gchhablani commited on
Commit
4f3d5e1
1 Parent(s): 72a9e56

Add README.md

Browse files
Files changed (1) hide show
  1. README.md +10 -3
README.md CHANGED
@@ -9,12 +9,13 @@ datasets:
9
  - wikipedia
10
  ---
11
  # MultiBERTs Seed 23 (uncased)
12
- Seed 23 pretrained BERT model on English language using a masked language modeling (MLM) objective. It was introduced in
13
  [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in
14
  [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This model is uncased: it does not make a difference
15
  between english and English.
16
 
17
  Disclaimer: The team releasing MultiBERTs did not write a model card for this model so this model card has been written by [gchhablani](https://hf.co/gchhablani).
 
18
  ## Model description
19
  MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it
20
  was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of
@@ -30,14 +31,16 @@ was pretrained with two objectives:
30
  predict if the two sentences were following each other or not.
31
  This way, the model learns an inner representation of the English language that can then be used to extract features
32
  useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard
33
- classifier using the features produced by the BERT model as inputs.
 
34
  ## Intended uses & limitations
35
  You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to
36
- be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=bert) to look for
37
  fine-tuned versions on a task that interests you.
38
  Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked)
39
  to make decisions, such as sequence classification, token classification or question answering. For tasks such as text
40
  generation you should look at model like GPT2.
 
41
  ### How to use
42
  Here is how to use this model to get the features of a given text in PyTorch:
43
  ```python
@@ -48,6 +51,7 @@ text = "Replace me by any text you'd like."
48
  encoded_input = tokenizer(text, return_tensors='pt')
49
  output = model(**encoded_input)
50
  ```
 
51
  ### Limitations and bias
52
  Even if the training data used for this model could be characterized as fairly neutral, this model can have biased
53
  predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular
@@ -58,6 +62,7 @@ The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com
58
  unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and
59
  headers).
60
  ## Training procedure
 
61
  ### Preprocessing
62
  The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are
63
  then of the form:
@@ -73,11 +78,13 @@ The details of the masking procedure for each sentence are the following:
73
  - In 80% of the cases, the masked tokens are replaced by `[MASK]`.
74
  - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace.
75
  - In the 10% remaining cases, the masked tokens are left as is.
 
76
  ### Pretraining
77
  The model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size
78
  of 256. The sequence length was set to 512 throughout. The optimizer
79
  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,
80
  learning rate warmup for 10,000 steps and linear decay of the learning rate after.
 
81
  ### BibTeX entry and citation info
82
  ```bibtex
83
  @article{DBLP:journals/corr/abs-2106-16163,
 
9
  - wikipedia
10
  ---
11
  # MultiBERTs Seed 23 (uncased)
12
+ Seed 23 MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in
13
  [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in
14
  [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This model is uncased: it does not make a difference
15
  between english and English.
16
 
17
  Disclaimer: The team releasing MultiBERTs did not write a model card for this model so this model card has been written by [gchhablani](https://hf.co/gchhablani).
18
+
19
  ## Model description
20
  MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it
21
  was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of
 
31
  predict if the two sentences were following each other or not.
32
  This way, the model learns an inner representation of the English language that can then be used to extract features
33
  useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard
34
+ classifier using the features produced by the MultiBERTs model as inputs.
35
+
36
  ## Intended uses & limitations
37
  You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to
38
+ be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for
39
  fine-tuned versions on a task that interests you.
40
  Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked)
41
  to make decisions, such as sequence classification, token classification or question answering. For tasks such as text
42
  generation you should look at model like GPT2.
43
+
44
  ### How to use
45
  Here is how to use this model to get the features of a given text in PyTorch:
46
  ```python
 
51
  encoded_input = tokenizer(text, return_tensors='pt')
52
  output = model(**encoded_input)
53
  ```
54
+
55
  ### Limitations and bias
56
  Even if the training data used for this model could be characterized as fairly neutral, this model can have biased
57
  predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular
 
62
  unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and
63
  headers).
64
  ## Training procedure
65
+
66
  ### Preprocessing
67
  The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are
68
  then of the form:
 
78
  - In 80% of the cases, the masked tokens are replaced by `[MASK]`.
79
  - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace.
80
  - In the 10% remaining cases, the masked tokens are left as is.
81
+
82
  ### Pretraining
83
  The model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size
84
  of 256. The sequence length was set to 512 throughout. The optimizer
85
  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,
86
  learning rate warmup for 10,000 steps and linear decay of the learning rate after.
87
+
88
  ### BibTeX entry and citation info
89
  ```bibtex
90
  @article{DBLP:journals/corr/abs-2106-16163,