kuotient's picture
Update README.md
6e910fd verified
metadata
license: cc-by-nc-sa-4.0
datasets:
  - squarelike/sharegpt_deepl_ko_translation
language:
  - ko
pipeline_tag: translation
tags:
  - translate
  - awq

Seagull-13b-translation-AWQ ๐Ÿ“‡

Seagull-typewriter

This is quantized version of original model: Seagull-13b-translation.

Seagull-13b-translation is yet another translator model, but carefully considered the following issues from existing translation models.

  • newline or space not matching the original text
  • Using translated dataset with first letter removed for training
  • Codes
  • Markdown format
  • LaTeX format
  • etc

์ด๋Ÿฐ ์ด์Šˆ๋“ค์„ ์ถฉ๋ถ„ํžˆ ์ฒดํฌํ•˜๊ณ  ํ•™์Šต์„ ์ง„ํ–‰ํ•˜์˜€์ง€๋งŒ, ๋ชจ๋ธ์„ ์‚ฌ์šฉํ•  ๋•Œ๋Š” ์ด๋Ÿฐ ๋ถ€๋ถ„์— ๋Œ€ํ•œ ๊ฒฐ๊ณผ๋ฅผ ๋ฉด๋ฐ€ํ•˜๊ฒŒ ์‚ดํŽด๋ณด๋Š” ๊ฒƒ์„ ์ถ”์ฒœํ•ฉ๋‹ˆ๋‹ค(์ฝ”๋“œ๊ฐ€ ํฌํ•จ๋œ ํ…์ŠคํŠธ ๋“ฑ).

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. For a coffee chat or if you have any questions, please do not hesitate to contact me as well! - [email protected]

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

License

From original model author:

Model Details

Developed by

Jisoo Kim(kuotient)

Base Model

beomi/llama-2-koen-13b

Datasets

  • sharegpt_deepl_ko_translation
  • AIHUB
    • ๊ธฐ์ˆ ๊ณผํ•™ ๋ถ„์•ผ ํ•œ-์˜ ๋ฒˆ์—ญ ๋ณ‘๋ ฌ ๋ง๋ญ‰์น˜ ๋ฐ์ดํ„ฐ
    • ์ผ์ƒ์ƒํ™œ ๋ฐ ๊ตฌ์–ด์ฒด ํ•œ-์˜ ๋ฒˆ์—ญ ๋ณ‘๋ ฌ ๋ง๋ญ‰์น˜ ๋ฐ์ดํ„ฐ

Usage

Format

It follows only ChatML format.

<|im_start|>system
์ฃผ์–ด์ง„ ๋ฌธ์žฅ์„ ํ•œ๊ตญ์–ด๋กœ ๋ฒˆ์—ญํ•˜์„ธ์š”.<|im_end|>
<|im_start|>user
{instruction}<|im_end|>
<|im_start|>assistant
# Don't miss newline here
<|im_start|>system
์ฃผ์–ด์ง„ ๋ฌธ์žฅ์„ ์˜์–ด๋กœ ๋ฒˆ์—ญํ•˜์„ธ์š”.<|im_end|>
<|im_start|>user
{instruction}<|im_end|>
<|im_start|>assistant
# Don't miss newline here

Example

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.

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", "์ฃผ์–ด์ง„ ๋ฌธ์žฅ์„ ํ•œ๊ตญ์–ด๋กœ ๋ฒˆ์—ญํ•˜์„ธ์š”."}
    {"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])