Source code for mindnlp.transforms.tokenizers.luke_tokenizer

# # Copyright 2023 Huawei Technologies Co., Ltd
# #
# # 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.
# # ============================================================================
"""
LukeTokenizer
"""
# pylint: disable=C0301
# pylint: disable=C0302
# pylint: disable=W0511
# pylint: disable=C0103
# pylint: disable=R0914
# pylint: disable=R1702
# pylint: disable=R0912
# pylint: disable=R0915
# pylint: disable=R0902
# pylint: disable=R0913
# pylint: disable=E1101

import itertools
import json
import warnings
import logging
from enum import Enum
from collections import OrderedDict
from collections.abc import Mapping
from functools import lru_cache
from typing import List, Optional, Tuple, Union

import regex as re
from tokenizers import AddedToken
from mindnlp.abc import PreTrainedTokenizer

logger = logging.getLogger(__name__)

VERY_LARGE_INTEGER = int(1e30)
LARGE_INTEGER = int(1e20)

EntitySpan = Tuple[int, int]
EntitySpanInput = List[EntitySpan]
Entity = str
EntityInput = List[Entity]
EncodedInput = List[int]
TextInput = str
TextInputPair = Tuple[str, str]
PreTokenizedInput = List[str]
PreTokenizedInputPair = Tuple[List[str], List[str]]

VOCAB_FILES_NAMES = {
    "vocab_file": "vocab.json",
    "merges_file": "merges.txt",
    "entity_vocab_file": "entity_vocab.json",
}

PRETRAINED_VOCAB_FILES_MAP = {
    "vocab_file": {
        "studio-ousia/luke-base": "https://huggingface.co/studio-ousia/luke-base/resolve/main/vocab.json",
        "studio-ousia/luke-large": "https://huggingface.co/studio-ousia/luke-large/resolve/main/vocab.json",
    },
    "merges_file": {
        "studio-ousia/luke-base": "https://huggingface.co/studio-ousia/luke-base/resolve/main/merges.txt",
        "studio-ousia/luke-large": "https://huggingface.co/studio-ousia/luke-large/resolve/main/merges.txt",
    },
    "entity_vocab_file": {
        "studio-ousia/luke-base": "https://huggingface.co/studio-ousia/luke-base/resolve/main/entity_vocab.json",
        "studio-ousia/luke-large": "https://huggingface.co/studio-ousia/luke-large/resolve/main/entity_vocab.json",
    },
}

PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES = {
    "studio-ousia/luke-base": 512,
    "studio-ousia/luke-large": 512,
}

ENCODE_PLUS_ADDITIONAL_KWARGS_DOCSTRING = r"""
            return_token_type_ids (`bool`, *optional*):
                Whether to return token type IDs. If left to the default, will return the token type IDs according to
                the specific tokenizer's default, defined by the `return_outputs` attribute.

                [What are token type IDs?](../glossary#token-type-ids)
            return_attention_mask (`bool`, *optional*):
                Whether to return the attention mask. If left to the default, will return the attention mask according
                to the specific tokenizer's default, defined by the `return_outputs` attribute.

                [What are attention masks?](../glossary#attention-mask)
            return_overflowing_tokens (`bool`, *optional*, defaults to `False`):
                Whether or not to return overflowing token sequences. If a pair of sequences of input ids (or a batch
                of pairs) is provided with `truncation_strategy = longest_first` or `True`, an error is raised instead
                of returning overflowing tokens.
            return_special_tokens_mask (`bool`, *optional*, defaults to `False`):
                Whether or not to return special tokens mask information.
            return_offsets_mapping (`bool`, *optional*, defaults to `False`):
                Whether or not to return `(char_start, char_end)` for each token.

                This is only available on fast tokenizers inheriting from [`PreTrainedTokenizerFast`], if using
                Python's tokenizer, this method will raise `NotImplementedError`.
            return_length  (`bool`, *optional*, defaults to `False`):
                Whether or not to return the lengths of the encoded inputs.
            verbose (`bool`, *optional*, defaults to `True`):
                Whether or not to print more information and warnings.
            **kwargs: passed to the `self.tokenize_()` method

        Return:
            [`BatchEncoding`]: A [`BatchEncoding`] with the following fields:

            - **input_ids** -- List of token ids to be fed to a model.

              [What are input IDs?](../glossary#input-ids)

            - **token_type_ids** -- List of token type ids to be fed to a model (when `return_token_type_ids=True` or
              if *"token_type_ids"* is in `self.model_input_names`).

              [What are token type IDs?](../glossary#token-type-ids)

            - **attention_mask** -- List of indices specifying which tokens should be attended to by the model (when
              `return_attention_mask=True` or if *"attention_mask"* is in `self.model_input_names`).

              [What are attention masks?](../glossary#attention-mask)

            - **entity_ids** -- List of entity ids to be fed to a model.

              [What are input IDs?](../glossary#input-ids)

            - **entity_position_ids** -- List of entity positions in the input sequence to be fed to a model.

            - **entity_token_type_ids** -- List of entity token type ids to be fed to a model (when
              `return_token_type_ids=True` or if *"entity_token_type_ids"* is in `self.model_input_names`).

              [What are token type IDs?](../glossary#token-type-ids)

            - **entity_attention_mask** -- List of indices specifying which entities should be attended to by the model
              (when `return_attention_mask=True` or if *"entity_attention_mask"* is in `self.model_input_names`).

              [What are attention masks?](../glossary#attention-mask)

            - **entity_start_positions** -- List of the start positions of entities in the word token sequence (when
              `task="entity_span_classification"`).
            - **entity_end_positions** -- List of the end positions of entities in the word token sequence (when
              `task="entity_span_classification"`).
            - **overflowing_tokens** -- List of overflowing tokens sequences (when a `max_length` is specified and
              `return_overflowing_tokens=True`).
            - **num_truncated_tokens** -- Number of tokens truncated (when a `max_length` is specified and
              `return_overflowing_tokens=True`).
            - **special_tokens_mask** -- List of 0s and 1s, with 1 specifying added special tokens and 0 specifying
              regular sequence tokens (when `add_special_tokens=True` and `return_special_tokens_mask=True`).
            - **length** -- The length of the inputs (when `return_length=True`)

"""


class ExplicitEnum(str, Enum):
    """
    Enum with more explicit error message for missing values.
    """

    @classmethod
    def _missing_(cls, value):
        raise ValueError(
            f"{value} is not a valid {cls.__name__}, please select one of {list(cls._value2member_map_.keys())}"
        )


class PaddingStrategy(ExplicitEnum):
    """
    Possible values for the `padding` argument in [`PreTrainedTokenizerBase.__call__`]. Useful for tab-completion in an
    IDE.
    """

    LONGEST = "longest"
    MAX_LENGTH = "max_length"
    DO_NOT_PAD = "do_not_pad"


class TensorType(ExplicitEnum):
    """
    Possible values for the `return_tensors` argument in [`PreTrainedTokenizerBase.__call__`]. Useful for
    tab-completion in an IDE.
    """

    PYTORCH = "pt"
    TENSORFLOW = "tf"
    NUMPY = "np"
    JAX = "jax"


class TruncationStrategy(ExplicitEnum):
    """
    Possible values for the `truncation` argument in [`PreTrainedTokenizerBase.__call__`]. Useful for tab-completion in
    an IDE.
    """

    ONLY_FIRST = "only_first"
    ONLY_SECOND = "only_second"
    LONGEST_FIRST = "longest_first"
    DO_NOT_TRUNCATE = "do_not_truncate"


@lru_cache()
# Copied from transformers.models.roberta.tokenization_roberta.bytes_to_unicode
def bytes_to_unicode():
    """
    Returns list of utf-8 byte and a mapping to unicode strings. We specifically avoids mapping to whitespace/control
    characters the bpe code barfs on.

    The reversible bpe codes work on unicode strings. This means you need a large # of unicode characters in your vocab
    if you want to avoid UNKs. When you're at something like a 10B token dataset you end up needing around 5K for
    decent coverage. This is a significant percentage of your normal, say, 32K bpe vocab. To avoid that, we want lookup
    tables between utf-8 bytes and unicode strings.
    """
    bs = (
            list(range(ord("!"), ord("~") + 1)) + list(range(ord("¡"), ord("¬") + 1)) + list(
        range(ord("®"), ord("ÿ") + 1))
    )
    cs = bs[:]
    n = 0
    for b in range(2 ** 8):
        if b not in bs:
            bs.append(b)
            cs.append(2 ** 8 + n)
            n += 1
    cs = [chr(n) for n in cs]
    return dict(zip(bs, cs))


# Copied from transformers.models.roberta.tokenization_roberta.get_pairs
def get_pairs(word):
    """
    Return set of symbol pairs in a word.

    Word is represented as tuple of symbols (symbols being variable-length strings).
    """
    pairs = set()
    prev_char = word[0]
    for char in word[1:]:
        pairs.add((prev_char, char))
        prev_char = char
    return pairs


class Trie:
    """
    Trie in Python. Creates a Trie out of a list of words. The trie is used to split on `added_tokens` in one pass
    Loose reference https://en.wikipedia.org/wiki/Trie
    """

    def __init__(self):
        self.data = {}

    def split(self, text: str) -> List[str]:
        """
        Will look for the words added to the trie within `text`. Output is the original string splitted along the
        boundaries of the words found.

        This trie will match the longest possible word first !

        Example:

        ```python
        >>> trie = Trie()
        >>> trie.split("[CLS] This is a extra_id_100")
        ["[CLS] This is a extra_id_100"]

        >>> trie.add("[CLS]")
        >>> trie.add("extra_id_1")
        >>> trie.add("extra_id_100")
        >>> trie.split("[CLS] This is a extra_id_100")
        ["[CLS]", " This is a ", "extra_id_100"]
        ```
        """
        # indexes are counted left of the chars index.
        # "hello", index 0, is left of h, index 1 is between h and e.
        # index 5 is right of the "o".

        # States are going to capture every possible start (indexes as above)
        # as keys, and have as values, a pointer to the position in the trie
        # where we're at. This is a partial match for now.
        # This enables to keep track of multiple matches while we're iterating
        # the string
        # If the trie contains, "blowing", and "lower" and we encounter the
        # string "blower", we need to split into ["b", "lower"].
        # This is where we need to keep track of multiple possible starts.
        states = OrderedDict()

        # This will contain every indices where we need
        # to cut.
        # We force to cut at offset 0 and len(text) (added later)
        offsets = [0]

        # This is used by the lookahead which needs to skip over
        # some text where the full match exceeded the place in the initial
        # for loop
        skip = 0
        # Main loop, Giving this algorithm O(n) complexity
        for current, current_char in enumerate(text):
            if skip and current < skip:
                # Prevents the lookahead for matching twice
                # like extra_id_100 and id_100
                continue

            # This will track every state
            # that stop matching, we need to stop tracking them.
            # If we look at "lowball", we're going to match "l" (add it to states), "o", "w", then
            # fail on "b", we need to remove 0 from the valid states.
            to_remove = set()
            # Whenever we found a match, we need to drop everything
            # this is a greedy algorithm, it will match on the first found token
            reset = False

            # In this case, we already have partial matches (But unfinished)
            for start, trie_pointer in states.items():
                if "" in trie_pointer:
                    # This is a final match, we need to reset and
                    # store the results in `offsets`.

                    # Lookahead to match longest first
                    # Important in case of extra_id_1 vs extra_id_100
                    # Here we are also actively looking for other earlier partial
                    # matches
                    # "[CLS]", "L", we need to match CLS even if L is special
                    for lookstart, looktrie_pointer in states.items():
                        if lookstart > start:
                            # This partial match is later, we can stop looking
                            break
                        if lookstart < start:
                            # This partial match is earlier, the trie pointer
                            # was already updated, so index is + 1
                            lookahead_index = current + 1
                            end = current + 1
                        else:
                            # Here lookstart == start and
                            #      looktrie_pointer == trie_pointer
                            # It wasn't updated yet so indices are current ones
                            lookahead_index = current
                            end = current
                        next_char = text[lookahead_index] if lookahead_index < len(text) else None
                        if "" in looktrie_pointer:
                            start = lookstart
                            end = lookahead_index
                            skip = lookahead_index

                        while next_char in looktrie_pointer:
                            looktrie_pointer = looktrie_pointer[next_char]
                            lookahead_index += 1
                            if "" in looktrie_pointer:
                                start = lookstart
                                end = lookahead_index
                                skip = lookahead_index

                            if lookahead_index == len(text):
                                # End of string
                                break
                            next_char = text[lookahead_index]
                        # End lookahead

                    # Storing and resetting
                    offsets.append(start)
                    offsets.append(end)
                    reset = True
                    break
                if current_char in trie_pointer:
                    # The current character being looked at has a match within the trie
                    # update the pointer (it will be stored back into states later).
                    trie_pointer = trie_pointer[current_char]

                    # Storing back the new pointer into the states.
                    # Partial matches got longer by one.
                    states[start] = trie_pointer
                else:
                    # The new character has not match in the trie, we need
                    # to stop keeping track of this partial match.
                    # We can't do it directly within the loop because of how
                    # python iteration works
                    to_remove.add(start)

            # Either clearing the full start (we found a real match)
            # Or clearing only the partial matches that didn't work.
            if reset:
                states = {}
            else:
                for start in to_remove:
                    del states[start]

            # If this character is a starting character within the trie
            # start keeping track of this partial match.
            if current >= skip and current_char in self.data:
                states[current] = self.data[current_char]

        # We have a cut at the end with states.
        for start, trie_pointer in states.items():
            if "" in trie_pointer:
                # This is a final match, we need to reset and
                # store the results in `offsets`.
                end = len(text)
                offsets.append(start)
                offsets.append(end)
                # Longest cut is always the one with lower start so the first
                # item so we need to break.
                break

        return self.cut_text(text, offsets)

    def cut_text(self, text, offsets):
        """
        cut_text
        """
        # We have all the offsets now, we just need to do the actual splitting.
        # We need to eventually add the first part of the string and the eventual
        # last part.
        offsets.append(len(text))
        tokens = []
        start = 0
        for end in offsets:
            if start > end:
                logger.error(
                    "There was a bug in Trie algorithm in tokenization. Attempting to recover. Please report it"
                    " anyway."
                )
                continue
            if start == end:
                # This might happen if there's a match at index 0
                # we're also preventing zero-width cuts in case of two
                # consecutive matches
                continue
            tokens.append(text[start:end])
            start = end

        return tokens


[docs]class LukeTokenizer(PreTrainedTokenizer): """ Constructs a LUKE tokenizer, derived from the GPT-2 tokenizer, using byte-level Byte-Pair-Encoding. This tokenizer has been trained to treat spaces like parts of the tokens (a bit like sentencepiece) so a word will be encoded differently whether it is at the beginning of the sentence (without space) or not: ```python >>> from transformers import LukeTokenizer >>> tokenizer = LukeTokenizer.from_pretrained("studio-ousia/luke-base") >>> tokenizer("Hello world")["input_ids"] [0, 31414, 232, 2] >>> tokenizer(" Hello world")["input_ids"] [0, 20920, 232, 2] ``` You can get around that behavior by passing `add_prefix_space=True` when instantiating this tokenizer or when you call it on some text, but since the model was not pretrained this way, it might yield a decrease in performance. <Tip> When used with `is_split_into_words=True`, this tokenizer will add a space before each word (even the first one). </Tip> This tokenizer inherits from [`PreTrainedTokenizer`] which contains most of the main methods. Users should refer to this superclass for more information regarding those methods. It also creates entity sequences, namely `entity_ids`, `entity_attention_mask`, `entity_token_type_ids`, and `entity_position_ids` to be used by the LUKE model. Args: vocab_file (`str`): Path to the vocabulary file. merges_file (`str`): Path to the merges file. entity_vocab_file (`str`): Path to the entity vocabulary file. task (`str`, *optional*): Task for which you want to prepare sequences. One of `"entity_classification"`, `"entity_pair_classification"`, or `"entity_span_classification"`. If you specify this argument, the entity sequence is automatically created based on the given entity span(s). max_entity_length (`int`, *optional*, defaults to 32): The maximum length of `entity_ids`. max_mention_length (`int`, *optional*, defaults to 30): The maximum number of tokens inside an entity span. entity_token_1 (`str`, *optional*, defaults to `<ent>`): The special token used to represent an entity span in a word token sequence. This token is only used when `task` is set to `"entity_classification"` or `"entity_pair_classification"`. entity_token_2 (`str`, *optional*, defaults to `<ent2>`): The special token used to represent an entity span in a word token sequence. This token is only used when `task` is set to `"entity_pair_classification"`. errors (`str`, *optional*, defaults to `"replace"`): Paradigm to follow when decoding bytes to UTF-8. See [bytes.decode](https://docs.python.org/3/library/stdtypes.html#bytes.decode) for more information. bos_token (`str`, *optional*, defaults to `"<s>"`): The beginning of sequence token that was used during pretraining. Can be used a sequence classifier token. <Tip> When building a sequence using special tokens, this is not the token that is used for the beginning of sequence. The token used is the `cls_token`. </Tip> eos_token (`str`, *optional*, defaults to `"</s>"`): The end of sequence token. <Tip> When building a sequence using special tokens, this is not the token that is used for the end of sequence. The token used is the `sep_token`. </Tip> sep_token (`str`, *optional*, defaults to `"</s>"`): The separator token, which is used when building a sequence from multiple sequences, e.g. two sequences for sequence classification or for a text and a question for question answering. It is also used as the last token of a sequence built with special tokens. cls_token (`str`, *optional*, defaults to `"<s>"`): The classifier token which is used when doing sequence classification (classification of the whole sequence instead of per-token classification). It is the first token of the sequence when built with special tokens. unk_token (`str`, *optional*, defaults to `"<unk>"`): The unknown token. A token that is not in the vocabulary cannot be converted to an ID and is set to be this token instead. pad_token (`str`, *optional*, defaults to `"<pad>"`): The token used for padding, for example when batching sequences of different lengths. mask_token (`str`, *optional*, defaults to `"<mask>"`): The token used for masking values. This is the token used when training this model with masked language modeling. This is the token which the model will try to predict. add_prefix_space (`bool`, *optional*, defaults to `False`): Whether or not to add an initial space to the input. This allows to treat the leading word just as any other word. (LUKE tokenizer detect beginning of words by the preceding space). """ vocab_files_names = VOCAB_FILES_NAMES pretrained_vocab_files_map = PRETRAINED_VOCAB_FILES_MAP max_model_input_sizes = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES model_input_names = ["input_ids", "attention_mask"] def __init__( self, vocab_file, merges_file, entity_vocab_file, task=None, max_entity_length=32, max_mention_length=30, entity_token_1="<ent>", entity_token_2="<ent2>", entity_unk_token="[UNK]", entity_pad_token="[PAD]", entity_mask_token="[MASK]", entity_mask2_token="[MASK2]", errors="replace", bos_token="<s>", eos_token="</s>", sep_token="</s>", cls_token="<s>", unk_token="<unk>", pad_token="<pad>", mask_token="<mask>", add_prefix_space=False, **kwargs, ): bos_token = AddedToken(bos_token, lstrip=False, rstrip=False) if isinstance(bos_token, str) else bos_token eos_token = AddedToken(eos_token, lstrip=False, rstrip=False) if isinstance(eos_token, str) else eos_token sep_token = AddedToken(sep_token, lstrip=False, rstrip=False) if isinstance(sep_token, str) else sep_token cls_token = AddedToken(cls_token, lstrip=False, rstrip=False) if isinstance(cls_token, str) else cls_token unk_token = AddedToken(unk_token, lstrip=False, rstrip=False) if isinstance(unk_token, str) else unk_token pad_token = AddedToken(pad_token, lstrip=False, rstrip=False) if isinstance(pad_token, str) else pad_token # Mask token behave like a normal word, i.e. include the space before it mask_token = AddedToken(mask_token, lstrip=True, rstrip=False) if isinstance(mask_token, str) else mask_token super().__init__( errors=errors, bos_token=bos_token, eos_token=eos_token, unk_token=unk_token, sep_token=sep_token, cls_token=cls_token, pad_token=pad_token, mask_token=mask_token, add_prefix_space=add_prefix_space, task=task, max_entity_length=32, max_mention_length=30, entity_token_1="<ent>", entity_token_2="<ent2>", entity_unk_token=entity_unk_token, entity_pad_token=entity_pad_token, entity_mask_token=entity_mask_token, entity_mask2_token=entity_mask2_token, **kwargs, ) with open(vocab_file, encoding="utf-8") as vocab_handle: self.encoder = json.load(vocab_handle) self.decoder = {v: k for k, v in self.encoder.items()} self.errors = errors # how to handle errors in decoding self.byte_encoder = bytes_to_unicode() self.byte_decoder = {v: k for k, v in self.byte_encoder.items()} with open(merges_file, encoding="utf-8") as merges_handle: bpe_merges = merges_handle.read().split("\n")[1:-1] bpe_merges = [tuple(merge.split()) for merge in bpe_merges] self.bpe_ranks = dict(zip(bpe_merges, range(len(bpe_merges)))) self.cache = {} self.add_prefix_space = add_prefix_space # Should have added re.IGNORECASE so BPE merges can happen for capitalized versions of contractions self.pat = re.compile(r"""'s|'t|'re|'ve|'m|'ll|'d| ?\p{L}+| ?\p{N}+| ?[^\s\p{L}\p{N}]+|\s+(?!\S)|\s+""") # we add 2 special tokens for downstream tasks # for more information about lstrip and rstrip, see https://github.com/huggingface/transformers/pull/2778 entity_token_1 = ( AddedToken(entity_token_1, lstrip=False, rstrip=False) if isinstance(entity_token_1, str) else entity_token_1 ) entity_token_2 = ( AddedToken(entity_token_2, lstrip=False, rstrip=False) if isinstance(entity_token_2, str) else entity_token_2 ) kwargs["additional_special_tokens"] = kwargs.get("additional_special_tokens", []) kwargs["additional_special_tokens"] += [entity_token_1, entity_token_2] with open(entity_vocab_file, encoding="utf-8") as entity_vocab_handle: self.entity_vocab = json.load(entity_vocab_handle) for entity_special_token in [entity_unk_token, entity_pad_token, entity_mask_token, entity_mask2_token]: if entity_special_token not in self.entity_vocab: raise ValueError( f"Specified entity special token ``{entity_special_token}`` is not found in entity_vocab. " f"Probably an incorrect entity vocab file is loaded: {entity_vocab_file}." ) self.entity_unk_token_id = self.entity_vocab[entity_unk_token] self.entity_pad_token_id = self.entity_vocab[entity_pad_token] self.entity_mask_token_id = self.entity_vocab[entity_mask_token] self.entity_mask2_token_id = self.entity_vocab[entity_mask2_token] self.task = task if task is None or task == "entity_span_classification": self.max_entity_length = max_entity_length elif task == "entity_classification": self.max_entity_length = 1 elif task == "entity_pair_classification": self.max_entity_length = 2 else: raise ValueError( f"Task {task} not supported. Select task from ['entity_classification', 'entity_pair_classification'," " 'entity_span_classification'] only." ) self.max_mention_length = max_mention_length
[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 (`str`): The sequence to be encoded. **kwargs (additional keyword arguments): Passed along to the model-specific `prepare_for_tokenization` preprocessing method. Returns: `List[str]`: The list of tokens. """ # Simple mapping string => AddedToken for special tokens with specific tokenization behaviors all_special_tokens_extended = { 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("Keyword arguments %s not recognized.", kwargs) 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.unique_no_split_tokens + 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) no_split_token = set(self.unique_no_split_tokens) tokens = Trie().split(text) # ["This is something", "<special_token_1>", " else"] for i, token in enumerate(tokens): if token in no_split_token: tok_extended = all_special_tokens_extended.get(token, None) left = tokens[i - 1] if i > 0 else None right = tokens[i + 1] if i < len(tokens) - 1 else None if isinstance(tok_extended, AddedToken): if tok_extended.rstrip and right: # 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 tokens[i + 1] = right.lstrip() # Strip white spaces on the left if tok_extended.lstrip and left: tokens[i - 1] = left.rstrip() # Opposite here else: # We strip left and right by default if right: tokens[i + 1] = right.lstrip() if left: tokens[i - 1] = left.rstrip() # ["This is something", "<special_token_1>", "else"] tokenized_text = [] for token in tokens: # Need to skip eventual empty (fully stripped) tokens if not token: continue if token in no_split_token: tokenized_text.append(token) else: tokenized_text.extend(self._tokenize(token)) # ["This", " is", " something", "<special_token_1>", "else"] return tokenized_text
def _convert_token_to_id(self, token): """Converts a token (str) in an id using the vocab.""" return self.encoder.get(token, self.encoder.get(self.unk_token)) def _convert_id_to_token(self, index: int) -> str: pass # Copied from transformers.models.roberta.tokenization_roberta.RobertaTokenizer.bpe with Roberta->Luke, RoBERTa->LUKE
[docs] def bpe(self, token): """ bpe """ if token in self.cache: return self.cache[token] word = tuple(token) pairs = get_pairs(word) if not pairs: return token while True: bigram = min(pairs, key=lambda pair: self.bpe_ranks.get(pair, float("inf"))) if bigram not in self.bpe_ranks: break first, second = bigram new_word = [] i = 0 while i < len(word): try: j = word.index(first, i) except ValueError: new_word.extend(word[i:]) break else: new_word.extend(word[i:j]) i = j if word[i] == first and i < len(word) - 1 and word[i + 1] == second: new_word.append(first + second) i += 2 else: new_word.append(word[i]) i += 1 new_word = tuple(new_word) word = new_word if len(word) == 1: break # else: pairs = get_pairs(word) word = " ".join(word) self.cache[token] = word return word
# Copied from transformers.models.roberta.tokenization_roberta.RobertaTokenizer._tokenize with Roberta->Luke, RoBERTa->LUKE def _tokenize(self, text): """Tokenize a string.""" bpe_tokens = [] for token in re.findall(self.pat, text): token = "".join( self.byte_encoder[b] for b in token.encode("utf-8") ) # Maps all our bytes to unicode strings, avoiding control tokens of the BPE (spaces in our case) bpe_tokens.extend(bpe_token for bpe_token in self.bpe(token).split(" ")) return bpe_tokens
[docs] def create_token_type_ids_from_sequences( self, token_ids_0: List[int], token_ids_1: Optional[List[int]] = None ) -> List[int]: """ Create a mask from the two sequences passed to be used in a sequence-pair classification task. LUKE does not make use of token type ids, therefore a list of zeros is returned. Args: token_ids_0 (`List[int]`): List of IDs. token_ids_1 (`List[int]`, *optional*): Optional second list of IDs for sequence pairs. Returns: `List[int]`: List of zeros. """ sep = [self.sep_token_id] cls = [self.cls_token_id] if token_ids_1 is None: return len(cls + token_ids_0 + sep) * [0] return len(cls + token_ids_0 + sep + sep + token_ids_1 + sep) * [0]
# Copied from transformers.models.roberta.tokenization_roberta.RobertaTokenizer.prepare_for_tokenization with Roberta->Luke, RoBERTa->LUKE
[docs] def prepare_for_tokenization(self, text, is_split_into_words=False, **kwargs): """ prepare_for_tokenization """ add_prefix_space = kwargs.pop("add_prefix_space", self.add_prefix_space) if (is_split_into_words or add_prefix_space) and (len(text) > 0 and not text[0].isspace()): text = " " + text return (text, kwargs)
def __call__( self, text: Union[TextInput, List[TextInput]], text_pair: Optional[Union[TextInput, List[TextInput]]] = None, entity_spans: Optional[Union[EntitySpanInput, List[EntitySpanInput]]] = None, entity_spans_pair: Optional[Union[EntitySpanInput, List[EntitySpanInput]]] = None, entities: Optional[Union[EntityInput, List[EntityInput]]] = None, entities_pair: Optional[Union[EntityInput, List[EntityInput]]] = None, add_special_tokens: bool = True, padding: Union[bool, str, PaddingStrategy] = False, truncation: Union[bool, str, TruncationStrategy] = None, max_length: Optional[int] = None, max_entity_length: Optional[int] = None, stride: int = 0, is_split_into_words: Optional[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, ): """ Main method to tokenize and prepare for the model one or several sequence(s) or one or several pair(s) of sequences, depending on the task you want to prepare them for. Args: text (`str`, `List[str]`, `List[List[str]]`): The sequence or batch of sequences to be encoded. Each sequence must be a string. Note that this tokenizer does not support tokenization based on pretokenized strings. text_pair (`str`, `List[str]`, `List[List[str]]`): The sequence or batch of sequences to be encoded. Each sequence must be a string. Note that this tokenizer does not support tokenization based on pretokenized strings. entity_spans (`List[Tuple[int, int]]`, `List[List[Tuple[int, int]]]`, *optional*): The sequence or batch of sequences of entity spans to be encoded. Each sequence consists of tuples each with two integers denoting character-based start and end positions of entities. If you specify `"entity_classification"` or `"entity_pair_classification"` as the `task` argument in the constructor, the length of each sequence must be 1 or 2, respectively. If you specify `entities`, the length of each sequence must be equal to the length of each sequence of `entities`. entity_spans_pair (`List[Tuple[int, int]]`, `List[List[Tuple[int, int]]]`, *optional*): The sequence or batch of sequences of entity spans to be encoded. Each sequence consists of tuples each with two integers denoting character-based start and end positions of entities. If you specify the `task` argument in the constructor, this argument is ignored. If you specify `entities_pair`, the length of each sequence must be equal to the length of each sequence of `entities_pair`. entities (`List[str]`, `List[List[str]]`, *optional*): The sequence or batch of sequences of entities to be encoded. Each sequence consists of strings representing entities, i.e., special entities (e.g., [MASK]) or entity titles of Wikipedia (e.g., Los Angeles). This argument is ignored if you specify the `task` argument in the constructor. The length of each sequence must be equal to the length of each sequence of `entity_spans`. If you specify `entity_spans` without specifying this argument, the entity sequence or the batch of entity sequences is automatically constructed by filling it with the [MASK] entity. entities_pair (`List[str]`, `List[List[str]]`, *optional*): The sequence or batch of sequences of entities to be encoded. Each sequence consists of strings representing entities, i.e., special entities (e.g., [MASK]) or entity titles of Wikipedia (e.g., Los Angeles). This argument is ignored if you specify the `task` argument in the constructor. The length of each sequence must be equal to the length of each sequence of `entity_spans_pair`. If you specify `entity_spans_pair` without specifying this argument, the entity sequence or the batch of entity sequences is automatically constructed by filling it with the [MASK] entity. max_entity_length (`int`, *optional*): The maximum length of `entity_ids`. """ # Input type checking for clearer error is_valid_single_text = isinstance(text, str) is_valid_batch_text = isinstance(text, (list, tuple)) and (len(text) == 0 or (isinstance(text[0], str))) if not (is_valid_single_text or is_valid_batch_text): raise ValueError("text input must be of type `str` (single example) or `List[str]` (batch).") is_valid_single_text_pair = isinstance(text_pair, str) is_valid_batch_text_pair = isinstance(text_pair, (list, tuple)) and ( len(text_pair) == 0 or isinstance(text_pair[0], str) ) if not (text_pair is None or is_valid_single_text_pair or is_valid_batch_text_pair): raise ValueError("text_pair input must be of type `str` (single example) or `List[str]` (batch).") is_batched = bool(isinstance(text, (list, tuple))) if is_batched: batch_text_or_text_pairs = list(zip(text, text_pair)) if text_pair is not None else text if entities is None: batch_entities_or_entities_pairs = None else: batch_entities_or_entities_pairs = ( list(zip(entities, entities_pair)) if entities_pair is not None else entities ) if entity_spans is None: batch_entity_spans_or_entity_spans_pairs = None else: batch_entity_spans_or_entity_spans_pairs = ( list(zip(entity_spans, entity_spans_pair)) if entity_spans_pair is not None else entity_spans ) return self.batch_encode_plus( batch_text_or_text_pairs=batch_text_or_text_pairs, batch_entity_spans_or_entity_spans_pairs=batch_entity_spans_or_entity_spans_pairs, batch_entities_or_entities_pairs=batch_entities_or_entities_pairs, add_special_tokens=add_special_tokens, padding=padding, truncation=truncation, max_length=max_length, max_entity_length=max_entity_length, stride=stride, is_split_into_words=is_split_into_words, pad_to_multiple_of=pad_to_multiple_of, return_tensors=return_tensors, return_token_type_ids=return_token_type_ids, return_attention_mask=return_attention_mask, return_overflowing_tokens=return_overflowing_tokens, return_special_tokens_mask=return_special_tokens_mask, return_offsets_mapping=return_offsets_mapping, return_length=return_length, verbose=verbose, **kwargs, ) return self.encode_plus( text=text, text_pair=text_pair, entity_spans=entity_spans, entity_spans_pair=entity_spans_pair, entities=entities, entities_pair=entities_pair, add_special_tokens=add_special_tokens, padding=padding, truncation=truncation, max_length=max_length, max_entity_length=max_entity_length, stride=stride, is_split_into_words=is_split_into_words, pad_to_multiple_of=pad_to_multiple_of, return_tensors=return_tensors, return_token_type_ids=return_token_type_ids, return_attention_mask=return_attention_mask, return_overflowing_tokens=return_overflowing_tokens, return_special_tokens_mask=return_special_tokens_mask, return_offsets_mapping=return_offsets_mapping, return_length=return_length, verbose=verbose, **kwargs, ) def _create_input_sequence( self, text: Union[TextInput], text_pair: Optional[Union[TextInput]] = None, entities: Optional[EntityInput] = None, entities_pair: Optional[EntityInput] = None, entity_spans: Optional[EntitySpanInput] = None, entity_spans_pair: Optional[EntitySpanInput] = None, **kwargs, ) -> Tuple[list, list, list, list, list, list]: def get_input_ids(text): tokens = self.tokenize_(text, **kwargs) return self.convert_tokens_to_ids(tokens) def get_input_ids_and_entity_token_spans(text, entity_spans): if entity_spans is None: return get_input_ids(text), None cur = 0 input_ids = [] entity_token_spans = [None] * len(entity_spans) split_char_positions = sorted(frozenset(itertools.chain(*entity_spans))) char_pos2token_pos = {} for split_char_position in split_char_positions: orig_split_char_position = split_char_position if ( split_char_position > 0 and text[split_char_position - 1] == " " ): # whitespace should be prepended to the following token split_char_position -= 1 if cur != split_char_position: input_ids += get_input_ids(text[cur:split_char_position]) cur = split_char_position char_pos2token_pos[orig_split_char_position] = len(input_ids) input_ids += get_input_ids(text[cur:]) entity_token_spans = [ (char_pos2token_pos[char_start], char_pos2token_pos[char_end]) for char_start, char_end in entity_spans ] return input_ids, entity_token_spans first_ids, second_ids = None, None first_entity_ids, second_entity_ids = None, None first_entity_token_spans, second_entity_token_spans = None, None if self.task is None: if entity_spans is None: first_ids = get_input_ids(text) else: self._check_entity_input_format(entities, entity_spans) first_ids, first_entity_token_spans = get_input_ids_and_entity_token_spans(text, entity_spans) if entities is None: first_entity_ids = [self.entity_mask_token_id] * len(entity_spans) else: first_entity_ids = [self.entity_vocab.get(entity, self.entity_unk_token_id) for entity in entities] if text_pair is not None: if entity_spans_pair is None: second_ids = get_input_ids(text_pair) else: self._check_entity_input_format(entities_pair, entity_spans_pair) second_ids, second_entity_token_spans = get_input_ids_and_entity_token_spans( text_pair, entity_spans_pair ) if entities_pair is None: second_entity_ids = [self.entity_mask_token_id] * len(entity_spans_pair) else: second_entity_ids = [ self.entity_vocab.get(entity, self.entity_unk_token_id) for entity in entities_pair ] elif self.task == "entity_classification": if not (isinstance(entity_spans, list) and len(entity_spans) == 1 and isinstance(entity_spans[0], tuple)): raise ValueError( "Entity spans should be a list containing a single tuple " "containing the start and end character indices of an entity" ) first_entity_ids = [self.entity_mask_token_id] first_ids, first_entity_token_spans = get_input_ids_and_entity_token_spans(text, entity_spans) # add special tokens to input ids entity_token_start, entity_token_end = first_entity_token_spans[0] first_ids = ( first_ids[:entity_token_end] + [self.additional_special_tokens_ids[0]] + first_ids[ entity_token_end:] ) first_ids = ( first_ids[:entity_token_start] + [self.additional_special_tokens_ids[0]] + first_ids[entity_token_start:] ) first_entity_token_spans = [(entity_token_start, entity_token_end + 2)] elif self.task == "entity_pair_classification": if not ( isinstance(entity_spans, list) and len(entity_spans) == 2 and isinstance(entity_spans[0], tuple) and isinstance(entity_spans[1], tuple) ): raise ValueError( "Entity spans should be provided as a list of two tuples, " "each tuple containing the start and end character indices of an entity" ) # head_span, tail_span = entity_spans first_entity_ids = [self.entity_mask_token_id, self.entity_mask2_token_id] first_ids, first_entity_token_spans = get_input_ids_and_entity_token_spans(text, entity_spans) head_token_span, tail_token_span = first_entity_token_spans token_span_with_special_token_ids = [ (head_token_span, self.additional_special_tokens_ids[0]), (tail_token_span, self.additional_special_tokens_ids[1]), ] if head_token_span[0] < tail_token_span[0]: first_entity_token_spans[0] = (head_token_span[0], head_token_span[1] + 2) first_entity_token_spans[1] = (tail_token_span[0] + 2, tail_token_span[1] + 4) token_span_with_special_token_ids = reversed(token_span_with_special_token_ids) else: first_entity_token_spans[0] = (head_token_span[0] + 2, head_token_span[1] + 4) first_entity_token_spans[1] = (tail_token_span[0], tail_token_span[1] + 2) for (entity_token_start, entity_token_end), special_token_id in token_span_with_special_token_ids: first_ids = first_ids[:entity_token_end] + [special_token_id] + first_ids[entity_token_end:] first_ids = first_ids[:entity_token_start] + [special_token_id] + first_ids[entity_token_start:] elif self.task == "entity_span_classification": if not (isinstance(entity_spans, list) and len(entity_spans) > 0 and isinstance(entity_spans[0], tuple)): raise ValueError( "Entity spans should be provided as a list of tuples, " "each tuple containing the start and end character indices of an entity" ) first_ids, first_entity_token_spans = get_input_ids_and_entity_token_spans(text, entity_spans) first_entity_ids = [self.entity_mask_token_id] * len(entity_spans) else: raise ValueError(f"Task {self.task} not supported") return ( first_ids, second_ids, first_entity_ids, second_entity_ids, first_entity_token_spans, second_entity_token_spans, ) def _get_padding_truncation_strategies( self, padding=False, truncation=None, max_length=None, pad_to_multiple_of=None, verbose=True, **kwargs ): """ Find the correct padding/truncation strategy with backward compatibility for old arguments (truncation_strategy and pad_to_max_length) and behaviors. """ old_truncation_strategy = kwargs.pop("truncation_strategy", "do_not_truncate") old_pad_to_max_length = kwargs.pop("pad_to_max_length", False) # Backward compatibility for previous behavior, maybe we should deprecate it: # If you only set max_length, it activates truncation for max_length if max_length is not None and padding is False and truncation is None: if verbose: if not self.deprecation_warnings.get("Truncation-not-explicitly-activated", False): logger.warning( "Truncation was not explicitly activated but `max_length` is provided a specific value, please" " use `truncation=True` to explicitly truncate examples to max length. Defaulting to" " 'longest_first' truncation strategy. If you encode pairs of sequences (GLUE-style) with the" " tokenizer you can select this strategy more precisely by providing a specific strategy to" " `truncation`." ) self.deprecation_warnings["Truncation-not-explicitly-activated"] = True truncation = "longest_first" # Get padding strategy if padding is False and old_pad_to_max_length: if verbose: warnings.warn( "The `pad_to_max_length` argument is deprecated and will be removed in a future version, " "use `padding=True` or `padding='longest'` to pad to the longest sequence in the batch, or " "use `padding='max_length'` to pad to a max length. In this case, you can give a specific " "length with `max_length` (e.g. `max_length=45`) or leave max_length to None to pad to the " "maximal input size of the model (e.g. 512 for Bert).", FutureWarning, ) if max_length is None: padding_strategy = PaddingStrategy.LONGEST else: padding_strategy = PaddingStrategy.MAX_LENGTH elif padding is not False: if padding is True: if verbose: if max_length is not None and ( truncation is None or truncation is False or truncation == "do_not_truncate" ): warnings.warn( "`max_length` is ignored when `padding`=`True` and there is no truncation strategy. " "To pad to max length, use `padding='max_length'`." ) if old_pad_to_max_length is not False: warnings.warn("Though `pad_to_max_length` = `True`, it is ignored because `padding`=`True`.") padding_strategy = PaddingStrategy.LONGEST # Default to pad to the longest sequence in the batch elif not isinstance(padding, PaddingStrategy): padding_strategy = PaddingStrategy(padding) elif isinstance(padding, PaddingStrategy): padding_strategy = padding else: padding_strategy = PaddingStrategy.DO_NOT_PAD # Get truncation strategy if truncation is None and old_truncation_strategy != "do_not_truncate": if verbose: warnings.warn( "The `truncation_strategy` argument is deprecated and will be removed in a future version, use" " `truncation=True` to truncate examples to a max length. You can give a specific length with" " `max_length` (e.g. `max_length=45`) or leave max_length to None to truncate to the maximal input" " size of the model (e.g. 512 for Bert). If you have pairs of inputs, you can give a specific" " truncation strategy selected among `truncation='only_first'` (will only truncate the first" " sentence in the pairs) `truncation='only_second'` (will only truncate the second sentence in the" " pairs) or `truncation='longest_first'` (will iteratively remove tokens from the longest sentence" " in the pairs).", FutureWarning, ) truncation_strategy = TruncationStrategy(old_truncation_strategy) elif truncation is not False and truncation is not None: if truncation is True: truncation_strategy = ( TruncationStrategy.LONGEST_FIRST ) # Default to truncate the longest sequences in pairs of inputs elif not isinstance(truncation, TruncationStrategy): truncation_strategy = TruncationStrategy(truncation) elif isinstance(truncation, TruncationStrategy): truncation_strategy = truncation else: truncation_strategy = TruncationStrategy.DO_NOT_TRUNCATE # Set max length if needed if max_length is None: if padding_strategy == PaddingStrategy.MAX_LENGTH: if self.model_max_length > LARGE_INTEGER: if verbose: if not self.deprecation_warnings.get("Asking-to-pad-to-max_length", False): logger.warning( "Asking to pad to max_length but no maximum length is provided and the model has no" " predefined maximum length. Default to no padding." ) self.deprecation_warnings["Asking-to-pad-to-max_length"] = True padding_strategy = PaddingStrategy.DO_NOT_PAD else: max_length = self.model_max_length if truncation_strategy != TruncationStrategy.DO_NOT_TRUNCATE: if self.model_max_length > LARGE_INTEGER: if verbose: if not self.deprecation_warnings.get("Asking-to-truncate-to-max_length", False): logger.warning( "Asking to truncate to max_length but no maximum length is provided and the model has" " no predefined maximum length. Default to no truncation." ) self.deprecation_warnings["Asking-to-truncate-to-max_length"] = True truncation_strategy = TruncationStrategy.DO_NOT_TRUNCATE else: max_length = self.model_max_length # Test if we have a padding token if padding_strategy != PaddingStrategy.DO_NOT_PAD and (not self.pad_token or self.pad_token_id < 0): raise ValueError( "Asking to pad but the tokenizer does not have a padding token. " "Please select a token to use as `pad_token` `(tokenizer.pad_token = tokenizer.eos_token e.g.)` " "or add a new pad token via `tokenizer.add_special_tokens({'pad_token': '[PAD]'})`." ) # Check that we will truncate to a multiple of pad_to_multiple_of if both are provided if ( truncation_strategy != TruncationStrategy.DO_NOT_TRUNCATE and padding_strategy != PaddingStrategy.DO_NOT_PAD and pad_to_multiple_of is not None and max_length is not None and (max_length % pad_to_multiple_of != 0) ): raise ValueError( "Truncation and padding are both activated but " f"truncation length ({max_length}) is not a multiple of pad_to_multiple_of ({pad_to_multiple_of})." ) return padding_strategy, truncation_strategy, max_length, kwargs
[docs] def encode_plus( self, text: Union[TextInput, PreTokenizedInput, EncodedInput], text_pair: Optional[Union[TextInput, PreTokenizedInput, EncodedInput]] = None, add_special_tokens: bool = True, padding: Union[bool, str, PaddingStrategy] = False, truncation: Union[bool, str, TruncationStrategy] = None, 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, ): """ Tokenize and prepare for the model a sequence or a pair of sequences. <Tip warning={true}> This method is deprecated, `__call__` should be used instead. </Tip> Args: text (`str`, `List[str]` or `List[int]` (the latter only for not-fast tokenizers)): The first sequence to be encoded. This can be a string, a list of strings (tokenized string using the `tokenize` method) or a list of integers (tokenized string ids using the `convert_tokens_to_ids` method). text_pair (`str`, `List[str]` or `List[int]`, *optional*): Optional second sequence to be encoded. This can be a string, a list of strings (tokenized string using the `tokenize` method) or a list of integers (tokenized string ids using the `convert_tokens_to_ids` method). """ # Backward compatibility for 'truncation_strategy', 'pad_to_max_length' padding_strategy, truncation_strategy, max_length, kwargs = self._get_padding_truncation_strategies( padding=padding, truncation=truncation, max_length=max_length, pad_to_multiple_of=pad_to_multiple_of, verbose=verbose, **kwargs, ) return self._encode_plus( text=text, text_pair=text_pair, add_special_tokens=add_special_tokens, padding_strategy=padding_strategy, truncation_strategy=truncation_strategy, max_length=max_length, stride=stride, is_split_into_words=is_split_into_words, pad_to_multiple_of=pad_to_multiple_of, return_tensors=return_tensors, return_token_type_ids=return_token_type_ids, return_attention_mask=return_attention_mask, return_overflowing_tokens=return_overflowing_tokens, return_special_tokens_mask=return_special_tokens_mask, return_offsets_mapping=return_offsets_mapping, return_length=return_length, verbose=verbose, **kwargs, )
def _encode_plus( self, text: Union[TextInput], text_pair: Optional[Union[TextInput]] = None, entity_spans: Optional[EntitySpanInput] = None, entity_spans_pair: Optional[EntitySpanInput] = None, entities: Optional[EntityInput] = None, entities_pair: Optional[EntityInput] = 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, max_entity_length: Optional[int] = None, stride: int = 0, is_split_into_words: Optional[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, ): 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" ) if is_split_into_words: raise NotImplementedError("is_split_into_words is not supported in this tokenizer.") ( first_ids, second_ids, first_entity_ids, second_entity_ids, first_entity_token_spans, second_entity_token_spans, ) = self._create_input_sequence( text=text, text_pair=text_pair, entities=entities, entities_pair=entities_pair, entity_spans=entity_spans, entity_spans_pair=entity_spans_pair, **kwargs, ) # prepare_for_model will create the attention_mask and token_type_ids return self.prepare_for_model( first_ids, pair_ids=second_ids, entity_ids=first_entity_ids, pair_entity_ids=second_entity_ids, entity_token_spans=first_entity_token_spans, pair_entity_token_spans=second_entity_token_spans, add_special_tokens=add_special_tokens, padding=padding_strategy.value, truncation=truncation_strategy.value, max_length=max_length, max_entity_length=max_entity_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, )
[docs] def build_inputs_with_special_tokens( self, token_ids_0: List[int], token_ids_1: Optional[List[int]] = None ) -> List[int]: """ Build model inputs from a sequence or a pair of sequence for sequence classification tasks by concatenating and adding special tokens. A LUKE sequence has the following format: - single sequence: `<s> X </s>` - pair of sequences: `<s> A </s></s> B </s>` Args: token_ids_0 (`List[int]`): List of IDs to which the special tokens will be added. token_ids_1 (`List[int]`, *optional*): Optional second list of IDs for sequence pairs. Returns: `List[int]`: List of [input IDs](../glossary#input-ids) with the appropriate special tokens. """ if token_ids_1 is None: return [self.cls_token_id] + token_ids_0 + [self.sep_token_id] cls = [self.cls_token_id] sep = [self.sep_token_id] return cls + token_ids_0 + sep + sep + token_ids_1 + sep
[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. <Tip> 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. </Tip> Args: pair (`bool`, *optional*, defaults to `False`): Whether the number of added tokens should be computed in the case of a sequence pair or a single sequence. Returns: `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 prepare_for_model( self, ids: List[int], pair_ids: Optional[List[int]] = None, entity_ids: Optional[List[int]] = None, pair_entity_ids: Optional[List[int]] = None, entity_token_spans: Optional[List[Tuple[int, int]]] = None, pair_entity_token_spans: Optional[List[Tuple[int, int]]] = None, add_special_tokens: bool = True, padding: Union[bool, str, PaddingStrategy] = False, truncation: Union[bool, str, TruncationStrategy] = None, max_length: Optional[int] = None, max_entity_length: Optional[int] = None, stride: int = 0, pad_to_multiple_of: Optional[int] = 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, **kwargs, ): """ Prepares a sequence of input id, entity id and entity span, or a pair of sequences of inputs ids, entity ids, entity spans 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. Please Note, for *pair_ids* different than `None` and *truncation_strategy = longest_first* or `True`, it is not possible to return overflowing tokens. Such a combination of arguments will raise an error. Args: ids (`List[int]`): Tokenized input ids of the first sequence. pair_ids (`List[int]`, *optional*): Tokenized input ids of the second sequence. entity_ids (`List[int]`, *optional*): Entity ids of the first sequence. pair_entity_ids (`List[int]`, *optional*): Entity ids of the second sequence. entity_token_spans (`List[Tuple[int, int]]`, *optional*): Entity spans of the first sequence. pair_entity_token_spans (`List[Tuple[int, int]]`, *optional*): Entity spans of the second sequence. max_entity_length (`int`, *optional*): The maximum length of the entity sequence. """ # Backward compatibility for 'truncation_strategy', 'pad_to_max_length' padding_strategy, truncation_strategy, max_length, kwargs = self._get_padding_truncation_strategies( padding=padding, truncation=truncation, max_length=max_length, pad_to_multiple_of=pad_to_multiple_of, verbose=verbose, **kwargs, ) # Compute lengths pair = bool(pair_ids is not None) len_ids = len(ids) len_pair_ids = len(pair_ids) if pair else 0 if return_token_type_ids and not add_special_tokens: raise ValueError( "Asking to return token_type_ids while setting add_special_tokens to False " "results in an undefined behavior. Please set add_special_tokens to True or " "set return_token_type_ids to None." ) if ( return_overflowing_tokens and truncation_strategy == TruncationStrategy.LONGEST_FIRST and pair_ids is not None ): raise ValueError( "Not possible to return overflowing tokens for pair of sequences with the " "`longest_first`. Please select another truncation strategy than `longest_first`, " "for instance `only_second` or `only_first`." ) # Load from model defaults if return_token_type_ids is None: return_token_type_ids = "token_type_ids" in self.model_input_names if return_attention_mask is None: return_attention_mask = "attention_mask" in self.model_input_names encoded_inputs = {} # Compute the total size of the returned word encodings total_len = len_ids + len_pair_ids + (self.num_special_tokens_to_add(pair=pair) if add_special_tokens else 0) # Truncation: Handle max sequence length and max_entity_length overflowing_tokens = [] if truncation_strategy != TruncationStrategy.DO_NOT_TRUNCATE and max_length and total_len > max_length: # truncate words up to max_length ids, pair_ids, overflowing_tokens = self.truncate_sequences( ids, pair_ids=pair_ids, num_tokens_to_remove=total_len - max_length, truncation_strategy=truncation_strategy, stride=stride, ) if return_overflowing_tokens: encoded_inputs["overflowing_tokens"] = overflowing_tokens encoded_inputs["num_truncated_tokens"] = total_len - max_length # Add special tokens if add_special_tokens: sequence = self.build_inputs_with_special_tokens(ids, pair_ids) token_type_ids = self.create_token_type_ids_from_sequences(ids, pair_ids) entity_token_offset = 1 # 1 * <s> token pair_entity_token_offset = len(ids) + 3 # 1 * <s> token & 2 * <sep> tokens else: sequence = ids + pair_ids if pair else ids token_type_ids = [0] * len(ids) + ([0] * len(pair_ids) if pair else []) entity_token_offset = 0 pair_entity_token_offset = len(ids) # Build output dictionary encoded_inputs["input_ids"] = sequence if return_token_type_ids: encoded_inputs["token_type_ids"] = token_type_ids if return_special_tokens_mask: if add_special_tokens: encoded_inputs["special_tokens_mask"] = self.get_special_tokens_mask(ids, pair_ids) else: encoded_inputs["special_tokens_mask"] = [0] * len(sequence) # Set max entity length if not max_entity_length: max_entity_length = self.max_entity_length if entity_ids is not None: total_entity_len = 0 num_invalid_entities = 0 valid_entity_ids = [ent_id for ent_id, span in zip(entity_ids, entity_token_spans) if span[1] <= len(ids)] valid_entity_token_spans = [span for span in entity_token_spans if span[1] <= len(ids)] total_entity_len += len(valid_entity_ids) num_invalid_entities += len(entity_ids) - len(valid_entity_ids) valid_pair_entity_ids, valid_pair_entity_token_spans = None, None if pair_entity_ids is not None: valid_pair_entity_ids = [ ent_id for ent_id, span in zip(pair_entity_ids, pair_entity_token_spans) if span[1] <= len(pair_ids) ] valid_pair_entity_token_spans = [span for span in pair_entity_token_spans if span[1] <= len(pair_ids)] total_entity_len += len(valid_pair_entity_ids) num_invalid_entities += len(pair_entity_ids) - len(valid_pair_entity_ids) if num_invalid_entities != 0: logger.warning( "%d entities are ignored because their entity spans are " "invalid due to the truncation of input tokens", num_invalid_entities ) if truncation_strategy != TruncationStrategy.DO_NOT_TRUNCATE and total_entity_len > max_entity_length: # truncate entities up to max_entity_length valid_entity_ids, valid_pair_entity_ids, overflowing_entities = self.truncate_sequences( valid_entity_ids, pair_ids=valid_pair_entity_ids, num_tokens_to_remove=total_entity_len - max_entity_length, truncation_strategy=truncation_strategy, stride=stride, ) valid_entity_token_spans = valid_entity_token_spans[: len(valid_entity_ids)] if valid_pair_entity_token_spans is not None: valid_pair_entity_token_spans = valid_pair_entity_token_spans[: len(valid_pair_entity_ids)] if return_overflowing_tokens: encoded_inputs["overflowing_entities"] = overflowing_entities encoded_inputs["num_truncated_entities"] = total_entity_len - max_entity_length final_entity_ids = valid_entity_ids + valid_pair_entity_ids if valid_pair_entity_ids else valid_entity_ids encoded_inputs["entity_ids"] = list(final_entity_ids) entity_position_ids = [] entity_start_positions = [] entity_end_positions = [] for token_spans, offset in ( (valid_entity_token_spans, entity_token_offset), (valid_pair_entity_token_spans, pair_entity_token_offset), ): if token_spans is not None: for start, end in token_spans: start += offset end += offset position_ids = list(range(start, end))[: self.max_mention_length] position_ids += [-1] * (self.max_mention_length - end + start) entity_position_ids.append(position_ids) entity_start_positions.append(start) entity_end_positions.append(end - 1) encoded_inputs["entity_position_ids"] = entity_position_ids if self.task == "entity_span_classification": encoded_inputs["entity_start_positions"] = entity_start_positions encoded_inputs["entity_end_positions"] = entity_end_positions if return_token_type_ids: encoded_inputs["entity_token_type_ids"] = [0] * len(encoded_inputs["entity_ids"]) # Padding if padding_strategy != PaddingStrategy.DO_NOT_PAD or return_attention_mask: encoded_inputs = self.pad( encoded_inputs, max_length=max_length, max_entity_length=max_entity_length, padding=padding_strategy.value, pad_to_multiple_of=pad_to_multiple_of, return_attention_mask=return_attention_mask, ) if return_length: encoded_inputs["length"] = len(encoded_inputs["input_ids"]) return encoded_inputs
[docs] def pad( self, encoded_inputs, padding: Union[bool, str, PaddingStrategy] = True, max_length: Optional[int] = None, max_entity_length: Optional[int] = None, pad_to_multiple_of: Optional[int] = None, return_attention_mask: Optional[bool] = None, verbose: bool = True, ): """ Pad a single encoded input or a batch of encoded inputs up to predefined length or to the max sequence length in the batch. Padding side (left/right) padding token ids are defined at the tokenizer level (with `self.padding_side`, `self.pad_token_id` and `self.pad_token_type_id`) .. note:: If the `encoded_inputs` passed are dictionary of numpy arrays, PyTorch tensors or TensorFlow tensors, the result will use the same type unless you provide a different tensor type with `return_tensors`. In the case of PyTorch tensors, you will lose the specific device of your tensors however. Args: encoded_inputs ([`BatchEncoding`], list of [`BatchEncoding`], `Dict[str, List[int]]`, `Dict[str, List[List[int]]` or `List[Dict[str, List[int]]]`): Tokenized inputs. Can represent one input ([`BatchEncoding`] or `Dict[str, List[int]]`) or a batch of tokenized inputs (list of [`BatchEncoding`], *Dict[str, List[List[int]]]* or *List[Dict[str, List[int]]]*) so you can use this method during preprocessing as well as in a PyTorch Dataloader collate function. Instead of `List[int]` you can have tensors (numpy arrays, PyTorch tensors or TensorFlow tensors), see the note above for the return type. padding (`bool`, `str` or [`~utils.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). max_entity_length (`int`, *optional*): The maximum length of the entity sequence. 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). return_attention_mask (`bool`, *optional*): Whether to return the attention mask. If left to the default, will return the attention mask according to the specific tokenizer's default, defined by the `return_outputs` attribute. [What are attention masks?](../glossary#attention-mask) return_tensors (`str` or [`~utils.TensorType`], *optional*): If set, will return tensors instead of list of python integers. Acceptable values are: - `'tf'`: Return TensorFlow `tf.constant` objects. - `'pt'`: Return PyTorch `torch.Tensor` objects. - `'np'`: Return Numpy `np.ndarray` objects. verbose (`bool`, *optional*, defaults to `True`): Whether or not to print more information and warnings. """ # If we have a list of dicts, let's convert it in a dict of lists # We do this to allow using this method as a collate_fn function in PyTorch Dataloader if isinstance(encoded_inputs, (list, tuple)) and isinstance(encoded_inputs[0], Mapping): encoded_inputs = {key: [example[key] for example in encoded_inputs] for key in encoded_inputs[0].keys()} # The model's main input name, usually `input_ids`, has be passed for padding if self.model_input_names[0] not in encoded_inputs: raise ValueError( "You should supply an encoding or a list of encodings to this method " f"that includes {self.model_input_names[0]}, but you provided {list(encoded_inputs.keys())}" ) required_input = encoded_inputs[self.model_input_names[0]] if not required_input: if return_attention_mask: encoded_inputs["attention_mask"] = [] return encoded_inputs # If we have PyTorch/TF/NumPy tensors/arrays as inputs, we cast them as python objects # and rebuild them afterwards if no return_tensors is specified # Note that we lose the specific device the tensor may be on for PyTorch # first_element = required_input[0] # Convert padding_strategy in PaddingStrategy padding_strategy, _, max_length, _ = self._get_padding_truncation_strategies( padding=padding, max_length=max_length, verbose=verbose ) if max_entity_length is None: max_entity_length = self.max_entity_length required_input = encoded_inputs[self.model_input_names[0]] if required_input and not isinstance(required_input[0], (list, tuple)): encoded_inputs = self._pad( encoded_inputs, max_length=max_length, max_entity_length=max_entity_length, padding_strategy=padding_strategy, pad_to_multiple_of=pad_to_multiple_of, return_attention_mask=return_attention_mask, ) return encoded_inputs batch_size = len(required_input) if any(len(v) != batch_size for v in encoded_inputs.values()): raise ValueError("Some items in the output dictionary have a different batch size than others.") if padding_strategy == PaddingStrategy.LONGEST: max_length = max(len(inputs) for inputs in required_input) max_entity_length = ( max(len(inputs) for inputs in encoded_inputs["entity_ids"]) if "entity_ids" in encoded_inputs else 0 ) padding_strategy = PaddingStrategy.MAX_LENGTH batch_outputs = {} for i in range(batch_size): inputs = {k: v[i] for k, v in encoded_inputs.items()} outputs = self._pad( inputs, max_length=max_length, max_entity_length=max_entity_length, padding_strategy=padding_strategy, pad_to_multiple_of=pad_to_multiple_of, return_attention_mask=return_attention_mask, ) for key, value in outputs.items(): if key not in batch_outputs: batch_outputs[key] = [] batch_outputs[key].append(value) return batch_outputs
def _pad( self, encoded_inputs, max_length: Optional[int] = None, max_entity_length: Optional[int] = None, padding_strategy: PaddingStrategy = PaddingStrategy.DO_NOT_PAD, pad_to_multiple_of: Optional[int] = None, return_attention_mask: Optional[bool] = None, ) -> dict: """ Pad encoded inputs (on left/right and up to predefined length or max length in the batch) Args: encoded_inputs: Dictionary of tokenized inputs (`List[int]`) or batch of tokenized inputs (`List[List[int]]`). max_length: maximum length of the returned list and optionally padding length (see below). Will truncate by taking into account the special tokens. max_entity_length: The maximum length of the entity sequence. padding_strategy: PaddingStrategy to use for padding. - PaddingStrategy.LONGEST Pad to the longest sequence in the batch - PaddingStrategy.MAX_LENGTH: Pad to the max length (default) - PaddingStrategy.DO_NOT_PAD: Do not pad The tokenizer padding sides are defined in self.padding_side: - 'left': pads on the left of the sequences - 'right': pads on the right of the sequences pad_to_multiple_of: (optional) Integer if set will pad the sequence to a multiple of the provided value. This is especially useful to enable the use of Tensor Core on NVIDIA hardware with compute capability `>= 7.5` (Volta). return_attention_mask: (optional) Set to False to avoid returning attention mask (default: set to model specifics) """ entities_provided = bool("entity_ids" in encoded_inputs) # Load from model defaults if return_attention_mask is None: return_attention_mask = "attention_mask" in self.model_input_names if padding_strategy == PaddingStrategy.LONGEST: max_length = len(encoded_inputs["input_ids"]) if entities_provided: max_entity_length = len(encoded_inputs["entity_ids"]) if max_length is not None and pad_to_multiple_of is not None and (max_length % pad_to_multiple_of != 0): max_length = ((max_length // pad_to_multiple_of) + 1) * pad_to_multiple_of if ( entities_provided and max_entity_length is not None and pad_to_multiple_of is not None and (max_entity_length % pad_to_multiple_of != 0) ): max_entity_length = ((max_entity_length // pad_to_multiple_of) + 1) * pad_to_multiple_of needs_to_be_padded = padding_strategy != PaddingStrategy.DO_NOT_PAD and ( len(encoded_inputs["input_ids"]) != max_length or (entities_provided and len(encoded_inputs["entity_ids"]) != max_entity_length) ) # Initialize attention mask if not present. if return_attention_mask and "attention_mask" not in encoded_inputs: encoded_inputs["attention_mask"] = [1] * len(encoded_inputs["input_ids"]) if entities_provided and return_attention_mask and "entity_attention_mask" not in encoded_inputs: encoded_inputs["entity_attention_mask"] = [1] * len(encoded_inputs["entity_ids"]) if needs_to_be_padded: difference = max_length - len(encoded_inputs["input_ids"]) if entities_provided: entity_difference = max_entity_length - len(encoded_inputs["entity_ids"]) if self.padding_side == "right": if return_attention_mask: encoded_inputs["attention_mask"] = encoded_inputs["attention_mask"] + [0] * difference if entities_provided: encoded_inputs["entity_attention_mask"] = ( encoded_inputs["entity_attention_mask"] + [0] * entity_difference ) if "token_type_ids" in encoded_inputs: encoded_inputs["token_type_ids"] = encoded_inputs["token_type_ids"] + [0] * difference if entities_provided: encoded_inputs["entity_token_type_ids"] = ( encoded_inputs["entity_token_type_ids"] + [0] * entity_difference ) if "special_tokens_mask" in encoded_inputs: encoded_inputs["special_tokens_mask"] = encoded_inputs["special_tokens_mask"] + [1] * difference encoded_inputs["input_ids"] = encoded_inputs["input_ids"] + [self.pad_token_id] * difference if entities_provided: encoded_inputs["entity_ids"] = ( encoded_inputs["entity_ids"] + [self.entity_pad_token_id] * entity_difference ) encoded_inputs["entity_position_ids"] = ( encoded_inputs["entity_position_ids"] + [[-1] * self.max_mention_length] * entity_difference ) if self.task == "entity_span_classification": encoded_inputs["entity_start_positions"] = ( encoded_inputs["entity_start_positions"] + [0] * entity_difference ) encoded_inputs["entity_end_positions"] = ( encoded_inputs["entity_end_positions"] + [0] * entity_difference ) elif self.padding_side == "left": if return_attention_mask: encoded_inputs["attention_mask"] = [0] * difference + encoded_inputs["attention_mask"] if entities_provided: encoded_inputs["entity_attention_mask"] = [0] * entity_difference + encoded_inputs[ "entity_attention_mask" ] if "token_type_ids" in encoded_inputs: encoded_inputs["token_type_ids"] = [0] * difference + encoded_inputs["token_type_ids"] if entities_provided: encoded_inputs["entity_token_type_ids"] = [0] * entity_difference + encoded_inputs[ "entity_token_type_ids" ] if "special_tokens_mask" in encoded_inputs: encoded_inputs["special_tokens_mask"] = [1] * difference + encoded_inputs["special_tokens_mask"] encoded_inputs["input_ids"] = [self.pad_token_id] * difference + encoded_inputs["input_ids"] if entities_provided: encoded_inputs["entity_ids"] = [self.entity_pad_token_id] * entity_difference + encoded_inputs[ "entity_ids" ] encoded_inputs["entity_position_ids"] = [ [-1] * self.max_mention_length ] * entity_difference + encoded_inputs[ "entity_position_ids"] if self.task == "entity_span_classification": encoded_inputs["entity_start_positions"] = [0] * entity_difference + encoded_inputs[ "entity_start_positions" ] encoded_inputs["entity_end_positions"] = [0] * entity_difference + encoded_inputs[ "entity_end_positions" ] else: raise ValueError("Invalid padding strategy:" + str(self.padding_side)) return encoded_inputs