Source code for transformers.tokenization_utils

# coding=utf-8
# Copyright 2020 The HuggingFace Inc. team.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
 Tokenization classes for python tokenizers. For fast tokenizers (provided by HuggingFace's tokenizers library) see
 tokenization_utils_fast.py
"""
import itertools
import re
import unicodedata
from typing import Any, Dict, List, Optional, Tuple, Union, overload

from .file_utils import add_end_docstrings
from .tokenization_utils_base import (
    ENCODE_KWARGS_DOCSTRING,
    ENCODE_PLUS_ADDITIONAL_KWARGS_DOCSTRING,
    INIT_TOKENIZER_DOCSTRING,
    AddedToken,
    BatchEncoding,
    EncodedInput,
    EncodedInputPair,
    PaddingStrategy,
    PreTokenizedInput,
    PreTokenizedInputPair,
    PreTrainedTokenizerBase,
    TensorType,
    TextInput,
    TextInputPair,
    TruncationStrategy,
)
from .utils import logging


logger = logging.get_logger(__name__)

# Slow tokenizers are saved in a vocabulary plus three separated files
SPECIAL_TOKENS_MAP_FILE = "special_tokens_map.json"
ADDED_TOKENS_FILE = "added_tokens.json"
TOKENIZER_CONFIG_FILE = "tokenizer_config.json"


def _is_whitespace(char):
    """Checks whether `char` is a whitespace character."""
    # \t, \n, and \r are technically control characters but we treat them
    # as whitespace since they are generally considered as such.
    if char == " " or char == "\t" or char == "\n" or char == "\r":
        return True
    cat = unicodedata.category(char)
    if cat == "Zs":
        return True
    return False


def _is_control(char):
    """Checks whether `char` is a control character."""
    # These are technically control characters but we count them as whitespace
    # characters.
    if char == "\t" or char == "\n" or char == "\r":
        return False
    cat = unicodedata.category(char)
    if cat.startswith("C"):
        return True
    return False


def _is_punctuation(char):
    """Checks whether `char` is a punctuation character."""
    cp = ord(char)
    # We treat all non-letter/number ASCII as punctuation.
    # Characters such as "^", "$", and "`" are not in the Unicode
    # Punctuation class but we treat them as punctuation anyways, for
    # consistency.
    if (cp >= 33 and cp <= 47) or (cp >= 58 and cp <= 64) or (cp >= 91 and cp <= 96) or (cp >= 123 and cp <= 126):
        return True
    cat = unicodedata.category(char)
    if cat.startswith("P"):
        return True
    return False


def _is_end_of_word(text):
    """Checks whether the last character in text is one of a punctuation, control or whitespace character."""
    last_char = text[-1]
    return bool(_is_control(last_char) | _is_punctuation(last_char) | _is_whitespace(last_char))


def _is_start_of_word(text):
    """Checks whether the first character in text is one of a punctuation, control or whitespace character."""
    first_char = text[0]
    return bool(_is_control(first_char) | _is_punctuation(first_char) | _is_whitespace(first_char))


[docs]@add_end_docstrings(INIT_TOKENIZER_DOCSTRING, """ .. automethod:: __call__""") class PreTrainedTokenizer(PreTrainedTokenizerBase): """ Base class for all slow tokenizers. Inherits from :class:`~transformers.tokenization_utils_base.PreTrainedTokenizerBase`. Handle all the shared methods for tokenization and special tokens as well as methods downloading/caching/loading pretrained tokenizers as well as adding tokens to the vocabulary. This class also contain the added tokens in a unified way on top of all tokenizers so we don't have to handle the specific vocabulary augmentation methods of the various underlying dictionary structures (BPE, sentencepiece...). """ def __init__(self, **kwargs): super().__init__(**kwargs) # Added tokens - We store this for both slow and fast tokenizers # until the serialization of Fast tokenizers is updated self.added_tokens_encoder: Dict[str, int] = {} self.added_tokens_decoder: Dict[int, str] = {} self.unique_no_split_tokens: List[str] = [] @property def is_fast(self) -> bool: return False @property def vocab_size(self) -> int: """ :obj:`int`: Size of the base vocabulary (without the added tokens). """ raise NotImplementedError
[docs] def get_added_vocab(self) -> Dict[str, int]: """ Returns the added tokens in the vocabulary as a dictionary of token to index. Returns: :obj:`Dict[str, int]`: The added tokens. """ return self.added_tokens_encoder
def __len__(self): """ Size of the full vocabulary with the added tokens. """ return self.vocab_size + len(self.added_tokens_encoder) def _add_tokens(self, new_tokens: Union[List[str], List[AddedToken]], special_tokens: bool = False) -> int: """ Add a list of new tokens to the tokenizer class. If the new tokens are not in the vocabulary, they are added to it with indices starting from length of the current vocabulary. Args: new_tokens (:obj:`List[str]`or :obj:`List[tokenizers.AddedToken]`): Token(s) to add in vocabulary. A token is only added if it's not already in the vocabulary (tested by checking if the tokenizer assign the index of the ``unk_token`` to them). special_tokens (:obj:`bool`, `optional`, defaults to :obj:`False`): Whether or not the tokens should be added as special tokens. Returns: :obj:`int`: The number of tokens actually added to the vocabulary. Examples:: # Let's see how to increase the vocabulary of Bert model and tokenizer tokenizer = BertTokenizer.from_pretrained('bert-base-uncased') model = BertModel.from_pretrained('bert-base-uncased') num_added_toks = tokenizer.add_tokens(['new_tok1', 'my_new-tok2']) print('We have added', num_added_toks, 'tokens') # Note: resize_token_embeddings expects to receive the full size of the new vocabulary, i.e. the length of the tokenizer. model.resize_token_embeddings(len(tokenizer)) """ new_tokens = [str(tok) for tok in new_tokens] tokens_to_add = [] for token in new_tokens: assert isinstance(token, str) if not special_tokens and hasattr(self, "do_lower_case") and self.do_lower_case: token = token.lower() if ( token != self.unk_token and self.convert_tokens_to_ids(token) == self.convert_tokens_to_ids(self.unk_token) and token not in tokens_to_add ): tokens_to_add.append(token) if self.verbose: logger.info("Adding %s to the vocabulary", token) added_tok_encoder = dict((tok, len(self) + i) for i, tok in enumerate(tokens_to_add)) added_tok_decoder = {v: k for k, v in added_tok_encoder.items()} self.added_tokens_encoder.update(added_tok_encoder) self.added_tokens_decoder.update(added_tok_decoder) # Make sure we don't split on any special tokens (even they were already in the vocab before e.g. for Albert) if special_tokens: self.unique_no_split_tokens = sorted(set(self.unique_no_split_tokens).union(set(new_tokens))) else: # Or on the newly added tokens self.unique_no_split_tokens = sorted(set(self.unique_no_split_tokens).union(set(tokens_to_add))) return len(tokens_to_add)
[docs] def num_special_tokens_to_add(self, pair: bool = False) -> int: """ Returns the number of added tokens when encoding a sequence with special tokens. .. note:: This encodes a dummy input and checks the number of added tokens, and is therefore not efficient. Do not put this inside your training loop. Args: pair (:obj:`bool`, `optional`, defaults to :obj:`False`): Whether the number of added tokens should be computed in the case of a sequence pair or a single sequence. Returns: :obj:`int`: Number of special tokens added to sequences. """ token_ids_0 = [] token_ids_1 = [] return len(self.build_inputs_with_special_tokens(token_ids_0, token_ids_1 if pair else None))
[docs] def tokenize(self, text: TextInput, **kwargs) -> List[str]: """ Converts a string in a sequence of tokens, using the tokenizer. Split in words for word-based vocabulary or sub-words for sub-word-based vocabularies (BPE/SentencePieces/WordPieces). Takes care of added tokens. Args: text (:obj:`str`): The sequence to be encoded. **kwargs (additional keyword arguments): Passed along to the model-specific ``prepare_for_tokenization`` preprocessing method. Returns: :obj:`List[str]`: The list of tokens. """ # Simple mapping string => AddedToken for special tokens with specific tokenization behaviors all_special_tokens_extended = dict( (str(t), t) for t in self.all_special_tokens_extended if isinstance(t, AddedToken) ) text, kwargs = self.prepare_for_tokenization(text, **kwargs) if kwargs: logger.warning(f"Keyword arguments {kwargs} not recognized.") # TODO: should this be in the base class? if hasattr(self, "do_lower_case") and self.do_lower_case: # convert non-special tokens to lowercase escaped_special_toks = [re.escape(s_tok) for s_tok in self.all_special_tokens] pattern = r"(" + r"|".join(escaped_special_toks) + r")|" + r"(.+?)" text = re.sub(pattern, lambda m: m.groups()[0] or m.groups()[1].lower(), text) def split_on_token(tok, text): result = [] tok_extended = all_special_tokens_extended.get(tok, None) split_text = text.split(tok) full_word = "" for i, sub_text in enumerate(split_text): # AddedToken can control whitespace stripping around them. # We use them for GPT2 and Roberta to have different behavior depending on the special token # Cf. https://github.com/huggingface/transformers/pull/2778 # and https://github.com/huggingface/transformers/issues/3788 if isinstance(tok_extended, AddedToken): if tok_extended.single_word: # Try to avoid splitting on token if ( i < len(split_text) - 1 and not _is_end_of_word(sub_text) and not _is_start_of_word(split_text[i + 1]) ): # Don't extract the special token full_word += sub_text + tok elif full_word: full_word += sub_text result.append(full_word) full_word = "" continue # Strip white spaces on the right if tok_extended.rstrip and i > 0: # A bit counter-intuitive but we strip the left of the string # since tok_extended.rstrip means the special token is eating all white spaces on its right sub_text = sub_text.lstrip() # Strip white spaces on the left if tok_extended.lstrip and i < len(split_text) - 1: sub_text = sub_text.rstrip() # Opposite here else: # We strip left and right by default if i < len(split_text) - 1: sub_text = sub_text.rstrip() if i > 0: sub_text = sub_text.lstrip() if i == 0 and not sub_text: result.append(tok) elif i == len(split_text) - 1: if sub_text: result.append(sub_text) else: pass else: if sub_text: result.append(sub_text) result.append(tok) return result def split_on_tokens(tok_list, text): if not text.strip(): return [] if not tok_list: return self._tokenize(text) tokenized_text = [] text_list = [text] for tok in tok_list: tokenized_text = [] for sub_text in text_list: if sub_text not in self.unique_no_split_tokens: tokenized_text.extend(split_on_token(tok, sub_text)) else: tokenized_text.append(sub_text) text_list = tokenized_text return list( itertools.chain.from_iterable( ( self._tokenize(token) if token not in self.unique_no_split_tokens else [token] for token in tokenized_text ) ) ) no_split_token = self.unique_no_split_tokens tokenized_text = split_on_tokens(no_split_token, text) return tokenized_text
def _tokenize(self, text, **kwargs): """ Converts a string in a sequence of tokens (string), using the tokenizer. Split in words for word-based vocabulary or sub-words for sub-word-based vocabularies (BPE/SentencePieces/WordPieces). Do NOT take care of added tokens. """ raise NotImplementedError
[docs] def convert_tokens_to_ids(self, tokens: Union[str, List[str]]) -> Union[int, List[int]]: """ Converts a token string (or a sequence of tokens) in a single integer id (or a sequence of ids), using the vocabulary. Args: tokens (:obj:`str` or :obj:`List[str]`): One or several token(s) to convert to token id(s). Returns: :obj:`int` or :obj:`List[int]`: The token id or list of token ids. """ if tokens is None: return None if isinstance(tokens, str): return self._convert_token_to_id_with_added_voc(tokens) ids = [] for token in tokens: ids.append(self._convert_token_to_id_with_added_voc(token)) return ids
def _convert_token_to_id_with_added_voc(self, token): if token is None: return None if token in self.added_tokens_encoder: return self.added_tokens_encoder[token] return self._convert_token_to_id(token) def _convert_token_to_id(self, token): raise NotImplementedError def _encode_plus( self, text: Union[TextInput, PreTokenizedInput, EncodedInput], text_pair: Optional[Union[TextInput, PreTokenizedInput, EncodedInput]] = None, add_special_tokens: bool = True, padding_strategy: PaddingStrategy = PaddingStrategy.DO_NOT_PAD, truncation_strategy: TruncationStrategy = TruncationStrategy.DO_NOT_TRUNCATE, max_length: Optional[int] = None, stride: int = 0, is_split_into_words: bool = False, pad_to_multiple_of: Optional[int] = None, return_tensors: Optional[Union[str, TensorType]] = None, return_token_type_ids: Optional[bool] = None, return_attention_mask: Optional[bool] = None, return_overflowing_tokens: bool = False, return_special_tokens_mask: bool = False, return_offsets_mapping: bool = False, return_length: bool = False, verbose: bool = True, **kwargs ) -> BatchEncoding: def get_input_ids(text): if isinstance(text, str): tokens = self.tokenize(text, **kwargs) return self.convert_tokens_to_ids(tokens) elif isinstance(text, (list, tuple)) and len(text) > 0 and isinstance(text[0], str): if is_split_into_words: tokens = list( itertools.chain(*(self.tokenize(t, is_split_into_words=True, **kwargs) for t in text)) ) return self.convert_tokens_to_ids(tokens) else: return self.convert_tokens_to_ids(text) elif isinstance(text, (list, tuple)) and len(text) > 0 and isinstance(text[0], int): return text else: if is_split_into_words: raise ValueError( f"Input {text} is not valid. Should be a string or a list/tuple of strings when `is_split_into_words=True`." ) else: raise ValueError( f"Input {text} is not valid. Should be a string, a list/tuple of strings or a list/tuple of integers." ) if return_offsets_mapping: raise NotImplementedError( "return_offset_mapping is not available when using Python tokenizers." "To use this feature, change your tokenizer to one deriving from " "transformers.PreTrainedTokenizerFast." "More information on available tokenizers at " "https://github.com/huggingface/transformers/pull/2674" ) first_ids = get_input_ids(text) second_ids = get_input_ids(text_pair) if text_pair is not None else None return self.prepare_for_model( first_ids, pair_ids=second_ids, add_special_tokens=add_special_tokens, padding=padding_strategy.value, truncation=truncation_strategy.value, max_length=max_length, stride=stride, pad_to_multiple_of=pad_to_multiple_of, return_tensors=return_tensors, prepend_batch_axis=True, return_attention_mask=return_attention_mask, return_token_type_ids=return_token_type_ids, return_overflowing_tokens=return_overflowing_tokens, return_special_tokens_mask=return_special_tokens_mask, return_length=return_length, verbose=verbose, ) def _batch_encode_plus( self, batch_text_or_text_pairs: Union[ List[TextInput], List[TextInputPair], List[PreTokenizedInput], List[PreTokenizedInputPair], List[EncodedInput], List[EncodedInputPair], ], add_special_tokens: bool = True, padding_strategy: PaddingStrategy = PaddingStrategy.DO_NOT_PAD, truncation_strategy: TruncationStrategy = TruncationStrategy.DO_NOT_TRUNCATE, max_length: Optional[int] = None, stride: int = 0, is_split_into_words: bool = False, pad_to_multiple_of: Optional[int] = None, return_tensors: Optional[Union[str, TensorType]] = None, return_token_type_ids: Optional[bool] = None, return_attention_mask: Optional[bool] = None, return_overflowing_tokens: bool = False, return_special_tokens_mask: bool = False, return_offsets_mapping: bool = False, return_length: bool = False, verbose: bool = True, **kwargs ) -> BatchEncoding: def get_input_ids(text): if isinstance(text, str): tokens = self.tokenize(text, **kwargs) return self.convert_tokens_to_ids(tokens) elif isinstance(text, (list, tuple)) and len(text) > 0 and isinstance(text[0], str): if is_split_into_words: tokens = list( itertools.chain(*(self.tokenize(t, is_split_into_words=True, **kwargs) for t in text)) ) return self.convert_tokens_to_ids(tokens) else: return self.convert_tokens_to_ids(text) elif isinstance(text, (list, tuple)) and len(text) > 0 and isinstance(text[0], int): return text else: raise ValueError( "Input is not valid. Should be a string, a list/tuple of strings or a list/tuple of integers." ) if return_offsets_mapping: raise NotImplementedError( "return_offset_mapping is not available when using Python tokenizers." "To use this feature, change your tokenizer to one deriving from " "transformers.PreTrainedTokenizerFast." ) input_ids = [] for ids_or_pair_ids in batch_text_or_text_pairs: if not isinstance(ids_or_pair_ids, (list, tuple)): ids, pair_ids = ids_or_pair_ids, None elif is_split_into_words and not isinstance(ids_or_pair_ids[0], (list, tuple)): ids, pair_ids = ids_or_pair_ids, None else: ids, pair_ids = ids_or_pair_ids first_ids = get_input_ids(ids) second_ids = get_input_ids(pair_ids) if pair_ids is not None else None input_ids.append((first_ids, second_ids)) batch_outputs = self._batch_prepare_for_model( input_ids, add_special_tokens=add_special_tokens, padding_strategy=padding_strategy, truncation_strategy=truncation_strategy, max_length=max_length, stride=stride, pad_to_multiple_of=pad_to_multiple_of, return_attention_mask=return_attention_mask, return_token_type_ids=return_token_type_ids, return_overflowing_tokens=return_overflowing_tokens, return_special_tokens_mask=return_special_tokens_mask, return_length=return_length, return_tensors=return_tensors, verbose=verbose, ) return BatchEncoding(batch_outputs) @add_end_docstrings(ENCODE_KWARGS_DOCSTRING, ENCODE_PLUS_ADDITIONAL_KWARGS_DOCSTRING) def _batch_prepare_for_model( self, batch_ids_pairs: List[Union[PreTokenizedInputPair, Tuple[List[int], None]]], add_special_tokens: bool = True, padding_strategy: PaddingStrategy = PaddingStrategy.DO_NOT_PAD, truncation_strategy: TruncationStrategy = TruncationStrategy.DO_NOT_TRUNCATE, max_length: Optional[int] = None, stride: int = 0, pad_to_multiple_of: Optional[int] = None, return_tensors: Optional[str] = None, return_token_type_ids: Optional[bool] = None, return_attention_mask: Optional[bool] = None, return_overflowing_tokens: bool = False, return_special_tokens_mask: bool = False, return_length: bool = False, verbose: bool = True, ) -> BatchEncoding: """ Prepares a sequence of input id, or a pair of sequences of inputs ids so that it can be used by the model. It adds special tokens, truncates sequences if overflowing while taking into account the special tokens and manages a moving window (with user defined stride) for overflowing tokens Args: batch_ids_pairs: list of tokenized input ids or input ids pairs """ batch_outputs = {} for first_ids, second_ids in batch_ids_pairs: outputs = self.prepare_for_model( first_ids, second_ids, add_special_tokens=add_special_tokens, padding=PaddingStrategy.DO_NOT_PAD.value, # we pad in batch afterward truncation=truncation_strategy.value, max_length=max_length, stride=stride, pad_to_multiple_of=None, # we pad in batch afterward return_attention_mask=False, # we pad in batch afterward return_token_type_ids=return_token_type_ids, return_overflowing_tokens=return_overflowing_tokens, return_special_tokens_mask=return_special_tokens_mask, return_length=return_length, return_tensors=None, # We convert the whole batch to tensors at the end prepend_batch_axis=False, verbose=verbose, ) for key, value in outputs.items(): if key not in batch_outputs: batch_outputs[key] = [] batch_outputs[key].append(value) batch_outputs = self.pad( batch_outputs, padding=padding_strategy.value, max_length=max_length, pad_to_multiple_of=pad_to_multiple_of, return_attention_mask=return_attention_mask, ) batch_outputs = BatchEncoding(batch_outputs, tensor_type=return_tensors) return batch_outputs
[docs] def prepare_for_tokenization( self, text: str, is_split_into_words: bool = False, **kwargs ) -> Tuple[str, Dict[str, Any]]: """ Performs any necessary transformations before tokenization. This method should pop the arguments from kwargs and return the remaining :obj:`kwargs` as well. We test the :obj:`kwargs` at the end of the encoding process to be sure all the arguments have been used. Args: text (:obj:`str`): The text to prepare. is_split_into_words (:obj:`bool`, `optional`, defaults to :obj:`False`): Whether or not the text has been pretokenized. kwargs: Keyword arguments to use for the tokenization. Returns: :obj:`Tuple[str, Dict[str, Any]]`: The prepared text and the unused kwargs. """ return (text, kwargs)
[docs] def get_special_tokens_mask( self, token_ids_0: List, token_ids_1: Optional[List] = None, already_has_special_tokens: bool = False ) -> List[int]: """ Retrieves sequence ids from a token list that has no special tokens added. This method is called when adding special tokens using the tokenizer ``prepare_for_model`` or ``encode_plus`` methods. Args: token_ids_0 (:obj:`List[int]`): List of ids of the first sequence. token_ids_1 (:obj:`List[int]`, `optional`): List of ids of the second sequence. already_has_special_tokens (:obj:`bool`, `optional`, defaults to :obj:`False`): Whether or not the token list is already formatted with special tokens for the model. Returns: A list of integers in the range [0, 1]: 1 for a special token, 0 for a sequence token. """ return [0] * ((len(token_ids_1) if token_ids_1 else 0) + len(token_ids_0))
@overload def convert_ids_to_tokens(self, ids: int, skip_special_tokens: bool = False) -> str: ... @overload def convert_ids_to_tokens(self, ids: List[int], skip_special_tokens: bool = False) -> List[str]: ...
[docs] def convert_ids_to_tokens( self, ids: Union[int, List[int]], skip_special_tokens: bool = False ) -> Union[str, List[str]]: """ Converts a single index or a sequence of indices in a token or a sequence of tokens, using the vocabulary and added tokens. Args: ids (:obj:`int` or :obj:`List[int]`): The token id (or token ids) to convert to tokens. skip_special_tokens (:obj:`bool`, `optional`, defaults to :obj:`False`): Whether or not to remove special tokens in the decoding. Returns: :obj:`str` or :obj:`List[str]`: The decoded token(s). """ if isinstance(ids, int): if ids in self.added_tokens_decoder: return self.added_tokens_decoder[ids] else: return self._convert_id_to_token(ids) tokens = [] for index in ids: index = int(index) if skip_special_tokens and index in self.all_special_ids: continue if index in self.added_tokens_decoder: tokens.append(self.added_tokens_decoder[index]) else: tokens.append(self._convert_id_to_token(index)) return tokens
def _convert_id_to_token(self, index: int) -> str: raise NotImplementedError
[docs] def convert_tokens_to_string(self, tokens: List[str]) -> str: return " ".join(tokens)
def _decode( self, token_ids: List[int], skip_special_tokens: bool = False, clean_up_tokenization_spaces: bool = True, spaces_between_special_tokens: bool = True, ) -> str: filtered_tokens = self.convert_ids_to_tokens(token_ids, skip_special_tokens=skip_special_tokens) # To avoid mixing byte-level and unicode for byte-level BPT # we need to build string separately for added tokens and byte-level tokens # cf. https://github.com/huggingface/transformers/issues/1133 sub_texts = [] current_sub_text = [] for token in filtered_tokens: if skip_special_tokens and token in self.all_special_ids: continue if token in self.added_tokens_encoder: if current_sub_text: sub_texts.append(self.convert_tokens_to_string(current_sub_text)) current_sub_text = [] sub_texts.append(token) else: current_sub_text.append(token) if current_sub_text: sub_texts.append(self.convert_tokens_to_string(current_sub_text)) if spaces_between_special_tokens: text = " ".join(sub_texts) else: text = "".join(sub_texts) if clean_up_tokenization_spaces: clean_text = self.clean_up_tokenization(text) return clean_text else: return text