failspy commited on
Commit
5b657d1
1 Parent(s): 2bd1d53

Upload tokenizer

Browse files
added_tokens.json ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "</box>": 92552,
3
+ "</img>": 92545,
4
+ "</quad>": 92548,
5
+ "</ref>": 92550,
6
+ "<IMG_CONTEXT>": 92546,
7
+ "<box>": 92551,
8
+ "<img>": 92544,
9
+ "<quad>": 92547,
10
+ "<ref>": 92549
11
+ }
special_tokens_map.json ADDED
@@ -0,0 +1,41 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": [
3
+ "<img>",
4
+ "</img>",
5
+ "<IMG_CONTEXT>",
6
+ "<quad>",
7
+ "</quad>",
8
+ "<ref>",
9
+ "</ref>",
10
+ "<box>",
11
+ "</box>"
12
+ ],
13
+ "bos_token": {
14
+ "content": "<s>",
15
+ "lstrip": false,
16
+ "normalized": false,
17
+ "rstrip": false,
18
+ "single_word": false
19
+ },
20
+ "eos_token": {
21
+ "content": "</s>",
22
+ "lstrip": false,
23
+ "normalized": false,
24
+ "rstrip": false,
25
+ "single_word": false
26
+ },
27
+ "pad_token": {
28
+ "content": "</s>",
29
+ "lstrip": false,
30
+ "normalized": false,
31
+ "rstrip": false,
32
+ "single_word": false
33
+ },
34
+ "unk_token": {
35
+ "content": "<unk>",
36
+ "lstrip": false,
37
+ "normalized": false,
38
+ "rstrip": false,
39
+ "single_word": false
40
+ }
41
+ }
tokenization_internlm2.py ADDED
@@ -0,0 +1,235 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Copyright (c) The InternLM team and The HuggingFace Inc. team. All rights reserved.
2
+ #
3
+ # This code is based on transformers/src/transformers/models/llama/tokenization_llama.py
4
+ #
5
+ # Licensed under the Apache License, Version 2.0 (the "License");
6
+ # you may not use this file except in compliance with the License.
7
+ # You may obtain a copy of the License at
8
+ #
9
+ # http://www.apache.org/licenses/LICENSE-2.0
10
+ #
11
+ # Unless required by applicable law or agreed to in writing, software
12
+ # distributed under the License is distributed on an "AS IS" BASIS,
13
+ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
14
+ # See the License for the specific language governing permissions and
15
+ # limitations under the License.
16
+
17
+ """Tokenization classes for InternLM."""
18
+ import os
19
+ from shutil import copyfile
20
+ from typing import Any, Dict, List, Optional, Tuple
21
+
22
+ import sentencepiece as spm
23
+ from transformers.tokenization_utils import PreTrainedTokenizer
24
+ from transformers.utils import logging
25
+
26
+ logger = logging.get_logger(__name__)
27
+
28
+ VOCAB_FILES_NAMES = {'vocab_file': './tokenizer.model'}
29
+
30
+ PRETRAINED_VOCAB_FILES_MAP = {}
31
+
32
+
33
+ # Modified from transformers.model.llama.tokenization_llama.LlamaTokenizer
34
+ class InternLM2Tokenizer(PreTrainedTokenizer):
35
+ """
36
+ Construct a InternLM2 tokenizer. Based on byte-level Byte-Pair-Encoding.
37
+
38
+ Args:
39
+ vocab_file (`str`):
40
+ Path to the vocabulary file.
41
+ """
42
+
43
+ vocab_files_names = VOCAB_FILES_NAMES
44
+ pretrained_vocab_files_map = PRETRAINED_VOCAB_FILES_MAP
45
+ model_input_names = ['input_ids', 'attention_mask']
46
+ _auto_class = 'AutoTokenizer'
47
+
48
+ def __init__(
49
+ self,
50
+ vocab_file,
51
+ unk_token='<unk>',
52
+ bos_token='<s>',
53
+ eos_token='</s>',
54
+ pad_token='</s>',
55
+ sp_model_kwargs: Optional[Dict[str, Any]] = None,
56
+ add_bos_token=True,
57
+ add_eos_token=False,
58
+ decode_with_prefix_space=False,
59
+ clean_up_tokenization_spaces=False,
60
+ **kwargs,
61
+ ):
62
+ self.sp_model_kwargs = {} if sp_model_kwargs is None else sp_model_kwargs
63
+ self.vocab_file = vocab_file
64
+ self.add_bos_token = add_bos_token
65
+ self.add_eos_token = add_eos_token
66
+ self.decode_with_prefix_space = decode_with_prefix_space
67
+ self.sp_model = spm.SentencePieceProcessor(**self.sp_model_kwargs)
68
+ self.sp_model.Load(vocab_file)
69
+ self._no_prefix_space_tokens = None
70
+ super().__init__(
71
+ bos_token=bos_token,
72
+ eos_token=eos_token,
73
+ unk_token=unk_token,
74
+ pad_token=pad_token,
75
+ clean_up_tokenization_spaces=clean_up_tokenization_spaces,
76
+ **kwargs,
77
+ )
78
+
79
+ @property
80
+ def no_prefix_space_tokens(self):
81
+ if self._no_prefix_space_tokens is None:
82
+ vocab = self.convert_ids_to_tokens(list(range(self.vocab_size)))
83
+ self._no_prefix_space_tokens = {i for i, tok in enumerate(vocab) if not tok.startswith('▁')}
84
+ return self._no_prefix_space_tokens
85
+
86
+ @property
87
+ def vocab_size(self):
88
+ """Returns vocab size"""
89
+ return self.sp_model.get_piece_size()
90
+
91
+ @property
92
+ def bos_token_id(self) -> Optional[int]:
93
+ return self.sp_model.bos_id()
94
+
95
+ @property
96
+ def eos_token_id(self) -> Optional[int]:
97
+ return self.sp_model.eos_id()
98
+
99
+ def get_vocab(self):
100
+ """Returns vocab as a dict"""
101
+ vocab = {self.convert_ids_to_tokens(i): i for i in range(self.vocab_size)}
102
+ vocab.update(self.added_tokens_encoder)
103
+ return vocab
104
+
105
+ def _tokenize(self, text):
106
+ """Returns a tokenized string."""
107
+ return self.sp_model.encode(text, out_type=str)
108
+
109
+ def _convert_token_to_id(self, token):
110
+ """Converts a token (str) in an id using the vocab."""
111
+ return self.sp_model.piece_to_id(token)
112
+
113
+ def _convert_id_to_token(self, index):
114
+ """Converts an index (integer) in a token (str) using the vocab."""
115
+ token = self.sp_model.IdToPiece(index)
116
+ return token
117
+
118
+ def _maybe_add_prefix_space(self, tokens, decoded):
119
+ if tokens and tokens[0] not in self.no_prefix_space_tokens:
120
+ return ' ' + decoded
121
+ else:
122
+ return decoded
123
+
124
+ def convert_tokens_to_string(self, tokens):
125
+ """Converts a sequence of tokens (string) in a single string."""
126
+ current_sub_tokens = []
127
+ out_string = ''
128
+ prev_is_special = False
129
+ for token in tokens:
130
+ # make sure that special tokens are not decoded using sentencepiece model
131
+ if token in self.all_special_tokens:
132
+ if not prev_is_special:
133
+ out_string += ' '
134
+ out_string += self.sp_model.decode(current_sub_tokens) + token
135
+ prev_is_special = True
136
+ current_sub_tokens = []
137
+ else:
138
+ current_sub_tokens.append(token)
139
+ prev_is_special = False
140
+ out_string += self.sp_model.decode(current_sub_tokens)
141
+ out_string = self.clean_up_tokenization(out_string)
142
+ out_string = self._maybe_add_prefix_space(tokens=tokens, decoded=out_string)
143
+ return out_string[1:]
144
+
145
+ def save_vocabulary(self, save_directory, filename_prefix: Optional[str] = None) -> Tuple[str]:
146
+ """
147
+ Save the vocabulary and special tokens file to a directory.
148
+
149
+ Args:
150
+ save_directory (`str`):
151
+ The directory in which to save the vocabulary.
152
+
153
+ Returns:
154
+ `Tuple(str)`: Paths to the files saved.
155
+ """
156
+ if not os.path.isdir(save_directory):
157
+ logger.error(f'Vocabulary path ({save_directory}) should be a directory')
158
+ return
159
+ out_vocab_file = os.path.join(
160
+ save_directory, (filename_prefix + '-' if filename_prefix else '') + VOCAB_FILES_NAMES['vocab_file']
161
+ )
162
+
163
+ if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file) and os.path.isfile(self.vocab_file):
164
+ copyfile(self.vocab_file, out_vocab_file)
165
+ elif not os.path.isfile(self.vocab_file):
166
+ with open(out_vocab_file, 'wb') as fi:
167
+ content_spiece_model = self.sp_model.serialized_model_proto()
168
+ fi.write(content_spiece_model)
169
+
170
+ return (out_vocab_file,)
171
+
172
+ def build_inputs_with_special_tokens(self, token_ids_0, token_ids_1=None):
173
+ if self.add_bos_token:
174
+ bos_token_ids = [self.bos_token_id]
175
+ else:
176
+ bos_token_ids = []
177
+
178
+ output = bos_token_ids + token_ids_0
179
+
180
+ if token_ids_1 is not None:
181
+ output = output + token_ids_1
182
+
183
+ if self.add_eos_token:
184
+ output = output + [self.eos_token_id]
185
+
186
+ return output
187
+
188
+ def get_special_tokens_mask(
189
+ self, token_ids_0: List[int], token_ids_1: Optional[List[int]] = None, already_has_special_tokens: bool = False
190
+ ) -> List[int]:
191
+ """
192
+ Retrieve sequence ids from a token list that has no special tokens added. This method is called when adding
193
+ special tokens using the tokenizer `prepare_for_model` method.
194
+
195
+ Args:
196
+ token_ids_0 (`List[int]`):
197
+ List of IDs.
198
+ token_ids_1 (`List[int]`, *optional*):
199
+ Optional second list of IDs for sequence pairs.
200
+ already_has_special_tokens (`bool`, *optional*, defaults to `False`):
201
+ Whether or not the token list is already formatted with special tokens for the model.
202
+
203
+ Returns:
204
+ `List[int]`: A list of integers in the range [0, 1]: 1 for a special token, 0 for a sequence token.
205
+ """
206
+ if already_has_special_tokens:
207
+ return super().get_special_tokens_mask(
208
+ token_ids_0=token_ids_0, token_ids_1=token_ids_1, already_has_special_tokens=True
209
+ )
210
+
211
+ if token_ids_1 is None:
212
+ return [1] + ([0] * len(token_ids_0)) + [1]
213
+ return [1] + ([0] * len(token_ids_0)) + [1, 1] + ([0] * len(token_ids_1)) + [1]
214
+
215
+ def create_token_type_ids_from_sequences(
216
+ self, token_ids_0: List[int], token_ids_1: Optional[List[int]] = None
217
+ ) -> List[int]:
218
+ """
219
+ Create a mask from the two sequences passed to be used in a sequence-pair classification task. T5 does not make
220
+ use of token type ids, therefore a list of zeros is returned.
221
+
222
+ Args:
223
+ token_ids_0 (`List[int]`):
224
+ List of IDs.
225
+ token_ids_1 (`List[int]`, *optional*):
226
+ Optional second list of IDs for sequence pairs.
227
+
228
+ Returns:
229
+ `List[int]`: List of zeros.
230
+ """
231
+ eos = [self.eos_token_id]
232
+
233
+ if token_ids_1 is None:
234
+ return len(token_ids_0 + eos) * [0]
235
+ return len(token_ids_0 + eos + token_ids_1 + eos) * [0]
tokenizer.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f868398fc4e05ee1e8aeba95ddf18ddcc45b8bce55d5093bead5bbf80429b48b
3
+ size 1477754
tokenizer_config.json ADDED
@@ -0,0 +1,173 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "added_tokens_decoder": {
3
+ "0": {
4
+ "content": "<unk>",
5
+ "lstrip": false,
6
+ "normalized": false,
7
+ "rstrip": false,
8
+ "single_word": false,
9
+ "special": true
10
+ },
11
+ "1": {
12
+ "content": "<s>",
13
+ "lstrip": false,
14
+ "normalized": false,
15
+ "rstrip": false,
16
+ "single_word": false,
17
+ "special": true
18
+ },
19
+ "2": {
20
+ "content": "</s>",
21
+ "lstrip": false,
22
+ "normalized": false,
23
+ "rstrip": false,
24
+ "single_word": false,
25
+ "special": true
26
+ },
27
+ "92538": {
28
+ "content": "<|plugin|>",
29
+ "lstrip": false,
30
+ "normalized": false,
31
+ "rstrip": false,
32
+ "single_word": false,
33
+ "special": true
34
+ },
35
+ "92539": {
36
+ "content": "<|interpreter|>",
37
+ "lstrip": false,
38
+ "normalized": false,
39
+ "rstrip": false,
40
+ "single_word": false,
41
+ "special": true
42
+ },
43
+ "92540": {
44
+ "content": "<|action_end|>",
45
+ "lstrip": false,
46
+ "normalized": false,
47
+ "rstrip": false,
48
+ "single_word": false,
49
+ "special": true
50
+ },
51
+ "92541": {
52
+ "content": "<|action_start|>",
53
+ "lstrip": false,
54
+ "normalized": false,
55
+ "rstrip": false,
56
+ "single_word": false,
57
+ "special": true
58
+ },
59
+ "92542": {
60
+ "content": "<|im_end|>",
61
+ "lstrip": false,
62
+ "normalized": false,
63
+ "rstrip": false,
64
+ "single_word": false,
65
+ "special": true
66
+ },
67
+ "92543": {
68
+ "content": "<|im_start|>",
69
+ "lstrip": false,
70
+ "normalized": false,
71
+ "rstrip": false,
72
+ "single_word": false,
73
+ "special": true
74
+ },
75
+ "92544": {
76
+ "content": "<img>",
77
+ "lstrip": false,
78
+ "normalized": false,
79
+ "rstrip": false,
80
+ "single_word": false,
81
+ "special": true
82
+ },
83
+ "92545": {
84
+ "content": "</img>",
85
+ "lstrip": false,
86
+ "normalized": false,
87
+ "rstrip": false,
88
+ "single_word": false,
89
+ "special": true
90
+ },
91
+ "92546": {
92
+ "content": "<IMG_CONTEXT>",
93
+ "lstrip": false,
94
+ "normalized": false,
95
+ "rstrip": false,
96
+ "single_word": false,
97
+ "special": true
98
+ },
99
+ "92547": {
100
+ "content": "<quad>",
101
+ "lstrip": false,
102
+ "normalized": false,
103
+ "rstrip": false,
104
+ "single_word": false,
105
+ "special": true
106
+ },
107
+ "92548": {
108
+ "content": "</quad>",
109
+ "lstrip": false,
110
+ "normalized": false,
111
+ "rstrip": false,
112
+ "single_word": false,
113
+ "special": true
114
+ },
115
+ "92549": {
116
+ "content": "<ref>",
117
+ "lstrip": false,
118
+ "normalized": false,
119
+ "rstrip": false,
120
+ "single_word": false,
121
+ "special": true
122
+ },
123
+ "92550": {
124
+ "content": "</ref>",
125
+ "lstrip": false,
126
+ "normalized": false,
127
+ "rstrip": false,
128
+ "single_word": false,
129
+ "special": true
130
+ },
131
+ "92551": {
132
+ "content": "<box>",
133
+ "lstrip": false,
134
+ "normalized": false,
135
+ "rstrip": false,
136
+ "single_word": false,
137
+ "special": true
138
+ },
139
+ "92552": {
140
+ "content": "</box>",
141
+ "lstrip": false,
142
+ "normalized": false,
143
+ "rstrip": false,
144
+ "single_word": false,
145
+ "special": true
146
+ }
147
+ },
148
+ "additional_special_tokens": [
149
+ "<img>",
150
+ "</img>",
151
+ "<IMG_CONTEXT>",
152
+ "<quad>",
153
+ "</quad>",
154
+ "<ref>",
155
+ "</ref>",
156
+ "<box>",
157
+ "</box>"
158
+ ],
159
+ "auto_map": {
160
+ "AutoTokenizer": [
161
+ "tokenization_internlm2.InternLM2Tokenizer",
162
+ null
163
+ ]
164
+ },
165
+ "bos_token": "<s>",
166
+ "chat_template": "{{ bos_token }}{% for message in messages %}{{'<|im_start|>' + message['role'] + '\n' + message['content'] + '<|im_end|>' + '\n'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>assistant\n' }}{% endif %}",
167
+ "clean_up_tokenization_spaces": false,
168
+ "eos_token": "</s>",
169
+ "model_max_length": 4096,
170
+ "pad_token": "</s>",
171
+ "tokenizer_class": "InternLM2Tokenizer",
172
+ "unk_token": "<unk>"
173
+ }