metadata
language:
- en
- ko
license: llama3
library_name: transformers
base_model:
- meta-llama/Meta-Llama-3-8B
Bllossom | Demo | Homepage | Github | Colab-tutorial |
The Bllossom language model is a Korean-English bilingual language model based on the open-source LLama3. It enhances the connection of knowledge between Korean and English. It has the following features:
- Knowledge Linking: Linking Korean and English knowledge through additional training
- Vocabulary Expansion: Expansion of Korean vocabulary to enhance Korean expressiveness.
- Instruction Tuning: Tuning using custom-made instruction following data specialized for Korean language and Korean culture
- Human Feedback: DPO has been applied
- Vision-Language Alignment: Aligning the vision transformer with this language model
This model developed by MLPLab at Seoultech, Teddysum and Yonsei Univ
Demo Video
NEWS
- [2024/04] We released Bllossom v2.0, based on llama-3
- [2023/12] We released Bllossom-Vision v1.0, based on Bllossom
- [2023/08] We released Bllossom v1.0, based on llama-2.
- [2023/07] We released Bllossom v0.7, based on polyglot-ko.
์ ํฌ ์์ธ๊ณผ๊ธฐ๋ MLP์ฐ๊ตฌ์ค์์ ํ๊ตญ์ด-์์ด ์ด์ค ์ธ์ด๋ชจ๋ธ์ธ Bllossom์ ๊ณต๊ฐํ์ต๋๋ค!
- LLama3-8B ๊ธฐ๋ฐ์ ๊ฒฝ๋ํ๋ ์ฌ์ด์ฆ
- ํ๊ตญ์ด-์์ด ์ง์์ฐ๊ฒฐ์ ํตํ ํ๊ตญ์ด ์ง์ ๊ฐํ
- ํ๊ตญ์ด ์ดํ์ถ๊ฐ
- ํ๊ตญ์ด ๋ฌธํ, ์ธ์ด๋ฅผ ๊ณ ๋ คํ ์์ฒด์ ์ ๋ฐ์ดํฐ ๊ธฐ๋ฐ ๋ฏธ์ธ์กฐ์
- ๊ฐํํ์ต (DPO)
- ์๊ฐ-์ธ์ด ๋ชจ๋ธํ์ฅ
1. Bllossom์ ์์ธ๊ณผ๊ธฐ๋, ํ
๋์ธ, ์ฐ์ธ๋ ์ธ์ด์์ ์ฐ๊ตฌ์ค์ ์ธ์ดํ์์ ํ์
ํด ๋ง๋ ์ค์ฉ์ฃผ์๊ธฐ๋ฐ ์ธ์ด๋ชจ๋ธ์
๋๋ค! ์์ผ๋ก ์ง์์ ์ธ ์
๋ฐ์ดํธ๋ฅผ ํตํด ๊ด๋ฆฌํ๊ฒ ์ต๋๋ค ๋ง์ด ํ์ฉํด์ฃผ์ธ์ ๐
2. Bllossom70B๋ชจ๋ธ, ์ดํํ์ฅ๋ชจ๋ธ, ์๊ฐ-์ธ์ด๋ชจ๋ธ์ ์ถํ ๊ณต๊ฐํ ์์ ์
๋๋ค. (๊ถ๊ธํ์ ๋ถ์ ๊ฐ๋ณ ์ฐ๋ฝ์ฃผ์ธ์, GPU๋ง ์ง์ํด์ฃผ์๋ฉด ๋ฌด๋ฃ๋ก ๋๋ฆฝ๋๋ค!)
3. Bllossom์ NAACL2024, LREC-COLING2024 (๊ตฌ๋) ๋ฐํ๋ก ์ฑํ๋์์ต๋๋ค.
4. ์ข์ ์ธ์ด๋ชจ๋ธ ๊ณ์ ์
๋ฐ์ดํธ ํ๊ฒ ์ต๋๋ค!! ํ๊ตญ์ด ๊ฐํ๋ฅผ์ํด ๊ณต๋ ์ฐ๊ตฌํ์ค๋ถ ์ธ์ ๋ ํ์ํฉ๋๋ค!!
Example code
Colab Tutorial
Install Dependencies
pip install torch transformers==4.40.0 accelerate
Python code with Pipeline
import transformers
import torch
model_id = "MLP-KTLim/llama3-Bllossom"
pipeline = transformers.pipeline(
"text-generation",
model=model_id,
model_kwargs={"torch_dtype": torch.bfloat16},
device_map="auto",
)
pipeline.model.eval()
PROMPT = '''๋น์ ์ ์ ์ฉํ AI ์ด์์คํดํธ์
๋๋ค. ์ฌ์ฉ์์ ์ง์์ ๋ํด ์น์ ํ๊ณ ์ ํํ๊ฒ ๋ต๋ณํด์ผ ํฉ๋๋ค.'''
instruction = "์์ธ๊ณผํ๊ธฐ์ ๋ํ๊ต MLP์ฐ๊ตฌ์ค์ ๋ํด ์๊ฐํด์ค"
messages = [
{"role": "system", "content": f"{PROMPT}"},
{"role": "user", "content": f"{instruction}"}
]
prompt = pipeline.tokenizer.apply_chat_template(
messages,
tokenize=False,
add_generation_prompt=True
)
terminators = [
pipeline.tokenizer.eos_token_id,
pipeline.tokenizer.convert_tokens_to_ids("<|eot_id|>")
]
outputs = pipeline(
prompt,
max_new_tokens=2048,
eos_token_id=terminators,
do_sample=True,
temperature=0.6,
top_p=0.9,
repetition_penalty = 1.1
)
print(outputs[0]["generated_text"][len(prompt):])
# ์์ธ๊ณผํ๊ธฐ์ ๋ํ๊ต MLP์ฐ๊ตฌ์ค์ ๋ฉํฐ๋ชจ๋ฌ ์์ฐ์ด์ฒ๋ฆฌ ์ฐ๊ตฌ๋ฅผ ํ๊ณ ์์ต๋๋ค. ๊ตฌ์ฑ์์ ์๊ฒฝํ ๊ต์์ ๊น๋ฏผ์ค, ๊น์๋ฏผ, ์ต์ฐฝ์, ์์ธํธ, ์ ํ๊ฒฐ, ์ํ์, ์ก์น์ฐ, ์ก์ ํ, ์ ๋์ฌ ํ์์ด ์์ต๋๋ค.
Python code with AutoModel
import os
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM
model_id = 'MLP-KTLim/llama3-Bllossom'
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
model_id,
torch_dtype=torch.bfloat16,
device_map="auto",
)
model.eval()
PROMPT = '''๋น์ ์ ์ ์ฉํ AI ์ด์์คํดํธ์
๋๋ค. ์ฌ์ฉ์์ ์ง์์ ๋ํด ์น์ ํ๊ณ ์ ํํ๊ฒ ๋ต๋ณํด์ผ ํฉ๋๋ค.'''
instruction = "์์ธ๊ณผํ๊ธฐ์ ๋ํ๊ต MLP์ฐ๊ตฌ์ค์ ๋ํด ์๊ฐํด์ค"
messages = [
{"role": "system", "content": f"{PROMPT}"},
{"role": "user", "content": f"{instruction}"}
]
input_ids = tokenizer.apply_chat_template(
messages,
add_generation_prompt=True,
return_tensors="pt"
).to(model.device)
terminators = [
tokenizer.eos_token_id,
tokenizer.convert_tokens_to_ids("<|eot_id|>")
]
outputs = model.generate(
input_ids,
max_new_tokens=2048,
eos_token_id=terminators,
do_sample=True,
temperature=0.6,
top_p=0.9,
repetition_penalty = 1.1
)
print(tokenizer.decode(outputs[0][input_ids.shape[-1]:], skip_special_tokens=True))
# ์์ธ๊ณผํ๊ธฐ์ ๋ํ๊ต MLP์ฐ๊ตฌ์ค์ ๋ฉํฐ๋ชจ๋ฌ ์์ฐ์ด์ฒ๋ฆฌ ์ฐ๊ตฌ๋ฅผ ํ๊ณ ์์ต๋๋ค. ๊ตฌ์ฑ์์ ์๊ฒฝํ ๊ต์์ ๊น๋ฏผ์ค, ๊น์๋ฏผ, ์ต์ฐฝ์, ์์ธํธ, ์ ํ๊ฒฐ, ์ํ์, ์ก์น์ฐ, ์ก์ ํ, ์ ๋์ฌ ํ์์ด ์์ต๋๋ค.
Citation
Language Model
@misc{bllossom,
author = {ChangSu Choi, Yongbin Jeong, Seoyoon Park, InHo Won, HyeonSeok Lim, SangMin Kim, Yejee Kang, Chanhyuk Yoon, Jaewan Park, Yiseul Lee, HyeJin Lee, Younggyun Hahm, Hansaem Kim, KyungTae Lim},
title = {Optimizing Language Augmentation for Multilingual Large Language Models: A Case Study on Korean},
year = {2024},
journal = {LREC-COLING 2024},
paperLink = {\url{https://arxiv.org/pdf/2403.10882}},
},
}
Vision-Language Model
@misc{bllossom-V,
author = {Dongjae Shin, Hyunseok Lim, Inho Won, Changsu Choi, Minjun Kim, Seungwoo Song, Hangyeol Yoo, Sangmin Kim, Kyungtae Lim},
title = {X-LLaVA: Optimizing Bilingual Large Vision-Language Alignment},
year = {2024},
publisher = {GitHub},
journal = {NAACL 2024 findings},
paperLink = {\url{https://arxiv.org/pdf/2403.11399}},
},
}
Contact
- ์๊ฒฝํ(KyungTae Lim), Professor at Seoultech.
[email protected]
- ํจ์๊ท (Younggyun Hahm), CEO of Teddysum.
[email protected]
- ๊นํ์(Hansaem Kim), Professor at Yonsei.
[email protected]
Contributor
- ์ต์ฐฝ์(Chansu Choi), [email protected]
- ๊น์๋ฏผ(Sangmin Kim), [email protected]
- ์์ธํธ(Inho Won), [email protected]
- ๊น๋ฏผ์ค(Minjun Kim), [email protected]
- ์ก์น์ฐ(Seungwoo Song), [email protected]
- ์ ๋์ฌ(Dongjae Shin), [email protected]
- ์ํ์(Hyeonseok Lim), [email protected]
- ์ก์ ํ(Jeonghun Yuk), [email protected]
- ์ ํ๊ฒฐ(Hangyeol Yoo), [email protected]
- ์ก์ํ(Seohyun Song), [email protected]