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",
}
```