README.md

---

datasets:
- mc4
license: apache-2.0
---


# ByT5-Korean - large

ByT5-Korean is a Korean specific extension of Google's [ByT5](https://github.com/google-research/byt5).

A Korean syllable has three components (called Jamo): a beginning consonant, a middle vowel, and an optional final consonant; they are like individual characters of alphabet.
While the ByT5's utf-8 encoding allows generic encoding for multiple languages, it is unnatural for Korean because it splits the bits representation of each Jamo in the middle.

ByT5-Korean extends ByT5's utf-8 encoding with special care for Korean syllables; each Jamo is represented with a extra token.
ByT5-Korean was pre-trained on [mC4](https://www.tensorflow.org/datasets/catalog/c4#c4multilingual) with 70% Korean and 30% English.

## Encoding Scheme
```text

id: token

0: <pad>

1: <unk>

2: <eos>

3~258: utf-8 encoding

259~277: beginning consonants(초성), from ㄱ to ㅎ

279~299: middle vowel(중성), from ㅏ to ㅣ

300~327: final consonant(종성), None, from ㄱ to ㅎ

328~384: from <extra_id_0> to <extra_id_56>

```
## Example Inference

```python

import torch

from tokenizer import ByT5KoreanTokenizer # https://github.com/everdoubling/byt5-Korean

from transformers import T5ForConditionalGeneration



tokenizer_jamo = ByT5KoreanTokenizer()

model = T5ForConditionalGeneration.from_pretrained('everdoubling/byt5-Korean-large')



input_sentence = '한국어 위키백과(영어: Korean Wikipedia)는 한국어로 운영되는 위키백과의 다언어판 가운데 하나로서, 2002년 10월 11일에 <extra_id_0>. 또한 현재 한국어 위키백과에는 넘겨주기, 토론, 그림 등 페이지로 불리는 모든 문서를 포함하면 총 2,629,860개가 <extra_id_1>되어 있으며, 넘겨주기를 포함한 일반 문서 수는 1,278,560개,[1] 그중 넘겨주기, 막다른 문서를 제외한 일반 문서 수는 573,149개이다.'



input_ids_jamo = tokenizer_jamo(input_sentence).input_ids

outputs_jamo = model_jamo.generate(torch.tensor([input_ids_jamo]))

print(tokenizer_jamo.decode(outputs_jamo[0]))

# <pad><extra_id_0>설립되었다<extra_id_1>đě

```

Additional information coming soon...