SAVRN
Search Contact SAVRN

Open-weight model · Summarization

ptt5-base-summ

by Recogna NLP recogna-nlp/ptt5-base-summ

PTT5 Summ is a fine-tuned PTT5 model to perform Abstractive Summarization in Brazilian Portuguese texts. This model was fine-tuned on the datasets: RecognaSumm, WikiLingua, XL-Sum, TeMário.pdf) and CSTNews.

Parameters223M
Context
Weights1.8 GB
Licensemit
AccessOpen weights
Monthly Downloads3.8k

Runs On

What it takes to serve ptt5-base-summ (223M parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them.

PrecisionWeightsMemory neededCheapest setupPer hourAlso fits
16-bit 0.4 GB 0.5 GB 1x MI300X (192 GB)
Vultr
$1.85 1x H100 $1.99 · 1x MI325X $2.00
8-bit 0.2 GB 0.3 GB 1x MI300X (192 GB)
Vultr
$1.85 1x H100 $1.99 · 1x MI325X $2.00
4-bit 0.1 GB 0.1 GB 1x MI300X (192 GB)
Vultr
$1.85 1x H100 $1.99 · 1x MI325X $2.00

Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the SAVRN Index, read Sep 18, 2026.

Model Card

By Recogna NLP, published under mit, revision 84509bf8ec61.

PTT5 Summ is a fine-tuned PTT5 model to perform Abstractive Summarization in Brazilian Portuguese texts. This model was fine-tuned on the datasets: RecognaSumm, WikiLingua, XL-Sum, TeMário.pdf) and CSTNews. For further information, please go to PTT5 Summ repository.

Read Recogna NLP's full model card

Portuguese T5 for Abstractive Summarization (PTT5 Summ)

Introduction

PTT5 Summ is a fine-tuned PTT5 model to perform Abstractive Summarization in Brazilian Portuguese texts. This model was fine-tuned on the datasets: RecognaSumm, WikiLingua, XL-Sum, TeMário and CSTNews.

For further information, please go to PTT5 Summ repository.

Available models

Model Dataset used in fine-tuning
recogna-nlp/ptt5-base-summ RecognaSumm
recogna-nlp/ptt5-base-summ-wikilingua WikiLingua
recogna-nlp/ptt5-base-summ-xlsum XL-Sum
recogna-nlp/ptt5-base-summ-temario 1st phase: WikiLingua. 2nd phase: TeMario
recogna-nlp/ptt5-base-summ-cstnews 1st phase: WikiLingua. 2nd phase: CSTNews

Usage example

# Tokenizer 
from transformers import T5Tokenizer

# PyTorch model 
from transformers import T5Model, T5ForConditionalGeneration

token_name = 'unicamp-dl/ptt5-base-portuguese-vocab'
model_name = 'recogna-nlp/ptt5-base-summ'

tokenizer = T5Tokenizer.from_pretrained(token_name )
model_pt = T5ForConditionalGeneration.from_pretrained(model_name)

text = '''
“A tendência de queda da taxa de juros no Brasil é real, é visível”, disse Meirelles, que participou na capital americana de uma série de reuniões e encontros com banqueiros e investidores que aconteceram paralelamente às reuniões do Fundo Monetário Internacional (FMI) e do Banco Mundial (Bird) no fim de semana.
Para o presidente do BC, a atual política econômica do governo e a manutenção da taxa de inflação dentro da meta são fatores que garantem queda na taxa de juros a longo prazo.
“Mas é importante que nós não olhemos para isso apenas no curto prazo. Temos que olhar no médio e longo prazos”, disse Meirelles.
Para ele, o trabalho que o Banco Central tem feito para conter a inflação dentro da meta vai gerar queda gradual da taxa de juros.
BC do ano
Neste domingo, Meirelles participou da cerimônia de entrega do prêmio “Banco Central do ano”, oferecido pela revista The Banker à instituição que preside.
“Este é um sinal importante de reconhecimento do nosso trabalho, de que o Brasil está indo na direção correta”, disse ele.
Segundo Meirelles, o Banco Central do Brasil está sendo percebido como uma instituição comprometida com a meta de inflação.
“Isso tem um ganho importante, na medida em que os agentes formadores de preços começam a apostar que a inflação vai estar na meta, que isso é levado a sério no Brasil”, completou.
O presidente do Banco Central disse ainda que a crise política brasileira não foi um assunto de interesse prioritário dos investidores que encontrou no fim de semana.
'''

inputs = tokenizer.encode(text, max_length=512, truncation=True, return_tensors='pt')
summary_ids = model_pt.generate(inputs, max_length=256, min_length=32, num_beams=5, no_repeat_ngram_size=3, early_stopping=True)
summary = tokenizer.decode(summary_ids[0])
print(summary)
#<pad> Meirelles diz que tendência de queda da taxa de juros no Brasil é real e é visível. Presidente do Banco Central do Brasil participou de cerimônia de entrega do prêmio ‘Banco Central do ano’ à instituição que preside.</s>

Citation

Deep Learning-Based Abstractive Summarization for Brazilian Portuguese Texts (PROPOR 2022)

@aInProceedings{ptt5summ_bracis,
  author="Paiola, Pedro H.
    and de Rosa, Gustavo H.
    and Papa, Jo{\~a}o P.",
  editor="Xavier-Junior, Jo{\~a}o Carlos
    and Rios, Ricardo Ara{\'u}jo",
  title="Deep Learning-Based Abstractive Summarization for Brazilian Portuguese Texts",
  booktitle="BRACIS 2022: Intelligent Systems",
  year="2022",
  publisher="Springer International Publishing",
  address="Cham",
  pages="479--493",
  isbn="978-3-031-21689-3"}

RecognaSumm: A Novel Brazilian Summarization Dataset (PROPOR 2024)

@inproceedings{paiola-etal-2024-recognasumm,
    title = "{R}ecogna{S}umm: A Novel {B}razilian Summarization Dataset",
    author = "Paiola, Pedro Henrique  and
      Garcia, Gabriel Lino  and
      Jodas, Danilo Samuel  and
      Correia, Jo{\~a}o Vitor Mariano  and
      Sugi, Luis Afonso  and
      Papa, Jo{\~a}o Paulo",
    editor = "Gamallo, Pablo  and
      Claro, Daniela  and
      Teixeira, Ant{\'o}nio  and
      Real, Livy  and
      Garcia, Marcos  and
      Oliveira, Hugo Gon{\c{c}}alo  and
      Amaro, Raquel",
    booktitle = "Proceedings of the 16th International Conference on Computational Processing of Portuguese - Vol. 1",
    month = mar,
    year = "2024",
    address = "Santiago de Compostela, Galicia/Spain",
    publisher = "Association for Computational Lingustics",
    url = "https://aclanthology.org/2024.propor-1.63",
    pages = "575--579",
}

Configuration

Architecture
T5ForConditionalGeneration
Vocabulary size
32,128
Stored precision
float32
Model type
t5

Identity and Version

Repository
recogna-nlp/ptt5-base-summ
Publisher
Recogna NLP
Task
Summarization
Modality
Text
Library
transformers
Parameters
223M parameters
Languages
pt
Revision
84509bf8ec61f26f6f9403a8c3ab55f4cfe2cc66
First published
2023-12-22
Last updated
2024-10-31

Files and Weights

9 files, 1.8 GB in total. The weights are 2 files totalling 1.8 GB in bin, safetensors.

Weights2 files · 1.8 GB
Configuration3 files · 3.1 KB
Tokenizer2 files · 758.0 KB
Documentation1 file · 12.2 KB
Repository1 file · 1.5 KB
Every file
FileTypeSizeSHA-256
model.safetensorsWeights891.6 MB 2e8d150b3b08
pytorch_model.binWeights891.7 MB 603281b7ef3d
config.jsonConfiguration780 B
generation_config.jsonConfiguration142 B
special_tokens_map.jsonConfiguration2.2 KB
README.mdDocumentation12.2 KB
.gitattributesRepository1.5 KB
spiece.modelTokenizer755.6 KB fcb25b1d67f0
tokenizer_config.jsonTokenizer2.4 KB

License and Download

License
mit
Access
Open weights, no gate
Download size
1.8 GB
Download from Recogna NLP

Released by Recogna NLP through its official repository on Hugging Face. Read the license.

Built From

  • Trained on (disclosed) recogna-nlp/recognasumm

Memory Requirements

PrecisionWeights in memory
As published1.8 GB
16-bit0.4 GB
8-bit0.2 GB
4-bit0.1 GB

Weights only, from the published parameter count; the key-value cache and runtime add to this.

Questions About ptt5-base-summ

How much GPU memory does ptt5-base-summ need?

About 0.5 GB at 16-bit and 0.1 GB at 4-bit: the weights (223M parameters) plus a working margin. A long context needs more.

What is the cheapest GPU to run ptt5-base-summ on?

At 16-bit, 1x MI300X from $1.85 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.

Can I use ptt5-base-summ commercially?

Yes. ptt5-base-summ is released under MIT License. The MIT License is a short permissive license. It permits commercial use, modification and redistribution, provided the copyright notice and permission notice are included.

Similar Models

Model · Summarization

rut5-base-summ

Dmitry Balobin

Finetuned ai-forever/ruT5-base for text and dialogue summarization. All 'train' subsets was concatenated and shuffled with seed 1000 - 7. Train subset = 155678 rows. Evaluation on 10% of concatenated 'validation' subsets = 1458 rows. See WandB logs. See report at REPORT WIP.

Open weights 223M parameters transformers

Model · Summarization

T5-base-10K-summarization

Yatharth Sant

This model is a fine-tuned version of Google's T5-Base model tailored for summarizing financial 10K report sections. T5-Base-10K-Summarization is optimized to condense lengthy 10K reports into manageable summaries, enabling quick insights into financial data and trends. Ideal for use by financial analysts and regulatory agencies needing rapid insights from 10K reports. It may not be suited for summarizing non-financial documents or informal texts. Trained on a diverse collection of 10K reports from various industries, annotated for summarization to ensure broad applicability and accuracy. The following hyperparameters were used during training: - learningrate: 5e-05 - trainbatchsize: 8…

Open weights 223M parameters transformers

Model · Summarization

ptt5-base-summ-xlsum

Recogna NLP

PTT5 Summ is a fine-tuned PTT5 model to perform Abstractive Summarization in Brazilian Portuguese texts. This model was fine-tuned on the datasets: RecognaSumm, WikiLingua, XL-Sum, TeMário.pdf) and CSTNews. For further information, please go to PTT5 Summ repository. author="Paiola, Pedro H. and de Rosa, Gustavo H. and Papa, Jo{\~a}o P.", editor="Xavier-Junior, Jo{\~a}o Carlos and Rios, Ricardo Ara{\'u}jo", title="Deep Learning-Based Abstractive Summarization for Brazilian Portuguese Texts", booktitle="BRACIS 2022: Intelligent Systems", year="2022", publisher="Springer International Publishing", address="Cham", pages="479--493", isbn="978-3-031-21689-3"} This model was fine-tuned using the…

Open weights cc-by-nc-sa-4.0 223M parameters transformers

Model · Summarization

led-base-book-summary

Peter Szemraj

The Longformer Encoder-Decoder (LED) for Narrative-Esque Long Text Summarization is a model I fine-tuned from allenai/led-base-16384 to condense extensive technical, academic, and narrative content in a fairly generalizable way. - Ideal for summarizing long narratives, articles, papers, textbooks, and other documents. - the sparknotes-esque style leads to 'explanations' in the summarized content, offering insightful output. The model was trained on the BookSum dataset released by SalesForce, which leads to the bsd-3-clause license. The training process involved 16 epochs with parameters tweaked to facilitate very fine-tuning-type training (super low learning rate). This model is the…

Open weights apache-2.0 162M parameters transformers

Model · Summarization

distilbart-news-summarizer

Sachin Rao

Imagine you have a very long news article, and you want someone to read it and tell you the key points in just a few sentences. That's exactly what this model does! This model takes a long news article and turns it into a short, easy-to-read summary. - It writes back a 3-4 sentence summary that captures all the important information The special thing about this model is that it's: 1. Very accurate - It understands news writing style very well 2. Very fast - It works quickly even on regular computers (not just expensive AI servers) 3. Specialized in news - It was trained specifically on news articles, so it understands how journalists write 4. Good with financial news - It knows market…

Open weights agpl-3.0 306M parameters 1,024 tokens transformers

Model · Summarization

kobart-summary-v3

EbanLee

이 모델은 kobart모델을 문서요약, 도서자료요약, 요약문 및 레포트 생성 데이터로 fine-tuning한 모델입니다.

Open weights 124M parameters 1,026 tokens transformers