bert-base-uncased / README.md
1
---
2
language: en
3
tags:
4
- exbert
5
license: apache-2.0
6
datasets:
7
- bookcorpus
8
- wikipedia
9
---
10
11
# BERT base model (uncased)
12
13
Pretrained model on English language using a masked language modeling (MLM) objective. It was introduced in
14
[this paper](https://arxiv.org/abs/1810.04805) and first released in
15
[this repository](https://github.com/google-research/bert). This model is uncased: it does not make a difference
16
between english and English.
17
18
Disclaimer: The team releasing BERT did not write a model card for this model so this model card has been written by
19
the Hugging Face team.
20
21
## Model description
22
23
BERT is a transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it
24
was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of
25
publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it
26
was pretrained with two objectives:
27
28
- Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run
29
  the entire masked sentence through the model and has to predict the masked words. This is different from traditional
30
  recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like
31
  GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the
32
  sentence.
33
- Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes
34
  they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to
35
  predict if the two sentences were following each other or not.
36
37
This way, the model learns an inner representation of the English language that can then be used to extract features
38
useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard
39
classifier using the features produced by the BERT model as inputs.
40
41
## Intended uses & limitations
42
43
You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to
44
be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=bert) to look for
45
fine-tuned versions on a task that interests you.
46
47
Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked)
48
to make decisions, such as sequence classification, token classification or question answering. For tasks such as text
49
generation you should look at model like GPT2.
50
51
### How to use
52
53
You can use this model directly with a pipeline for masked language modeling:
54
55
```python
56
>>> from transformers import pipeline
57
>>> unmasker = pipeline('fill-mask', model='bert-base-uncased')
58
>>> unmasker("Hello I'm a [MASK] model.")
59
60
[{'sequence': "[CLS] hello i'm a fashion model. [SEP]",
61
  'score': 0.1073106899857521,
62
  'token': 4827,
63
  'token_str': 'fashion'},
64
 {'sequence': "[CLS] hello i'm a role model. [SEP]",
65
  'score': 0.08774490654468536,
66
  'token': 2535,
67
  'token_str': 'role'},
68
 {'sequence': "[CLS] hello i'm a new model. [SEP]",
69
  'score': 0.05338378623127937,
70
  'token': 2047,
71
  'token_str': 'new'},
72
 {'sequence': "[CLS] hello i'm a super model. [SEP]",
73
  'score': 0.04667217284440994,
74
  'token': 3565,
75
  'token_str': 'super'},
76
 {'sequence': "[CLS] hello i'm a fine model. [SEP]",
77
  'score': 0.027095865458250046,
78
  'token': 2986,
79
  'token_str': 'fine'}]
80
```
81
82
Here is how to use this model to get the features of a given text in PyTorch:
83
84
```python
85
from transformers import BertTokenizer, BertModel
86
tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')
87
model = BertModel.from_pretrained("bert-base-uncased")
88
text = "Replace me by any text you'd like."
89
encoded_input = tokenizer(text, return_tensors='pt')
90
output = model(**encoded_input)
91
```
92
93
and in TensorFlow:
94
95
```python
96
from transformers import BertTokenizer, TFBertModel
97
tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')
98
model = TFBertModel.from_pretrained("bert-base-uncased")
99
text = "Replace me by any text you'd like."
100
encoded_input = tokenizer(text, return_tensors='tf')
101
output = model(encoded_input)
102
```
103
104
### Limitations and bias
105
106
Even if the training data used for this model could be characterized as fairly neutral, this model can have biased
107
predictions:
108
109
```python
110
>>> from transformers import pipeline
111
>>> unmasker = pipeline('fill-mask', model='bert-base-uncased')
112
>>> unmasker("The man worked as a [MASK].")
113
114
[{'sequence': '[CLS] the man worked as a carpenter. [SEP]',
115
  'score': 0.09747550636529922,
116
  'token': 10533,
117
  'token_str': 'carpenter'},
118
 {'sequence': '[CLS] the man worked as a waiter. [SEP]',
119
  'score': 0.0523831807076931,
120
  'token': 15610,
121
  'token_str': 'waiter'},
122
 {'sequence': '[CLS] the man worked as a barber. [SEP]',
123
  'score': 0.04962705448269844,
124
  'token': 13362,
125
  'token_str': 'barber'},
126
 {'sequence': '[CLS] the man worked as a mechanic. [SEP]',
127
  'score': 0.03788609802722931,
128
  'token': 15893,
129
  'token_str': 'mechanic'},
130
 {'sequence': '[CLS] the man worked as a salesman. [SEP]',
131
  'score': 0.037680890411138535,
132
  'token': 18968,
133
  'token_str': 'salesman'}]
134
135
>>> unmasker("The woman worked as a [MASK].")
136
137
[{'sequence': '[CLS] the woman worked as a nurse. [SEP]',
138
  'score': 0.21981462836265564,
139
  'token': 6821,
140
  'token_str': 'nurse'},
141
 {'sequence': '[CLS] the woman worked as a waitress. [SEP]',
142
  'score': 0.1597415804862976,
143
  'token': 13877,
144
  'token_str': 'waitress'},
145
 {'sequence': '[CLS] the woman worked as a maid. [SEP]',
146
  'score': 0.1154729500412941,
147
  'token': 10850,
148
  'token_str': 'maid'},
149
 {'sequence': '[CLS] the woman worked as a prostitute. [SEP]',
150
  'score': 0.037968918681144714,
151
  'token': 19215,
152
  'token_str': 'prostitute'},
153
 {'sequence': '[CLS] the woman worked as a cook. [SEP]',
154
  'score': 0.03042375110089779,
155
  'token': 5660,
156
  'token_str': 'cook'}]
157
```
158
159
This bias will also affect all fine-tuned versions of this model.
160
161
## Training data
162
163
The BERT model was pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038
164
unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and
165
headers).
166
167
## Training procedure
168
169
### Preprocessing
170
171
The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are
172
then of the form:
173
174
```
175
[CLS] Sentence A [SEP] Sentence B [SEP]
176
```
177
178
With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in
179
the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a
180
consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two
181
"sentences" has a combined length of less than 512 tokens.
182
183
The details of the masking procedure for each sentence are the following:
184
- 15% of the tokens are masked.
185
- In 80% of the cases, the masked tokens are replaced by `[MASK]`.
186
- In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace.
187
- In the 10% remaining cases, the masked tokens are left as is.
188
189
### Pretraining
190
191
The model was trained on 4 cloud TPUs in Pod configuration (16 TPU chips total) for one million steps with a batch size
192
of 256. The sequence length was limited to 128 tokens for 90% of the steps and 512 for the remaining 10%. The optimizer
193
used is Adam with a learning rate of 1e-4, \\(\beta_{1} = 0.9\\) and \\(\beta_{2} = 0.999\\), a weight decay of 0.01,
194
learning rate warmup for 10,000 steps and linear decay of the learning rate after.
195
196
## Evaluation results
197
198
When fine-tuned on downstream tasks, this model achieves the following results:
199
200
Glue test results:
201
202
| Task | MNLI-(m/mm) | QQP  | QNLI | SST-2 | CoLA | STS-B | MRPC | RTE  | Average |
203
|:----:|:-----------:|:----:|:----:|:-----:|:----:|:-----:|:----:|:----:|:-------:|
204
|      | 84.6/83.4   | 71.2 | 90.5 | 93.5  | 52.1 | 85.8  | 88.9 | 66.4 | 79.6    |
205
206
207
### BibTeX entry and citation info
208
209
```bibtex
210
@article{DBLP:journals/corr/abs-1810-04805,
211
  author    = {Jacob Devlin and
212
               Ming{-}Wei Chang and
213
               Kenton Lee and
214
               Kristina Toutanova},
215
  title     = {{BERT:} Pre-training of Deep Bidirectional Transformers for Language
216
               Understanding},
217
  journal   = {CoRR},
218
  volume    = {abs/1810.04805},
219
  year      = {2018},
220
  url       = {http://arxiv.org/abs/1810.04805},
221
  archivePrefix = {arXiv},
222
  eprint    = {1810.04805},
223
  timestamp = {Tue, 30 Oct 2018 20:39:56 +0100},
224
  biburl    = {https://dblp.org/rec/journals/corr/abs-1810-04805.bib},
225
  bibsource = {dblp computer science bibliography, https://dblp.org}
226
}
227
```
228
229
<a href="https://huggingface.co/exbert/?model=bert-base-uncased">
230
	<img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png">
231
</a>
232