Split (1)

train · 58.9k rows

repo_id stringlengths 4 122	author stringlengths 2 38 ⌀	model_type stringlengths 2 33 ⌀	files_per_repo int64 2 39k	downloads_30d int64 0 33.7M	library stringlengths 2 37 ⌀	pipeline stringlengths 5 30 ⌀	pytorch bool 2 classes	tensorflow bool 2 classes	jax bool 2 classes	license stringlengths 2 33 ⌀	languages stringlengths 2 1.63k ⌀	datasets stringlengths 2 2.58k ⌀	co2 stringlengths 6 258 ⌀	tags stringlengths 2 513	has_model_index bool 2 classes	has_metadata bool 2 classes	has_text bool 1 class	text_length int64 201 598k	readme stringlengths 0 598k
MultiBertGunjanPatrick/multiberts-seed-11	MultiBertGunjanPatrick	bert	6	2	transformers	null	false	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts']	false	true	true	6,323	# MultiBERTs Seed 11 (uncased) Seed 11 MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-11') model = BertModel.from_pretrained("multiberts-seed-11") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-12	MultiBertGunjanPatrick	bert	7	2	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts']	false	true	true	6,323	# MultiBERTs Seed 12 (uncased) Seed 12 MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-12') model = BertModel.from_pretrained("multiberts-seed-12") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-13	MultiBertGunjanPatrick	bert	7	3	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts']	false	true	true	6,323	# MultiBERTs Seed 13 (uncased) Seed 13 MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-13') model = BertModel.from_pretrained("multiberts-seed-13") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-14	MultiBertGunjanPatrick	bert	7	2	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts']	false	true	true	6,323	# MultiBERTs Seed 14 (uncased) Seed 14 MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-14') model = BertModel.from_pretrained("multiberts-seed-14") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-15	MultiBertGunjanPatrick	bert	7	2	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts']	false	true	true	6,323	# MultiBERTs Seed 15 (uncased) Seed 15 MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-15') model = BertModel.from_pretrained("multiberts-seed-15") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-16	MultiBertGunjanPatrick	bert	7	2	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts']	false	true	true	6,323	# MultiBERTs Seed 16 (uncased) Seed 16 MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-16') model = BertModel.from_pretrained("multiberts-seed-16") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-17	MultiBertGunjanPatrick	bert	7	2	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts']	false	true	true	6,323	# MultiBERTs Seed 17 (uncased) Seed 17 MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-17') model = BertModel.from_pretrained("multiberts-seed-17") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-18	MultiBertGunjanPatrick	bert	7	2	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts']	false	true	true	6,323	# MultiBERTs Seed 18 (uncased) Seed 18 MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-18') model = BertModel.from_pretrained("multiberts-seed-18") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-19	MultiBertGunjanPatrick	bert	7	2	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts']	false	true	true	6,323	# MultiBERTs Seed 19 (uncased) Seed 19 MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-19') model = BertModel.from_pretrained("multiberts-seed-19") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-2-0k	MultiBertGunjanPatrick	bert	7	2	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-2']	false	true	true	6,475	# MultiBERTs Seed 2 Checkpoint 0k (uncased) Seed 2 intermediate checkpoint 0k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-2](https://hf.co/multberts-seed-2). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-2-0k') model = BertModel.from_pretrained("multiberts-seed-2-0k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-2-1000k	MultiBertGunjanPatrick	bert	7	2	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-2']	false	true	true	6,487	# MultiBERTs Seed 2 Checkpoint 1000k (uncased) Seed 2 intermediate checkpoint 1000k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-2](https://hf.co/multberts-seed-2). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-2-1000k') model = BertModel.from_pretrained("multiberts-seed-2-1000k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-2-100k	MultiBertGunjanPatrick	bert	7	5	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-2']	false	true	true	6,483	# MultiBERTs Seed 2 Checkpoint 100k (uncased) Seed 2 intermediate checkpoint 100k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-2](https://hf.co/multberts-seed-2). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-2-100k') model = BertModel.from_pretrained("multiberts-seed-2-100k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-2-1100k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-2']	false	true	true	6,487	# MultiBERTs Seed 2 Checkpoint 1100k (uncased) Seed 2 intermediate checkpoint 1100k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-2](https://hf.co/multberts-seed-2). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-2-1100k') model = BertModel.from_pretrained("multiberts-seed-2-1100k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-2-1200k	MultiBertGunjanPatrick	bert	7	5	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-2']	false	true	true	6,487	# MultiBERTs Seed 2 Checkpoint 1200k (uncased) Seed 2 intermediate checkpoint 1200k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-2](https://hf.co/multberts-seed-2). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-2-1200k') model = BertModel.from_pretrained("multiberts-seed-2-1200k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-2-120k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-2']	false	true	true	6,483	# MultiBERTs Seed 2 Checkpoint 120k (uncased) Seed 2 intermediate checkpoint 120k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-2](https://hf.co/multberts-seed-2). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-2-120k') model = BertModel.from_pretrained("multiberts-seed-2-120k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-2-1300k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-2']	false	true	true	6,487	# MultiBERTs Seed 2 Checkpoint 1300k (uncased) Seed 2 intermediate checkpoint 1300k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-2](https://hf.co/multberts-seed-2). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-2-1300k') model = BertModel.from_pretrained("multiberts-seed-2-1300k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-2-1400k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-2']	false	true	true	6,487	# MultiBERTs Seed 2 Checkpoint 1400k (uncased) Seed 2 intermediate checkpoint 1400k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-2](https://hf.co/multberts-seed-2). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-2-1400k') model = BertModel.from_pretrained("multiberts-seed-2-1400k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-2-140k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-2']	false	true	true	6,483	# MultiBERTs Seed 2 Checkpoint 140k (uncased) Seed 2 intermediate checkpoint 140k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-2](https://hf.co/multberts-seed-2). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-2-140k') model = BertModel.from_pretrained("multiberts-seed-2-140k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-2-1500k	MultiBertGunjanPatrick	bert	7	5	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-2']	false	true	true	6,487	# MultiBERTs Seed 2 Checkpoint 1500k (uncased) Seed 2 intermediate checkpoint 1500k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-2](https://hf.co/multberts-seed-2). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-2-1500k') model = BertModel.from_pretrained("multiberts-seed-2-1500k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-2-1600k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-2']	false	true	true	6,487	# MultiBERTs Seed 2 Checkpoint 1600k (uncased) Seed 2 intermediate checkpoint 1600k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-2](https://hf.co/multberts-seed-2). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-2-1600k') model = BertModel.from_pretrained("multiberts-seed-2-1600k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-2-160k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-2']	false	true	true	6,483	# MultiBERTs Seed 2 Checkpoint 160k (uncased) Seed 2 intermediate checkpoint 160k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-2](https://hf.co/multberts-seed-2). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-2-160k') model = BertModel.from_pretrained("multiberts-seed-2-160k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-2-1700k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-2']	false	true	true	6,487	# MultiBERTs Seed 2 Checkpoint 1700k (uncased) Seed 2 intermediate checkpoint 1700k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-2](https://hf.co/multberts-seed-2). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-2-1700k') model = BertModel.from_pretrained("multiberts-seed-2-1700k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-2-1800k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-2']	false	true	true	6,487	# MultiBERTs Seed 2 Checkpoint 1800k (uncased) Seed 2 intermediate checkpoint 1800k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-2](https://hf.co/multberts-seed-2). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-2-1800k') model = BertModel.from_pretrained("multiberts-seed-2-1800k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-2-180k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-2']	false	true	true	6,483	# MultiBERTs Seed 2 Checkpoint 180k (uncased) Seed 2 intermediate checkpoint 180k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-2](https://hf.co/multberts-seed-2). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-2-180k') model = BertModel.from_pretrained("multiberts-seed-2-180k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-2-1900k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-2']	false	true	true	6,487	# MultiBERTs Seed 2 Checkpoint 1900k (uncased) Seed 2 intermediate checkpoint 1900k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-2](https://hf.co/multberts-seed-2). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-2-1900k') model = BertModel.from_pretrained("multiberts-seed-2-1900k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-2-2000k	MultiBertGunjanPatrick	bert	7	5	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-2']	false	true	true	6,487	# MultiBERTs Seed 2 Checkpoint 2000k (uncased) Seed 2 intermediate checkpoint 2000k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-2](https://hf.co/multberts-seed-2). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-2-2000k') model = BertModel.from_pretrained("multiberts-seed-2-2000k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-2-200k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-2']	false	true	true	6,483	# MultiBERTs Seed 2 Checkpoint 200k (uncased) Seed 2 intermediate checkpoint 200k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-2](https://hf.co/multberts-seed-2). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-2-200k') model = BertModel.from_pretrained("multiberts-seed-2-200k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-2-20k	MultiBertGunjanPatrick	bert	7	5	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-2']	false	true	true	6,479	# MultiBERTs Seed 2 Checkpoint 20k (uncased) Seed 2 intermediate checkpoint 20k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-2](https://hf.co/multberts-seed-2). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-2-20k') model = BertModel.from_pretrained("multiberts-seed-2-20k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-2-300k	MultiBertGunjanPatrick	bert	7	6	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-2']	false	true	true	6,483	# MultiBERTs Seed 2 Checkpoint 300k (uncased) Seed 2 intermediate checkpoint 300k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-2](https://hf.co/multberts-seed-2). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-2-300k') model = BertModel.from_pretrained("multiberts-seed-2-300k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-2-400k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-2']	false	true	true	6,483	# MultiBERTs Seed 2 Checkpoint 400k (uncased) Seed 2 intermediate checkpoint 400k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-2](https://hf.co/multberts-seed-2). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-2-400k') model = BertModel.from_pretrained("multiberts-seed-2-400k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-2-40k	MultiBertGunjanPatrick	bert	7	5	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-2']	false	true	true	6,479	# MultiBERTs Seed 2 Checkpoint 40k (uncased) Seed 2 intermediate checkpoint 40k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-2](https://hf.co/multberts-seed-2). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-2-40k') model = BertModel.from_pretrained("multiberts-seed-2-40k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-2-500k	MultiBertGunjanPatrick	bert	7	5	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-2']	false	true	true	6,483	# MultiBERTs Seed 2 Checkpoint 500k (uncased) Seed 2 intermediate checkpoint 500k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-2](https://hf.co/multberts-seed-2). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-2-500k') model = BertModel.from_pretrained("multiberts-seed-2-500k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-2-600k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-2']	false	true	true	6,483	# MultiBERTs Seed 2 Checkpoint 600k (uncased) Seed 2 intermediate checkpoint 600k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-2](https://hf.co/multberts-seed-2). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-2-600k') model = BertModel.from_pretrained("multiberts-seed-2-600k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-2-60k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-2']	false	true	true	6,479	# MultiBERTs Seed 2 Checkpoint 60k (uncased) Seed 2 intermediate checkpoint 60k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-2](https://hf.co/multberts-seed-2). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-2-60k') model = BertModel.from_pretrained("multiberts-seed-2-60k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-2-700k	MultiBertGunjanPatrick	bert	7	5	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-2']	false	true	true	6,483	# MultiBERTs Seed 2 Checkpoint 700k (uncased) Seed 2 intermediate checkpoint 700k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-2](https://hf.co/multberts-seed-2). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-2-700k') model = BertModel.from_pretrained("multiberts-seed-2-700k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-2-800k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-2']	false	true	true	6,483	# MultiBERTs Seed 2 Checkpoint 800k (uncased) Seed 2 intermediate checkpoint 800k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-2](https://hf.co/multberts-seed-2). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-2-800k') model = BertModel.from_pretrained("multiberts-seed-2-800k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-2-80k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-2']	false	true	true	6,479	# MultiBERTs Seed 2 Checkpoint 80k (uncased) Seed 2 intermediate checkpoint 80k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-2](https://hf.co/multberts-seed-2). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-2-80k') model = BertModel.from_pretrained("multiberts-seed-2-80k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-2-900k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-2']	false	true	true	6,483	# MultiBERTs Seed 2 Checkpoint 900k (uncased) Seed 2 intermediate checkpoint 900k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-2](https://hf.co/multberts-seed-2). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-2-900k') model = BertModel.from_pretrained("multiberts-seed-2-900k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-2	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts']	false	true	true	6,319	# MultiBERTs Seed 0 (uncased) Seed 0 MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-0') model = BertModel.from_pretrained("multiberts-seed-0") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-20	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts']	false	true	true	6,323	# MultiBERTs Seed 20 (uncased) Seed 20 MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-20') model = BertModel.from_pretrained("multiberts-seed-20") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-21	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts']	false	true	true	6,323	# MultiBERTs Seed 21 (uncased) Seed 21 MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-21') model = BertModel.from_pretrained("multiberts-seed-21") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-22	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts']	false	true	true	6,323	# MultiBERTs Seed 22 (uncased) Seed 22 MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-22') model = BertModel.from_pretrained("multiberts-seed-22") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-23	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts']	false	true	true	6,323	# MultiBERTs Seed 23 (uncased) Seed 23 MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-23') model = BertModel.from_pretrained("multiberts-seed-23") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-24	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts']	false	true	true	6,323	# MultiBERTs Seed 24 (uncased) Seed 24 MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-24') model = BertModel.from_pretrained("multiberts-seed-24") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-3-0k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-3']	false	true	true	6,475	# MultiBERTs Seed 3 Checkpoint 0k (uncased) Seed 3 intermediate checkpoint 0k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-3](https://hf.co/multberts-seed-3). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-3-0k') model = BertModel.from_pretrained("multiberts-seed-3-0k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-3-1000k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-3']	false	true	true	6,487	# MultiBERTs Seed 3 Checkpoint 1000k (uncased) Seed 3 intermediate checkpoint 1000k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-3](https://hf.co/multberts-seed-3). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-3-1000k') model = BertModel.from_pretrained("multiberts-seed-3-1000k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-3-100k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-3']	false	true	true	6,483	# MultiBERTs Seed 3 Checkpoint 100k (uncased) Seed 3 intermediate checkpoint 100k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-3](https://hf.co/multberts-seed-3). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-3-100k') model = BertModel.from_pretrained("multiberts-seed-3-100k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-3-1100k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-3']	false	true	true	6,487	# MultiBERTs Seed 3 Checkpoint 1100k (uncased) Seed 3 intermediate checkpoint 1100k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-3](https://hf.co/multberts-seed-3). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-3-1100k') model = BertModel.from_pretrained("multiberts-seed-3-1100k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-3-1200k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-3']	false	true	true	6,487	# MultiBERTs Seed 3 Checkpoint 1200k (uncased) Seed 3 intermediate checkpoint 1200k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-3](https://hf.co/multberts-seed-3). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-3-1200k') model = BertModel.from_pretrained("multiberts-seed-3-1200k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-3-120k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-3']	false	true	true	6,483	# MultiBERTs Seed 3 Checkpoint 120k (uncased) Seed 3 intermediate checkpoint 120k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-3](https://hf.co/multberts-seed-3). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-3-120k') model = BertModel.from_pretrained("multiberts-seed-3-120k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-3-1300k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-3']	false	true	true	6,487	# MultiBERTs Seed 3 Checkpoint 1300k (uncased) Seed 3 intermediate checkpoint 1300k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-3](https://hf.co/multberts-seed-3). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-3-1300k') model = BertModel.from_pretrained("multiberts-seed-3-1300k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-3-1400k	MultiBertGunjanPatrick	bert	7	6	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-3']	false	true	true	6,487	# MultiBERTs Seed 3 Checkpoint 1400k (uncased) Seed 3 intermediate checkpoint 1400k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-3](https://hf.co/multberts-seed-3). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-3-1400k') model = BertModel.from_pretrained("multiberts-seed-3-1400k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-3-140k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-3']	false	true	true	6,483	# MultiBERTs Seed 3 Checkpoint 140k (uncased) Seed 3 intermediate checkpoint 140k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-3](https://hf.co/multberts-seed-3). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-3-140k') model = BertModel.from_pretrained("multiberts-seed-3-140k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-3-1500k	MultiBertGunjanPatrick	bert	7	5	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-3']	false	true	true	6,487	# MultiBERTs Seed 3 Checkpoint 1500k (uncased) Seed 3 intermediate checkpoint 1500k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-3](https://hf.co/multberts-seed-3). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-3-1500k') model = BertModel.from_pretrained("multiberts-seed-3-1500k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-3-1600k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-3']	false	true	true	6,487	# MultiBERTs Seed 3 Checkpoint 1600k (uncased) Seed 3 intermediate checkpoint 1600k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-3](https://hf.co/multberts-seed-3). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-3-1600k') model = BertModel.from_pretrained("multiberts-seed-3-1600k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-3-160k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-3']	false	true	true	6,483	# MultiBERTs Seed 3 Checkpoint 160k (uncased) Seed 3 intermediate checkpoint 160k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-3](https://hf.co/multberts-seed-3). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-3-160k') model = BertModel.from_pretrained("multiberts-seed-3-160k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-3-1700k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-3']	false	true	true	6,487	# MultiBERTs Seed 3 Checkpoint 1700k (uncased) Seed 3 intermediate checkpoint 1700k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-3](https://hf.co/multberts-seed-3). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-3-1700k') model = BertModel.from_pretrained("multiberts-seed-3-1700k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-3-1800k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-3']	false	true	true	6,487	# MultiBERTs Seed 3 Checkpoint 1800k (uncased) Seed 3 intermediate checkpoint 1800k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-3](https://hf.co/multberts-seed-3). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-3-1800k') model = BertModel.from_pretrained("multiberts-seed-3-1800k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-3-180k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-3']	false	true	true	6,483	# MultiBERTs Seed 3 Checkpoint 180k (uncased) Seed 3 intermediate checkpoint 180k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-3](https://hf.co/multberts-seed-3). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-3-180k') model = BertModel.from_pretrained("multiberts-seed-3-180k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-3-1900k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-3']	false	true	true	6,487	# MultiBERTs Seed 3 Checkpoint 1900k (uncased) Seed 3 intermediate checkpoint 1900k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-3](https://hf.co/multberts-seed-3). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-3-1900k') model = BertModel.from_pretrained("multiberts-seed-3-1900k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-3-2000k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-3']	false	true	true	6,487	# MultiBERTs Seed 3 Checkpoint 2000k (uncased) Seed 3 intermediate checkpoint 2000k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-3](https://hf.co/multberts-seed-3). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-3-2000k') model = BertModel.from_pretrained("multiberts-seed-3-2000k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-3-200k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-3']	false	true	true	6,483	# MultiBERTs Seed 3 Checkpoint 200k (uncased) Seed 3 intermediate checkpoint 200k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-3](https://hf.co/multberts-seed-3). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-3-200k') model = BertModel.from_pretrained("multiberts-seed-3-200k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-3-20k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-3']	false	true	true	6,479	# MultiBERTs Seed 3 Checkpoint 20k (uncased) Seed 3 intermediate checkpoint 20k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-3](https://hf.co/multberts-seed-3). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-3-20k') model = BertModel.from_pretrained("multiberts-seed-3-20k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-3-300k	MultiBertGunjanPatrick	bert	7	5	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-3']	false	true	true	6,483	# MultiBERTs Seed 3 Checkpoint 300k (uncased) Seed 3 intermediate checkpoint 300k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-3](https://hf.co/multberts-seed-3). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-3-300k') model = BertModel.from_pretrained("multiberts-seed-3-300k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-3-400k	MultiBertGunjanPatrick	bert	7	5	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-3']	false	true	true	6,483	# MultiBERTs Seed 3 Checkpoint 400k (uncased) Seed 3 intermediate checkpoint 400k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-3](https://hf.co/multberts-seed-3). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-3-400k') model = BertModel.from_pretrained("multiberts-seed-3-400k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-3-40k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-3']	false	true	true	6,479	# MultiBERTs Seed 3 Checkpoint 40k (uncased) Seed 3 intermediate checkpoint 40k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-3](https://hf.co/multberts-seed-3). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-3-40k') model = BertModel.from_pretrained("multiberts-seed-3-40k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-3-500k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-3']	false	true	true	6,483	# MultiBERTs Seed 3 Checkpoint 500k (uncased) Seed 3 intermediate checkpoint 500k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-3](https://hf.co/multberts-seed-3). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-3-500k') model = BertModel.from_pretrained("multiberts-seed-3-500k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-3-600k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-3']	false	true	true	6,483	# MultiBERTs Seed 3 Checkpoint 600k (uncased) Seed 3 intermediate checkpoint 600k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-3](https://hf.co/multberts-seed-3). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-3-600k') model = BertModel.from_pretrained("multiberts-seed-3-600k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-3-60k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-3']	false	true	true	6,479	# MultiBERTs Seed 3 Checkpoint 60k (uncased) Seed 3 intermediate checkpoint 60k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-3](https://hf.co/multberts-seed-3). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-3-60k') model = BertModel.from_pretrained("multiberts-seed-3-60k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-3-700k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-3']	false	true	true	6,483	# MultiBERTs Seed 3 Checkpoint 700k (uncased) Seed 3 intermediate checkpoint 700k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-3](https://hf.co/multberts-seed-3). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-3-700k') model = BertModel.from_pretrained("multiberts-seed-3-700k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-3-800k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-3']	false	true	true	6,483	# MultiBERTs Seed 3 Checkpoint 800k (uncased) Seed 3 intermediate checkpoint 800k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-3](https://hf.co/multberts-seed-3). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-3-800k') model = BertModel.from_pretrained("multiberts-seed-3-800k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-3-80k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-3']	false	true	true	6,479	# MultiBERTs Seed 3 Checkpoint 80k (uncased) Seed 3 intermediate checkpoint 80k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-3](https://hf.co/multberts-seed-3). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-3-80k') model = BertModel.from_pretrained("multiberts-seed-3-80k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-3-900k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-3']	false	true	true	6,483	# MultiBERTs Seed 3 Checkpoint 900k (uncased) Seed 3 intermediate checkpoint 900k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-3](https://hf.co/multberts-seed-3). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-3-900k') model = BertModel.from_pretrained("multiberts-seed-3-900k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-3	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts']	false	true	true	6,319	# MultiBERTs Seed 0 (uncased) Seed 0 MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-0') model = BertModel.from_pretrained("multiberts-seed-0") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-4-0k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-4']	false	true	true	6,475	# MultiBERTs Seed 4 Checkpoint 0k (uncased) Seed 4 intermediate checkpoint 0k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-4](https://hf.co/multberts-seed-4). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-4-0k') model = BertModel.from_pretrained("multiberts-seed-4-0k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-4-1000k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-4']	false	true	true	6,487	# MultiBERTs Seed 4 Checkpoint 1000k (uncased) Seed 4 intermediate checkpoint 1000k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-4](https://hf.co/multberts-seed-4). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-4-1000k') model = BertModel.from_pretrained("multiberts-seed-4-1000k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-4-100k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-4']	false	true	true	6,483	# MultiBERTs Seed 4 Checkpoint 100k (uncased) Seed 4 intermediate checkpoint 100k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-4](https://hf.co/multberts-seed-4). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-4-100k') model = BertModel.from_pretrained("multiberts-seed-4-100k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-4-1100k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-4']	false	true	true	6,487	# MultiBERTs Seed 4 Checkpoint 1100k (uncased) Seed 4 intermediate checkpoint 1100k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-4](https://hf.co/multberts-seed-4). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-4-1100k') model = BertModel.from_pretrained("multiberts-seed-4-1100k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-4-1200k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-4']	false	true	true	6,487	# MultiBERTs Seed 4 Checkpoint 1200k (uncased) Seed 4 intermediate checkpoint 1200k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-4](https://hf.co/multberts-seed-4). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-4-1200k') model = BertModel.from_pretrained("multiberts-seed-4-1200k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-4-120k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-4']	false	true	true	6,483	# MultiBERTs Seed 4 Checkpoint 120k (uncased) Seed 4 intermediate checkpoint 120k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-4](https://hf.co/multberts-seed-4). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-4-120k') model = BertModel.from_pretrained("multiberts-seed-4-120k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-4-1300k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-4']	false	true	true	6,487	# MultiBERTs Seed 4 Checkpoint 1300k (uncased) Seed 4 intermediate checkpoint 1300k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-4](https://hf.co/multberts-seed-4). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-4-1300k') model = BertModel.from_pretrained("multiberts-seed-4-1300k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-4-1400k	MultiBertGunjanPatrick	bert	7	5	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-4']	false	true	true	6,487	# MultiBERTs Seed 4 Checkpoint 1400k (uncased) Seed 4 intermediate checkpoint 1400k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-4](https://hf.co/multberts-seed-4). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-4-1400k') model = BertModel.from_pretrained("multiberts-seed-4-1400k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-4-140k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-4']	false	true	true	6,483	# MultiBERTs Seed 4 Checkpoint 140k (uncased) Seed 4 intermediate checkpoint 140k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-4](https://hf.co/multberts-seed-4). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-4-140k') model = BertModel.from_pretrained("multiberts-seed-4-140k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-4-1500k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-4']	false	true	true	6,487	# MultiBERTs Seed 4 Checkpoint 1500k (uncased) Seed 4 intermediate checkpoint 1500k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-4](https://hf.co/multberts-seed-4). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-4-1500k') model = BertModel.from_pretrained("multiberts-seed-4-1500k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-4-1600k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-4']	false	true	true	6,487	# MultiBERTs Seed 4 Checkpoint 1600k (uncased) Seed 4 intermediate checkpoint 1600k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-4](https://hf.co/multberts-seed-4). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-4-1600k') model = BertModel.from_pretrained("multiberts-seed-4-1600k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-4-160k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-4']	false	true	true	6,483	# MultiBERTs Seed 4 Checkpoint 160k (uncased) Seed 4 intermediate checkpoint 160k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-4](https://hf.co/multberts-seed-4). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-4-160k') model = BertModel.from_pretrained("multiberts-seed-4-160k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-4-1700k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-4']	false	true	true	6,487	# MultiBERTs Seed 4 Checkpoint 1700k (uncased) Seed 4 intermediate checkpoint 1700k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-4](https://hf.co/multberts-seed-4). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-4-1700k') model = BertModel.from_pretrained("multiberts-seed-4-1700k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-4-1800k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-4']	false	true	true	6,487	# MultiBERTs Seed 4 Checkpoint 1800k (uncased) Seed 4 intermediate checkpoint 1800k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-4](https://hf.co/multberts-seed-4). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-4-1800k') model = BertModel.from_pretrained("multiberts-seed-4-1800k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-4-180k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-4']	false	true	true	6,483	# MultiBERTs Seed 4 Checkpoint 180k (uncased) Seed 4 intermediate checkpoint 180k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-4](https://hf.co/multberts-seed-4). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-4-180k') model = BertModel.from_pretrained("multiberts-seed-4-180k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-4-1900k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-4']	false	true	true	6,487	# MultiBERTs Seed 4 Checkpoint 1900k (uncased) Seed 4 intermediate checkpoint 1900k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-4](https://hf.co/multberts-seed-4). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-4-1900k') model = BertModel.from_pretrained("multiberts-seed-4-1900k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-4-2000k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-4']	false	true	true	6,487	# MultiBERTs Seed 4 Checkpoint 2000k (uncased) Seed 4 intermediate checkpoint 2000k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-4](https://hf.co/multberts-seed-4). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-4-2000k') model = BertModel.from_pretrained("multiberts-seed-4-2000k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-4-200k	MultiBertGunjanPatrick	bert	7	5	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-4']	false	true	true	6,483	# MultiBERTs Seed 4 Checkpoint 200k (uncased) Seed 4 intermediate checkpoint 200k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-4](https://hf.co/multberts-seed-4). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-4-200k') model = BertModel.from_pretrained("multiberts-seed-4-200k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-4-20k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-4']	false	true	true	6,479	# MultiBERTs Seed 4 Checkpoint 20k (uncased) Seed 4 intermediate checkpoint 20k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-4](https://hf.co/multberts-seed-4). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-4-20k') model = BertModel.from_pretrained("multiberts-seed-4-20k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-4-300k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-4']	false	true	true	6,483	# MultiBERTs Seed 4 Checkpoint 300k (uncased) Seed 4 intermediate checkpoint 300k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-4](https://hf.co/multberts-seed-4). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-4-300k') model = BertModel.from_pretrained("multiberts-seed-4-300k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-4-400k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-4']	false	true	true	6,483	# MultiBERTs Seed 4 Checkpoint 400k (uncased) Seed 4 intermediate checkpoint 400k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-4](https://hf.co/multberts-seed-4). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-4-400k') model = BertModel.from_pretrained("multiberts-seed-4-400k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-4-40k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-4']	false	true	true	6,479	# MultiBERTs Seed 4 Checkpoint 40k (uncased) Seed 4 intermediate checkpoint 40k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-4](https://hf.co/multberts-seed-4). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-4-40k') model = BertModel.from_pretrained("multiberts-seed-4-40k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-4-500k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-4']	false	true	true	6,483	# MultiBERTs Seed 4 Checkpoint 500k (uncased) Seed 4 intermediate checkpoint 500k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-4](https://hf.co/multberts-seed-4). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-4-500k') model = BertModel.from_pretrained("multiberts-seed-4-500k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-4-600k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-4']	false	true	true	6,483	# MultiBERTs Seed 4 Checkpoint 600k (uncased) Seed 4 intermediate checkpoint 600k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-4](https://hf.co/multberts-seed-4). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-4-600k') model = BertModel.from_pretrained("multiberts-seed-4-600k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-4-60k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-4']	false	true	true	6,479	# MultiBERTs Seed 4 Checkpoint 60k (uncased) Seed 4 intermediate checkpoint 60k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-4](https://hf.co/multberts-seed-4). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-4-60k') model = BertModel.from_pretrained("multiberts-seed-4-60k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>
MultiBertGunjanPatrick/multiberts-seed-4-700k	MultiBertGunjanPatrick	bert	7	4	transformers	null	true	false	false	apache-2.0	['en']	['bookcorpus', 'wikipedia']	null	['exbert', 'multiberts', 'multiberts-seed-4']	false	true	true	6,483	# MultiBERTs Seed 4 Checkpoint 700k (uncased) Seed 4 intermediate checkpoint 700k MultiBERTs (pretrained BERT) model on English language using a masked language modeling (MLM) objective. It was introduced in [this paper](https://arxiv.org/pdf/2106.16163.pdf) and first released in [this repository](https://github.com/google-research/language/tree/master/language/multiberts). This is an intermediate checkpoint. The final checkpoint can be found at [multiberts-seed-4](https://hf.co/multberts-seed-4). This model is uncased: it does not make a difference between english and English. 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). ## Model description MultiBERTs models are transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was pretrained with two objectives: - Masked language modeling (MLM): taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence. - Next sentence prediction (NSP): the models concatenates two masked sentences as inputs during pretraining. Sometimes they correspond to sentences that were next to each other in the original text, sometimes not. The model then has to predict if the two sentences were following each other or not. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the MultiBERTs model as inputs. ## Intended uses & limitations You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=multiberts) to look for fine-tuned versions on a task that interests you. Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT2. ### How to use Here is how to use this model to get the features of a given text in PyTorch: ```python from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('multiberts-seed-4-700k') model = BertModel.from_pretrained("multiberts-seed-4-700k") text = "Replace me by any text you'd like." encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) ``` ### Limitations and bias Even if the training data used for this model could be characterized as fairly neutral, this model can have biased predictions. This bias will also affect all fine-tuned versions of this model. For an understanding of bias of this particular checkpoint, please try out this checkpoint with the snippet present in the [Limitation and bias section](https://huggingface.co/bert-base-uncased#limitations-and-bias) of the [bert-base-uncased](https://huggingface.co/bert-base-uncased) checkpoint. ## Training data The MultiBERTs models were pretrained on [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers). ## Training procedure ### Preprocessing The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are then of the form: ``` [CLS] Sentence A [SEP] Sentence B [SEP] ``` With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus and in the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two "sentences" has a combined length of less than 512 tokens. The details of the masking procedure for each sentence are the following: - 15% of the tokens are masked. - In 80% of the cases, the masked tokens are replaced by `[MASK]`. - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace. - In the 10% remaining cases, the masked tokens are left as is. ### Pretraining The full model was trained on 16 Cloud TPU v2 chips for two million steps with a batch size of 256. The sequence length was set to 512 throughout. The optimizer 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, learning rate warmup for 10,000 steps and linear decay of the learning rate after. ### BibTeX entry and citation info ```bibtex @article{DBLP:journals/corr/abs-2106-16163, author = {Thibault Sellam and Steve Yadlowsky and Jason Wei and Naomi Saphra and Alexander D'Amour and Tal Linzen and Jasmijn Bastings and Iulia Turc and Jacob Eisenstein and Dipanjan Das and Ian Tenney and Ellie Pavlick}, title = {The MultiBERTs: {BERT} Reproductions for Robustness Analysis}, journal = {CoRR}, volume = {abs/2106.16163}, year = {2021}, url = {https://arxiv.org/abs/2106.16163}, eprinttype = {arXiv}, eprint = {2106.16163}, timestamp = {Mon, 05 Jul 2021 15:15:50 +0200}, biburl = {https://dblp.org/rec/journals/corr/abs-2106-16163.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} } ``` <a href="https://huggingface.co/exbert/?model=multiberts"> <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png"> </a>

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.