-
Notifications
You must be signed in to change notification settings - Fork 28.4k
/
Copy pathtest_tokenization_big_bird.py
235 lines (191 loc) · 10.5 KB
/
test_tokenization_big_bird.py
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
# coding=utf-8
# Copyright 2020 The HuggingFace 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.
import os
import unittest
from transformers import BigBirdTokenizer, BigBirdTokenizerFast
from transformers.file_utils import cached_property
from transformers.testing_utils import require_sentencepiece, require_tokenizers, require_torch, slow
from .test_tokenization_common import TokenizerTesterMixin
SPIECE_UNDERLINE = "▁"
SAMPLE_VOCAB = os.path.join(os.path.dirname(os.path.abspath(__file__)), "fixtures/test_sentencepiece.model")
@require_sentencepiece
@require_tokenizers
class BigBirdTokenizationTest(TokenizerTesterMixin, unittest.TestCase):
tokenizer_class = BigBirdTokenizer
rust_tokenizer_class = BigBirdTokenizerFast
test_rust_tokenizer = True
test_sentencepiece = True
def setUp(self):
super().setUp()
tokenizer = self.tokenizer_class(SAMPLE_VOCAB, keep_accents=True)
tokenizer.save_pretrained(self.tmpdirname)
def test_convert_token_and_id(self):
"""Test ``_convert_token_to_id`` and ``_convert_id_to_token``."""
token = "<s>"
token_id = 1
self.assertEqual(self.get_tokenizer()._convert_token_to_id(token), token_id)
self.assertEqual(self.get_tokenizer()._convert_id_to_token(token_id), token)
def test_get_vocab(self):
vocab_keys = list(self.get_tokenizer().get_vocab().keys())
self.assertEqual(vocab_keys[0], "<unk>")
self.assertEqual(vocab_keys[1], "<s>")
self.assertEqual(vocab_keys[-1], "[MASK]")
self.assertEqual(len(vocab_keys), 1_004)
def test_vocab_size(self):
self.assertEqual(self.get_tokenizer().vocab_size, 1_000)
def test_rust_and_python_full_tokenizers(self):
if not self.test_rust_tokenizer:
return
tokenizer = self.get_tokenizer()
rust_tokenizer = self.get_rust_tokenizer()
sequence = "I was born in 92000, and this is falsé."
tokens = tokenizer.tokenize(sequence)
rust_tokens = rust_tokenizer.tokenize(sequence)
self.assertListEqual(tokens, rust_tokens)
ids = tokenizer.encode(sequence, add_special_tokens=False)
rust_ids = rust_tokenizer.encode(sequence, add_special_tokens=False)
self.assertListEqual(ids, rust_ids)
rust_tokenizer = self.get_rust_tokenizer()
ids = tokenizer.encode(sequence)
rust_ids = rust_tokenizer.encode(sequence)
self.assertListEqual(ids, rust_ids)
def test_full_tokenizer(self):
tokenizer = BigBirdTokenizer(SAMPLE_VOCAB, keep_accents=True)
tokens = tokenizer.tokenize("This is a test")
self.assertListEqual(tokens, ["▁This", "▁is", "▁a", "▁t", "est"])
self.assertListEqual(
tokenizer.convert_tokens_to_ids(tokens),
[285, 46, 10, 170, 382],
)
tokens = tokenizer.tokenize("I was born in 92000, and this is falsé.")
self.assertListEqual(
tokens,
[
SPIECE_UNDERLINE + "I",
SPIECE_UNDERLINE + "was",
SPIECE_UNDERLINE + "b",
"or",
"n",
SPIECE_UNDERLINE + "in",
SPIECE_UNDERLINE + "",
"9",
"2",
"0",
"0",
"0",
",",
SPIECE_UNDERLINE + "and",
SPIECE_UNDERLINE + "this",
SPIECE_UNDERLINE + "is",
SPIECE_UNDERLINE + "f",
"al",
"s",
"é",
".",
],
)
ids = tokenizer.convert_tokens_to_ids(tokens)
self.assertListEqual(
ids,
[8, 21, 84, 55, 24, 19, 7, 0, 602, 347, 347, 347, 3, 12, 66, 46, 72, 80, 6, 0, 4],
)
back_tokens = tokenizer.convert_ids_to_tokens(ids)
self.assertListEqual(
back_tokens,
[
SPIECE_UNDERLINE + "I",
SPIECE_UNDERLINE + "was",
SPIECE_UNDERLINE + "b",
"or",
"n",
SPIECE_UNDERLINE + "in",
SPIECE_UNDERLINE + "",
"<unk>",
"2",
"0",
"0",
"0",
",",
SPIECE_UNDERLINE + "and",
SPIECE_UNDERLINE + "this",
SPIECE_UNDERLINE + "is",
SPIECE_UNDERLINE + "f",
"al",
"s",
"<unk>",
".",
],
)
@cached_property
def big_tokenizer(self):
return BigBirdTokenizer.from_pretrained("google/bigbird-roberta-base")
@slow
def test_tokenization_base_easy_symbols(self):
symbols = "Hello World!"
original_tokenizer_encodings = [65, 18536, 2260, 101, 66]
self.assertListEqual(original_tokenizer_encodings, self.big_tokenizer.encode(symbols))
@slow
def test_tokenization_base_hard_symbols(self):
symbols = 'This is a very long text with a lot of weird characters, such as: . , ~ ? ( ) " [ ] ! : - . Also we will add words that should not exsist and be tokenized to <unk>, such as saoneuhaoesuth'
# fmt: off
original_tokenizer_encodings = [65, 871, 419, 358, 946, 991, 2521, 452, 358, 1357, 387, 7751, 3536, 112, 985, 456, 126, 865, 938, 5400, 5734, 458, 1368, 467, 786, 2462, 5246, 1159, 633, 865, 4519, 457, 582, 852, 2557, 427, 916, 508, 405, 34324, 497, 391, 408, 11342, 1244, 385, 100, 938, 985, 456, 574, 362, 12597, 3200, 3129, 1172, 66] # noqa: E231
# fmt: on
self.assertListEqual(original_tokenizer_encodings, self.big_tokenizer.encode(symbols))
@require_torch
@slow
def test_torch_encode_plus_sent_to_model(self):
import torch
from transformers import BigBirdConfig, BigBirdModel
# Build sequence
first_ten_tokens = list(self.big_tokenizer.get_vocab().keys())[:10]
sequence = " ".join(first_ten_tokens)
encoded_sequence = self.big_tokenizer.encode_plus(sequence, return_tensors="pt", return_token_type_ids=False)
batch_encoded_sequence = self.big_tokenizer.batch_encode_plus(
[sequence + " " + sequence], return_tensors="pt", return_token_type_ids=False
)
config = BigBirdConfig(attention_type="original_full")
model = BigBirdModel(config)
assert model.get_input_embeddings().weight.shape[0] >= self.big_tokenizer.vocab_size
with torch.no_grad():
model(**encoded_sequence)
model(**batch_encoded_sequence)
@slow
def test_special_tokens(self):
"""
To reproduce:
$ wget https://github.com/google-research/bigbird/blob/master/bigbird/vocab/gpt2.model?raw=true
$ mv gpt2.model?raw=true gpt2.model
```
import tensorflow_text as tft
import tensorflow as tf
vocab_model_file = "./gpt2.model"
tokenizer = tft.SentencepieceTokenizer(model=tf.io.gfile.GFile(vocab_model_file, "rb").read()))
ids = tokenizer.tokenize("Paris is the [MASK].")
ids = tf.concat([tf.constant([65]), ids, tf.constant([66])], axis=0)
detokenized = tokenizer.detokenize(ids) # should give [CLS] Paris is the [MASK].[SEP]
"""
tokenizer = BigBirdTokenizer.from_pretrained("google/bigbird-roberta-base")
decoded_text = tokenizer.decode(tokenizer("Paris is the [MASK].").input_ids)
self.assertTrue(decoded_text == "[CLS] Paris is the[MASK].[SEP]")
@slow
def test_tokenizer_integration(self):
# fmt: off
expected_encoding = {'input_ids': [[65, 39286, 458, 36335, 2001, 456, 13073, 13266, 455, 113, 7746, 1741, 11157, 391, 13073, 13266, 455, 113, 3967, 35412, 113, 4936, 109, 3870, 2377, 113, 30084, 45720, 458, 134, 17496, 112, 503, 11672, 113, 118, 112, 5665, 13347, 38687, 112, 1496, 31389, 112, 3268, 47264, 134, 962, 112, 16377, 8035, 23130, 430, 12169, 15518, 28592, 458, 146, 41697, 109, 391, 12169, 15518, 16689, 458, 146, 41358, 109, 452, 726, 4034, 111, 763, 35412, 5082, 388, 1903, 111, 9051, 391, 2870, 48918, 1900, 1123, 550, 998, 112, 9586, 15985, 455, 391, 410, 22955, 37636, 114, 66], [65, 448, 17496, 419, 3663, 385, 763, 113, 27533, 2870, 3283, 13043, 1639, 24713, 523, 656, 24013, 18550, 2521, 517, 27014, 21244, 420, 1212, 1465, 391, 927, 4833, 388, 578, 11786, 114, 66, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [65, 484, 2169, 7687, 21932, 18146, 726, 363, 17032, 3391, 114, 66, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]], 'attention_mask': [[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1], [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]]} # noqa: E501
# fmt: on
self.tokenizer_integration_test_util(
expected_encoding=expected_encoding,
model_name="google/bigbird-roberta-base",
revision="215c99f1600e06f83acce68422f2035b2b5c3510",
)