tokenization_vortex.py

"""SentencePiece tokenizer wrapper for Hugging Face AutoTokenizer."""

from __future__ import annotations

import shutil
from pathlib import Path

from transformers import PreTrainedTokenizer


class VortexTokenizer(PreTrainedTokenizer):
    vocab_files_names = {"vocab_file": "tokenizer.model"}
    model_input_names = ["input_ids", "attention_mask"]

    def __init__(
        self,
        vocab_file: str,
        unk_token: str = "<unk>",
        eos_token: str = "</s>",
        **kwargs,
    ) -> None:
        import sentencepiece as spm

        self.vocab_file = str(vocab_file)
        self.sp_model = spm.SentencePieceProcessor(model_file=self.vocab_file)
        super().__init__(unk_token=unk_token, eos_token=eos_token, **kwargs)

    @property
    def vocab_size(self) -> int:
        return int(self.sp_model.vocab_size())

    def get_vocab(self) -> dict[str, int]:
        vocab = {self.sp_model.id_to_piece(i): i for i in range(self.vocab_size)}
        vocab.update(self.added_tokens_encoder)
        return vocab

    def _tokenize(self, text: str, **kwargs) -> list[str]:
        return list(self.sp_model.encode(text, out_type=str))

    def _convert_token_to_id(self, token: str) -> int:
        token_id = int(self.sp_model.piece_to_id(token))
        return self.unk_token_id if token_id < 0 else token_id

    def _convert_id_to_token(self, index: int) -> str:
        return self.sp_model.id_to_piece(int(index))

    def convert_tokens_to_string(self, tokens: list[str]) -> str:
        return self.sp_model.decode_pieces(tokens)

    def build_inputs_with_special_tokens(
        self,
        token_ids_0: list[int],
        token_ids_1: list[int] | None = None,
    ) -> list[int]:
        if token_ids_1 is None:
            return list(token_ids_0)
        return list(token_ids_0) + list(token_ids_1)

    def get_special_tokens_mask(
        self,
        token_ids_0: list[int],
        token_ids_1: list[int] | None = None,
        already_has_special_tokens: bool = False,
    ) -> list[int]:
        if already_has_special_tokens:
            return [0] * len(token_ids_0)
        total = len(token_ids_0) + (len(token_ids_1) if token_ids_1 is not None else 0)
        return [0] * total

    def save_vocabulary(self, save_directory: str, filename_prefix: str | None = None) -> tuple[str]:
        directory = Path(save_directory)
        directory.mkdir(parents=True, exist_ok=True)
        name = (filename_prefix + "-" if filename_prefix else "") + "tokenizer.model"
        output = directory / name
        if Path(self.vocab_file).resolve() != output.resolve():
            shutil.copyfile(self.vocab_file, output)
        return (str(output),)