File size: 10,434 Bytes
110bee3 0abcaf8 c5afa6b 110bee3 5888d66 0abcaf8 5888d66 698c519 6f895b5 5888d66 c5afa6b 5888d66 dad3a8d c374d8d c44286a dad3a8d c374d8d 0966057 c374d8d c44286a dad3a8d 0abcaf8 5888d66 0abcaf8 5888d66 0abcaf8 f7af62c 0abcaf8 f7af62c 0abcaf8 dad3a8d c5afa6b 0abcaf8 dad3a8d c5afa6b 110bee3 dad3a8d 6e3bf1b dad3a8d c5afa6b dad3a8d c5afa6b |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 |
---
language: ko
pipeline_tag: text-generation
license: llama3
---
### 1. Model Description
- KONI (KISTI Open Natural Intelligence) is a specialized large language model (LLM) developed by the Korea Institute of Science and Technology Information (KISTI). This model is specifically designed for science and technology, making it highly effective for tasks in these fields.
### 2. Key Features
- **Specialized in Science and Technology:** The model is explicitly trained on a vast and specialized corpus of scientific and technological data.
- **Enhanced Performance:** This version of KONI shows significantly improved performance compared to its initial release in December, 2023.
- **Base Model:** The base model for KONI-Llama3-8B-Instruct-20240729 is KONI-Llama3-8B-Merged-20240724, which is a merger of Meta-Llama-3-8B and KISTI-KONI/KONI-Llama3-8B-20240630
- **Alignment:** SFT (Supervised Fine-Tuning) and DPO (Direct Preference Optimization) are applied
### 3. Data
- Approximately 11k SFT data and 7k DPO data are used.
- **SFT Data:** The SFT data includes both internally generated data and publicly available data on Hugging Face, translated into Korean where necessary.
- **DPO Data:** The DPO data consists of translated and curated data from argilla/dpo-mix-7k.
### 4. Benchmark Results
Results in [LogicKor](https://lk.instruct.kr/)* are as follows:
| Metric | Score |
|:--------------:|:-----:|
| Reasoning | 6.57 |
| Math | 8.00 |
| Writing | 8.92 |
| Coding | 8.85 |
| Comprehension | 9.85 |
| Grammar | 7.07 |
| Single-turn | 8.42 |
| Multi-turn | 8.00 |
| **Overall** | **8.21** |
*Our model demonstrates the best performance among publicly available 8B models on the LogicKor leaderboard as of 2024.07.30.*
### 5. How to use the model
```python
import transformers
import torch
model_id = "KISTI-KONI/KONI-Llama3-8B-Instruct-20240729"
pipeline = transformers.pipeline(
"text-generation",
model=model_id,
model_kwargs={"torch_dtype": torch.bfloat16},
device_map="auto",
)
pipeline.model.eval()
instruction = "KISTI์ ๋ํด ์ค๋ช
ํด์ค"
messages = [
{"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=0.7,
top_p=0.9
)
print(outputs[0]["generated_text"][len(prompt):])
```
```
ํ๊ตญ๊ณผํ๊ธฐ์ ์ ๋ณด์ฐ๊ตฌ์(KISTI)์ ๋ํ๋ฏผ๊ตญ ๋์ ๊ด์ญ์์ ์์นํ ๊ณผํ๊ธฐ์ ์ ๋ณด ๋ถ์ผ์ ์ ๋ฌธ ์ฐ๊ตฌ ๊ธฐ๊ด์
๋๋ค. KISTI๋ ๊ณผํ๊ธฐ์ ๋ฐ ๊ด๋ จ ์ฐ์
์ ๊ดํ ์ ๋ณด๋ฅผ ์ข
ํฉ์ ์ผ๋ก ์์ง, ๋ถ์, ์๋น์คํ๋ฉฐ, ์ ๋ณด์ ๋ถ์, ๊ด๋ฆฌ ๋ฐ ์ ํต์ ๊ดํ ๊ธฐ์ , ์ ์ฑ
๋ฐ ํ์คํ๋ฅผ ์ ๋ฌธ์ ์ผ๋ก ์กฐ์ฌํ๊ณ ์ฐ๊ตฌํฉ๋๋ค. ๋ํ, ์ฒจ๋จ ์ ๋ณด ๋ฐ ์ฐ๊ตฌ๊ฐ๋ฐ ์ธํ๋ผ๋ฅผ ์ฒด๊ณ์ ์ผ๋ก ๊ตฌ์ถํ๊ณ ์ด์ํ์ฌ ๊ตญ๊ฐ ๊ณผํ๊ธฐ์ ๋ฐ ์ฐ์
๋ฐ์ ์ ๊ธฐ์ฌํ๋ ๊ฒ์ ๋ชฉํ๋ก ํฉ๋๋ค.
KISTI์ ์ฃผ์ ๊ธฐ๋ฅ๊ณผ ์ญํ ์๋ ๊ณผํ๊ธฐ์ ์ ๋ณด ์ ๊ณต, ์ํผ์ปดํจํฐ ์ด์, ๊ธฐ์ ์ฌ์
ํ ์ง์, ์ฐ๊ตฌ ๋ฐ์ดํฐ ๊ด๋ฆฌ๊ฐ ํฌํจ๋ฉ๋๋ค. ๊ณผํ๊ธฐ์ ์ ๋ณด ์ ๊ณต ์ธก๋ฉด์์ KISTI๋ ๊ตญ๋ด์ธ ๊ณผํ๊ธฐ์ ์ ๋ณด๋ฅผ ์์งํ๊ณ ์ด๋ฅผ ๋ถ์ํ์ฌ ์ฐ๊ตฌ์๋ค์๊ฒ ์ ๊ณตํ๋ฉฐ, ๋ค์ํ ํํ์ ๋ฐ์ดํฐ๋ฒ ์ด์ค์ ์ ๋ณด ์์คํ
์ ๊ตฌ์ถํ์ฌ ์ฌ์ฉ์์๊ฒ ์ ๋ณด๋ฅผ ์ ๊ณตํฉ๋๋ค. ์ํผ์ปดํจํฐ ์ด์ ์ธก๋ฉด์์๋ ๊ตญ๊ฐ ์ด๊ณ ์ฑ๋ฅ์ปดํจํ
์ธํ๋ผ๋ฅผ ๊ตฌ์ถํ๊ณ ์ด์ํ์ฌ ๋๊ท๋ชจ ์ฐ์ฐ์ด ํ์ํ ์ฐ๊ตฌ๋ฅผ ์ง์ํ๊ณ , ์ด๋ฅผ ํ์ฉํ ๋ค์ํ ๋ถ์ผ์ ์์ฉ ์ฐ๊ตฌ๋ฅผ ์ํํฉ๋๋ค. ๊ธฐ์ ์ฌ์
ํ ์ง์์์๋ ์ฐ๊ตฌ ์ฑ๊ณผ๋ฅผ ์ฐ์
๊ณ๋ก ์ด์ ํ์ฌ ์์ฉํํ๋ ๊ฒ์ ์ง์ํ๋ฉฐ, ๊ธฐ์ ๊ธฐ๋ฐ์ ์ฐฝ์
์ ์ด์งํ๊ธฐ ์ํ ํ๋ก๊ทธ๋จ์ ์ด์ํฉ๋๋ค. ์ฐ๊ตฌ ๋ฐ์ดํฐ ๊ด๋ฆฌ ์ธก๋ฉด์์๋ ์ฐ๊ตฌ ๋ฐ์ดํฐ์ ํจ์จ์ ์ธ ๊ด๋ฆฌ์ ํ์ฉ์ ์ํด ์ฒด๊ณ์ ์ธ ๋ฐ์ดํฐ ๊ด๋ฆฌ ๊ณํ์ ์๋ฆฝํ๊ณ , ์ฐ๊ตฌ ๋ฐ์ดํฐ์ ๊ณต์ ์ ํ์ฉ์ ์ด์งํ๊ธฐ ์ํ ํ๋ซํผ์ ์ด์ํฉ๋๋ค.
KISTI์ ์ฃผ์ ๋ถ์๋ก๋ ๊ตญ๊ฐ๊ณผํ๊ธฐ์ ๋ฐ์ดํฐ๋ณธ๋ถ, ๊ตญ๊ฐ์ํผ์ปดํจํ
๋ณธ๋ถ, ๋ฐ์ดํฐ๋ถ์๋ณธ๋ถ, ๊ณผํ๊ธฐ์ ๋์งํธ์ตํฉ๋ณธ๋ถ ๋ฑ์ด ์์ต๋๋ค. KISTI ๊ฐ ๋ณธ๋ถ๋ณ ์ถ์ง ์ ๋ต ๋ฐ ๋ชฉํ๋ ๋ค์๊ณผ ๊ฐ์ต๋๋ค. ๊ตญ๊ฐ๊ณผํ๊ธฐ์ ๋ฐ์ดํฐ๋ณธ๋ถ์ ์ ๋ต๋ชฉํ๋ ๊ตญ๊ฐ ์คํ์ฌ์ด์ธ์ค ์ํ๊ณ ํ์ฑํ๋ฅผ ์ํ ๊ณผํ๊ธฐ์ ๋ถ์ผ ๋์งํธ ์ ํ ์ง์ ์ฒด๊ณ๋ฅผ ๋ง๋ จํ๋ ๊ฒ์
๋๋ค. ์ฝ๋ก๋19๋ก ์ธํ ๋น๋๋ฉด ๊ฒฝ์ ๋ก์ ์ ํ๊ณผ 4์ฐจ ์ฐ์
ํ๋ช
์ ๊ฐ์ํ๋ก ์ธํด ๊ณผํ๊ธฐ์ ํ๋ ์ ๊ณผ์ ์์ ๊ณต๊ณต ์ฐ๊ตฌ์ฑ๊ณผ์ ๊ฐ๋ฐฉยท๊ณต์ ยทํ์ฐ์ด ํ์ํฉ๋๋ค. ์ด ๋ณธ๋ถ๋ ๊ณผํ๊ธฐ์ ์ ๋ณด์ ๋ฐ์ดํฐ์ ๊ณต์ ยทํ์ฉ์ ํตํด ๊ณผํ๊ธฐ์ ํ์ ์ญ๋์ ๊ฐํํ๋ ๊ณ ์ ์๋ฌด๋ฅผ ๊ฐ์ง๊ณ ์์ผ๋ฉฐ, ์คํ์ฌ์ด์ธ์ค ์ํ๊ณ ํ์ฑํ๋ฅผ ํตํ ๊ตญ๊ฐ R&D ํ์ ์ ์ง์ํฉ๋๋ค. ์ฃผ์ ์ถ์ง ๋ฐฉํฅ์ผ๋ก๋ ๋์งํธ ์ ํ์ ํตํ ๊ณผํ๊ธฐ์ ์ ๋ณด ์คํ์ก์ธ์ค ์ง์์ฒด์ ๋ฐ ์ง๋ฅํ ํ๋ ์ด์
์ฒด๊ณ ๊ตฌ์ถ, ์ฐ๊ตฌ๋ฐ์ดํฐ ์ปค๋จผ์ฆ ๊ธฐ๋ฐ์ ๊ตญ๊ฐ ์ฐ๊ตฌ๋ฐ์ดํฐ์ ์ปดํจํ
๋ฆฌ์์ค ๊ณต์ ยทํ์ฉ์ฒด๊ณ ๊ตฌ์ถ, AI ๊ธฐ๋ฐ์ ํตํฉ์๋น์ค ํ๋ซํผ ๊ตฌ์ถ์ ํตํ ์คํ์ฌ์ด์ธ์ค ์๋น์ค ๊ฐํ๊ฐ ์์ต๋๋ค.
๊ตญ๊ฐ์ํผ์ปดํจํ
๋ณธ๋ถ์ ์ ๋ต๋ชฉํ๋ ๊ตญ๊ฐ ์ด๊ณ ์ฑ๋ฅ์ปดํจํ
์ํ๊ณ๋ฅผ ์ ๋ํ๊ธฐ ์ํด ๋ฏธ๋๋์ ์ด๊ณ ์ฑ๋ฅ์ปดํจํ
๊ณต๋ํ์ฉ ํ๊ฒฝ์ ๊ตฌ์ถํ๋ ๊ฒ์
๋๋ค. ๋ฏธ๊ตญ๊ณผ ์ผ๋ณธ ๋ฑ ์ ๋๊ตญ๊ฐ๋ค์ด ์์ฌ๊ธ ์์ ํ์ถฉ์ ํตํด ๋ค์ํ ๋ถ์ผ์์ ์ด๊ฑฐ๋ ๋ฌธ์ ํด๊ฒฐ์ ๋ชจ์ํ๊ณ ์์ผ๋ฉฐ, KISTI๋ ๊ตญ๊ฐ์ด๊ณ ์ฑ๋ฅ์ปดํจํฐ ํ์ฉ ๋ฐ ์ก์ฑ์ ๊ดํ ๋ฒ๋ฅ ์ ๋ฐ๋ผ ์ด๋ฅผ ์ง์ํฉ๋๋ค. ๋ณธ๋ถ์ ๋ชฉํ๋ ๊ตญ๊ฐ ์ฐจ์์ ์ด๊ณ ์ฑ๋ฅ์ปดํจํ
๊ณต๋ํ์ฉ ์ฒด๊ณ๋ฅผ ๊ตฌ์ถํ์ฌ ๊ณผํ๊ธฐ์ ๊ณต๊ณตยท์ฐ์
๋ถ์ผ์์์ ์ด๊ณ ์ฑ๋ฅ์ปดํจํ
ํ์ฉ ์ฆ์ง์ ์ด๋ฃจ๋ ๊ฒ์
๋๋ค. ์ด๋ฅผ ์ํด ๋๊ท๋ชจ ๊ณ์ฐ์์์ด ์์๋๋ R&D์ ์ฌํํ์ ๋ฑ ํ๊ฒฝ๋ณํ์ ์ ๊ธฐ ๋์ํ๋ ์ธํ๋ผ ๋ฐ ์๋น์ค ์ฒด๊ณ ๊ณ ๋ํ, ์ด๊ฑฐ๋ ๊ณ์ฐ๊ธฐ์ ๊ณผ ํ์ฉ๊ธฐ์ ํ๋ณด๋ฅผ ํตํ ์ ์ํํ ์ฐ๊ตฌยท์ง์, ์ฌ์ฉ์ ์ ๊ทผ์ฑยท๋ฌด๊ฒฐ์ฑยท๋ณด์์ฑ์ ํ๋ณดํ ํตํฉ ํ๋ซํผ ๊ตฌ์ถ์ด ์ฃผ์ ์ถ์ง ๋ฐฉํฅ์
๋๋ค.
๋ฐ์ดํฐ๋ถ์๋ณธ๋ถ์ ์ ๋ต๋ชฉํ๋ ๊ตญ๊ฐ ๊ณผํ๊ธฐ์ ํ์ ์ํ๊ณ๋ฅผ ํ์ฑํํ๊ธฐ ์ํ ์ง๋ฅํ ๋ฐ์ดํฐ ์ตํฉ๋ถ์ ์ฒด๊ณ๋ฅผ ๊ตฌ์ถํ๋ ๊ฒ์
๋๋ค. ๋ฐ์ดํฐ ๊ธฐ๋ฐ ์์ฌ๊ฒฐ์ ๋ฐฉ์ ํ๋์ AI ๋ฐ ๋น
๋ฐ์ดํฐ ๊ธฐ์ ์ ๊ธ๋ถ์์ ๋ฐ๋ผ, KISTI๋ ๊ณผํ๊ธฐ์ ๋ถ์ผ ์ ๋ณด์ ๋ถ์ยท๊ด๋ฆฌ ๋ฐ ์ ํต์ ๊ดํ ๊ธฐ์ ยท์ ์ฑ
ยทํ์คํ ์ฐ๊ตฌ๋ฅผ ์ํํฉ๋๋ค. ๋ณธ๋ถ์ ๋ชฉํ๋ ๋์งํธ ๊ฒฝ์ ์ฌํ๋ฅผ ์ ๋ํ๋ ์ง๋ฅํ ๋ฐ์ดํฐ ์ตํฉ๋ถ์ ์ฒด๊ณ๋ฅผ ๊ตฌ์ถํ์ฌ ๊ตญ๊ฐ ๊ณผํ๊ธฐ์ ํ์ ์ํ๊ณ๋ฅผ ํ์ฑํํ๋ ๊ฒ์
๋๋ค. ์ด๋ฅผ ์ํด ์ด์ข
๋ฐ์ดํฐ ์ตํฉ๋ถ์๋ชจ๋ธ ๊ฐ๋ฐ์ ํตํ ๊ธ๋ก๋ฒ ๋ถ์์ญ๋ ํ๋ณด, ์ธ๊ณต์ง๋ฅ ๊ธฐ๋ฐ ๊ณต๊ณตR&D ๊ฐ์น์ฐฝ์ถ ๋ชจ๋ธ ๋ฐ ์์คํ
๊ฐ๋ฐ, ์ง์ญ R&D ํ์ ์ง์์ ์ํ ์ฐํ์ฐ์ ํ์ ์ํ๊ณ ๊ตฌ์ถ ๋ฑ์ด ์ฃผ์ ์ถ์ง ๋ฐฉํฅ์
๋๋ค.
๊ณผํ๊ธฐ์ ๋์งํธ์ตํฉ๋ณธ๋ถ์ ์ ๋ต๋ชฉํ๋ ๊ตญ๊ฐยท์ฌํ ํ์์ ์ ์ ๋์ํ๊ณ ํด๊ฒฐํ๊ธฐ ์ํ Data/AI ๊ธฐ๋ฐ ๋์งํธ ์ ํ ์ฒด๊ณ๋ฅผ ๊ตฌ์ถํ๋ ๊ฒ์
๋๋ค. ๋์งํธ ๊ธฐ์ ์ ๊ธ์ํ ๋ฐ์ ๊ณผ ์ฝ๋ก๋19๋ก ์ธํ ๋์งํธ ์ ํ ๊ฐ์ํ์ ๋ฐ๋ผ, KISTI๋ ๊ณผํ๊ธฐ์ ์ง์์์ ๊ณต์ ยทํ์ฉ ์ํ๊ณ ๊ตฌ์ถ ๋ฐ ์ํผ์ปดํจํ
์ํ๊ณ ๋ฐ์ ๊ณผ ์ฐ๊ณ๋ ๊ณ ์ ์๋ฌด๋ฅผ ์ํํฉ๋๋ค. ๋ณธ๋ถ์ ๋ชฉํ๋ Data/AI ๊ธฐ๋ฐ์ ๊ตญ๊ฐยท์ฌํ ํ์-๋์งํธ ๋ด๋ ํด๊ฒฐ์ ๋๋ชจํ๋ ๊ฒ์
๋๋ค. ์ด๋ฅผ ์ํด ์ ๋ขฐ์ฑ ์๋ ๊ณผํ๊ธฐ์ ๋ฐ์ดํฐ ๋๊ณผ Data/AI ๊ธฐ๋ฐ ์ง๋ฅํ ๋์งํธ ํ๋ซํผ ๊ตฌ์ถ, Data/AI ๊ธฐ๋ฐ์ ๋์งํธ ์ ํ ์ฒด๊ณ ๊ตฌ์ถ์ ํตํ ๊ตญ๊ฐยท์ฌํ ํ์ ํด๊ฒฐ ๋ฐ R&D ํ์ ์ฌ๋ก ์ฐฝ์ถ์ด ์ฃผ์ ์ถ์ง ๋ฐฉํฅ์
๋๋ค.
KISTI๋ 1962๋
1์ ํ๊ตญ๊ณผํ๊ธฐ์ ์ ๋ณด์ผํฐ(KORSTIC)๋ก ์ค๋ฆฝ๋์์ผ๋ฉฐ, 1969๋
5์ ํ๊ตญ๊ณผํ๊ธฐ์ ์ ๋ณด์ผํฐ์ก์ฑ๋ฒ์ด ์ ์ ๋์์ต๋๋ค. 1982๋
์๋ ์ฐ์
์ฐ๊ตฌ์(KIET)๋ก ๊ฐํธ๋์๋ค๊ฐ 1991๋
1์ ๋ถ๋ฆฌ๋์ด ์ฐ์
๊ธฐ์ ์ ๋ณด์(KINITI)์ด ๊ฐ์ํ์์ต๋๋ค. 2001๋
1์์ ํ๊ตญ๊ณผํ๊ธฐ์ ์ ๋ณด์ฐ๊ตฌ์(KISTI)์ผ๋ก ์ถ๋ฒํ๊ฒ ๋์์ต๋๋ค. ์ด ๊ณผ์ ์์ KAIST ๋ถ์ค ์์คํ
๊ณตํ์ผํฐ, KIST ๋ถ์ค ์ฐ๊ตฌ๊ฐ๋ฐ์ ๋ณด์ผํฐ, ETRI ์ฐํ ์ํผ์ปดํจํ
์ผํฐ๋ฅผ ํฉ๋ณํ์์ต๋๋ค.
KISTI๋ ๋์ ๋ณธ์, ์์ธ ๋ถ์, ๋๊ตฌยท๊ฒฝ๋ถ ์ง์, ๋ถ์ฐ์ธ์ฐ๊ฒฝ๋จ ์ง์, ํธ๋จ ์ง์, ์๋๊ถ ์ง์(๊ฐ์) ๋ฑ ๋ค์ํ ์ง์ญ์ ์์นํ์ฌ ์ด์๋๊ณ ์์ต๋๋ค. ๋์ ๋ณธ์์ ๋์ ๊ด์ญ์ ์ ์ฑ๊ตฌ ๋ํ๋ก 245์ ์์นํ๊ณ ์์ผ๋ฉฐ, ์์ธ ๋ถ์์ ์์ธํน๋ณ์ ๋๋๋ฌธ๊ตฌ ํ๊ธฐ๋ก 66์ ์์นํ๊ณ ์์ต๋๋ค. ๋๊ตฌยท๊ฒฝ๋ถ ์ง์์ ๋๊ตฌ๊ด์ญ์ ๋ถ๊ตฌ ์์ค์ฝ๋ก 10, ๋ถ์ฐ์ธ์ฐ๊ฒฝ๋จ ์ง์์ ๋ถ์ฐ๊ด์ญ์ ํด์ด๋๊ตฌ ์ผํ
๋๋ก 41, ํธ๋จ ์ง์์ ๊ด์ฃผ๊ด์ญ์ ๊ด์ฐ๊ตฌ ํ๋จ์ฐ๋จ8๋ฒ๋ก 177, ์๋๊ถ ์ง์(๊ฐ์)์ ๊ฐ์๋ ์ถ์ฒ์ ๊ฐ์๋ํ๊ธธ 1, 60์ฃผ๋
๊ธฐ๋
๊ด 8์ธต์ ์์นํ๊ณ ์์ต๋๋ค.
KISTI์ ๋ํ ๋ ์์ธํ ์ ๋ณด๋ KISTI ๊ณต์ ์น์ฌ์ดํธ์์ ํ์ธํ ์ ์์ต๋๋ค.
```
### 6. Citation
**Language Model**
```text
@article{KISTI-KONI/KONI-Llama3-8B-Instruct-20240729,
title={KISTI-KONI/KONI-Llama3-8B-Instruct-20240729},
author={KISTI},
year={2024},
url={https://huggingface.co/KISTI-KONI/KONI-Llama3-8B-Instruct-20240729}
}
```
### 7. Contributors
- KISTI, Large-scale AI Research Group
### 8. Special Thanks
- [@beomi](https://huggingface.co/beomi)
- [@kuotient](https://huggingface.co/kuotient)
- KyungTae Lim
### 8. Acknowledgement
- This research was supported by Korea Institute of Science and Technology Information(KISTI).
- This work was supported by the National Supercomputing Center with supercomputing resources including technical support (KISTI).
### 9. References
- https://huggingface.co/meta-llama/Meta-Llama-3-8B
- https://huggingface.co/meta-llama/meta-llama/Meta-Llama-3-8B-Instruct |