|
--- |
|
license: cc-by-4.0 |
|
language: |
|
- en |
|
- ko |
|
tags: |
|
- translation |
|
--- |
|
|
|
I only did ctranslate2 convert to the original. |
|
|
|
cmd> ct2-transformers-converter --model NHNDQ/nllb-finetuned-en2ko --quantization float16 --output_dir NHNDQ-nllb-finetuned-en2ko-ct2 |
|
|
|
All copyrights belong to the original authors and the CT model may be deleted upon request. Below is the original model information. |
|
|
|
Original URL : https://huggingface.co/NHNDQ/nllb-finetuned-en2ko |
|
|
|
## Model Details |
|
* Model Description: Fine-tuned facebook/nllb-200-distilled-600M ct2 model |
|
* Developed by: DanielHeo |
|
* |
|
## Original Model Details |
|
* Model Description: Fine-tuned facebook/nllb-200-distilled-600M model |
|
* Developed by: Jisu Kim, Juhwan Lee, TakSung Heo, and Minsu Jeong |
|
* Model Type: Translation |
|
* Language(s): |
|
* Source Language: English |
|
* Target Language: Korean |
|
* License: CC-BY-4.0 |
|
|
|
## Dataset |
|
* [AI-hub dataset](https://www.aihub.or.kr/) |
|
|
|
## BLEU Score |
|
* Deepl translation: 22.83 |
|
* Fine-tune nllb: 33.66 |
|
|
|
## Uses |
|
This model can be used for translation and text-to-text generation |
|
|
|
## Data Augmentation with backtranslation task |
|
You can exercise korean data augmentation task with python package [KoTAN](https://github.com/KoJLabs/KoTAN/tree/main) |
|
|
