README.md

---
license: cc-by-nc-sa-4.0
datasets:
- squarelike/sharegpt_deepl_ko_translation
language:
- ko
pipeline_tag: translation
tags:
- translate
---
## **Seagull-13b-translation 📇**
![Seagull-typewriter](./Seagull-typewriter.png)
**Seagull-13b-translation** is yet another translator model, but carefully considered the following issues from existing translation models.
- Exact match of `newline` or `space`
- Not using dataset with first letter removed
- Code
- Markdown format
- LaTeX format
- etc

이런 이슈들을 충분히 체크하고 학습을 진행하였지만, 모델을 사용할 때는 이런 부분에 대한 결과를 면밀하게 살펴보는 것을 추천합니다(코드가 포함된 텍스트 등).

> If you're interested in building large-scale language models to solve a wide variety of problems in a wide variety of domains, you should consider joining [Allganize](https://allganize.career.greetinghr.com/o/65146).
For a coffee chat or if you have any questions, please do not hesitate to contact me as well! - kuotient.dev@gmail.com

This model was created as a personal experiment, unrelated to the organization I work for.

## **License**
## From original model author:
- Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International Public License, under LLAMA 2 COMMUNITY LICENSE AGREEMENT
- Full License available at: https://huggingface.co/beomi/llama-2-koen-13b/blob/main/LICENSE

# **Model Details**
#### **Developed by**
Jisoo Kim(kuotient)
#### **Base Model**  
[beomi/llama-2-koen-13b](https://huggingface.co/beomi/llama-2-koen-13b)
#### **Datasets**
- [sharegpt_deepl_ko_translation](https://huggingface.co/datasets/squarelike/sharegpt_deepl_ko_translation)
- [KOR-OpenOrca-Platypus-v3](https://huggingface.co/datasets/kyujinpy/KOR-OpenOrca-Platypus-v3)
- AIHUB
  - 기술과학 분야 한-영 번역 병렬 말뭉치 데이터
  - 일상생활 및 구어체 한-영 번역 병렬 말뭉치 데이터

## **Usage**
#### **Format**
It follows only **ChatML** format.

```python
<|im_start|>system
주어진 문장을 한국어로 번역하세요.<|im_end|>
<|im_start|>user
{instruction}<|im_end|>
<|im_start|>assistant
# Don't miss newline here
```
```python
<|im_start|>system
주어진 문장을 영어로 번역하세요.<|im_end|>
<|im_start|>user
{instruction}<|im_end|>
<|im_start|>assistant
# Don't miss newline here
```

#### Output example
원문:
> A particle's wave function, $\psi(x)$, is given by $$\psi(x)=\begin{cases} 3x & \text{if } -1 \leq x \leq 0 \\ 3(1-x) & \text{if } 0 < x \leq 1 \\ 0 & \text{otherwise} \end{cases}$$ Compute the Fourier transform, $\tilde{\psi}(k)$, of the wave function $\psi(x)$ and show that it satisfies the Fourier inversion theorem, i.e., $\psi(x) = \frac{1}{\sqrt{2\pi}} \int_{-\infty}^{\infty} \tilde{\psi}(k) e^{ikx} \mathrm{d}k$.

Seagull-13b-translation:
> 입자의 파동 함수 $\psi(x)$는 다음과 같이 주어집니다. $$\psi(x)=\begin{cases} 3x & \text{if } -1 \leq x \leq 0 \\ 3(1-x) & \text{if } 0 < x \leq 1 \\ 0 & \text{otherwise} \end{cases}$$ 파동 함수 $\psi(x)$의 푸리에 변환 $\tilde{\psi}(k)$를 계산하고 푸리에 반전 정리, 즉 $\psi(x) = \frac{1}{\sqrt{2\pi}} \int_{-\infty}^{\infty} \tilde{\psi}(k) e^{ikx} \mathrm{d}k$를 만족합니다.

DeepL:
> 입자의 파동 함수 $\psi(x)$는 $$\psi(x)=\begin{cases}로 주어집니다. 3x & \text{if } -1 \leq x \leq 0 \\ 3(1-x) & \text{if } 0 < x \leq 1 \\ 0 & \text{기타} \end{cases}$$ 파동 함수 $\psi(x)$의 푸리에 변환인 $\tilde{\psi}(k)$를 계산하고 푸리에 반전 정리, 즉 $\psi(x) = \frac{1}{\sqrt{2\pi}}를 만족함을 증명합니다. \int_{-\infty}^{\infty} \물결표{\psi}(k) e^{ikx} \mathrm{d}k$.

...and much more awesome cases with SQL query, code and markdown!

#### **How to**
**I highly recommend to inference model with vllm. I will write a guide for quick and easy inference if requested.**
Since, chat_template already contains insturction format above.
You can use the code below.
```python
from transformers import AutoModelForCausalLM, AutoTokenizer
device = "cuda" # the device to load the model onto
model = AutoModelForCausalLM.from_pretrained("kuotient/Seagull-13B-translation")
tokenizer = AutoTokenizer.from_pretrained("kuotient/Seagull-13B-translation")
messages = [
    {"role": "system", "content", "주어진 문장을 한국어로 번역하세요."}
    {"role": "user", "content": "Here are five examples of nutritious foods to serve your kids."},
]
encodeds = tokenizer.apply_chat_template(messages, return_tensors="pt")

model_inputs = encodeds.to(device)
model.to(device)
generated_ids = model.generate(model_inputs, max_new_tokens=1000, do_sample=True)
decoded = tokenizer.batch_decode(generated_ids)
print(decoded[0])
```