kiranr commited on
Commit
d8c7aa3
1 Parent(s): f84e741

Upload tokenizer

Browse files
special_tokens_map.json ADDED
@@ -0,0 +1,30 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token": {
3
+ "content": "<s>",
4
+ "lstrip": false,
5
+ "normalized": false,
6
+ "rstrip": false,
7
+ "single_word": false
8
+ },
9
+ "eos_token": {
10
+ "content": "</s>",
11
+ "lstrip": false,
12
+ "normalized": false,
13
+ "rstrip": false,
14
+ "single_word": false
15
+ },
16
+ "pad_token": {
17
+ "content": "</s>",
18
+ "lstrip": false,
19
+ "normalized": false,
20
+ "rstrip": false,
21
+ "single_word": false
22
+ },
23
+ "unk_token": {
24
+ "content": "<unk>",
25
+ "lstrip": false,
26
+ "normalized": false,
27
+ "rstrip": false,
28
+ "single_word": false
29
+ }
30
+ }
tokenization_internlm.py ADDED
@@ -0,0 +1,240 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # coding=utf-8
2
+ # Copyright (c) InternLM. All rights reserved.
3
+ #
4
+ # This code is based on EleutherAI's GPT-NeoX library and the GPT-NeoX
5
+ # and OPT implementations in this library. It has been modified from its
6
+ # original forms to accommodate minor architectural differences compared
7
+ # to GPT-NeoX and OPT used by the Meta AI team that trained the model.
8
+ #
9
+ # Licensed under the Apache License, Version 2.0 (the "License");
10
+ # you may not use this file except in compliance with the License.
11
+ # You may obtain a copy of the License at
12
+ #
13
+ # http://www.apache.org/licenses/LICENSE-2.0
14
+ #
15
+ # Unless required by applicable law or agreed to in writing, software
16
+ # distributed under the License is distributed on an "AS IS" BASIS,
17
+ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
18
+ # See the License for the specific language governing permissions and
19
+ # limitations under the License.
20
+
21
+ """Tokenization classes for IntermLM."""
22
+ import os
23
+ from shutil import copyfile
24
+ from typing import Any, Dict, List, Optional, Tuple
25
+
26
+ import sentencepiece as spm
27
+ from transformers.tokenization_utils import PreTrainedTokenizer
28
+ from transformers.utils import logging
29
+
30
+ logger = logging.get_logger(__name__)
31
+
32
+ VOCAB_FILES_NAMES = {"vocab_file": "./tokenizer.model"}
33
+
34
+ PRETRAINED_VOCAB_FILES_MAP = {}
35
+
36
+
37
+ class InternLMTokenizer(PreTrainedTokenizer):
38
+ """
39
+ Construct a InternLM tokenizer. Based on byte-level Byte-Pair-Encoding.
40
+
41
+ Args:
42
+ vocab_file (`str`):
43
+ Path to the vocabulary file.
44
+ """
45
+
46
+ vocab_files_names = VOCAB_FILES_NAMES
47
+ pretrained_vocab_files_map = PRETRAINED_VOCAB_FILES_MAP
48
+ model_input_names = ["input_ids", "attention_mask"]
49
+ _auto_class = "AutoTokenizer"
50
+
51
+ def __init__(
52
+ self,
53
+ vocab_file,
54
+ unk_token="<unk>",
55
+ bos_token="<s>",
56
+ eos_token="</s>",
57
+ pad_token="</s>",
58
+ sp_model_kwargs: Optional[Dict[str, Any]] = None,
59
+ add_bos_token=True,
60
+ add_eos_token=False,
61
+ decode_with_prefix_space=False,
62
+ clean_up_tokenization_spaces=False,
63
+ **kwargs,
64
+ ):
65
+ self.sp_model_kwargs = {} if sp_model_kwargs is None else sp_model_kwargs
66
+ self.vocab_file = vocab_file
67
+ self.add_bos_token = add_bos_token
68
+ self.add_eos_token = add_eos_token
69
+ self.decode_with_prefix_space = decode_with_prefix_space
70
+ self.sp_model = spm.SentencePieceProcessor(**self.sp_model_kwargs)
71
+ self.sp_model.Load(vocab_file)
72
+ self._no_prefix_space_tokens = None
73
+ super().__init__(
74
+ bos_token=bos_token,
75
+ eos_token=eos_token,
76
+ unk_token=unk_token,
77
+ pad_token=pad_token,
78
+ clean_up_tokenization_spaces=clean_up_tokenization_spaces,
79
+ **kwargs,
80
+ )
81
+
82
+ """ Initialization"""
83
+
84
+ @property
85
+ def no_prefix_space_tokens(self):
86
+ if self._no_prefix_space_tokens is None:
87
+ vocab = self.convert_ids_to_tokens(list(range(self.vocab_size)))
88
+ self._no_prefix_space_tokens = {i for i, tok in enumerate(vocab) if not tok.startswith("▁")}
89
+ return self._no_prefix_space_tokens
90
+
91
+ @property
92
+ def vocab_size(self):
93
+ """Returns vocab size"""
94
+ return self.sp_model.get_piece_size()
95
+
96
+ @property
97
+ def bos_token_id(self) -> Optional[int]:
98
+ return self.sp_model.bos_id()
99
+
100
+ @property
101
+ def eos_token_id(self) -> Optional[int]:
102
+ return self.sp_model.eos_id()
103
+
104
+ def get_vocab(self):
105
+ """Returns vocab as a dict"""
106
+ vocab = {self.convert_ids_to_tokens(i): i for i in range(self.vocab_size)}
107
+ vocab.update(self.added_tokens_encoder)
108
+ return vocab
109
+
110
+ def _tokenize(self, text):
111
+ """Returns a tokenized string."""
112
+ return self.sp_model.encode(text, out_type=str)
113
+
114
+ def _convert_token_to_id(self, token):
115
+ """Converts a token (str) in an id using the vocab."""
116
+ return self.sp_model.piece_to_id(token)
117
+
118
+ def _convert_id_to_token(self, index):
119
+ """Converts an index (integer) in a token (str) using the vocab."""
120
+ token = self.sp_model.IdToPiece(index)
121
+ return token
122
+
123
+ def _maybe_add_prefix_space(self, tokens, decoded):
124
+ if tokens and tokens[0] not in self.no_prefix_space_tokens:
125
+ return " " + decoded
126
+ else:
127
+ return decoded
128
+
129
+ def convert_tokens_to_string(self, tokens):
130
+ """Converts a sequence of tokens (string) in a single string."""
131
+ current_sub_tokens = []
132
+ out_string = ""
133
+ prev_is_special = False
134
+ for token in tokens:
135
+ # make sure that special tokens are not decoded using sentencepiece model
136
+ if token in self.all_special_tokens:
137
+ if not prev_is_special:
138
+ out_string += " "
139
+ out_string += self.sp_model.decode(current_sub_tokens) + token
140
+ prev_is_special = True
141
+ current_sub_tokens = []
142
+ else:
143
+ current_sub_tokens.append(token)
144
+ prev_is_special = False
145
+ out_string += self.sp_model.decode(current_sub_tokens)
146
+ out_string = self.clean_up_tokenization(out_string)
147
+ out_string = self._maybe_add_prefix_space(tokens=tokens, decoded=out_string)
148
+ return out_string[1:]
149
+
150
+ def save_vocabulary(self, save_directory, filename_prefix: Optional[str] = None) -> Tuple[str]:
151
+ """
152
+ Save the vocabulary and special tokens file to a directory.
153
+
154
+ Args:
155
+ save_directory (`str`):
156
+ The directory in which to save the vocabulary.
157
+
158
+ Returns:
159
+ `Tuple(str)`: Paths to the files saved.
160
+ """
161
+ if not os.path.isdir(save_directory):
162
+ logger.error(f"Vocabulary path ({save_directory}) should be a directory")
163
+ return
164
+ out_vocab_file = os.path.join(
165
+ save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"]
166
+ )
167
+
168
+ if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file) and os.path.isfile(self.vocab_file):
169
+ copyfile(self.vocab_file, out_vocab_file)
170
+ elif not os.path.isfile(self.vocab_file):
171
+ with open(out_vocab_file, "wb") as fi:
172
+ content_spiece_model = self.sp_model.serialized_model_proto()
173
+ fi.write(content_spiece_model)
174
+
175
+ return (out_vocab_file,)
176
+
177
+ def build_inputs_with_special_tokens(self, token_ids_0, token_ids_1=None):
178
+ if self.add_bos_token:
179
+ bos_token_ids = [self.bos_token_id]
180
+ else:
181
+ bos_token_ids = []
182
+
183
+ output = bos_token_ids + token_ids_0
184
+
185
+ if token_ids_1 is not None:
186
+ output = output + token_ids_1
187
+
188
+ if self.add_eos_token:
189
+ output = output + [self.eos_token_id]
190
+
191
+ return output
192
+
193
+ def get_special_tokens_mask(
194
+ self, token_ids_0: List[int], token_ids_1: Optional[List[int]] = None, already_has_special_tokens: bool = False
195
+ ) -> List[int]:
196
+ """
197
+ Retrieve sequence ids from a token list that has no special tokens added. This method is called when adding
198
+ special tokens using the tokenizer `prepare_for_model` method.
199
+
200
+ Args:
201
+ token_ids_0 (`List[int]`):
202
+ List of IDs.
203
+ token_ids_1 (`List[int]`, *optional*):
204
+ Optional second list of IDs for sequence pairs.
205
+ already_has_special_tokens (`bool`, *optional*, defaults to `False`):
206
+ Whether or not the token list is already formatted with special tokens for the model.
207
+
208
+ Returns:
209
+ `List[int]`: A list of integers in the range [0, 1]: 1 for a special token, 0 for a sequence token.
210
+ """
211
+ if already_has_special_tokens:
212
+ return super().get_special_tokens_mask(
213
+ token_ids_0=token_ids_0, token_ids_1=token_ids_1, already_has_special_tokens=True
214
+ )
215
+
216
+ if token_ids_1 is None:
217
+ return [1] + ([0] * len(token_ids_0)) + [1]
218
+ return [1] + ([0] * len(token_ids_0)) + [1, 1] + ([0] * len(token_ids_1)) + [1]
219
+
220
+ def create_token_type_ids_from_sequences(
221
+ self, token_ids_0: List[int], token_ids_1: Optional[List[int]] = None
222
+ ) -> List[int]:
223
+ """
224
+ Create a mask from the two sequences passed to be used in a sequence-pair classification task. T5 does not make
225
+ use of token type ids, therefore a list of zeros is returned.
226
+
227
+ Args:
228
+ token_ids_0 (`List[int]`):
229
+ List of IDs.
230
+ token_ids_1 (`List[int]`, *optional*):
231
+ Optional second list of IDs for sequence pairs.
232
+
233
+ Returns:
234
+ `List[int]`: List of zeros.
235
+ """
236
+ eos = [self.eos_token_id]
237
+
238
+ if token_ids_1 is None:
239
+ return len(token_ids_0 + eos) * [0]
240
+ 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 +1,90 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ },
76
+ "auto_map": {
77
+ "AutoTokenizer": [
78
+ "tokenization_internlm.InternLMTokenizer",
79
+ null
80
+ ]
81
+ },
82
+ "bos_token": "<s>",
83
+ "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 %}",
84
+ "clean_up_tokenization_spaces": false,
85
+ "eos_token": "</s>",
86
+ "model_max_length": 1000000000000000019884624838656,
87
+ "pad_token": "</s>",
88
+ "tokenizer_class": "InternLMTokenizer",
89
+ "unk_token": "<unk>"
90
+ }