src/LlamaRetriever/transformers/models/code_llama/tokenization_code_llama.py

# coding=utf-8
# Copyright 2023 MetaAI and the HuggingFace Inc. team. 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.

"""Tokenization classes for Code LLaMA."""

import os
from shutil import copyfile
from typing import Any, Dict, List, Optional, Tuple

import sentencepiece as spm

from ...convert_slow_tokenizer import import_protobuf
from ...tokenization_utils import AddedToken, PreTrainedTokenizer
from ...utils import logging, requires_backends


logger = logging.get_logger(__name__)

VOCAB_FILES_NAMES = {"vocab_file": "tokenizer.model"}

SPIECE_UNDERLINE = "▁"

B_INST, E_INST = "[INST]", "[/INST]"
B_SYS, E_SYS = "<<SYS>>\n", "\n<</SYS>>\n\n"

# fmt: off
DEFAULT_SYSTEM_PROMPT = """You are a helpful, respectful and honest assistant. Always answer as helpfully as possible, while being safe. Your \
answers should not include any harmful, unethical, racist, sexist, toxic, dangerous, or illegal content. Please ensure\
 that your responses are socially unbiased and positive in nature.

If a question does not make any sense, or is not factually coherent, explain why instead of answering something not \
correct. If you don't know the answer to a question, please don't share false information."""
# fmt: on


class CodeLlamaTokenizer(PreTrainedTokenizer):
    """
    Construct a CodeLlama tokenizer. Based on byte-level Byte-Pair-Encoding. The default padding token is unset as
    there is no padding token in the original model.

    The default configuration match that of
    [codellama/CodeLlama-7b-Instruct-hf](https://huggingface.co/meta-llama/CodeLlama-7b-Instruct-hf/blob/main/tokenizer_config.json)
    which supports prompt infilling.

    Args:
        vocab_file (`str`):
            Path to the vocabulary file.
        unk_token (`str`, *optional*, defaults to `"<unk>"`):
            The unknown token. A token that is not in the vocabulary cannot be converted to an ID and is set to be this
            token instead.
        bos_token (`str`, *optional*, defaults to `"<s>"`):
            The beginning of sequence token that was used during pretraining. Can be used a sequence classifier token.
        eos_token (`str`, *optional*, defaults to `"</s>"`):
            The end of sequence token.

            <Tip>

            When building a sequence using special tokens, this is not the token that is used for the end of sequence.
            The token used is the `sep_token`.

            </Tip>

        prefix_token (`str`, *optional*, defaults to `"▁<PRE>"`):
            Prefix token used for infilling.
        middle_token (`str`, *optional*, defaults to `"▁<MID>"`):
            Middle token used for infilling.
        suffix_token (`str`, *optional*, defaults to `"▁<SUF>"`):
            Suffix token used for infilling.
        eot_token (`str`, *optional*, defaults to `"▁<EOT>"`):
            End of text token used for infilling.
        fill_token (`str`, *optional*, defaults to `"<FILL_ME>"`):
            The token used to split the input between the prefix and suffix.
        suffix_first (`bool`, *optional*, defaults to `False`):
            Whether the input prompt and suffix should be formatted with the suffix first.
        sp_model_kwargs (`dict`, *optional*):
            Will be passed to the `SentencePieceProcessor.__init__()` method. The [Python wrapper for
            SentencePiece](https://github.com/google/sentencepiece/tree/master/python) can be used, among other things,
            to set:

            - `enable_sampling`: Enable subword regularization.
            - `nbest_size`: Sampling parameters for unigram. Invalid for BPE-Dropout.

              - `nbest_size = {0,1}`: No sampling is performed.
              - `nbest_size > 1`: samples from the nbest_size results.
              - `nbest_size < 0`: assuming that nbest_size is infinite and samples from the all hypothesis (lattice)
                using forward-filtering-and-backward-sampling algorithm.

            - `alpha`: Smoothing parameter for unigram sampling, and dropout probability of merge operations for
              BPE-dropout.
        add_bos_token (`bool`, *optional*, defaults to `True`):
            Whether to add a beginning of sequence token at the start of sequences.
        add_eos_token (`bool`, *optional*, defaults to `False`):
            Whether to add an end of sequence token at the end of sequences.
        clean_up_tokenization_spaces (`bool`, *optional*, defaults to `False`):
            Whether or not to clean up the tokenization spaces.
        additional_special_tokens (`List[str]`, *optional*):
            Additional special tokens used by the tokenizer.
        use_default_system_prompt (`bool`, *optional*, defaults to `False`):
            Whether or not the default system prompt for Llama should be used.
    """

    vocab_files_names = VOCAB_FILES_NAMES
    model_input_names = ["input_ids", "attention_mask"]

    def __init__(
        self,
        vocab_file,
        unk_token="<unk>",
        bos_token="<s>",
        eos_token="</s>",
        prefix_token="▁<PRE>",
        middle_token="▁<MID>",
        suffix_token="▁<SUF>",
        eot_token="▁<EOT>",
        fill_token="<FILL_ME>",
        suffix_first=False,
        sp_model_kwargs: Optional[Dict[str, Any]] = None,
        add_bos_token=True,
        add_eos_token=False,
        clean_up_tokenization_spaces=False,
        additional_special_tokens=None,
        use_default_system_prompt=False,
        **kwargs,
    ):
        requires_backends(self, "protobuf")
        self.sp_model_kwargs = {} if sp_model_kwargs is None else sp_model_kwargs
        bos_token = AddedToken(bos_token, normalized=False, special=True) if isinstance(bos_token, str) else bos_token
        eos_token = AddedToken(eos_token, normalized=False, special=True) if isinstance(eos_token, str) else eos_token
        unk_token = AddedToken(unk_token, normalized=False, special=True) if isinstance(unk_token, str) else unk_token

        self.use_default_system_prompt = use_default_system_prompt
        # mark tokens special to skip them
        additional_special_tokens = additional_special_tokens or []
        for token in [prefix_token, middle_token, suffix_token, eot_token]:
            additional_special_tokens += [token] if token is not None else []

        self.vocab_file = vocab_file
        self.add_bos_token = add_bos_token
        self.add_eos_token = add_eos_token
        self._prefix_token = prefix_token
        self._middle_token = middle_token
        self._suffix_token = suffix_token
        self._eot_token = eot_token
        self.fill_token = fill_token
        self.suffix_first = suffix_first
        self.sp_model = self.get_spm_processor()

        super().__init__(
            bos_token=bos_token,
            eos_token=eos_token,
            unk_token=unk_token,
            add_bos_token=add_bos_token,
            add_eos_token=add_eos_token,
            prefix_token=prefix_token,
            middle_token=middle_token,
            suffix_token=suffix_token,
            eot_token=eot_token,
            fill_token=fill_token,
            sp_model_kwargs=self.sp_model_kwargs,
            suffix_first=suffix_first,
            clean_up_tokenization_spaces=clean_up_tokenization_spaces,
            additional_special_tokens=additional_special_tokens,
            use_default_system_prompt=use_default_system_prompt,
            **kwargs,
        )

    @property
    def unk_token_length(self):
        return len(self.sp_model.encode(str(self.unk_token)))

    def get_spm_processor(self):
        tokenizer = spm.SentencePieceProcessor(**self.sp_model_kwargs)
        with open(self.vocab_file, "rb") as f:
            sp_model = f.read()
            model_pb2 = import_protobuf()
            model = model_pb2.ModelProto.FromString(sp_model)
            normalizer_spec = model_pb2.NormalizerSpec()
            normalizer_spec.add_dummy_prefix = False
            model.normalizer_spec.MergeFrom(normalizer_spec)
            sp_model = model.SerializeToString()
            tokenizer.LoadFromSerializedProto(sp_model)
        return tokenizer

    @property
    def prefix_token(self):
        return self._prefix_token

    @property
    def prefix_id(self):
        if self._prefix_token is None:
            return None
        return self.convert_tokens_to_ids(self.prefix_token)

    @property
    def middle_token(self):
        return self._middle_token

    @property
    def middle_id(self):
        if self._middle_token is None:
            return None
        return self.convert_tokens_to_ids(self.middle_token)

    @property
    def suffix_token(self):
        return self._suffix_token

    @property
    def suffix_id(self):
        if self._suffix_token is None:
            return None
        return self.convert_tokens_to_ids(self.suffix_token)

    @property
    def eot_token(self):
        return self._eot_token

    @property
    def eot_id(self):
        if self._eot_token is None:
            return None
        return self.convert_tokens_to_ids(self.eot_token)

    @property
    def vocab_size(self):
        """Returns vocab size"""
        return self.sp_model.get_piece_size()

    # Copied from transformers.models.llama.tokenization_llama.LlamaTokenizer.get_vocab
    def get_vocab(self):
        """Returns vocab as a dict"""
        vocab = {self.convert_ids_to_tokens(i): i for i in range(self.vocab_size)}
        vocab.update(self.added_tokens_encoder)
        return vocab

    def tokenize(self, prefix, suffix=None, suffix_first=False, **kwargs) -> List[int]:
        # add a prefix space to `prefix`
        if self.fill_token is not None and self.fill_token in prefix and suffix is None:
            prefix, suffix = prefix.split(self.fill_token)

        if len(prefix) > 0:
            prefix = SPIECE_UNDERLINE + prefix.replace(SPIECE_UNDERLINE, " ")

        if suffix is None or len(suffix) < 1:
            tokens = super().tokenize(prefix, **kwargs)
            if len(tokens) > 1 and tokens[0] == SPIECE_UNDERLINE and tokens[1] in self.all_special_tokens:
                tokens = tokens[1:]
            return tokens

        prefix_tokens = self._tokenize(prefix)  # prefix has an extra `SPIECE_UNDERLINE`

        if None in (self.prefix_id, self.middle_id, self.suffix_id):
            raise ValueError(
                "The input either includes a `prefix` and a `suffix` used for the infilling task,"
                f"  or can be split on the {self.fill_token} token, creating a suffix and prefix,"
                " but the model does not support `infilling`."
            )
        suffix_tokens = self._tokenize(suffix)  # make sure CodeLlama sp model does not mess up

        suffix_first = suffix_first if suffix_first is not None else self.suffix_first
        if suffix_first:
            # format as " <PRE> <SUF>{suf} <MID> {pre}"
            return [self.prefix_token, self.suffix_token] + suffix_tokens + [self.middle_token] + prefix_tokens
        else:
            # format as " <PRE> {pre} <SUF>{suf} <MID>"
            return [self.prefix_token] + prefix_tokens + [self.suffix_token] + suffix_tokens + [self.middle_token]

    def _tokenize(self, text, **kwargs):
        """
        Returns a tokenized string.

        We de-activated the `add_dummy_prefix` option, thus the sentencepiece internals will always strip any
        SPIECE_UNDERLINE. For example: `self.sp_model.encode(f"{SPIECE_UNDERLINE}Hey", out_type = str)` will give
        `['H', 'e', 'y']` instead of `['▁He', 'y']`. Thus we always encode `f"{unk_token}text"` and strip the
        `unk_token`. Here is an example with `unk_token = "<unk>"` and `unk_token_length = 4`.
        `self.tokenizer.sp_model.encode("<unk> Hey", out_type = str)[4:]`.
        """
        tokens = self.sp_model.encode(text, out_type=str)
        if not text.startswith((SPIECE_UNDERLINE, " ")):
            return tokens
        # 1. Encode string + prefix ex: "<unk> Hey"
        tokens = self.sp_model.encode(self.unk_token + text, out_type=str)
        # 2. Remove self.unk_token from ['<','unk','>', '▁Hey']
        return tokens[self.unk_token_length :] if len(tokens) >= self.unk_token_length else tokens

    # Copied from transformers.models.llama.tokenization_llama.LlamaTokenizer._convert_token_to_id
    def _convert_token_to_id(self, token):
        """Converts a token (str) in an id using the vocab."""
        return self.sp_model.piece_to_id(token)

    # Copied from transformers.models.llama.tokenization_llama.LlamaTokenizer._convert_id_to_token
    def _convert_id_to_token(self, index):
        """Converts an index (integer) in a token (str) using the vocab."""
        token = self.sp_model.IdToPiece(index)
        return token

    def convert_tokens_to_string(self, tokens):
        """Converts a sequence of tokens (string) in a single string."""
        # since we manually add the prefix space, we have to remove it when decoding
        if tokens[0].startswith(SPIECE_UNDERLINE):
            tokens[0] = tokens[0][1:]

        current_sub_tokens = []
        out_string = ""
        for _, token in enumerate(tokens):
            # make sure that special tokens are not decoded using sentencepiece model
            if token in self.all_special_tokens:
                out_string += self.sp_model.decode(current_sub_tokens) + token
                current_sub_tokens = []
            else:
                current_sub_tokens.append(token)
        out_string += self.sp_model.decode(current_sub_tokens)
        return out_string

    # Copied from transformers.models.llama.tokenization_llama.LlamaTokenizer.save_vocabulary
    def save_vocabulary(self, save_directory, filename_prefix: Optional[str] = None) -> Tuple[str]:
        """
        Save the vocabulary and special tokens file to a directory.

        Args:
            save_directory (`str`):
                The directory in which to save the vocabulary.

        Returns:
            `Tuple(str)`: Paths to the files saved.
        """
        if not os.path.isdir(save_directory):
            logger.error(f"Vocabulary path ({save_directory}) should be a directory")
            return
        out_vocab_file = os.path.join(
            save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"]
        )

        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file) and os.path.isfile(self.vocab_file):
            copyfile(self.vocab_file, out_vocab_file)
        elif not os.path.isfile(self.vocab_file):
            with open(out_vocab_file, "wb") as fi:
                content_spiece_model = self.sp_model.serialized_model_proto()
                fi.write(content_spiece_model)

        return (out_vocab_file,)

    # Copied from transformers.models.llama.tokenization_llama.LlamaTokenizer.build_inputs_with_special_tokens
    def build_inputs_with_special_tokens(self, token_ids_0, token_ids_1=None):
        bos_token_id = [self.bos_token_id] if self.add_bos_token else []
        eos_token_id = [self.eos_token_id] if self.add_eos_token else []

        output = bos_token_id + token_ids_0 + eos_token_id

        if token_ids_1 is not None:
            output = output + bos_token_id + token_ids_1 + eos_token_id

        return output

    # Copied from transformers.models.llama.tokenization_llama.LlamaTokenizer.get_special_tokens_mask
    def get_special_tokens_mask(
        self, token_ids_0: List[int], token_ids_1: Optional[List[int]] = None, already_has_special_tokens: bool = False
    ) -> List[int]:
        """
        Retrieve sequence ids from a token list that has no special tokens added. This method is called when adding
        special tokens using the tokenizer `prepare_for_model` method.

        Args:
            token_ids_0 (`List[int]`):
                List of IDs.
            token_ids_1 (`List[int]`, *optional*):
                Optional second list of IDs for sequence pairs.
            already_has_special_tokens (`bool`, *optional*, defaults to `False`):
                Whether or not the token list is already formatted with special tokens for the model.

        Returns:
            `List[int]`: A list of integers in the range [0, 1]: 1 for a special token, 0 for a sequence token.
        """
        if already_has_special_tokens:
            return super().get_special_tokens_mask(
                token_ids_0=token_ids_0, token_ids_1=token_ids_1, already_has_special_tokens=True
            )

        bos_token_id = [1] if self.add_bos_token else []
        eos_token_id = [1] if self.add_eos_token else []

        if token_ids_1 is None:
            return bos_token_id + ([0] * len(token_ids_0)) + eos_token_id
        return (
            bos_token_id
            + ([0] * len(token_ids_0))
            + eos_token_id
            + bos_token_id
            + ([0] * len(token_ids_1))
            + eos_token_id
        )

    # Copied from transformers.models.llama.tokenization_llama.LlamaTokenizer.create_token_type_ids_from_sequences
    def create_token_type_ids_from_sequences(
        self, token_ids_0: List[int], token_ids_1: Optional[List[int]] = None
    ) -> List[int]:
        """
        Creates a mask from the two sequences passed to be used in a sequence-pair classification task. An ALBERT
        sequence pair mask has the following format:

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

        if token_ids_1 is None, only returns the first portion of the mask (0s).

        Args:
            token_ids_0 (`List[int]`):
                List of ids.
            token_ids_1 (`List[int]`, *optional*):
                Optional second list of IDs for sequence pairs.

        Returns:
            `List[int]`: List of [token type IDs](../glossary#token-type-ids) according to the given sequence(s).
        """
        bos_token_id = [self.bos_token_id] if self.add_bos_token else []
        eos_token_id = [self.eos_token_id] if self.add_eos_token else []

        output = [0] * len(bos_token_id + token_ids_0 + eos_token_id)

        if token_ids_1 is not None:
            output += [1] * len(bos_token_id + token_ids_1 + eos_token_id)

        return output

    def __getstate__(self):
        state = self.__dict__.copy()
        state["sp_model"] = None
        state["sp_model_proto"] = self.sp_model.serialized_model_proto()
        return state

    def __setstate__(self, d):
        self.__dict__ = d
        self.sp_model = spm.SentencePieceProcessor(**self.sp_model_kwargs)
        self.sp_model.LoadFromSerializedProto(self.sp_model_proto)