File size: 1,819 Bytes
02844c5 6a7b744 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 |
---
language:
- en
- pt
datasets:
- EMEA
- ParaCrawl 99k
- CAPES
- Scielo
- JRC-Acquis
- Biomedical Domain Corpora
tags:
- translation
metrics:
- bleu
---
# Introduction
This repository brings an implementation of T5 for translation in PT-EN tasks using a modest hardware setup. We propose some changes in tokenizator and post-processing that improves the result and used a Portuguese pretrained model for the translation. You can collect more informations in [our repository](https://github.com/unicamp-dl/Lite-T5-Translation). Also, check [our paper](https://aclanthology.org/2020.wmt-1.90.pdf)!
# Usage
Just follow "Use in Transformers" instructions. It is necessary to add a few words before to define the task to T5.
You can also create a pipeline for it. An example with the phrase " Eu gosto de comer arroz" is:
```python
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline
tokenizer = AutoTokenizer.from_pretrained("unicamp-dl/translation-pt-en-t5")
model = AutoModelForSeq2SeqLM.from_pretrained("unicamp-dl/translation-pt-en-t5")
pten_pipeline = pipeline('text2text-generation', model=model, tokenizer=tokenizer)
pten_pipeline("translate Portuguese to English: Eu gosto de comer arroz.")
```
# Citation
```bibtex
@inproceedings{lopes-etal-2020-lite,
title = "Lite Training Strategies for {P}ortuguese-{E}nglish and {E}nglish-{P}ortuguese Translation",
author = "Lopes, Alexandre and
Nogueira, Rodrigo and
Lotufo, Roberto and
Pedrini, Helio",
booktitle = "Proceedings of the Fifth Conference on Machine Translation",
month = nov,
year = "2020",
address = "Online",
publisher = "Association for Computational Linguistics",
url = "https://www.aclweb.org/anthology/2020.wmt-1.90",
pages = "833--840",
}
``` |