tokenizer.py

import os
from typing import Optional, Dict, Any, Tuple

import sentencepiece
from torch import TensorType
from transformers import PreTrainedTokenizer
from transformers.models.deberta_v2.tokenization_deberta_v2 import SPMTokenizer
from transformers.tokenization_utils_base import TextInput, PreTokenizedInput, TruncationStrategy
from transformers.utils import PaddingStrategy


class QiDeBERTaTokenizer(PreTrainedTokenizer):
    vocab_files_names = {"spm_model_file": "spm.model"}

    def __init__(
        self,
        spm_model_file: str,
        bos_token: str = '<s>',
        eos_token: str = '</s>',
        unk_token: str = '<unk>',
        sep_token: str = '</s>',
        pad_token: str = '<pad>',
        cls_token: str = '<s>',
        mask_token: str = '<mask>',
        do_lower_case=False,
        split_by_punct=False,
        sp_model_kwargs: Optional[Dict[str, Any]] = None,
        **kwargs,
    ):
        self.sp_model_kwargs = {} if sp_model_kwargs is None else sp_model_kwargs

        if not os.path.isfile(spm_model_file):
            raise ValueError(
                f"Can't find a vocabulary file at path '{spm_model_file}'. To load the vocabulary from a Google pretrained"
                " model use `tokenizer = AutoTokenizer.from_pretrained(PRETRAINED_MODEL_NAME)`"
            )
        self.do_lower_case = do_lower_case
        self.split_by_punct = split_by_punct
        self.spm_model_file = spm_model_file
        self._tokenizer = SPMTokenizer(
            spm_model_file, None, split_by_punct=split_by_punct, sp_model_kwargs=self.sp_model_kwargs
        )

        super().__init__(
            do_lower_case=do_lower_case,
            bos_token=bos_token,
            eos_token=eos_token,
            unk_token=unk_token,
            sep_token=sep_token,
            pad_token=pad_token,
            cls_token=cls_token,
            mask_token=mask_token,
            split_by_punct=split_by_punct,
            sp_model_kwargs=self.sp_model_kwargs,
            **kwargs,
        )
        self._tokenizer.special_tokens = self.all_special_tokens
        self.space_token_id = self._tokenizer.spm.PieceToId('▁')

    @property
    def vocab_size(self):
        return len(self.vocab)

    @property
    def vocab(self):
        return self._tokenizer.vocab

    def __call__(
        self,
        text: TextInput|PreTokenizedInput|list[TextInput]|list[PreTokenizedInput],
        text_pair: Optional[TextInput|PreTokenizedInput|list[TextInput]|list[PreTokenizedInput]] = None,
        text_target: Optional[TextInput|PreTokenizedInput|list[TextInput]|list[PreTokenizedInput]] = None,
        text_pair_target: Optional[TextInput|PreTokenizedInput|list[TextInput]|list[PreTokenizedInput]] = None,
        add_special_tokens: bool = True,
        padding: bool|str|PaddingStrategy = False,
        truncation: Optional[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,
        padding_side: Optional[str] = None,
        return_tensors: str|TensorType = 'pt',
        return_token_type_ids: bool = False,
        return_attention_mask: bool = True,
        return_overflowing_tokens: bool = False,
        return_special_tokens_mask: bool = False,
        return_offsets_mapping: bool = False,
        return_length: bool = False,
        verbose: bool = True,
        **kwargs,
    ):
        return super().__call__(
            text=text,
            text_pair=text_pair,
            text_target=text_target,
            text_pair_target=text_pair_target,
            add_special_tokens=add_special_tokens,
            padding=padding,
            truncation=truncation,
            max_length=max_length,
            stride=stride,
            is_split_into_words=is_split_into_words,
            pad_to_multiple_of=pad_to_multiple_of,
            padding_side=padding_side,
            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 get_vocab(self):
        vocab = self.vocab.copy()
        vocab.update(self.get_added_vocab())
        return vocab

    def _tokenize(self, text: str) -> list[str]:
        """Take as input a string and return a list of strings (tokens) for words/sub-words"""
        if self.do_lower_case:
            text = text.lower()
        return self._tokenizer.tokenize(text)

    def tokenize(self, text: TextInput, **kwargs) -> list[str]:
        result = super().tokenize(text, **kwargs)
        return result[1:] if result[0] == '▁' else result

    def _convert_token_to_id(self, token: str):
        """Converts a token (str) in an id using the vocab."""
        return self._tokenizer.spm.PieceToId(token)

    def _convert_id_to_token(self, index: int):
        """Converts an index (integer) in a token (str) using the vocab."""
        return self._tokenizer.spm.IdToPiece(index) if index < self.vocab_size else self.unk_token

    def convert_tokens_to_string(self, tokens):
        """Converts a sequence of tokens (string) in a single string."""
        return self._tokenizer.decode(tokens)

    def build_inputs_with_special_tokens(self, token_ids_0: list[int]|str, token_ids_1: Optional[list[int]|str] = None):
        """
        Build model inputs from a sequence or a pair of sequence for sequence classification tasks by concatenating and
        adding special tokens. A DeBERTa sequence has the following format:

        - 单个序列: [CLS] X [SEP]
        - 序列对: [CLS] A [SEP] B [SEP]

        Args:
            token_ids_0 (`List[int]`):
                将向其添加特殊令牌的 ID 列表。
            token_ids_1 (`List[int]`, *optional*):
                序列对的可选第二个 ID 列表。

        Returns:
            `List[int]`: List of [input IDs](../glossary#input-ids) with the appropriate special tokens.
        """
        cls = [self.cls_token_id]
        sep = [self.sep_token_id]

        if isinstance(token_ids_0, str):
            token_ids_0 = self._tokenizer.spm.encode_as_ids(token_ids_0)[1:]

        if token_ids_1 is None:
            return cls + token_ids_0 + sep
        else:
            if isinstance(token_ids_1, str):
                token_ids_1 = self._tokenizer.spm.encode_as_ids(token_ids_1)[1:]

        return cls + token_ids_0 + sep + token_ids_1 + sep

    def get_special_tokens_mask(self, token_ids_0: list[int], token_ids_1: Optional[list[int]] = None, already_has_special_tokens=False):
        """
        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 (`List[int]`):
                List of IDs.
            token_ids_1 (`List[int]`, *optional*):
                Optional second list of IDs for sequence pairs.
            already_has_special_tokens (`bool`, *optional*, defaults to `False`):
                token列表是否已使用模型的bos、eos特殊令牌进行格式化。

        Returns:
            `List[int]`: 范围 [0， 1] 中的整数列表：1 表示特殊token，0 表示正文序列token。
        """

        if already_has_special_tokens:
            return super().get_special_tokens_mask(
                token_ids_0=token_ids_0, token_ids_1=token_ids_1, already_has_special_tokens=True
            )

        if token_ids_1 is not None:
            return [1] + ([0] * len(token_ids_0)) + [1] + ([0] * len(token_ids_1)) + [1]
        return [1] + ([0] * len(token_ids_0)) + [1]

    def create_token_type_ids_from_sequences(self, token_ids_0: list[int], token_ids_1: Optional[list[int]] = None):
        """
        Create a mask from the two sequences passed to be used in a sequence-pair classification task. A DeBERTa
        sequence pair mask has the following format:

        ```
        0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1
        | first sequence    | second sequence |
        ```

        If `token_ids_1` is `None`, this method only returns the first portion of the mask (0s).

        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 [token type IDs](../glossary#token-type-ids) according to the given sequence(s).
        """
        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) * [0] + len(token_ids_1 + sep) * [1]

    def prepare_for_tokenization(self, text, is_split_into_words=False, **kwargs):
        add_prefix_space = kwargs.pop("add_prefix_space", False)
        if is_split_into_words or add_prefix_space:
            text = " " + text
        return (text, kwargs)

    def save_vocabulary(self, save_directory: str, filename_prefix: Optional[str] = None) -> Tuple[str]:
        return self._tokenizer.save_pretrained(save_directory, filename_prefix=filename_prefix)

    def _get_bos_piece(self) -> str:
        """
        获取BOS Piece
        :return:
        """
        return self._tokenizer.spm.IdToPiece(self._tokenizer.spm.bos_id())

    def _get_eos_piece(self) -> str:
        """
        获取EOS Piece
        :return:
        """
        return self._tokenizer.spm.IdToPiece(self._tokenizer.spm.eos_id())

    def processor(self) -> sentencepiece.SentencePieceProcessor:
        return self._tokenizer.spm