tokenizer添加新词-大模型词表扩充必备工具SentencePiece_added_tokens_decoder

一、目录

1. transformers 分词器添加新词
2. sentencepiece 训练自己的tokenizer
3. 将transformer 中tokenizer 修改为个人的。

二、实现

1. transformers 分词器添加新词
   借助sentencepiece进行分词器训练，并与源分词器合并成新的分词器。

#coding=utf8
#训练分词器
def train():
    import sentencepiece as spm
    spm.SentencePieceTrainer.Train(
        input="/home/chatglm6/sentences_vocab/tianlongbabu.txt",
        model_prefix="tokenizer",
        vocab_size=50000,
        user_defined_symbols=['foo', 'bar'],
        character_coverage=1.0,
        model_type="bpe",
    )
def practice_sentencepiece():
    import sentencepiece as spm
    sp_model=spm.SentencePieceProcessor()
    sp_model.Load("./tokenizer.model")
    #编码
    print(sp_model.EncodeAsPieces("你好是一个汉语词语"))
    print(sp_model.EncodeAsIds("你好是一个汉语词语"))
    #解码
    print(sp_model.DecodePieces(['▁', '你好', '是', '一个', '汉', '语', '词', '语']))
    print(sp_model.Decode([46706, 2382, 46699, 21, 47120, 47105, 48432, 47105]))

#practice_sentencepiece()
#transformer 分词器使用 加载
def tf_practice():

    import os
    from transformers import LlamaTokenizer
    from sentencepiece import sentencepiece_model_pb2 as sp_pb2_model
    import sentencepiece as spm

    chinese_sp_model_file = "./tokenizer.model"

    # 加载分词器模型
    chinese_sp_model = spm.SentencePieceProcessor()
    chinese_sp_model.Load(chinese_sp_model_file)

    chinese_spm = sp_pb2_model.ModelProto()
    chinese_spm.ParseFromString(chinese_sp_model.serialized_model_proto())   #序列化

    ## Save
    output_dir = './transformers_tokenizer/chinese/'
    os.makedirs(output_dir, exist_ok=True)
    with open(output_dir + 'chinese.model', 'wb') as f:
        f.write(chinese_spm.SerializeToString())        #模型保存

    #transformer 加载
    tokenizer = LlamaTokenizer(vocab_file=output_dir + 'chinese.model')
    tokenizer.save_pretrained(output_dir)                          #并保存到transformer格式

    #测试
    chinese_tokenizer = LlamaTokenizer.from_pretrained(output_dir)    #transformer 格式加载
    print(tokenizer.all_special_tokens)
    print(tokenizer.all_special_ids)
    print(tokenizer.special_tokens_map)
    text = '''白日依山尽，黄河入海流。欲穷千里目，更上一层楼。
    The primary use of LLaMA is research on large language models, including'''
    print("Test text:\n", text)
    print(f"Tokenized by Chinese-LLaMA tokenizer:{chinese_tokenizer.tokenize(text)}")

#tf_practice()
#transformer 分词器合并
def combine():
    import os
    from transformers import LlamaTokenizer
    from sentencepiece import sentencepiece_model_pb2 as sp_pb2_model
    import sentencepiece as spm

    llama_tokenizer_dir = "/home/llama_/CodeLlama-7b/tokenizer.model"
    chinese_sp_model_file = "./tokenizer.model"

    # load
    llama_tokenizer = LlamaTokenizer.from_pretrained(llama_tokenizer_dir)    #已有分词器

    chinese_sp_model = spm.SentencePieceProcessor()
    chinese_sp_model.Load(chinese_sp_model_file)    #新分词器

    llama_spm = sp_pb2_model.ModelProto()
    llama_spm.ParseFromString(llama_tokenizer.sp_model.serialized_model_proto())   #已有分词序列化

    chinese_spm = sp_pb2_model.ModelProto()
    chinese_spm.ParseFromString(chinese_sp_model.serialized_model_proto())    #现有分词序列化

    # print number of tokens
    print(len(llama_tokenizer), len(chinese_sp_model))
    print(llama_tokenizer.all_special_tokens)
    print(llama_tokenizer.all_special_ids)
    print(llama_tokenizer.special_tokens_map)

    ## Add Chinese tokens to LLaMA tokenizer
    llama_spm_tokens_set = set(p.piece for p in llama_spm.pieces)    #分词片段
    print(len(llama_spm_tokens_set))
    print(f"Before:{len(llama_spm_tokens_set)}")
    for p in chinese_spm.pieces:                         #将中文分词添加到已有分词器中
        piece = p.piece
        if piece not in llama_spm_tokens_set:
            new_p = sp_pb2_model.ModelProto().SentencePiece()   #创建词片对象
            new_p.piece = piece
            new_p.score = 0
            llama_spm.pieces.append(new_p)   #添加到词片段集合中
    print(f"New model pieces: {len(llama_spm.pieces)}")

    ## Save 保存
    output_sp_dir = 'transformers_tokenizer/llama_chinese'
    output_hf_dir = 'transformers_tokenizer/llama_chinese'  # the path to save Chinese-LLaMA tokenizer
    os.makedirs(output_sp_dir, exist_ok=True)
    with open(output_sp_dir + '/chinese_llama.model', 'wb') as f:
        f.write(llama_spm.SerializeToString())                     #序列号，并保存到新模型

    tokenizer = LlamaTokenizer(vocab_file=output_sp_dir + '/chinese_llama.model')   #加载新模型，并保存
    tokenizer.save_pretrained(output_hf_dir)
    print(f"Chinese-LLaMA tokenizer has been saved to {output_hf_dir}")

    # Test
    llama_tokenizer = LlamaTokenizer.from_pretrained(llama_tokenizer_dir)
    chinese_llama_tokenizer = LlamaTokenizer.from_pretrained(output_hf_dir)
    print(tokenizer.all_special_tokens)
    print(tokenizer.all_special_ids)
    print(tokenizer.special_tokens_map)
    text = '''白日依山尽，黄河入海流。欲穷千里目，更上一层楼。
    The primary use of LLaMA is research on large language models, including'''
    print("Test text:\n", text)
    print(f"Tokenized by LLaMA tokenizer:{llama_tokenizer.tokenize(text)}")
    print(f"Tokenized by Chinese-LLaMA tokenizer:{chinese_llama_tokenizer.tokenize(text)}")

#combine()

#单独使用合并后的transformers
def test():
    from transformers import LlamaTokenizer
    output_hf_dir = 'transformers_tokenizer/llama_chinese'

    tokenizer = LlamaTokenizer.from_pretrained(output_hf_dir)
    print(tokenizer.all_special_tokens)
    print(tokenizer.all_special_ids)
    print(tokenizer.special_tokens_map)
    text = '''白日依山尽，黄河入海流。欲穷千里目，更上一层楼。
       The primary use of LLaMA is research on large language models, including'''
    print("Test text:\n", text)
    print(f"Tokenized by Chinese-LLaMA tokenizer:{tokenizer.tokenize(text)}")
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141

2. sentencepiece 训练自己的tokenizer

def make_self_tokenizer():
    import os
    from sentences_vocab.tokenization import NewChineseTokenizer
    import sentencepiece as spm
    from sentencepiece import sentencepiece_model_pb2 as sp_pb2_model
    chinese_sp_model_file = "./tokenizer.model"
    # 加载分词器模型
    chinese_sp_model = spm.SentencePieceProcessor()
    chinese_sp_model.Load(chinese_sp_model_file)

    chinese_spm = sp_pb2_model.ModelProto()
    chinese_spm.ParseFromString(chinese_sp_model.serialized_model_proto())  # 序列化

    ## Save
    output_dir = './transformers_tokenizer/news/'
    os.makedirs(output_dir, exist_ok=True)
    with open(output_dir + 'chinese.model', 'wb') as f:
        f.write(chinese_spm.SerializeToString())  # 模型保存

    #加载
    tokenizer=NewChineseTokenizer(vocab_file=output_dir+"chinese.model")
    tokenizer.save_pretrained(output_dir)  # 并保存到transformer格式

#make_self_tokenizer()


def test():
    #手动导入类，加载
    from sentences_vocab.tokenization import NewChineseTokenizer
    output_dir = './transformers_tokenizer/news/'
    chinese_tokenizer=NewChineseTokenizer.from_pretrained(output_dir)
    text = '''白日依山尽，黄河入海流。欲穷千里目，更上一层楼。
        The primary use of LLaMA is research on large language models, including'''
    print("Test text:\n", text)
    print(f"Tokenized by Chinese-LLaMA tokenizer:{chinese_tokenizer.tokenize(text)}")

test()   #创建自己的tokenizer
print("======================================")

def test1():
    '''
    修改配置文件，让模型能够找到对应的类
            {
          "name_or_path": "",
          "remove_space": false,
          "do_lower_case": false,
          "tokenizer_class": "ChatGLMTokenizer",
          "auto_map": {
            "AutoTokenizer": [
              "tokenization_chatglm.ChatGLMTokenizer",
              null
              ]
               }
            }
    开始：
        {
      "add_bos_token": true,
      "add_eos_token": false,
      "added_tokens_decoder": {
       ...
      "tokenizer_class": "NewChineseTokenizer",
      "unk_token": "<unk>",
         }
    插入后：
     {
      "add_bos_token": true,
      "add_eos_token": false,
      "added_tokens_decoder": {
       ...
      "tokenizer_class": "NewChineseTokenizer",
      "unk_token": "<unk>",

      "name_or_path": "",
      "remove_space": false,
      "do_lower_case": false,
      "tokenizer_class": "ChatGLMTokenizer",
      "auto_map": {
        "AutoTokenizer": [
          "tokenization_chatglm.ChatGLMTokenizer",
          null
          ]
           }

         }
    '''
    #自动加载
    from transformers import AutoTokenizer
    output_dir = './transformers_tokenizer/news/'
    chinese_tokenizer=AutoTokenizer.from_pretrained(output_dir,trust_remote_code=True)
    text = '''白日依山尽，黄河入海流。欲穷千里目，更上一层楼。
        The primary use of LLaMA is research on large language models, including'''
    print("Test text:\n", text)
    print(f"Tokenized by Chinese-LLaMA tokenizer:{chinese_tokenizer.tokenize(text)}")

#test1()

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96

tokenization.py: 将llama 的tokenizer 改个名而已。

# coding=utf-8
# Copyright 2022 EleutherAI and the HuggingFace Inc. team. All rights reserved.
#
# This code is based on EleutherAI's GPT-NeoX library and the GPT-NeoX
# and OPT implementations in this library. It has been modified from its
# original forms to accommodate minor architectural differences compared
# to GPT-NeoX and OPT used by the Meta AI team that trained the model.
#
# 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 LLaMA."""
import os
from shutil import copyfile
from typing import Any, Dict, List, Optional, Tuple

import sentencepiece as spm

from transformers.tokenization_utils import AddedToken, PreTrainedTokenizer
from transformers.utils import logging


logger = logging.get_logger(__name__)

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

# PRETRAINED_VOCAB_FILES_MAP = {
#     "vocab_file": {
#         "hf-internal-testing/llama-tokenizer": "https://huggingface.co/hf-internal-testing/llama-tokenizer/resolve/main/tokenizer.model",
#     },
#     "tokenizer_file": {
#         "hf-internal-testing/llama-tokenizer": "https://huggingface.co/hf-internal-testing/llama-tokenizer/resolve/main/tokenizer_config.json",
#     },
# }
# PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES = {
#     "hf-internal-testing/llama-tokenizer": 2048,
# }


class NewChineseTokenizer(PreTrainedTokenizer):
    """
    Construct a Llama tokenizer. Based on byte-level Byte-Pair-Encoding.

    Args:
        vocab_file (`str`):
            Path to the vocabulary file.
    """

    vocab_files_names = VOCAB_FILES_NAMES
    # pretrained_vocab_files_map = PRETRAINED_VOCAB_FILES_MAP
    # max_model_input_sizes = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
    model_input_names = ["input_ids", "attention_mask"]

    def __init__(
        self,
        vocab_file,
        unk_token="<unk>",
        bos_token="<s>",
        eos_token="</s>",
        pad_token=None,
        sp_model_kwargs: Optional[Dict[str, Any]] = None,
        add_bos_token=True,
        add_eos_token=False,
        clean_up_tokenization_spaces=False,
        **kwargs,
    ):
        self.sp_model_kwargs = {} if sp_model_kwargs is None else sp_model_kwargs
        bos_token = AddedToken(bos_token, lstrip=False, rstrip=False) if isinstance(bos_token, str) else bos_token
        eos_token = AddedToken(eos_token, lstrip=False, rstrip=False) if isinstance(eos_token, str) else eos_token
        unk_token = AddedToken(unk_token, lstrip=False, rstrip=False) if isinstance(unk_token, str) else unk_token
        pad_token = AddedToken(pad_token, lstrip=False, rstrip=False) if isinstance(pad_token, str) else pad_token
        self.vocab_file = vocab_file
        self.add_bos_token = add_bos_token
        self.add_eos_token = add_eos_token
        self.sp_model = spm.SentencePieceProcessor(**self.sp_model_kwargs)
        self.sp_model.Load(vocab_file)

        super().__init__(
            bos_token=bos_token,
            eos_token=eos_token,
            unk_token=unk_token,
            pad_token=pad_token,
            add_bos_token=add_bos_token,
            add_eos_token=add_eos_token,
            sp_model_kwargs=self.sp_model_kwargs,
            clean_up_tokenization_spaces=clean_up_tokenization_spaces,
            **kwargs,
        )


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

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

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

    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, text):
        """Returns a tokenized string."""
        return self.sp_model.encode(text, out_type=str)

    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)

    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."""
        current_sub_tokens = []
        out_string = ""
        prev_is_special = False
        for i, token in enumerate(tokens):
            # make sure that special tokens are not decoded using sentencepiece model
            if token in self.all_special_tokens:
                if not prev_is_special and i != 0:
                    out_string += " "
                out_string += self.sp_model.decode(current_sub_tokens) + token
                prev_is_special = True
                current_sub_tokens = []
            else:
                current_sub_tokens.append(token)
                prev_is_special = False
        out_string += self.sp_model.decode(current_sub_tokens)
        return out_string

    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,)

    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

    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
        )

    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
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260

3. 将transformer 中tokenizer 修改为个人的。

'''复制已经训练过的tokenizer'''
#只需要修改一下配置文件的名字
from transformers import AutoTokenizer
tokenizer=AutoTokenizer.from_pretrained("../chatglm3-6b",trust_remote_code=True)
print(tokenizer.encode("白日依山尽，黄河入海流。欲穷千里目，更上一层楼。 The primary use of LLaMA is research on large language models, including"))
print(tokenizer.tokenize("白日依山尽，黄河入海流。欲穷千里目，更上一层楼。 The primary use of LLaMA is research on large language models, including"))
import os
output_dir = './transformers_tokenizer/news1/'
os.makedirs(output_dir, exist_ok=True)
# tokenizer.save_pretrained(output_dir)

#不知道什么原因，需要中tokenizer_config.py特殊字符的配置给删除掉
tokenizer1=AutoTokenizer.from_pretrained(output_dir,trust_remote_code=True)
print(tokenizer1.encode("白日依山尽，黄河入海流。欲穷千里目，更上一层楼。 The primary use of LLaMA is research on large language models, including"))
print(tokenizer1.tokenize("白日依山尽，黄河入海流。欲穷千里目，更上一层楼。 The primary use of LLaMA is research on large language models, including"))
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16