| 1 | # SPDX-License-Identifier: Apache-2.0 |
| 2 | # SPDX-FileCopyrightText: Copyright contributors to the vLLM project |
| 3 | from collections.abc import Sequence |
| 4 | from pathlib import Path |
| 5 | from typing import TYPE_CHECKING, Any, Protocol, overload |
| 6 | |
| 7 | if TYPE_CHECKING: |
| 8 | from transformers import BatchEncoding |
| 9 | |
| 10 | from vllm.entrypoints.chat_utils import ChatCompletionMessageParam |
| 11 | |
| 12 | |
| 13 | class TokenizerLike(Protocol): |
| 14 | @classmethod |
| 15 | def from_pretrained( |
| 16 | cls, |
| 17 | path_or_repo_id: str | Path, |
| 18 | *args, |
| 19 | trust_remote_code: bool = False, |
| 20 | revision: str | None = None, |
| 21 | download_dir: str | None = None, |
| 22 | **kwargs, |
| 23 | ) -> "TokenizerLike": |
| 24 | raise NotImplementedError |
| 25 | |
| 26 | def num_special_tokens_to_add(self) -> int: |
| 27 | raise NotImplementedError |
| 28 | |
| 29 | @property |
| 30 | def all_special_tokens(self) -> list[str]: |
| 31 | raise NotImplementedError |
| 32 | |
| 33 | @property |
| 34 | def all_special_ids(self) -> list[int]: |
| 35 | raise NotImplementedError |
| 36 | |
| 37 | @property |
| 38 | def bos_token_id(self) -> int: |
| 39 | raise NotImplementedError |
| 40 | |
| 41 | @property |
| 42 | def eos_token_id(self) -> int: |
| 43 | raise NotImplementedError |
| 44 | |
| 45 | @property |
| 46 | def pad_token_id(self) -> int: |
| 47 | raise NotImplementedError |
| 48 | |
| 49 | @property |
| 50 | def is_fast(self) -> bool: |
| 51 | raise NotImplementedError |
| 52 | |
| 53 | @property |
| 54 | def vocab_size(self) -> int: |
| 55 | raise NotImplementedError |
| 56 | |
| 57 | @property |
| 58 | def max_token_id(self) -> int: |
| 59 | raise NotImplementedError |
| 60 | |
| 61 | @property |
| 62 | def max_chars_per_token(self) -> int: |
| 63 | raise NotImplementedError |
| 64 | |
| 65 | @property |
| 66 | def truncation_side(self) -> str: |
| 67 | raise NotImplementedError |
| 68 | |
| 69 | def __hash__(self) -> int: |
| 70 | return hash(id(self)) |
| 71 | |
| 72 | def __len__(self) -> int: |
| 73 | return self.vocab_size |
| 74 | |
| 75 | def __call__( |
| 76 | self, |
| 77 | text: str | list[str], |
| 78 | text_pair: str | None = None, |
| 79 | add_special_tokens: bool = True, |
| 80 | truncation: bool = False, |
| 81 | max_length: int | None = None, |
| 82 | ) -> "BatchEncoding": |
| 83 | raise NotImplementedError |
| 84 | |
| 85 | def get_vocab(self) -> dict[str, int]: |
| 86 | raise NotImplementedError |
| 87 | |
| 88 | def get_added_vocab(self) -> dict[str, int]: |
| 89 | raise NotImplementedError |
| 90 | |
| 91 | def encode( |
| 92 | self, |
| 93 | text: str, |
| 94 | truncation: bool | None = None, |
| 95 | max_length: int | None = None, |
| 96 | add_special_tokens: bool = True, |
| 97 | ) -> list[int]: |
| 98 | raise NotImplementedError |
| 99 | |
| 100 | def apply_chat_template( |
| 101 | self, |
| 102 | messages: list["ChatCompletionMessageParam"], |
| 103 | tools: list[dict[str, Any]] | None = None, |
| 104 | **kwargs, |
| 105 | ) -> str | list[int]: |
| 106 | raise NotImplementedError |
| 107 | |
| 108 | @overload |
| 109 | def convert_tokens_to_ids(self, tokens: str) -> int: ... |
| 110 | |
| 111 | @overload |
| 112 | def convert_tokens_to_ids(self, tokens: list[str]) -> list[int]: ... |
| 113 | |
| 114 | def convert_tokens_to_ids(self, tokens: str | list[str]) -> int | list[int]: |
| 115 | raise NotImplementedError |
| 116 | |
| 117 | def convert_tokens_to_string(self, tokens: list[str]) -> str: |
| 118 | raise NotImplementedError |
| 119 | |
| 120 | def decode( |
| 121 | self, ids: Sequence[int] | int, skip_special_tokens: bool = False |
| 122 | ) -> str: |
| 123 | raise NotImplementedError |
| 124 | |
| 125 | def convert_ids_to_tokens( |
| 126 | self, |
| 127 | ids: Sequence[int], |
| 128 | skip_special_tokens: bool = False, |
| 129 | ) -> list[str]: |
| 130 | raise NotImplementedError |