File size: 8,360 Bytes
9ba8455 0b1aeb4 a7bd71a f48cbb6 0f05ab2 f48cbb6 a7bd71a 4d3013c a7bd71a f48cbb6 0f05ab2 f48cbb6 a7bd71a ae8ec45 a7bd71a 0f05ab2 a7bd71a f48cbb6 a7bd71a f48cbb6 9ec8ba5 f48cbb6 c025f22 a7bd71a 1e75918 0f05ab2 1e75918 a7bd71a f48cbb6 6ae5e71 d3ef354 6ae5e71 a7bd71a 4dbd174 a7bd71a c025f22 a7bd71a 4dbd174 c025f22 4dbd174 f48cbb6 c025f22 a7bd71a f3cd592 57d542a a7bd71a 57d542a a7bd71a |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 |
---
license: llama3
language:
- en
- ko
pipeline_tag: text-generation
tags:
- saltlux
- luxia
- meta
- llama-3
- pytorch
---
# Model Details
Saltlux, AI Labs ์ธ์ด๋ชจ๋ธํ์์ ํ์ต ๋ฐ ๊ณต๊ฐํ <b>Ko-Llama3-Luxia-8B</b> ๋ชจ๋ธ์ Meta์์ ์ถ์ํ Llama-3-8B ๋ชจ๋ธ์ <b>ํ๊ตญ์ด์ ํนํ</b>ํ ๋ชจ๋ธ์
๋๋ค.<br><br>
์์ฒด ๋ณด์ ํ๊ณ ์๋ 1TB ์ด์์ ํ๊ตญ์ด ํ์ต ๋ฐ์ดํฐ ์ค, ์ฝ 100GB ์ ๋์ ๋ฐ์ดํฐ๋ฅผ ์ ๋ณํ์ฌ ์ฌ์ ํ์ต์ ํ์ฉํ์์ต๋๋ค.<br><br>
๋ํ ๊ณต๊ฐ๋ Llama-3 Tokenizer๋ฅผ ํ๊ตญ์ด๋ก ํ์ฅํ๊ณ ์ฌ์ ํ์ต์ ํ์ฉํ์ต๋๋ค.
- **Meta Llama-3:** Meta developed and released the Meta Llama 3 family of large language models (LLMs), a collection of pretrained and instruction tuned generative text models in 8 and 70B sizes. The Llama 3 instruction tuned models are optimized for dialogue use cases and outperform many of the available open source chat models on common industry benchmarks. Further, in developing these models, we took great care to optimize helpfulness and safety.
- **License:** Llama3 License [https://llama.meta.com/llama3/license](https://llama.meta.com/llama3/license)
### Intended Use
Ko-Llama3-Luxia-8B๋ ์ฐ๊ตฌ์ฉ์ผ๋ก ์ ์๋์์ผ๋ฉฐ, ๋ค์ํ ์์ฐ์ด ์์ฑ ํ์คํฌ๋ฅผ ์ํด ์์ ๋กญ๊ฒ ํ์ต ๋ฐ ํ์ฉํ ์ ์์ต๋๋ค.
### How to Use
ํด๋น ๋ชจ๋ธ ์นด๋์๋ `Ko-Llama3-Luxia-8B` ๋ชจ๋ธ๊ณผ transformers ๋ผ์ด๋ธ๋ฌ๋ฆฌ ๊ธฐ๋ฐ์ ์์ ์ฝ๋๋ฅผ ์ ๊ณตํฉ๋๋ค.
```
import transformers
import torch
model_id = "saltlux/Ko-Llama3-Luxia-8B"
pipeline = transformers.pipeline(
"text-generation", model=model_id, model_kwargs={"torch_dtype": torch.bfloat16}, device_map="auto"
)
pipeline("<|begin_of_text|>์๋
ํ์ธ์. ์ํธ๋ฃฉ์ค AI Labs ์
๋๋ค.")
```
# Training Details
ํ๊ตญ์ด ํนํ๋ฅผ ์ํ ์ฌ์ ํ์ต ๋ฐ์ดํฐ๋ Saltlux์์ ๋ณด์ ํ ๋ด์ค, ๋ฒ๋ฅ , ํนํ, ์๋ฃ, ์ญ์ฌ, ์ฌํ, ๋ฌธํ, ๋ํ(๋ฌธ์ด/๊ตฌ์ด) ๋ฑ์ ๋๋ฉ์ธ์ผ๋ก ๊ตฌ์ฑ๋ 100GB ์์ค์ ์ฝํผ์ค(~2023๋
)๋ฅผ ํ์ฉํ์์ต๋๋ค.<br>
### Use Device
์ฌ์ ํ์ต์ NVIDIA H100 80GB * 8EA ์ฅ๋น๋ฅผ ํ์ฉํ์ฌ ์งํํ์์ต๋๋ค.
#### Training Hyperparameters
<table>
<tr>
<td><strong>Model</strong>
</td>
<td><strong>Params</strong>
</td>
<td><strong>Context length</strong>
</td>
<td><strong>GQA</strong>
</td>
<td><strong>Learning rate</strong>
</td>
<td><strong>Batch</strong>
</td>
<td><strong>Precision</strong>
</td>
</tr>
<tr>
<td>Ko-Llama3-Luxia-8B
</td>
<td>8B
</td>
<td>8k
</td>
<td>yes
</td>
<td>1e-5
</td>
<td>128
</td>
<td>bf16
</td>
</tr>
</table>
### Tokenizer
Llama-3-Tokenizer๋ฅผ ํ๊ตญ์ด ํนํํ๊ธฐ ์ํด ํ๊ตญ์ด ํ ํฐ 17,536๊ฐ๋ฅผ ์ถ๊ฐํ๊ณ ํ์ฉํ์์ต๋๋ค.
<table>
<tr>
<td><strong>Model</strong>
</td>
<td><strong>Vocab Size</strong>
</td>
</tr>
<tr>
<td>Llama-3
</td>
<td>128,256
</td>
</tr>
<tr>
<td>Ko-Llama3-Luxia-8B
</td>
<td>145,792
</td>
</tr>
</table>
#### Tokenizer Result
+ Ko
<table>
<tr>
<td><strong>์
๋ ฅ</strong>
</td>
<td><strong>Llama-3</strong>
</td>
<td><strong>Ko-Llama3-Luxia-8B</strong>
</td>
</tr>
<tr>
<td>์์ฆ ๋ ์จ๊ฐ ๋๋ฌด ์ค๋ฝ๊ฐ๋ฝํด์ ์์ง๋ ๊ฒจ์ธ์ท์ ๋ชป์น์ ์ด์..
</td>
<td>['์', '์ฆ', ' ๋ ', '์จ', '๊ฐ', ' ๋๋ฌด', ' ์ค', '๋ฝ', '๊ฐ', '๋ฝ', 'ํด์', ' ์์ง', '๋', ' ๊ฒจ', '์ธ', '๏ฟฝ', '๏ฟฝ', '์', ' ๋ชป', '์น', '์ ', '์ด์', '..']
</td>
<td>['์์ฆ', ' ๋ ์จ', '๊ฐ', ' ๋๋ฌด', ' ์ค๋ฝ', '๊ฐ๋ฝ', 'ํด์', ' ์์ง', '๋', ' ๊ฒจ์ธ', '์ท', '์', ' ๋ชป', '์น', '์ ', '์ด์', '..']
</td>
</tr>
<tr>
<td>๋ง์๋ ๋ฐฅ์ ๋์
จ์ต๋๊น? ๋ง์ด ๊ถ๊ธํ๋ค์.
</td>
<td>['๋ง', '์๋', ' ๏ฟฝ', '๏ฟฝ', '์', ' ๋', '์
จ', '์ต', '๋๊น', '?', ' ๋ง', '์ด', ' ๊ถ๊ธ', 'ํ', '๋ค์', '.']
</td>
<td>['๋ง', '์๋', ' ๋ฐฅ', '์', ' ๋์
จ', '์ต', '๋๊น', '?', ' ๋ง', '์ด', ' ๊ถ๊ธ', 'ํ', '๋ค์', '.']
</td>
</tr>
<tr>
<td>๋๋ฒ์๋ถํฐ ํ๊ธ์ฌ ํ๋ก๊น์ง ์ํ๋ ํ๋ก๋ฅผ ์ฐพ๋ ๊ฐ์ฅ ๋น ๋ฅธ ๋ฐฉ๋ฒ - ์๋ฉด ๊ฒ์, ์์ฒญ ํ๋ก, ์ ์ฌ ํ๋ก, AI ์ถ์ฒ, ํ๋ก ๋ฐ ๋ฒ๋ น ๊ฒ์.
</td>
<td>['๋', '๋ฒ', '์', '๋ถํฐ', ' ํ', '๊ธ', '์ฌ', ' ํ', '๋ก', '๊น์ง', ' ์', 'ํ๋', ' ํ', '๋ก', '๋ฅผ', ' ์ฐพ', '๋', ' ๊ฐ์ฅ', ' ๋น ', '๋ฅธ', ' ๋ฐฉ๋ฒ', ' -', ' ์', '๋ฉด', ' ๊ฒ์', ',', ' ์์ฒญ', ' ํ', '๋ก', ',', ' ์ ', '์ฌ', ' ํ', '๋ก', ',', ' AI', ' ์ถ์ฒ', ',', ' ํ', '๋ก', ' ๋ฐ', ' ๋ฒ', '๋ น', ' ๊ฒ์', '.']
</td>
<td>['๋', '๋ฒ', '์', '๋ถํฐ', ' ํ', '๊ธ', '์ฌ', ' ํ๋ก', '๊น์ง', ' ์', 'ํ๋', ' ํ๋ก', '๋ฅผ', ' ์ฐพ', '๋', ' ๊ฐ์ฅ', ' ๋น ๋ฅธ', ' ๋ฐฉ๋ฒ', ' -', ' ์๋ฉด', ' ๊ฒ์', ',', ' ์์ฒญ', ' ํ๋ก', ',', ' ์ ์ฌ', ' ํ๋ก', ',', ' AI', ' ์ถ์ฒ', ',', ' ํ๋ก', ' ๋ฐ', ' ๋ฒ๋ น', ' ๊ฒ์', '.']
</td>
</tr>
<tr>
<td>๋ณธ ๋ฐ๋ช
์ ๊ธ์ํ์ ๋ค์ ๋ถ๋ถ์ ์์นญ์์ผ ํน์ ๋ฌด๋ฌ๋ชจ์์ ํ์ฑํ๋ ๊ฑด์ถ์ฉ ๊ธ์์ฌ ์ฅ์ํ์ผ๋ก ์ด๋ฃจ์ด์ง ๊ฒ์ ํน์ง์ด ์๋ค.
</td>
<td>['๋ณธ', ' ๋ฐ', '๋ช
', '์', ' ๊ธ', '์', 'ํ', '์', ' ๋ค', '์', ' ๋ถ๋ถ', '์', ' ์', '์นญ', '์', '์ผ', ' ํน', '์ ', ' ๋ฌด', '๏ฟฝ', '๏ฟฝ', '๋ชจ', '์', '์', ' ํ', '์ฑ', 'ํ๋', ' ๊ฑด', '์ถ', '์ฉ', ' ๊ธ', '์', '์ฌ', ' ์ฅ', '์', 'ํ', '์ผ๋ก', ' ์ด๋ฃจ', '์ด์ง', ' ๊ฒ', '์', ' ํน', '์ง', '์ด', ' ์๋ค', '.']
</td>
<td>['๋ณธ', ' ๋ฐ๋ช
', '์', ' ๊ธ์', 'ํ', '์', ' ๋ค์', ' ๋ถ๋ถ', '์', ' ์์นญ', '์', '์ผ', ' ํน์ ', ' ๋ฌด๋ฌ', '๋ชจ', '์', '์', ' ํ์ฑ', 'ํ๋', ' ๊ฑด์ถ', '์ฉ', ' ๊ธ์', '์ฌ', ' ์ฅ์', 'ํ', '์ผ๋ก', ' ์ด๋ฃจ์ด์ง', ' ๊ฒ', '์', ' ํน์ง', '์ด', ' ์๋ค', '.']
</td>
</tr>
<tr>
<td>๊ณจ๋ค๊ณต์ฆ์ ์ ์๊ธฐ๋๊ฑฐ์์? ๊ทธ๋ฆฌ๊ณ ์น๋ฃํ๋ ค๋ฉด ์ด๋ป๊ฒํด์ผํ์ฃ ?
</td>
<td>['๊ณจ', '๋ค', '๊ณต', '์ฆ', '์', ' ์', ' ์', '๊ธฐ๋', '๊ฑฐ', '์', '์', '?', ' ๊ทธ๋ฆฌ๊ณ ', ' ์น', '๋ฃ', 'ํ๋ ค', '๋ฉด', ' ์ด๋ป๊ฒ', 'ํด์ผ', 'ํ', '์ฃ ', '?']
</td>
<td>['๊ณจ', '๋ค', '๊ณต์ฆ', '์', ' ์', ' ์', '๊ธฐ๋', '๊ฑฐ', '์', '์', '?', ' ๊ทธ๋ฆฌ๊ณ ', ' ์น๋ฃ', 'ํ๋ ค', '๋ฉด', ' ์ด๋ป๊ฒ', 'ํด์ผ', 'ํ', '์ฃ ', '?']
</td>
</tr>
</table>
+ En
<table>
<tr>
<td><strong>์
๋ ฅ</strong>
</td>
<td><strong>Llama-3</strong>
</td>
<td><strong>Ko-Llama3-Luxia-8B</strong>
</td>
</tr>
<tr>
<td>Korean cuisine, hanguk yori, or hansik, has evolved through centuries of social and political change.
</td>
<td>['K', 'orean', ' cuisine', ',', ' h', 'angu', 'k', ' y', 'ori', ',', ' or', ' hans', 'ik', ',', ' has', ' evolved', ' through', ' centuries', ' of', ' social', ' and', ' political', ' change', '.']
</td>
<td>['K', 'orean', ' cuisine', ',', ' h', 'angu', 'k', ' y', 'ori', ',', ' or', ' hans', 'ik', ',', ' has', ' evolved', ' through', ' centuries', ' of', ' social', ' and', ' political', ' change', '.']
</td>
</tr>
<tr>
<td>Son Heung-min is a South Korean professional footballer who plays as a forward for and captains both Premier League club Tottenham Hotspur and the South Korea national team.
</td>
<td>['Son', ' He', 'ung', '-min', ' is', ' a', ' South', ' Korean', ' professional', ' football', 'er', ' who', ' plays', ' as', ' a', ' forward', ' for', ' and', ' captains', ' both', ' Premier', ' League', ' club', ' Tottenham', ' Hot', 'sp', 'ur', ' and', ' the', ' South', ' Korea', ' national', ' team', '.']
</td>
<td>['Son', ' He', 'ung', '-min', ' is', ' a', ' South', ' Korean', ' professional', ' football', 'er', ' who', ' plays', ' as', ' a', ' forward', ' for', ' and', ' captains', ' both', ' Premier', ' League', ' club', ' Tottenham', ' Hot', 'sp', 'ur', ' and', ' the', ' South', ' Korea', ' national', ' team', '.']
</td>
</tr>
</table>
### Citation instructions
**Ko-Llama3-Luxia-8B**
```
@article{kollama3luxiamodelcard,
title={Ko Llama 3 Luxia Model Card},
author={AILabs@Saltux},
year={2024},
url={https://huggingface.co/saltlux/Ko-Llama3-Luxia-8B/blob/main/README.md}
}
```
**Original Llama-3**
```
@article{llama3modelcard,
title={Llama 3 Model Card},
author={AI@Meta},
year={2024},
url={https://github.com/meta-llama/llama3/blob/main/MODEL_CARD.md}
}
``` |