|
|
--- |
|
|
datasets: |
|
|
- Helsinki-NLP/tatoeba |
|
|
language: |
|
|
- ko |
|
|
- en |
|
|
metrics: |
|
|
- bleu |
|
|
- chrf |
|
|
pipeline_tag: translation |
|
|
library_name: transformers |
|
|
|
|
|
--- |
|
|
# Model info |
|
|
|
|
|
Forked from [odegiber/ko-en](https://huggingface.co/odegiber/ko-en), with a `.tflite` version of the model weights added. |
|
|
|
|
|
Distilled model from a Tatoeba-MT Teacher: [Tatoeba-MT-models/kor-eng/opusTCv20210807-sepvoc_transformer-big_2022-07-28](https://object.pouta.csc.fi/Tatoeba-MT-models/kor-eng/opusTCv20210807-sepvoc_transformer-big_2022-07-28.zip), which has been trained on the [Tatoeba](https://github.com/Helsinki-NLP/Tatoeba-Challenge/tree/master/data) dataset. |
|
|
|
|
|
We used the [OpusDistillery](https://github.com/Helsinki-NLP/OpusDistillery) to train new a new student with the tiny architecture, with a regular transformer decoder. |
|
|
For training data, we used [Tatoeba](https://github.com/Helsinki-NLP/Tatoeba-Challenge/tree/master/data). |
|
|
The configuration file fed into OpusDistillery can be found [here](https://github.com/Helsinki-NLP/OpusDistillery/blob/main/configs/hplt/config.hplt.kor-eng.yml). |
|
|
|
|
|
## How to run |
|
|
```python |
|
|
>>> from transformers import pipeline |
|
|
|
|
|
>>> pipe = pipeline("translation", model="odegiber/ko-en", max_length=256) |
|
|
>>> pipe("2017년 말, 시미노프는 쇼핑 텔레비젼 채널인 QVC에 출연했다.") |
|
|
[{'translation_text': 'At the end of 2017, Siminof appeared on the shopping television channel QVC.'}] |
|
|
|
|
|
``` |
|
|
|
|
|
## Benchmarks |
|
|
|
|
|
| testset | BLEU | chr-F | |
|
|
|-----------------------|-------|-------| |
|
|
| flores200 | 20.3 | 50.3 | |
|
|
|
