PaddleX/paddlex/inference/models/common/tokenizer/tokenizer_utils.py

# Copyright (c) 2024 PaddlePaddle Authors. All Rights Reserved.
#
# 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.

import bisect
import functools
import inspect
import io
import itertools
import json
import os
import re
import unicodedata
from collections import OrderedDict
from dataclasses import asdict, dataclass
from functools import lru_cache
from typing import Any, Dict, List, Optional, Tuple, Union

import numpy as np

from .....utils import logging
from .....utils.deps import (
    class_requires_deps,
    function_requires_deps,
    is_dep_available,
)
from .tokenizer_utils_base import (
    CHAT_TEMPLATE_CONFIG_NAME,
    AddedToken,
    BatchEncoding,
    EncodedInput,
    EncodedInputPair,
    PaddingStrategy,
    PreTokenizedInput,
    PreTokenizedInputPair,
    PretrainedTokenizerBase,
    TensorType,
    TextInput,
    TextInputPair,
    TruncationStrategy,
)
from .utils import convert_to_dict_message, fn_args_to_dict
from .vocab import Vocab

if is_dep_available("Jinja2"):
    from jinja2 import Template
    from jinja2.exceptions import TemplateError, TemplateSyntaxError
    from jinja2.sandbox import ImmutableSandboxedEnvironment

__all__ = [
    "ChatTemplate",
    "Trie",
    "ChatTemplateMixin",
    "PretrainedTokenizer",
    "InitTrackerMeta",
]


@class_requires_deps("Jinja2")
@dataclass
class ChatTemplate:
    conversation: Union[List[str], None] = None
    system: Union[str, None] = None
    query: str = None

    @staticmethod
    @lru_cache()
    def _compile_jinja_template(chat_template) -> "Template":
        def raise_exception(message):
            raise TemplateError(message)

        jinja_env = ImmutableSandboxedEnvironment(
            trim_blocks=True, lstrip_blocks=True, keep_trailing_newline=True
        )
        jinja_env.globals["raise_exception"] = raise_exception
        return jinja_env.from_string(chat_template)

    def render_conversation(
        self,
        conversation_data: Union[List[str], Dict[str, str]],
        index: int = 0,
        context_data: Dict[str, Any] = {},
    ) -> List[str]:
        """
        Args:
            conversation_data (list[str]): the conversation data which must be two parts
            index (int): the index of current conversation

        Returns:
            list[str]: the rendered conversation data
        """
        if self.conversation is None:
            raise ValueError(
                "The template for multi-turns is invalid, please check `conversation` filed in your chat-template."
            )

        if isinstance(conversation_data, (list, tuple)):
            assert (
                len(conversation_data) == 2
            ), "Each round/turn of conversation must be two participants, eg: [user-query, bot-query]"

            conversation_data = {
                "user": conversation_data[0],
                "bot": conversation_data[1],
                "index": index,
            }
        conversation_data.update(context_data)

        one_turn_conversation = []
        for conversation in self.conversation:
            template = self._compile_jinja_template(conversation)
            result = template.render(conversation_data)
            one_turn_conversation.append(result)
        return one_turn_conversation

    def render_query(
        self, query: str, index: int = 0, context_data: Dict[str, Union[int, str]] = {}
    ):
        if self.query is None:
            return query

        template = self._compile_jinja_template(self.query)
        return template.render(query=query, index=index, **context_data)

    def _init_context_data(
        self, context_data: Dict[str, Union[int, str]] = {}
    ) -> Dict[str, Union[int, str]]:
        """init the context data for chat-template"""
        context_data["is_training"] = context_data.get("is_training", False)
        return context_data

    def render_system(self, context_data: Dict[str, Union[int, str]] = {}) -> str:
        if self.system is None:
            return ""

        template = self._compile_jinja_template(self.system)
        return template.render(**context_data)

    def __call__(
        self,
        conversations: Union[List[List[str]], str],
        context_data: Dict[str, Union[int, str]] = {},
    ) -> str:
        """render the conversations by chat-template

        Args:
            conversations (list[list[str]]): the conversations of use and bot

        Returns:
            str: the result of conversation
        """
        if isinstance(conversations, str):
            conversations = [[conversations]]

        # [1 ... n-1] conversation
        final_query = self.render_system(context_data=context_data)
        context_data["length"] = len(conversations)
        for index, conversation in enumerate(conversations[:-1]):
            context_data["is_first"] = index == 0
            context_data["is_last"] = False
            final_query += "".join(
                self.render_conversation(
                    conversation, index=index, context_data=context_data
                )
            )

        if not isinstance(conversations[-1], list) and not len(conversations[-1]) != 1:
            raise ValueError(
                "The length of last conversation must be one, eg: [[user-query, bot-answer], [user-query, bot-answer], ..., [user-query]]"
            )
        if len(conversations[-1]) > 1:
            logging.warning(
                f"The last conversation is not a single-round, chat-template will skip the conversation: {conversations[-1][1:]}"
            )

        final_query += self.render_query(
            conversations[-1][0],
            index=len(conversations) - 1,
            context_data=context_data,
        )
        return final_query

    @classmethod
    def from_dict(cls, config: Dict):
        return cls(**config)

    @classmethod
    def from_file(cls, file: str):
        with open(file, "r", encoding="utf-8") as f:
            config = json.load(f)
        return cls.from_dict(config)


@function_requires_deps("paddlepaddle")
def adapt_stale_fwd_patch(self, name, value):
    """
    Since there are some monkey patches for forward of PretrainedModel, such as
    model compression, we make these patches compatible with the latest forward
    method.
    """

    if name == "forward":
        # NOTE(guosheng): In dygraph to static, `layer.forward` would be patched
        # by an instance of `StaticFunction`. And use string compare to avoid to
        # import fluid.
        if type(value).__name__.endswith(
            "StaticFunction"
        ) or self.forward.__class__.__name__.endswith("StaticFunction"):
            return value
        (
            patch_spec_args,
            patch_spec_varargs,
            patch_spec_varkw,
            patch_spec_defaults,
            _,
            _,
            _,
        ) = inspect.getfullargspec(value)
        (spec_args, spec_varargs, spec_varkw, spec_defaults, _, _, _) = (
            inspect.getfullargspec(self.forward)
        )
        new_args = [
            arg
            for arg in ("output_hidden_states", "output_attentions", "return_dict")
            if arg not in patch_spec_args and arg in spec_args
        ]

        if new_args:
            import paddle

            if self.__module__.startswith("paddlenlp"):
                logging.warning(
                    f"The `forward` method of {self.__class__ if isinstance(self, paddle.nn.Layer) else self} is patched and the patch "
                    "might be based on an old oversion which missing some "
                    f"arguments compared with the latest, such as {new_args}. "
                    "We automatically add compatibility on the patch for "
                    "these arguemnts, and maybe the patch should be updated."
                )
            else:
                logging.warning(
                    f"The `forward` method of {self.__class__ if isinstance(self, paddle.nn.Layer) else self} "
                    "is patched and the patch might be conflict with patches made "
                    f"by paddlenlp which seems have more arguments such as {new_args}. "
                    "We automatically add compatibility on the patch for "
                    "these arguemnts, and maybe the patch should be updated."
                )
            if isinstance(self, paddle.nn.Layer) and inspect.isfunction(value):

                @functools.wraps(value)
                def wrap_fwd(*args, **kwargs):
                    for arg in new_args:
                        kwargs.pop(arg, None)
                    return value(self, *args, **kwargs)

            else:

                @functools.wraps(value)
                def wrap_fwd(*args, **kwargs):
                    for arg in new_args:
                        kwargs.pop(arg, None)
                    return value(*args, **kwargs)

            return wrap_fwd
    return value


# NOTE:
# Modification:
#   class InitTrackerMeta(type(paddle.nn.Layer)) -> class InitTrackerMeta(type)
# Context:
#   1. In paddle 3.0rc, type(paddle.nn.Layer) == type
#   2. Solve the conflict between ultra-infer and paddle
class InitTrackerMeta(type):
    """
    This metaclass wraps the `__init__` method of a class to add `init_config`
    attribute for instances of that class, and `init_config` use a dict to track
    the initial configuration. If the class has `_pre_init` or `_post_init`
    method, it would be hooked before or after `__init__` and called as
    `_pre_init(self, init_fn, init_args)` or `_post_init(self, init_fn, init_args)`.
    Since InitTrackerMeta would be used as metaclass for pretrained model classes,
    which always are Layer and `type(Layer)` is not `type`, thus use `type(Layer)`
    rather than `type` as base class for it to avoid inheritance metaclass
    conflicts.
    """

    def __init__(cls, name, bases, attrs):
        init_func = cls.__init__
        # If attrs has `__init__`, wrap it using accessable `_pre_init, _post_init`.
        # Otherwise, no need to wrap again since the super cls has been wraped.
        # TODO: remove reduplicated tracker if using super cls `__init__`
        pre_init_func = getattr(cls, "_pre_init", None) if "__init__" in attrs else None
        post_init_func = (
            getattr(cls, "_post_init", None) if "__init__" in attrs else None
        )
        cls.__init__ = InitTrackerMeta.init_and_track_conf(
            init_func, pre_init_func, post_init_func
        )
        super(InitTrackerMeta, cls).__init__(name, bases, attrs)

    @staticmethod
    def init_and_track_conf(init_func, pre_init_func=None, post_init_func=None):
        """
        wraps `init_func` which is `__init__` method of a class to add `init_config`
        attribute for instances of that class.
        Args:
            init_func (callable): It should be the `__init__` method of a class.
                warning: `self` always is the class type of down-stream model, eg: BertForTokenClassification
            pre_init_func (callable, optional): If provided, it would be hooked after
                `init_func` and called as `pre_init_func(self, init_func, *init_args, **init_args)`.
                Default None.
            post_init_func (callable, optional): If provided, it would be hooked after
                `init_func` and called as `post_init_func(self, init_func, *init_args, **init_args)`.
                Default None.

        Returns:
            function: the wrapped function
        """

        @functools.wraps(init_func)
        def __impl__(self, *args, **kwargs):
            # registed helper by `pre_init_func`
            if pre_init_func:
                pre_init_func(self, init_func, *args, **kwargs)
            # keep full configuration
            init_func(self, *args, **kwargs)
            # registed helper by `post_init_func`
            if post_init_func:
                post_init_func(self, init_func, *args, **kwargs)
            self.init_config = kwargs
            if args:
                kwargs["init_args"] = args
            kwargs["init_class"] = self.__class__.__name__

        return __impl__

    def __setattr__(self, name, value):
        value = adapt_stale_fwd_patch(self, name, value)
        return super(InitTrackerMeta, self).__setattr__(name, value)


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 add(self, word: str):
        """
        Passes over every char (utf-8 char) on word and recursively adds it to the internal `data` trie representation.
        The special key `""` is used to represent termination.

        This function is idempotent, adding twice the same word will leave the trie unchanged

        Example:

        ```python
        >>> trie = Trie()
        >>> trie.add("Hello 友達")
        >>> trie.data
        {"H": {"e": {"l": {"l": {"o": {" ": {"友": {"達": {"": 1}}}}}}}}}

        >>> trie.add("Hello")
        >>> trie.data
        {"H": {"e": {"l": {"l": {"o": {"": 1, " ": {"友": {"達": {"": 1}}}}}}}}}
        ```
        """
        if not word:
            # Prevent empty string
            return
        ref = self.data
        for char in word:
            ref[char] = char in ref and ref[char] or {}
            ref = ref[char]
        ref[""] = 1

    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
                        elif 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
                elif 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):
        # 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:
                logging.error(
                    "There was a bug in Trie algorithm in tokenization. Attempting to recover. Please report it anyway."
                )
                continue
            elif 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


def _insert_one_token_to_ordered_list(token_list: List[str], new_token: str):
    """
    Inserts one token to an ordered list if it does not already exist. Note: token_list must be sorted.
    """
    insertion_idx = bisect.bisect_left(token_list, new_token)
    # Checks if new_token is already in the ordered token_list
    if insertion_idx < len(token_list) and token_list[insertion_idx] == new_token:
        # new_token is in token_list, don't add
        return
    else:
        token_list.insert(insertion_idx, new_token)


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


def _is_nonnormalized_char(char):
    """Check whther `chars` is a non-normalized character."""
    cp = ord(char)
    if (
        (0xFF00 <= cp <= 0xFFEF)
        or (0xFE50 <= cp <= 0xFE6B)  # Halfwidth and Fullwidth Forms
        or (0x3358 <= cp <= 0x33FF)  # Small Form Variants
        or (0x249C <= cp <= 0x24E9)  # CJK Compatibility
        or (0x3200 <= cp <= 0x32FF)  # Enclosed Alphanumerics: Ⓛ ⒰
    ):  # Enclosed CJK Letters and Months
        return True

    return False


def _is_nonnormalized_numeric(char):
    """Check whether `chars` is a non-normalized numeric character."""
    cp = ord(char)
    if (
        (0x2460 <= cp <= 0x249B)
        or (0x24EA <= cp <= 0x24FF)  #
        or (0x2776 <= cp <= 0x2793)  #
        or (0x2160 <= cp <= 0x217F)  # Enclosed Alphanumerics
    ):  # Number Forms
        return True

    return False


def normalize_chars(text):
    """
    Normalize the text for multiligual and chinese models. Unicode range:
    https://www.ling.upenn.edu/courses/Spring_2003/ling538/UnicodeRanges.html
    """
    output = []
    for char in text:
        if _is_nonnormalized_char(char):
            for c in unicodedata.normalize("NFKC", char):
                output.append(c)
        elif _is_nonnormalized_numeric(char):
            output.append(" ")
            for c in str(int(unicodedata.numeric(char))):
                output.append(c)
            output.append(" ")
        elif ord(char) == 0xF979:  # https://www.zhihu.com/question/20697984
            output.append("凉")
        else:
            output.append(char)
    return "".join(output)


@class_requires_deps("paddlepaddle", "Jinja2")
class ChatTemplateMixin:
    chat_template: Optional[ChatTemplate] = None

    def apply_chat_template(
        self,
        conversation: Union[Dict[str, str], str],
        tokenize: bool = True,
        context_data: Dict[str, Any] = {},
        **tokenizer_kwargs,
    ):
        """apply chat_template rules to conversation which should not be batched data

        Args:
            conversation (List[List[str, str]] | str): the conversation messages between user and bot
            context_data (Dict[str, Any]): the context data for chat_template.json
            tokenize (bool, optional): whether do tokenization. Defaults to True.

        Returns:
            str | dict[str, Union[numpy.ndarray, paddle.Tensor]]: return the result of applied data
        """
        if not self.chat_template:
            raise ValueError(
                "chat_template is not set, please set chat_template first."
            )
        elif isinstance(self.chat_template, Template):
            add_generation_prompt = tokenizer_kwargs.pop("add_generation_prompt", True)
            query = self._apply_chat_template(
                conversation, add_generation_prompt=add_generation_prompt
            )
        elif isinstance(self.chat_template, ChatTemplate):
            query = self._apply_chat_template_paddle(conversation, context_data)

        if not tokenize:
            return query

        # chat_template should not add special tokens
        tokenizer_kwargs["add_special_tokens"] = False
        return self(query, **tokenizer_kwargs)

    def _apply_chat_template_paddle(
        self,
        conversation: Union[List[Dict[str, str]], str],
        context_data: Dict[str, Any] = {},
    ):
        context_data = self.chat_template._init_context_data(context_data)

        if isinstance(conversation, str):
            conversation = [[conversation]]
        elif isinstance(conversation, list) and isinstance(conversation[0], str):
            raise ValueError(
                "apply_chat_template do not support appling batch conversations, "
                "so you should apply the conversation one by one."
            )

        query = self.chat_template(conversation, context_data=context_data)
        return query

    def _apply_chat_template(
        self,
        conversation: Union[Dict[str, str], str],
        add_generation_prompt=True,
    ):
        if isinstance(conversation, str):
            conversations = [{"role": "user", "content": conversation}]
        elif isinstance(conversation, list):
            assert len(conversation) > 0, "empty conversation is not allowed"
            if isinstance(conversation[0], list):
                conversations = convert_to_dict_message(conversation)
            elif isinstance(conversation[0], dict):
                conversations = conversation
            else:
                raise ValueError(
                    "apply_chat_template do not support appling batch conversations, "
                    "so you should apply the conversation one by one."
                )
        query = self.chat_template.render(
            messages=conversations,
            **self.special_tokens_map,
            add_generation_prompt=add_generation_prompt,
        )
        return query

    def encode_chat_inputs(
        self,
        conversations: List[Dict[str, str]],
        context_data: Dict[str, Any] = {},
        **kwargs,
    ):
        """Encodes conversation to pairs of token ids.
        Turn 0: bos + system + sep + user     bot + eos
        Turn t: sep + bot + query             bot + eos

        Args:
            conversation (List[Dict[str, str]]): the conversation of data
            context_data (Dict[str, Any]): the context data of conversation

        Returns:
            List[list[int], list[int]]: the pair of input_ids and target_ids
        """
        if not self.chat_template:
            raise ValueError(
                "chat_template is not set, please set chat_template first."
            )
        elif isinstance(self.chat_template, Template):
            add_generation_prompt = kwargs.pop("add_generation_prompt", True)
            query = self._encode_chat_inputs(
                conversations, context_data, add_generation_prompt=add_generation_prompt
            )
        elif isinstance(self.chat_template, ChatTemplate):
            query = self._encode_chat_inputs_paddle(conversations, context_data)
        return query

    def _encode_chat_inputs_paddle(
        self, conversations: List[Dict[str, str]], context_data: Dict[str, Any] = {}
    ):
        context_data = self.chat_template._init_context_data(context_data)
        # encode system
        result = {}
        if self.chat_template.system:
            system = self.chat_template.render_system(context_data)
            result["system"] = self.encode(system, add_special_tokens=False)[
                "input_ids"
            ]

        # encode conversation
        conversation_ids = []
        for index, conversation in enumerate(conversations):
            # give more control to chat_template
            context_data["is_first"] = index == 0
            context_data["is_last"] = index == len(conversations) - 1

            user_input, bot_output = self.chat_template.render_conversation(
                conversation, index=index, context_data=context_data
            )
            user_ids = self.encode(user_input, add_special_tokens=False)["input_ids"]
            bot_ids = self.encode(bot_output, add_special_tokens=False)["input_ids"]
            conversation_ids.append([user_ids, bot_ids])

        result["conversations"] = conversation_ids
        return result

    def _encode_chat_inputs(
        self,
        conversations: List[Dict[str, str]],
        context_data: Dict[str, Any] = {},
        system: str = None,
        add_generation_prompt=True,
    ):
        result = {}

        # Some template do not support system msg, so we need to check it first.
        if system:
            try:
                self.chat_template.render(
                    messages={"role": "system", "content": system}
                )
            except Exception as e:
                raise ValueError("System is not supported in this tokenizer.", e)

        # convert list msg to role dict msg
        conversation_dict = []
        origin_msg = []
        for round in conversations:
            round_role = [
                {"role": "user", "content": round[0]},
                {"role": "assistant", "content": round[1]},
            ]
            origin_msg.extend(round_role)
            conversation_dict.append(round_role)
        ans = []

        # get answer in single round, then compile the chat entirely and split by single round ans
        # attention: answer should include end token!
        for conv in conversation_dict:
            roundi = [system] + conv if system else conv
            roundi_str = self.chat_template.render(
                messages=roundi, add_generation_prompt=False, **self.special_tokens_map
            )
            roundi_no_ans = [system] + [conv[0]] if system else [conv[0]]
            roundi_no_ans_str = self.chat_template.render(
                messages=roundi_no_ans,
                add_generation_prompt=add_generation_prompt,
                **self.special_tokens_map,
            )
            ans_roundi = roundi_str[len(roundi_no_ans_str) :]
            ans.append(ans_roundi)

        non_learnable_parts = self._extract_non_learnable_parts(origin_msg, ans)
        assert len(non_learnable_parts) == len(ans)

        conversation_ids = []
        for i in range(len(non_learnable_parts)):
            conversation_ids.append(
                self.batch_encode(
                    [non_learnable_parts[i], ans[i]],
                    add_special_tokens=False,
                    padding=False,
                )["input_ids"]
            )

        result["conversations"] = conversation_ids
        return result

    def _extract_non_learnable_parts(
        self, origin_msg: List[Dict[str, str]], split_s: List[str]
    ):
        """Split the entire chat by specified words. Extract the non-learnable parts."""
        # distingish and replace the special words in original string to an uncompiled form: Like | -> \|
        regex_pattern = "|".join(map(re.escape, split_s))
        # splited by replaced specified words
        non_learnable_parts = re.split(
            r"(?:%s)" % regex_pattern,
            self.chat_template.render(
                messages=origin_msg,
                add_generation_prompt=False,
                **self.special_tokens_map,
            ),
        )
        if non_learnable_parts[-1] == "":
            non_learnable_parts.pop()
        return non_learnable_parts

    @classmethod
    def from_pretrained(cls, pretrained_model_name_or_path, *args, **kwargs):
        cache_dir = kwargs.pop("cache_dir", None)
        from_hf_hub = kwargs.pop("from_hf_hub", False)
        from_aistudio = kwargs.pop("from_aistudio", False)
        subfolder = kwargs.pop("subfolder", "")
        if subfolder is None:
            subfolder = ""

        kwargs["subfolder"] = subfolder
        kwargs["cache_dir"] = cache_dir
        kwargs["from_hf_hub"] = from_hf_hub
        kwargs["from_aistudio"] = from_aistudio
        kwargs["return_tokenizer_file_dir"] = True
        tokenizer, tokenizer_config_file_dir = super().from_pretrained(
            pretrained_model_name_or_path, *args, **kwargs
        )

        # load chat-template
        chat_template_file = os.path.join(
            tokenizer_config_file_dir, CHAT_TEMPLATE_CONFIG_NAME
        )
        if not os.path.exists(chat_template_file):
            return tokenizer

        if tokenizer.chat_template is not None:
            logging.warning(
                "Chat-template already exists in config file, it will be overwritten by chat_template.json file."
            )
            logging.warning(
                "`chat_template.json` will be deprecated in the future! Please set it in `tokenizer_config.json`."
            )
        tokenizer.init_chat_template(chat_template_file)
        return tokenizer

    def init_chat_template(self, chat_template: Union[str, Dict]):
        """init chat_tempalte by file_path or template dict data

        Args:
            chat_template (str | dict): file_path or template dict data
        """
        if isinstance(chat_template, str):
            if not os.path.exists(chat_template):
                try:
                    self.chat_template: Template = ChatTemplate._compile_jinja_template(
                        chat_template
                    )
                except TemplateSyntaxError:
                    # It is neither jinjia string nor path string
                    raise TemplateSyntaxError(
                        "The chat-template in json is not valid jinja string: {}".format(
                            chat_template
                        ),
                        lineno=0,  # fake lineno, useless required msg
                    )
            else:
                self.chat_template = ChatTemplate.from_file(chat_template)
        elif isinstance(chat_template, dict):
            self.chat_template = ChatTemplate.from_dict(chat_template)
        elif isinstance(chat_template, ChatTemplate):
            self.chat_template = chat_template
        else:
            raise ValueError("Receive error chat_template data: ", chat_template)

    def save_resources(self, save_directory):
        super().save_resources(save_directory)

        if isinstance(
            self.chat_template, ChatTemplate
        ):  # Future remove if ChatTemplate is deprecated
            chat_template_file = os.path.join(save_directory, CHAT_TEMPLATE_CONFIG_NAME)
            with open(chat_template_file, "w", encoding="utf-8") as f:
                json.dump(asdict(self.chat_template), f, ensure_ascii=False, indent=4)
            logging.info("Chat-template config file saved in " + chat_template_file)


class PretrainedTokenizer(
    ChatTemplateMixin, PretrainedTokenizerBase, metaclass=InitTrackerMeta
):
    """
    Base class for all tokenizers.

    Inherits from [`~tokenizer_utils_base.PretrainedTokenizerBase`].

    Handle all the shared methods for tokenization and special tokens as well as methods downloading/caching/loading
    pretrained tokenizers as well as adding tokens to the vocabulary.

    This class also contain the added tokens in a unified way on top of all tokenizers so we don't have to handle the
    specific vocabulary augmentation methods of the various underlying dictionary structures (BPE, sentencepiece...).

    - **resource_files_names** (`Dict[str, str]`) -- A dictionary with, as keys, the `__init__` keyword name of each
        vocabulary file required by the model, and as associated values, the filename for saving the associated file
        (string).
    - **pretrained_resource_files_map** (`Dict[str, Dict[str, str]]`) -- A dictionary of dictionaries, with the
        high-level keys being the `__init__` keyword name of each vocabulary file required by the model, the
        low-level being the `short-cut-names` of the pretrained models with, as associated values, the `url` to the
        associated pretrained vocabulary file.
    - **max_model_input_sizes** (`Dict[str, Optional[int]]`) -- A dictionary with, as keys, the `short-cut-names`
        of the pretrained models, and as associated values, the maximum length of the sequence inputs of this model,
        or `None` if the model has no maximum input size.
    - **pretrained_init_configuration** (`Dict[str, Dict[str, Any]]`) -- A dictionary with, as keys, the
        `short-cut-names` of the pretrained models, and as associated values, a dictionary of specific arguments to
        pass to the `__init__` method of the tokenizer class for this pretrained model when loading the tokenizer
        with the [`~tokenizer_utils_base.PretrainedTokenizerBase.from_pretrained`] method.
    - **model_input_names** (`List[str]`) -- A list of inputs expected in the forward pass of the model.
    - **padding_side** (`str`) -- The default value for the side on which the model should have padding applied.
        Should be `'right'` or `'left'`.
    - **truncation_side** (`str`) -- The default value for the side on which the model should have truncation
        applied. Should be `'right'` or `'left'`.

    Moreover, methods common to tokenizers for tokenization, token/id conversion
    and encoding as model inputs are also provided here.

    Besides, metaclass `InitTrackerMeta` is used to create `PretrainedTokenizer`,
    by which subclasses can track arguments for initialization automatically
    and expose special tokens initialization used as attributes.
    """

    added_tokens_encoder: Dict[str, int] = {}
    added_tokens_decoder: Dict[int, str] = {}
    unique_no_split_tokens: List[str] = []
    tokens_trie = Trie()

    _decode_use_source_tokenizer = False

    def _pre_init(self, original_init, *args, **kwargs):
        """
        It would be hooked before `__init__` to add specials tokens (arguments of
        `__init__` whose name ends with `_token`) as attributes of the tokenizer
        instance.
        """
        init_dict = fn_args_to_dict(original_init, *((self,) + args), **kwargs)
        init_dict.pop("self", None)
        super(PretrainedTokenizer, self).__init__(**init_dict)

        self.added_tokens_encoder: Dict[str, int] = {}
        self.added_tokens_decoder: Dict[int, str] = {}
        self.unique_no_split_tokens: List[str] = []
        self.tokens_trie = Trie()

        self._decode_use_source_tokenizer = False

    def _build_special_tokens_map_extended(self, **kwargs):
        for key, value in kwargs.items():
            if value is None:
                continue
            if key in self.SPECIAL_TOKENS_ATTRIBUTES:
                if key == "additional_special_tokens":
                    assert isinstance(
                        value, (list, tuple)
                    ), f"Value {value} is not a list or tuple"
                    assert all(
                        isinstance(t, (str, AddedToken)) for t in value
                    ), "One of the tokens is not a string or an AddedToken"
                    setattr(self, key, value)
                elif isinstance(value, (str, AddedToken)):
                    setattr(self, key, value)
                else:
                    raise TypeError(
                        f"special token {key} has to be either str or AddedToken but got: {type(value)}"
                    )

    @property
    def vocab_size(self) -> int:
        """
        `int`: Size of the base vocabulary (without the added tokens).
        """
        raise NotImplementedError

    @property
    def is_fast(self) -> bool:
        return False

    def get_added_vocab(self) -> Dict[str, int]:
        """
        Returns the added tokens in the vocabulary as a dictionary of token to index.

        Returns:
            `Dict[str, int]`: The added tokens.
        """
        return self.added_tokens_encoder

    def __len__(self):
        """
        Size of the full vocabulary with the added tokens.
        """
        return self.vocab_size + len(self.added_tokens_encoder)

    def _add_tokens(
        self,
        new_tokens: Union[List[str], List[AddedToken]],
        special_tokens: bool = False,
    ) -> int:
        """
        Add a list of new tokens to the tokenizer class. If the new tokens are not in the vocabulary, they are added to
        it with indices starting from length of the current vocabulary.

        Args:
            new_tokens (`List[str]`or `List[AddedToken]`):
                Token(s) to add in vocabulary. A token is only added if it's not already in the vocabulary (tested by
                checking if the tokenizer assign the index of the `unk_token` to them).
            special_tokens (`bool`, *optional*, defaults to `False`):
                Whether or not the tokens should be added as special tokens.

        Returns:
            `int`: The number of tokens actually added to the vocabulary.

        Examples:

        ```python
        # Let's see how to increase the vocabulary of Bert model and tokenizer
        tokenizer = BertTokenizer.from_pretrained("bert-base-uncased")
        model = BertModel.from_pretrained("bert-base-uncased")

        num_added_toks = tokenizer.add_tokens(["new_tok1", "my_new-tok2"])
        print("We have added", num_added_toks, "tokens")
        ```"""
        new_tokens = [str(tok) for tok in new_tokens]

        tokens_to_add = []
        for token in new_tokens:
            if not isinstance(token, str):
                raise TypeError(f"Token {token} is not a string but a {type(token)}.")
            if (
                not special_tokens
                and hasattr(self, "do_lower_case")
                and self.do_lower_case
            ):
                token = token.lower()
            if (
                token != self.unk_token
                and self.convert_tokens_to_ids(token)
                == self.convert_tokens_to_ids(self.unk_token)
                and token not in tokens_to_add
            ):
                tokens_to_add.append(token)
                if self.verbose:
                    logging.info(f"Adding {token} to the vocabulary")

        added_tok_encoder = dict(
            (tok, len(self) + i) for i, tok in enumerate(tokens_to_add)
        )
        added_tok_decoder = {v: k for k, v in added_tok_encoder.items()}
        self.added_tokens_encoder.update(added_tok_encoder)
        self.added_tokens_decoder.update(added_tok_decoder)

        # Make sure we don't split on any special tokens (even they were already in the vocab before e.g. for Albert)
        if special_tokens:
            if len(new_tokens) == 1:
                _insert_one_token_to_ordered_list(
                    self.unique_no_split_tokens, new_tokens[0]
                )
            else:
                self.unique_no_split_tokens = sorted(
                    set(self.unique_no_split_tokens).union(set(new_tokens))
                )
        else:
            # Or on the newly added tokens
            if len(tokens_to_add) == 1:
                _insert_one_token_to_ordered_list(
                    self.unique_no_split_tokens, tokens_to_add[0]
                )
            else:
                self.unique_no_split_tokens = sorted(
                    set(self.unique_no_split_tokens).union(set(tokens_to_add))
                )
        self._create_trie(self.unique_no_split_tokens)

        return len(tokens_to_add)

    def _create_trie(self, unique_no_split_tokens):
        trie = Trie()
        for token in unique_no_split_tokens:
            if (
                hasattr(self, "do_lower_case")
                and self.do_lower_case
                and token not in self.all_special_tokens
            ):
                trie.add(token.lower())
            else:
                trie.add(token)
        self.tokens_trie = trie

    def prepare_for_tokenization(self, text, is_split_into_words=False, **kwargs):
        """
        Performs any necessary transformations before tokenization.

        This method should pop the arguments from kwargs and return the remaining `kwargs` as well. We test the
        `kwargs` at the end of the encoding process to be sure all the arguments have been used.

        Args:
            text (`str`):
                The text to prepare.
            is_split_into_words (`bool`, *optional*, defaults to `False`):
                Whether or not the input is already pre-tokenized (e.g., split into words). If set to `True`, the
                tokenizer assumes the input is already split into words (for instance, by splitting it on whitespace)
                which it will tokenize. This is useful for NER or token classification.
            kwargs:
                Keyword arguments to use for the tokenization.

        Returns:
            `Tuple[str, Dict[str, Any]]`: The prepared text and the unused kwargs.
        """

        return (text, kwargs)

    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 = dict(
            (str(t), t)
            for t in self.all_special_tokens_extended
            if isinstance(t, AddedToken)
        )

        text, kwargs = self.prepare_for_tokenization(text, **kwargs)

        # TODO: should this be in the base class?
        if hasattr(self, "do_lower_case") and self.do_lower_case:
            # convert non-special tokens to lowercase
            escaped_special_toks = [
                re.escape(s_tok)
                for s_tok in (self.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 = self.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 _tokenize(self, text, **kwargs):
        """
        Converts a string in a sequence of tokens (string), using the tokenizer. Split in words for word-based
        vocabulary or sub-words for sub-word-based vocabularies (BPE/SentencePieces/WordPieces).

        Do NOT take care of added tokens.
        """
        raise NotImplementedError

    def convert_tokens_to_ids(self, tokens):
        if tokens is None:
            return None

        if isinstance(tokens, str):
            return self._convert_token_to_id_with_added_voc(tokens)

        ids = []
        for token in tokens:
            ids.append(self._convert_token_to_id_with_added_voc(token))

        return ids

    def _convert_token_to_id_with_added_voc(self, token):
        if token is None:
            return None

        if token in self.added_tokens_encoder:
            return self.added_tokens_encoder[token]
        return self._convert_token_to_id(token)

    def _convert_token_to_id(self, token):

        return self.vocab.to_indices(token)

    def convert_tokens_to_string(self, tokens):
        """
        Converts a sequence of tokens (list of string) to a single string by
        using ``' '.join(tokens)`` .

        Args:
            tokens (list[str]): A sequence of tokens.

        Returns:
            str: Converted string.
        """
        return " ".join(tokens)

    def convert_ids_to_tokens(self, ids, skip_special_tokens=False):
        if isinstance(ids, int):
            if ids in self.added_tokens_decoder:
                return self.added_tokens_decoder[ids]
            else:
                return self._convert_id_to_token(ids)
        tokens = []
        for index in ids:
            index = int(index)
            if skip_special_tokens and index in self.all_special_ids:
                continue
            if index in self.added_tokens_decoder:
                tokens.append(self.added_tokens_decoder[index])
            else:
                tokens.append(self._convert_id_to_token(index))
        return tokens

    def _convert_id_to_token(self, index):

        return self.vocab.to_tokens(index)

    @staticmethod
    def load_vocabulary(
        filepath,
        unk_token=None,
        pad_token=None,
        bos_token=None,
        eos_token=None,
        **kwargs,
    ):
        """
        Instantiate an instance of `Vocab` from a file reserving all tokens
        by using `Vocab.from_dict`. The file contains a token per line, and the
        line number would be the index of corresponding token.

        Args:
            filepath (str): path of file to construct vocabulary.
            unk_token (str): special token for unknown token. If no need, it also
                could be `None`. Defaults to `None`.
            pad_token (str): special token for padding token. If no need, it also
                could be `None`. Defaults to `None`.
            bos_token (str): special token for bos token. If no need, it also
                could be `None`. Defaults to `None`.
            eos_token (str): special token for eos token. If no need, it also
                could be `None`. Defaults to `None`.
            **kwargs (dict): keyword arguments for `Vocab.from_dict`.

        Returns:
            Vocab: An instance of `Vocab`.
        """
        token_to_idx = {}
        with io.open(filepath, "r", encoding="utf-8") as f:
            for index, line in enumerate(f):
                token = line.rstrip("\n")
                token_to_idx[token] = int(index)
        vocab = Vocab.from_dict(
            token_to_idx,
            unk_token=unk_token,
            pad_token=pad_token,
            bos_token=bos_token,
            eos_token=eos_token,
            **kwargs,
        )
        return vocab

    @staticmethod
    def save_vocabulary(filepath, vocab):
        """
        Save all tokens to a vocabulary file. The file contains a token per line,
        and the line number would be the index of corresponding token.

        Args:
            filepath (str): File path to be saved to.
            vocab (Vocab|dict): The `Vocab` or `dict` instance to be saved.
        """
        if isinstance(vocab, Vocab):
            tokens = vocab.idx_to_token
        else:
            tokens = sorted(vocab.keys(), key=lambda token: vocab[token])
        with io.open(filepath, "w", encoding="utf-8") as f:
            for token in tokens:
                f.write(token + "\n")

    def get_special_tokens_mask(
        self, token_ids_0, token_ids_1=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 ``encode`` methods.

        Args:
            token_ids_0 (List[int]): List of ids of the first sequence.
            token_ids_1 (List[int], optional): List of ids of the second sequence.
            already_has_special_tokens (bool, optional): Whether or not the token list is already
                formatted with special tokens for the model. Defaults to None.

        Returns:
            results (List[int]): The list of integers in the range [0, 1]:
                1 for a special token, 0 for a sequence token.
        """
        if already_has_special_tokens:
            if token_ids_1 is not None:
                raise ValueError(
                    "You should not supply a second sequence if the provided sequence of "
                    "ids is already formatted with special tokens for the model."
                )

            return super().get_special_tokens_mask(
                token_ids_0=token_ids_0,
                token_ids_1=token_ids_1,
                already_has_special_tokens=True,
            )
        return [0] * ((len(token_ids_1) if token_ids_1 else 0) + len(token_ids_0))

    def num_special_tokens_to_add(self, pair):
        """
        Returns the number of added tokens when encoding a sequence with special tokens.

        Args:
            pair (bool, optional):
                Whether the number of added tokens should be computed in the case of a sequence pair or a single
                sequence. Defaults to `False`.
        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
            )
        )

    def _encode_plus(
        self,
        text: Union[TextInput, PreTokenizedInput, EncodedInput],
        text_pair: Optional[Union[TextInput, PreTokenizedInput, EncodedInput]] = None,
        add_special_tokens: bool = True,
        padding_strategy: PaddingStrategy = PaddingStrategy.DO_NOT_PAD,
        truncation_strategy: TruncationStrategy = TruncationStrategy.DO_NOT_TRUNCATE,
        max_length: Optional[int] = None,
        stride: int = 0,
        is_split_into_words: bool = False,
        pad_to_multiple_of: Optional[int] = None,
        return_tensors: Optional[Union[str, TensorType]] = None,
        return_position_ids: Optional[bool] = None,
        return_token_type_ids: Optional[bool] = None,
        return_attention_mask: Optional[bool] = None,
        return_overflowing_tokens: bool = False,
        return_special_tokens_mask: bool = False,
        return_offsets_mapping: bool = False,
        return_length: bool = False,
        verbose: bool = True,
        **kwargs,
    ) -> BatchEncoding:
        def get_input_ids(text):
            if isinstance(text, str):
                tokens = self.tokenize(text, **kwargs)
                return self.convert_tokens_to_ids(tokens)
            elif (
                isinstance(text, (list, tuple))
                and len(text) > 0
                and isinstance(text[0], str)
            ):
                if is_split_into_words:
                    tokens = list(
                        itertools.chain(
                            *(
                                self.tokenize(t, is_split_into_words=True, **kwargs)
                                for t in text
                            )
                        )
                    )
                    return self.convert_tokens_to_ids(tokens)
                else:
                    return self.convert_tokens_to_ids(text)
            elif (
                isinstance(text, (list, tuple))
                and len(text) > 0
                and isinstance(text[0], int)
            ):
                return text
            else:
                if is_split_into_words:
                    raise ValueError(
                        f"Input {text} is not valid. Should be a string or a list/tuple of strings when `is_split_into_words=True`."
                    )
                else:
                    raise ValueError(
                        f"Input {text} is not valid. Should be a string, a list/tuple of strings or a list/tuple of integers."
                    )

        first_ids = get_input_ids(text)
        second_ids = get_input_ids(text_pair) if text_pair is not None else None

        if return_offsets_mapping:
            kwargs["text"] = text
            kwargs["text_pair"] = text_pair

        return self.prepare_for_model(
            first_ids,
            pair_ids=second_ids,
            add_special_tokens=add_special_tokens,
            padding=padding_strategy.value,
            truncation=truncation_strategy.value,
            max_length=max_length,
            stride=stride,
            pad_to_multiple_of=pad_to_multiple_of,
            return_tensors=return_tensors,
            prepend_batch_axis=True,
            return_position_ids=return_position_ids,
            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_offsets_mapping=return_offsets_mapping,
            return_length=return_length,
            verbose=verbose,
            **kwargs,
        )

    def _batch_encode_plus(
        self,
        batch_text_or_text_pairs: Union[
            List[TextInput],
            List[TextInputPair],
            List[PreTokenizedInput],
            List[PreTokenizedInputPair],
            List[EncodedInput],
            List[EncodedInputPair],
        ],
        add_special_tokens: bool = True,
        padding_strategy: PaddingStrategy = PaddingStrategy.DO_NOT_PAD,
        truncation_strategy: TruncationStrategy = TruncationStrategy.DO_NOT_TRUNCATE,
        max_length: Optional[int] = None,
        stride: int = 0,
        is_split_into_words: bool = False,
        pad_to_multiple_of: Optional[int] = None,
        return_position_ids: Optional[bool] = 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_dict: bool = True,
        return_offsets_mapping: bool = False,
        return_length: bool = False,
        verbose: bool = True,
        **kwargs,
    ) -> BatchEncoding:
        def get_input_ids(text):
            if isinstance(text, str):
                tokens = self.tokenize(text, **kwargs)
                return self.convert_tokens_to_ids(tokens)
            elif (
                isinstance(text, (list, tuple))
                and len(text) > 0
                and isinstance(text[0], str)
            ):
                if is_split_into_words:
                    tokens = list(
                        itertools.chain(
                            *(
                                self.tokenize(t, is_split_into_words=True, **kwargs)
                                for t in text
                            )
                        )
                    )
                    return self.convert_tokens_to_ids(tokens)
                else:
                    return self.convert_tokens_to_ids(text)
            elif (
                isinstance(text, (list, tuple))
                and len(text) > 0
                and isinstance(text[0], int)
            ):
                return text
            else:
                raise ValueError(
                    "Input is not valid. Should be a string, a list/tuple of strings or a list/tuple of integers."
                )

        input_ids = []
        for ids_or_pair_ids in batch_text_or_text_pairs:
            if not isinstance(ids_or_pair_ids, (list, tuple)):
                ids, pair_ids = ids_or_pair_ids, None
            elif is_split_into_words and not isinstance(
                ids_or_pair_ids[0], (list, tuple)
            ):
                ids, pair_ids = ids_or_pair_ids, None
            else:
                ids, pair_ids = ids_or_pair_ids

            first_ids = get_input_ids(ids)
            second_ids = get_input_ids(pair_ids) if pair_ids is not None else None
            input_ids.append((first_ids, second_ids))

        if stride > 0 and second_ids is not None:
            kwargs["batch_text_or_text_pairs"] = batch_text_or_text_pairs
        else:
            if return_offsets_mapping:
                has_pair = False
                if len(batch_text_or_text_pairs) > 0:
                    if isinstance(batch_text_or_text_pairs[0], (list, tuple)):
                        has_pair = True
                kwargs["texts"] = None
                kwargs["text_pairs"] = None
                if has_pair:
                    kwargs["texts"] = [text[0] for text in batch_text_or_text_pairs]
                    kwargs["text_pairs"] = [
                        text[1] for text in batch_text_or_text_pairs
                    ]
                else:
                    kwargs["texts"] = [text for text in batch_text_or_text_pairs]

        batch_outputs = self._batch_prepare_for_model(
            input_ids,
            add_special_tokens=add_special_tokens,
            padding_strategy=padding_strategy,
            truncation_strategy=truncation_strategy,
            max_length=max_length,
            stride=stride,
            pad_to_multiple_of=pad_to_multiple_of,
            return_position_ids=return_position_ids,
            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_dict=return_dict,
            return_offsets_mapping=return_offsets_mapping,
            return_length=return_length,
            return_tensors=return_tensors,
            verbose=verbose,
            **kwargs,
        )

        return batch_outputs

    def _batch_prepare_for_model(
        self,
        batch_ids_pairs: List[Union[PreTokenizedInputPair, Tuple[List[int], None]]],
        add_special_tokens: bool = True,
        padding_strategy: PaddingStrategy = PaddingStrategy.DO_NOT_PAD,
        truncation_strategy: TruncationStrategy = TruncationStrategy.DO_NOT_TRUNCATE,
        max_length: Optional[int] = None,
        stride: int = 0,
        pad_to_multiple_of: Optional[int] = None,
        return_position_ids: Optional[bool] = None,
        return_tensors: Optional[str] = None,
        return_token_type_ids: Optional[bool] = None,
        return_attention_mask: Optional[bool] = None,
        return_overflowing_tokens: bool = False,
        return_special_tokens_mask: bool = False,
        return_dict: bool = True,
        return_offsets_mapping: bool = False,
        return_length: bool = False,
        verbose: bool = True,
        **kwargs,
    ) -> BatchEncoding:
        """
        Prepares a sequence of input id, or a pair of sequences of inputs ids so that it can be used by the model. It
        adds special tokens, truncates sequences if overflowing while taking into account the special tokens and
        manages a moving window (with user defined stride) for overflowing tokens

        Args:
            batch_ids_pairs: list of tokenized input ids or input ids pairs
        """
        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."
            )

        batch_outputs = {}
        batch_outputs_list = []
        for example_id, (first_ids, second_ids) in enumerate(batch_ids_pairs):
            if stride > 0 and second_ids is not None:
                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

                max_len_for_pair = (
                    max_length
                    - len(first_ids)
                    - (
                        self.num_special_tokens_to_add(pair=True)
                        if add_special_tokens
                        else 0
                    )
                )

                text, text_pair = kwargs["batch_text_or_text_pairs"][example_id]
                token_offset_mapping = self.get_offset_mapping(text)
                token_pair_offset_mapping = self.get_offset_mapping(text_pair)

                offset = 0
                while offset < len(second_ids):
                    encoded_inputs = {}
                    length = len(second_ids) - offset
                    if length > max_len_for_pair:
                        length = max_len_for_pair

                    ids = first_ids
                    pair_ids = second_ids[offset : offset + length]
                    pair = bool(pair_ids is not None)
                    mapping = token_offset_mapping
                    pair_mapping = token_pair_offset_mapping[offset : offset + length]
                    if add_special_tokens:
                        offset_mapping = self.build_offset_mapping_with_special_tokens(
                            mapping, pair_mapping
                        )
                        sequence = self.build_inputs_with_special_tokens(ids, pair_ids)
                        token_type_ids = self.create_token_type_ids_from_sequences(
                            ids, pair_ids
                        )
                    else:
                        offset_mapping = mapping + pair_mapping
                        sequence = ids + pair_ids if pair else ids
                        token_type_ids = [0] * len(ids) + (
                            [0] * len(pair_ids) if pair else []
                        )
                    encoded_inputs["offset_mapping"] = offset_mapping
                    # Build output dictionnary
                    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)

                    # Check lengths
                    self._eventual_warn_about_too_long_sequence(
                        encoded_inputs["input_ids"], max_length, verbose
                    )
                    if return_position_ids:
                        encoded_inputs["position_ids"] = list(
                            range(len(encoded_inputs["input_ids"]))
                        )

                    if return_length:
                        encoded_inputs["length"] = len(encoded_inputs["input_ids"])
                        encoded_inputs["seq_len"] = encoded_inputs["length"]

                    encoded_inputs["overflow_to_sample"] = example_id

                    for key, value in encoded_inputs.items():
                        if key not in batch_outputs:
                            batch_outputs[key] = []
                        batch_outputs[key].append(value)

                    if offset + length == len(second_ids):
                        break
                    offset += min(length, stride)
            else:
                if return_offsets_mapping:
                    kwargs["text"] = kwargs["texts"][example_id]
                    kwargs["text_pair"] = None
                    if kwargs["text_pairs"] is not None:
                        kwargs["text_pair"] = kwargs["text_pairs"][example_id]

                encoded_inputs = self.prepare_for_model(
                    first_ids,
                    second_ids,
                    add_special_tokens=add_special_tokens,
                    padding=PaddingStrategy.DO_NOT_PAD.value,  # we pad in batch afterward
                    truncation=truncation_strategy.value,
                    max_length=max_length,
                    stride=stride,
                    pad_to_multiple_of=None,  # we pad in batch afterward
                    return_position_ids=return_position_ids,  # we pad in batch afterward
                    return_attention_mask=False,  # we pad in batch afterward
                    return_token_type_ids=return_token_type_ids,
                    return_overflowing_tokens=return_overflowing_tokens,
                    return_special_tokens_mask=return_special_tokens_mask,
                    return_offsets_mapping=return_offsets_mapping,
                    return_length=return_length,
                    return_tensors=None,  # We convert the whole batch to tensors at the end
                    prepend_batch_axis=False,
                    verbose=verbose,
                    **kwargs,
                )
                for key, value in encoded_inputs.items():
                    if key not in batch_outputs:
                        batch_outputs[key] = []
                    batch_outputs[key].append(value)

        batch_outputs = self.pad(
            batch_outputs,
            padding=padding_strategy.value,
            max_length=max_length,
            pad_to_multiple_of=pad_to_multiple_of,
            return_attention_mask=return_attention_mask,
        )
        if return_dict:
            batch_outputs = BatchEncoding(batch_outputs, tensor_type=return_tensors)
            return batch_outputs
        else:
            for k, v in batch_outputs.items():
                for i in range(len(v)):
                    if i >= len(batch_outputs_list):
                        batch_outputs_list.append({k: v[i]})
                    else:
                        batch_outputs_list[i][k] = v[i]
            return batch_outputs_list

    def _get_bert_like_offset_mapping(self, text: str):
        """
        Returns the map of tokens and the start and end index of their start and end character.
        Modified from https://github.com/bojone/bert4keras/blob/master/bert4keras/tokenizers.py#L372
        Args:
            text (str):
                Input text.
        Returns:
            list: The offset map of input text.

        """
        if text is None:
            return None
        split_tokens = self.tokenize(text)

        normalized_text, char_mapping = "", []

        for i, ch in enumerate(text):
            if hasattr(self, "do_lower_case") and self.do_lower_case:
                ch = ch.lower()
                if self.basic_tokenizer.strip_accents is not False:
                    ch = unicodedata.normalize("NFD", ch)
                    ch = "".join([c for c in ch if unicodedata.category(c) != "Mn"])
            elif self.basic_tokenizer.strip_accents:
                ch = unicodedata.normalize("NFD", ch)
                ch = "".join([c for c in ch if unicodedata.category(c) != "Mn"])

            ch = "".join(
                [
                    c
                    for c in ch
                    if not (ord(c) == 0 or ord(c) == 0xFFFD or _is_control(c))
                ]
            )
            normalized_text += ch

            char_mapping.extend([i] * len(ch))
        text, token_mapping, offset = normalized_text, [], 0

        char_mapping_indexes = []
        for index, token in enumerate(split_tokens):
            if token[:2] == "##":
                token = token[2:]
            if token in self.all_special_tokens:
                token = (
                    token.lower()
                    if hasattr(self, "do_lower_case") and self.do_lower_case
                    else token
                )
            # The greek letter "sigma" has 2 forms of lowercase, σ and ς respectively.
            # When used as a final letter of a word, the final form (ς) is used. Otherwise, the form (σ) is used.
            # https://latin.stackexchange.com/questions/6168/how-and-when-did-we-get-two-forms-of-sigma
            if "σ" in token or "ς" in token:
                start = (
                    text[offset:].replace("ς", "σ").index(token.replace("ς", "σ"))
                    + offset
                )
            else:

                # try to fix: https://github.com/PaddlePaddle/PaddleNLP/issues/3985
                if token not in text[offset:]:
                    # check whether there are consecutive UNK tokens, eg: ['好', '[UNK]', '[UNK]', 'good']
                    if (
                        index < len(split_tokens) - 1
                        and split_tokens[index + 1] in self.all_special_tokens
                    ):
                        start = offset
                        token = " "  # only contains one char
                    else:
                        start = -1
                else:
                    start = text[offset:].index(token) + offset

            end = start + len(token)
            char_mapping_indexes.append([start, end])

            if start != -1:
                offset = end

        token_mapping = []
        for index, (start, end) in enumerate(char_mapping_indexes):
            if start == -1:
                # init start
                if index == 0:
                    start = 0
                else:
                    start = char_mapping_indexes[index - 1][1]

                # init end
                if index == len(char_mapping_indexes) - 1:
                    end = len(char_mapping)
                else:
                    # next start
                    end = char_mapping_indexes[index + 1][0]

            token_mapping.append((char_mapping[start], char_mapping[end - 1] + 1))

        return token_mapping

    def get_offset_mapping(self, text: str, split_tokens: Optional[List[str]] = None):
        """
        Returns the map of tokens and the start and end index of their start and end character.
        Modified from https://github.com/bojone/bert4keras/blob/master/bert4keras/tokenizers.py#L372
        Args:
            text (str):
                Input text.
            split_tokens (Optional[List[str]]):
                the tokens which has been split which can accelerate the operation.

        Returns:
            list: The offset map of input text.

        """
        if text is None:
            return None
        split_tokens = self.tokenize(text)

        # bert-like tokenizer use the old-school code block
        if hasattr(self, "basic_tokenizer") or hasattr(self, "wordpiece_tokenizer"):
            return self._get_bert_like_offset_mapping(text)

        if not split_tokens:
            split_tokens = self.tokenize(text)

        normalized_text, char_mapping = "", []

        for i, ch in enumerate(text):
            normalized_text += normalize_chars(ch)
            char_mapping.extend([i] * len(ch))

        text, token_mapping, offset = normalized_text, [], 0
        do_lower_case = getattr(self, "do_lower_case", False)

        # lower the text if the token is lower-cased
        # keep align with token
        if do_lower_case:
            text = text.lower()

        char_mapping_indexes = []
        for token in split_tokens:

            # convert tokens into original string
            token: str = self.convert_tokens_to_string(token).strip()

            if token in self.all_special_tokens:
                if do_lower_case:
                    token = token.lower()

            # The greek letter "sigma" has 2 forms of lowercase, σ and ς respectively.
            # When used as a final letter of a word, the final form (ς) is used. Otherwise, the form (σ) is used.
            # https://latin.stackexchange.com/questions/6168/how-and-when-did-we-get-two-forms-of-sigma
            if "σ" in token or "ς" in token:
                start = (
                    text[offset:].replace("ς", "σ").index(token.replace("ς", "σ"))
                    + offset
                )
            else:

                # try to fix: https://github.com/PaddlePaddle/PaddleNLP/issues/3985
                if token not in text[offset:]:
                    start = -1
                else:
                    start = text[offset:].index(token) + offset

            end = start + len(token)
            char_mapping_indexes.append([start, end])

            if start != -1:
                offset = end

        token_mapping = []
        for index, (start, end) in enumerate(char_mapping_indexes):
            if start == -1:
                # init start
                if index == 0:
                    start = 0
                else:
                    start = char_mapping_indexes[index - 1][1]

                # init end
                if index == len(char_mapping_indexes) - 1:
                    end = len(char_mapping)
                else:
                    # next start
                    end = char_mapping_indexes[index + 1][0]

            token_mapping.append((char_mapping[start], char_mapping[end - 1] + 1))

        return token_mapping

    def _decode(
        self,
        token_ids: List[int],
        skip_special_tokens: bool = False,
        clean_up_tokenization_spaces: bool = True,
        spaces_between_special_tokens: bool = True,
        **kwargs,
    ) -> str:
        if isinstance(token_ids, np.ndarray):
            token_ids = token_ids.tolist()
        self._decode_use_source_tokenizer = kwargs.pop("use_source_tokenizer", False)
        filtered_tokens = self.convert_ids_to_tokens(
            token_ids, skip_special_tokens=skip_special_tokens
        )

        # To avoid mixing byte-level and unicode for byte-level BPT
        # we need to build string separately for added tokens and byte-level tokens
        # cf. https://github.com/huggingface/transformers/issues/1133
        sub_texts = []
        current_sub_text = []
        for token in filtered_tokens:
            if skip_special_tokens and token in self.all_special_ids:
                continue
            if token in self.added_tokens_encoder:
                if current_sub_text:
                    sub_texts.append(self.convert_tokens_to_string(current_sub_text))
                    current_sub_text = []
                sub_texts.append(token)
            else:
                current_sub_text.append(token)
        if current_sub_text:
            sub_texts.append(self.convert_tokens_to_string(current_sub_text))

        if spaces_between_special_tokens:
            text = " ".join(sub_texts)
        else:
            text = "".join(sub_texts)

        if clean_up_tokenization_spaces:
            clean_text = self.clean_up_tokenization(text)
            return clean_text
        else:
            return text

    def decode_token(
        self,
        all_input_ids: List[int],
        prefix_offset: int = 0,
        read_offset: int = 0,
    ) -> Tuple[str, int, int]:
        """tokenizer decoding for the streaming generation use case. This method can be overrided for tokenizer that doesn't follow this API"""
        # The prefix text is necessary only to defeat cleanup algorithms in the decode
        # which decide to add a space or not depending on the surrounding ids.
        prefix_text = self.decode(
            all_input_ids[prefix_offset:read_offset], skip_special_tokens=False
        )
        new_text = self.decode(all_input_ids[prefix_offset:], skip_special_tokens=False)

        if len(new_text) > len(prefix_text) and not new_text.endswith("�"):
            # utf-8 char at the end means it's a potential unfinished byte sequence
            # from byte fallback tokenization.
            # If it's in the middle, it's probably a real invalid id generated
            # by the model
            prefix_index = new_text.index(prefix_text)
            new_text = new_text[prefix_index + len(prefix_text) :]
            return new_text, read_offset, len(all_input_ids)
        else:
            return "", prefix_offset, read_offset


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


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


def _is_symbol(char):
    """Check whether CP is the codepoint of a Symbol character."""
    cp = ord(char)
    if unicodedata.category(char).startswith("S") or (
        cp in [0x00AD, 0x00B2, 0x00BA, 0x3007, 0x00B5, 0x00D8, 0x014B, 0x01B1]
    ):
        return True
    return False


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


def convert_to_unicode(text):
    """
    Converts `text` to Unicode (if it's not already), assuming utf-8 input.
    Args:
        text (str|bytes): Text to be converted to unicode.
    Returns:
        str: converted text.
    """
    if isinstance(text, str):
        return text
    elif isinstance(text, bytes):
        return text.decode("utf-8", "ignore")
    else:
        raise ValueError("Unsupported string type: %s" % (type(text)))


def whitespace_tokenize(text):
    """
    Runs basic whitespace cleaning and splitting on a peice of text.
    Args:
        text (str): Text to be tokenized.
    Returns:
        list(str): Token list.
    """
    text = text.strip()
    if not text:
        return []
    tokens = text.split()
    return tokens