dnotitia/Llama-DNA-1.0-8B-Instruct

DNA 1.0 8B Instruct

DNA 1.0 8B Instruct is a state-of-the-art (SOTA) bilingual language model based on Llama architecture, specifically optimized for Korean language understanding and generation, while also maintaining strong English capabilities. The model was developed through a sophisticated process involving model merging via spherical linear interpolation (SLERP) with Llama 3.1 8B Instruct, and underwent knowledge distillation (KD) using Llama 3.1 405B as the teacher model. It was extensively trained through continual pre-training (CPT) with a high-quality Korean dataset. The training pipeline was completed with supervised fine-tuning (SFT) and direct preference optimization (DPO) to align with human preferences and enhance instruction-following abilities.

DNA 1.0 8B Instruct was fine-tuned on approximately 10B tokens of carefully curated data and has undergone extensive instruction tuning to enhance its ability to follow complex instructions and engage in natural conversations.

• Developed by: Dnotitia Inc.
• Supported Languages: Korean, English
• Model Release Date: Dec 10, 2024
• Vocab Size: 128,256
• Context Length: 131,072 tokens (128k)
• License: CC BY-NC 4.0

NOTICE (Korean):

본 모델은 상업적 목적으로 활용하실 수 있습니다. 상업적 이용을 원하시는 경우, Contact us를 통해 문의해 주시기 바랍니다. 간단한 협의 절차를 거쳐 상업적 활용을 승인해 드리도록 하겠습니다.

Try DNA-powered Mnemos Assistant! Beta Open →

Training Procedure

Evaluation

We evaluated DNA 1.0 8B Instruct against other prominent language models of similar size across various benchmarks, including Korean-specific tasks and general language understanding metrics. More details will be provided in the upcoming Technical Report.

Language	Benchmark	dnotitia/Llama-DNA-1.0-8B-Instruct	LGAI-EXAONE/EXAONE-3.5-7.8B-Instruct	LGAI-EXAONE/EXAONE-3.0-7.8B-Instruct	yanolja/EEVE-Korean-Instruct-10.8B-v1.0	Qwen/Qwen2.5-7B-Instruct	meta-llama/Llama-3.1-8B-Instruct	mistralai/Mistral-7B-Instruct-v0.3	NCSOFT/Llama-VARCO-8B-Instruct	upstage/SOLAR-10.7B-Instruct-v1.0
Korean	KMMLU	53.26 (1st)	45.30	45.28	42.17	45.66	41.66	31.45	38.49	41.50
	KMMLU-hard	29.46 (1st)	23.17	20.78	19.25	24.78	20.49	17.86	19.83	20.61
	KoBEST	83.40 (1st)	79.05	80.13	81.67	78.51	67.56	63.77	72.99	73.26
	Belebele	57.99 (1st)	40.97	45.11	49.40	54.85	54.70	40.31	53.17	48.68
	CSATQA	43.32 (2nd)	40.11	34.76	39.57	45.45	36.90	27.27	32.62	34.22
English	MMLU	66.64 (3rd)	65.27	64.32	63.63	74.26	68.26	62.04	63.25	65.30
	MMLU-Pro	43.05 (1st)	40.73	38.90	32.79	42.5	40.92	33.49	37.11	30.25
	GSM8K	80.52 (1st)	65.96	80.06	56.18	75.74	75.82	49.66	64.14	69.22

• The highest scores are in bold form, and the second-highest scores are underlined.

Evaluation Protocol
For easy reproduction of our evaluation results, we list the evaluation tools and settings used below:

	Evaluation setting	Metric	Evaluation tool
KMMLU	5-shot	macro_avg / exact_match	lm-eval-harness
KMMLU Hard	5-shot	macro_avg / exact_match	lm-eval-harness
KoBEST	5-shot	macro_avg / f1	lm-eval-harness
Belebele	0-shot	acc	lm-eval-harness
CSATQA	0-shot	acc_norm	lm-eval-harness
MMLU	5-shot	macro_avg / acc	lm-eval-harness
MMLU Pro	5-shot	macro_avg / exact_match	lm-eval-harness
GSM8K	5-shot	acc, exact_match & strict_extract	lm-eval-harness

Quickstart

This model requires transformers >= 4.43.0.

from transformers import AutoModelForCausalLM, AutoTokenizer, TextStreamer

tokenizer = AutoTokenizer.from_pretrained('dnotitia/Llama-DNA-1.0-8B-Instruct')
model = AutoModelForCausalLM.from_pretrained('dnotitia/Llama-DNA-1.0-8B-Instruct', device_map='auto')
streamer = TextStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)

conversation = [
    {"role": "system", "content": "You are a helpful assistant, Dnotitia DNA."},
    {"role": "user", "content": "너의 이름은?"},
]
inputs = tokenizer.apply_chat_template(conversation,
                                       add_generation_prompt=True,
                                       return_dict=True,
                                       return_tensors="pt").to(model.device)
_ = model.generate(**inputs, streamer=streamer)

Limitations

While DNA 1.0 8B Instruct demonstrates strong performance, users should be aware of the following limitations:

• The model may occasionally generate biased or inappropriate content
• Responses are based on training data and may not reflect current information
• The model may sometimes produce factually incorrect or inconsistent answers
• Performance may vary depending on the complexity and domain of the task
• Generated content should be reviewed for accuracy and appropriateness

License

This model is released under CC BY-NC 4.0 license. For commercial usage inquiries, please Contact us.

Appendix

• KMMLU scores comparison chart:

  

• DNA 1.0 8B Instruct model architecture ¹:

  

• The median percentage of model’s weight difference between before and after the merge (our SFT model + Llama 3.1 8B Instruct):

Citation

If you use or discuss this model in your academic research, please cite the project to help spread awareness:

@article{dnotitiadna2024,
  title = {Dnotitia DNA 1.0 8B Instruct},
  author = {Jungyup Lee, Jemin Kim, Sang Park, Seungjae Lee},
  year = {2024},
  url = {https://huggingface.co./dnotitia/DNA-1.0-8B-Instruct},
  version = {1.0},
}