well as the labels.

Parameters

• tokenizer (PreTrainedTokenizer or PreTrainedTokenizerFast) – The tokenizer used for encoding the data.

• model (PreTrainedModel) –

  The model that is being trained. If set and has the prepare_decoder_input_ids_from_labels, use it to prepare the decoder_input_ids

  This is useful when using label_smoothing to avoid calculating loss twice.

• padding (bool, str or PaddingStrategy, optional, defaults to True) –

  Select a strategy to pad the returned sequences (according to the model’s padding side and padding index) among:

  • True or 'longest': Pad to the longest sequence in the batch (or no padding if only a single sequence is provided).

  • 'max_length': Pad to a maximum length specified with the argument max_length or to the maximum acceptable input length for the model if that argument is not provided.

  • False or 'do_not_pad' (default): No padding (i.e., can output a batch with sequences of different lengths).

• max_length (int, optional) – Maximum length of the returned list and optionally padding length (see above).

• pad_to_multiple_of (int, optional) –

  If set will pad the sequence to a multiple of the provided value.

  This is especially useful to enable the use of Tensor Cores on NVIDIA hardware with compute capability >= 7.5 (Volta).

• label_pad_token_id (int, optional, defaults to -100) – The id to use when padding the labels (-100 will be automatically ignored by PyTorch loss functions).

DataCollatorForLanguageModeling

class transformers.data.data_collator.DataCollatorForLanguageModeling(tokenizer: transformers.tokenization_utils_base.PreTrainedTokenizerBase, mlm: bool = True, mlm_probability: float = 0.15, pad_to_multiple_of: Optional[int] = None, tf_experimental_compile: bool = False, return_tensors: str = 'pt')

Data collator used for language modeling. Inputs are dynamically padded to the maximum length of a batch if they are not all of the same length.

Parameters

• tokenizer (PreTrainedTokenizer or PreTrainedTokenizerFast) – The tokenizer used for encoding the data.

• mlm (bool, optional, defaults to True) – Whether or not to use masked language modeling. If set to False, the labels are the same as the inputs with the padding tokens ignored (by setting them to -100). Otherwise, the labels are -100 for non-masked tokens and the value to predict for the masked token.

• mlm_probability (float, optional, defaults to 0.15) – The probability with which to (randomly) mask tokens in the input, when mlm is set to True.

• pad_to_multiple_of (int, optional) – If set will pad the sequence to a multiple of the provided value.

Note

For best performance, this data collator should be used with a dataset having items that are dictionaries or BatchEncoding, with the "special_tokens_mask" key, as returned by a PreTrainedTokenizer or a PreTrainedTokenizerFast with the argument return_special_tokens_mask=True.

numpy_mask_tokens(inputs: Any, special_tokens_mask: Optional[Any] = None) → Tuple[Any, Any]

Prepare masked tokens inputs/labels for masked language modeling: 80% MASK, 10% random, 10% original.

tf_mask_tokens(inputs: Any, vocab_size, mask_token_id, special_tokens_mask: Optional[Any] = None) → Tuple[Any, Any]

Prepare masked tokens inputs/labels for masked language modeling: 80% MASK, 10% random, 10% original.

torch_mask_tokens(inputs: Any, special_tokens_mask: Optional[Any] = None) → Tuple[Any, Any]

Prepare masked tokens inputs/labels for masked language modeling: 80% MASK, 10% random, 10% original.

DataCollatorForWholeWordMask

class transformers.data.data_collator.DataCollatorForWholeWordMask(tokenizer: transformers.tokenization_utils_base.PreTrainedTokenizerBase, mlm: bool = True, mlm_probability: float = 0.15, pad_to_multiple_of: Optional[int] = None, tf_experimental_compile: bool = False, return_tensors: str = 'pt')

Data collator used for language modeling that masks entire words.

• collates batches of tensors, honoring their tokenizer’s pad_token

• preprocesses batches for masked language modeling

Note

This collator relies on details of the implementation of subword tokenization by BertTokenizer, specifically that subword tokens are prefixed with ##. For tokenizers that do not adhere to this scheme, this collator will produce an output that is roughly equivalent to DataCollatorForLanguageModeling.

numpy_mask_tokens(inputs: Any, mask_labels: Any) → Tuple[Any, Any]

Prepare masked tokens inputs/labels for masked language modeling: 80% MASK, 10% random, 10% original. Set ‘mask_labels’ means we use whole word mask (wwm), we directly mask idxs according to it’s ref.

tf_mask_tokens(inputs: Any, mask_labels: Any) → Tuple[Any, Any]

Prepare masked tokens inputs/labels for masked language modeling: 80% MASK, 10% random, 10% original. Set ‘mask_labels’ means we use whole word mask (wwm), we directly mask idxs according to it’s ref.

torch_mask_tokens(inputs: Any, mask_labels: Any) → Tuple[Any, Any]

Prepare masked tokens inputs/labels for masked language modeling: 80% MASK, 10% random, 10% original. Set ‘mask_labels’ means we use whole word mask (wwm), we directly mask idxs according to it’s ref.

DataCollatorForPermutationLanguageModeling

class transformers.data.data_collator.DataCollatorForPermutationLanguageModeling(tokenizer: transformers.tokenization_utils_base.PreTrainedTokenizerBase, plm_probability: float = 0.16666666666666666, max_span_length: int = 5, return_tensors: str = 'pt')

Data collator used for permutation language modeling.

• collates batches of tensors, honoring their tokenizer’s pad_token

• preprocesses batches for permutation language modeling with procedures specific to XLNet

numpy_mask_tokens(inputs: Any) → Tuple[Any, Any, Any, Any]

The masked tokens to be predicted for a particular sequence are determined by the following algorithm:

0. Start from the beginning of the sequence by setting cur_len = 0 (number of tokens processed so far).

1. Sample a span_length from the interval [1, max_span_length] (length of span of tokens to be masked)

2. Reserve a context of length context_length = span_length / plm_probability to surround span to be masked

3. Sample a starting point start_index from the interval [cur_len, cur_len + context_length - span_length] and mask tokens start_index:start_index + span_length

4. Set cur_len = cur_len + context_length. If cur_len < max_len (i.e. there are tokens remaining in the sequence to be processed), repeat from Step 1.

tf_mask_tokens(inputs: Any) → Tuple[Any, Any, Any, Any]

The masked tokens to be predicted for a particular sequence are determined by the following algorithm:

0. Start from the beginning of the sequence by setting cur_len = 0 (number of tokens processed so far).

1. Sample a span_length from the interval [1, max_span_length] (length of span of tokens to be masked)

2. Reserve a context of length context_length = span_length / plm_probability to surround span to be masked

3. Sample a starting point start_index from the interval [cur_len, cur_len + context_length - span_length] and mask tokens start_index:start_index + span_length

4. Set cur_len = cur_len + context_length. If cur_len < max_len (i.e. there are tokens remaining in the sequence to be processed), repeat from Step 1.

torch_mask_tokens(inputs: Any) → Tuple[Any, Any, Any, Any]

The masked tokens to be predicted for a particular sequence are determined by the following algorithm:

0. Start from the beginning of the sequence by setting cur_len = 0 (number of tokens processed so far).

1. Sample a span_length from the interval [1, max_span_length] (length of span of tokens to be masked)

2. Reserve a context of length context_length = span_length / plm_probability to surround span to be masked

3. Sample a starting point start_index from the interval [cur_len, cur_len + context_length - span_length] and mask tokens start_index:start_index + span_length

4. Set cur_len = cur_len + context_length. If cur_len < max_len (i.e. there are tokens remaining in the sequence to be processed), repeat from Step 1.