Spaces:
Running
Running
File size: 4,150 Bytes
650c5f6 |
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 |
# WMT 20
This page provides pointers to the models of Facebook-FAIR's WMT'20 news translation task submission [(Chen et al., 2020)](https://arxiv.org/abs/2011.08298).
## Single best MT models (after finetuning on part of WMT20 news dev set)
Model | Description | Download
---|---|---
`transformer.wmt20.ta-en` | Ta->En | [download (.tar.gz)](https://dl.fbaipublicfiles.com/fairseq/models/wmt20.ta-en.single.tar.gz)
`transformer.wmt20.en-ta` | En->Ta | [download (.tar.gz)](https://dl.fbaipublicfiles.com/fairseq/models/wmt20.en-ta.single.tar.gz)
`transformer.wmt20.iu-en.news` | Iu->En (News domain) | [download (.tar.gz)](https://dl.fbaipublicfiles.com/fairseq/models/wmt20.iu-en.news.single.tar.gz)
`transformer.wmt20.en-iu.news` | En->Iu (News domain) | [download (.tar.gz)](https://dl.fbaipublicfiles.com/fairseq/models/wmt20.en-iu.news.single.tar.gz)
`transformer.wmt20.iu-en.nh` | Iu->En (Nunavut Hansard domain) | [download (.tar.gz)](https://dl.fbaipublicfiles.com/fairseq/models/wmt20.iu-en.nh.single.tar.gz)
`transformer.wmt20.en-iu.nh` | En->Iu (Nunavut Hansard domain) | [download (.tar.gz)](https://dl.fbaipublicfiles.com/fairseq/models/wmt20.en-iu.nh.single.tar.gz)
## Language models
Model | Description | Download
---|---|---
`transformer_lm.wmt20.en` | En Language Model | [download (.tar.gz)](https://dl.fbaipublicfiles.com/fairseq/models/wmt20.en.tar.gz)
`transformer_lm.wmt20.ta` | Ta Language Model | [download (.tar.gz)](https://dl.fbaipublicfiles.com/fairseq/models/wmt20.ta.tar.gz)
`transformer_lm.wmt20.iu.news` | Iu Language Model (News domain) | [download (.tar.gz)](https://dl.fbaipublicfiles.com/fairseq/models/wmt20.iu.news.tar.gz)
`transformer_lm.wmt20.iu.nh` | Iu Language Model (Nunavut Hansard domain) | [download (.tar.gz)](https://dl.fbaipublicfiles.com/fairseq/models/wmt20.iu.nh.tar.gz)
## Example usage (torch.hub)
#### Translation
```python
import torch
# English to Tamil translation
en2ta = torch.hub.load('pytorch/fairseq', 'transformer.wmt20.en-ta')
en2ta.translate("Machine learning is great!") # 'இயந்திரக் கற்றல் அருமை!'
# Tamil to English translation
ta2en = torch.hub.load('pytorch/fairseq', 'transformer.wmt20.ta-en')
ta2en.translate("இயந்திரக் கற்றல் அருமை!") # 'Machine learning is great!'
# English to Inuktitut translation
en2iu = torch.hub.load('pytorch/fairseq', 'transformer.wmt20.en-iu.news')
en2iu.translate("machine learning is great!") # 'ᖃᒧᑕᐅᔭᓄᑦ ᐃᓕᓐᓂᐊᕐᓂᖅ ᐱᐅᔪᒻᒪᕆᒃ!'
# Inuktitut to English translation
iu2en = torch.hub.load('pytorch/fairseq', 'transformer.wmt20.iu-en.news')
iu2en.translate("ᖃᒧᑕᐅᔭᓄᑦ ᐃᓕᓐᓂᐊᕐᓂᖅ ᐱᐅᔪᒻᒪᕆᒃ!") # 'Machine learning excellence!'
```
#### Language Modeling
```python
# Sample from the English LM
en_lm = torch.hub.load('pytorch/fairseq', 'transformer_lm.wmt20.en')
en_lm.sample("Machine learning is") # 'Machine learning is a type of artificial intelligence that uses machine learning to learn from data and make predictions.'
# Sample from the Tamil LM
ta_lm = torch.hub.load('pytorch/fairseq', 'transformer_lm.wmt20.ta')
ta_lm.sample("இயந்திரக் கற்றல் என்பது செயற்கை நுண்ணறிவின்") # 'இயந்திரக் கற்றல் என்பது செயற்கை நுண்ணறிவின் ஒரு பகுதியாகும்.'
# Sample from the Inuktitut LM
iu_lm = torch.hub.load('pytorch/fairseq', 'transformer_lm.wmt20.iu.news')
iu_lm.sample("ᖃᒧᑕᐅᔭᓄᑦ ᐃᓕᓐᓂᐊᕐᓂᖅ") # 'ᖃᒧᑕᐅᔭᓄᑦ ᐃᓕᓐᓂᐊᕐᓂᖅ, ᐊᒻᒪᓗ ᓯᓚᐅᑉ ᐊᓯᙳᖅᐸᓪᓕᐊᓂᖓᓄᑦ ᖃᓄᐃᓕᐅᕈᑎᒃᓴᑦ, ᐃᓚᖃᖅᖢᑎᒃ ᐅᑯᓂᖓ:'
```
## Citation
```bibtex
@inproceedings{chen2020facebook
title={Facebook AI's WMT20 News Translation Task Submission},
author={Peng-Jen Chen and Ann Lee and Changhan Wang and Naman Goyal and Angela Fan and Mary Williamson and Jiatao Gu},
booktitle={Proc. of WMT},
year={2020},
}
```
|