albertina-900m-portuguese-ptbr-encoder

---

<img align="left" width="40" height="40" src="https://github.githubassets.com/images/icons/emoji/unicode/1f917.png"> <p style="text-align: center;">&nbsp;&nbsp;&nbsp;&nbsp;This is the model card for Albertina 900m PT-PT No-brWaC You may be interested in some of the other models in the <a href="https://huggingface.co/PORTULAN">Albertina (encoders) and Gervásio (decoders) families</a>. </p>

---

Albertina PT-BR No-brWaC

Albertina PT-* is a foundation, large language model for the Portuguese language.

It is an encoder of the BERT family, based on the neural architecture Transformer and developed over the DeBERTa model, and with most competitive performance for this language. It has different versions that were trained for different variants of Portuguese (PT), namely the European variant from Portugal (PT-PT) and the American variant from Brazil (PT-BR), and it is distributed free of charge and under a most permissible license.

Albertina PT-BR No-brWaC is a version for American Portuguese from Brazil trained on data sets other than brWaC, and thus with a most permissive license.

You may be interested also in Albertina PT-BR, trained on brWaC. To the best of our knowledge, these are encoders specifically for this language and variant that set a new state of the art for it, and is made publicly available and distributed for reuse.

Albertina PT-BR No-brWaC is developed by a joint team from the University of Lisbon and the University of Porto, Portugal. For further details, check the respective publication:

@misc{albertina-pt,
      title={Advancing Neural Encoding of Portuguese
             with Transformer Albertina PT-*}, 
      author={João Rodrigues and Luís Gomes and João Silva and
              António Branco and Rodrigo Santos and
              Henrique Lopes Cardoso and Tomás Osório},
      year={2023},
      eprint={2305.06721},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}

Please use the above canonical reference when using or citing this model.

<br>

Model Description

This model card is for Albertina-PT-BR No-brWaC, with 900M parameters, 24 layers and a hidden size of 1536.

Albertina-PT-BR No-brWaC is distributed under an MIT license.

DeBERTa is distributed under an MIT license.

<br>

Training Data

Albertina PT-BR No-brWac was trained over a 3.7 billion token curated selection of documents from the OSCAR data set. The OSCAR data set includes documents in more than one hundred languages, including Portuguese, and it is widely used in the literature. It is the result of a selection performed over the Common Crawl data set, crawled from the Web, that retains only pages whose metadata indicates permission to be crawled, that performs deduplication, and that removes some boilerplate, among other filters. Given that it does not discriminate between the Portuguese variants, we performed extra filtering by retaining only documents whose meta-data indicate the Internet country code top-level domain of Brazil. We used the January 2023 version of OSCAR, which is based on the November/December 2022 version of Common Crawl.

Preprocessing

We filtered the PT-BR corpora using the BLOOM pre-processing pipeline. We skipped the default filtering of stopwords since it would disrupt the syntactic structure, and also the filtering for language identification given the corpus was pre-selected as Portuguese.

Training

As codebase, we resorted to the DeBERTa V2 XLarge, for English.

To train Albertina PT-PT No-brWac, the data set was tokenized with the original DeBERTa tokenizer with a 128 token sequence truncation and dynamic padding. The model was trained using the maximum available memory capacity resulting in a batch size of 896 samples (56 samples per GPU). We chose a learning rate of 1e-5 with linear decay and 10k warm-up steps. In total, around 200k training steps were taken across 50 epochs. The model was trained for 1 day and 13 hours on a2-megagpu-16gb Google Cloud A2 VMs with 16 GPUs, 96 vCPUs and 1.360 GB of RAM.

<br>

Evaluation

The two model versions were evaluated on downstream tasks organized into two groups.

In one group, we have the two data sets from the ASSIN 2 benchmark, namely STS and RTE, that were used to evaluate the previous state-of-the-art model BERTimbau Large. In the other group of data sets, we have the translations into PT-BR of the English data sets used for a few of the tasks in the widely-used GLUE benchmark, which allowed us to test both Albertina-PT-* variants on a wider variety of downstream tasks.

ASSIN 2

ASSIN 2 is a PT-BR data set of approximately 10.000 sentence pairs, split into 6.500 for training, 500 for validation, and 2.448 for testing, annotated with semantic relatedness scores (range 1 to 5) and with binary entailment judgments. This data set supports the task of semantic textual similarity (STS), which consists of assigning a score of how semantically related two sentences are; and the task of recognizing textual entailment (RTE), which given a pair of sentences, consists of determining whether the first entails the second.

GLUE tasks translated

We resort to PLUE (Portuguese Language Understanding Evaluation), a data set that was obtained by automatically translating GLUE into PT-BR. We address four tasks from those in PLUE, namely:

• two similarity tasks: MRPC, for detecting whether two sentences are paraphrases of each other, and STS-B, for semantic textual similarity;
• and two inference tasks: RTE, for recognizing textual entailment and WNLI, for coreference and natural language inference.

<br>

How to use

You can use this model directly with a pipeline for masked language modeling:

>>> from transformers import pipeline
>>> unmasker = pipeline('fill-mask', model='PORTULAN/albertina-ptbr-nobrwac')
>>> unmasker("A culinária brasileira é rica em sabores e [MASK], tornando-se um dos maiores patrimônios do país.")

[{'score': 0.3866911828517914, 'token': 23395, 'token_str': 'aromas', 'sequence': 'A culinária brasileira é rica em sabores e aromas, tornando-se um dos maiores patrimônios do país.'},
{'score': 0.2926434874534607, 'token': 10392, 'token_str': 'costumes', 'sequence': 'A culinária brasileira é rica em sabores e costumes, tornando-se um dos maiores patrimônios do país.'},
{'score': 0.1913347691297531, 'token': 21925, 'token_str': 'cores', 'sequence': 'A culinária brasileira é rica em sabores e cores, tornando-se um dos maiores patrimônios do país.'},
{'score': 0.06453365087509155, 'token': 117371, 'token_str': 'cultura', 'sequence': 'A culinária brasileira é rica em sabores e cultura, tornando-se um dos maiores patrimônios do país.'},
{'score': 0.019388679414987564, 'token': 22647, 'token_str': 'nuances', 'sequence': 'A culinária brasileira é rica em sabores e nuances, tornando-se um dos maiores patrimônios do país.'}]

The model can be used by fine-tuning it for a specific task:

>>> from transformers import AutoTokenizer, AutoModelForSequenceClassification, TrainingArguments, Trainer
>>> from datasets import load_dataset

>>> model = AutoModelForSequenceClassification.from_pretrained("PORTULAN/albertina-ptbr-nobrwac", num_labels=2)
>>> tokenizer = AutoTokenizer.from_pretrained("PORTULAN/albertina-ptbr-nobrwac")
>>> dataset = load_dataset("PORTULAN/glue-ptpt", "rte")

>>> def tokenize_function(examples):
...     return tokenizer(examples["sentence1"], examples["sentence2"], padding="max_length", truncation=True)

>>> tokenized_datasets = dataset.map(tokenize_function, batched=True)

>>> training_args = TrainingArguments(output_dir="albertina-ptbr-rte", evaluation_strategy="epoch")
>>> trainer = Trainer(
...     model=model,
...     args=training_args,
...     train_dataset=tokenized_datasets["train"],
...     eval_dataset=tokenized_datasets["validation"],
... )

>>> trainer.train()

<br>

Citation

When using or citing this model, kindly cite the following publication:

@misc{albertina-pt,
      title={Advancing Neural Encoding of Portuguese
             with Transformer Albertina PT-*}, 
      author={João Rodrigues and Luís Gomes and João Silva and
              António Branco and Rodrigo Santos and
              Henrique Lopes Cardoso and Tomás Osório},
      year={2023},
      eprint={2305.06721},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}

<br>

Acknowledgments

The research reported here was partially supported by: PORTULAN CLARIN—Research Infrastructure for the Science and Technology of Language, funded by Lisboa 2020, Alentejo 2020 and FCT—Fundação para a Ciência e Tecnologia under the grant PINFRA/22117/2016; research project ALBERTINA - Foundation Encoder Model for Portuguese and AI, funded by FCT—Fundação para a Ciência e Tecnologia under the grant CPCA-IAC/AV/478394/2022; innovation project ACCELERAT.AI - Multilingual Intelligent Contact Centers, funded by IAPMEI, I.P. - Agência para a Competitividade e Inovação under the grant C625734525-00462629, of Plano de Recuperação e Resiliência, call RE-C05-i01.01 – Agendas/Alianças Mobilizadoras para a Reindustrialização; and LIACC - Laboratory for AI and Computer Science, funded by FCT—Fundação para a Ciência e Tecnologia under the grant FCT/UID/CEC/0027/2020.