Update README.md
Browse files
README.md
CHANGED
@@ -1,47 +1,71 @@
|
|
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 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
library_name: transformers
|
3 |
+
license: apache-2.0
|
4 |
+
base_model:
|
5 |
+
- Saxo/Linkbricks-Horizon-AI-Korean-Advanced-27B
|
6 |
+
datasets:
|
7 |
+
- Saxo/ko_cn_translation_tech_social_science_linkbricks_single_dataset
|
8 |
+
- Saxo/ko_jp_translation_tech_social_science_linkbricks_single_dataset
|
9 |
+
- >-
|
10 |
+
Saxo/en_ko_translation_tech_science_linkbricks_single_dataset_with_prompt_text_huggingface
|
11 |
+
- >-
|
12 |
+
Saxo/en_ko_translation_social_science_linkbricks_single_dataset_with_prompt_text_huggingface
|
13 |
+
- >-
|
14 |
+
Saxo/ko_aspect_sentiment_sns_mall_sentiment_linkbricks_single_dataset_with_prompt_text_huggingface
|
15 |
+
- Saxo/ko_summarization_linkbricks_single_dataset_with_prompt_text_huggingface
|
16 |
+
- >-
|
17 |
+
Saxo/OpenOrca_cleaned_kor_linkbricks_single_dataset_with_prompt_text_huggingface
|
18 |
+
- >-
|
19 |
+
Saxo/ko_government_qa_total_linkbricks_single_dataset_with_prompt_text_huggingface_sampled
|
20 |
+
- Saxo/ko-news-corpus-1
|
21 |
+
- Saxo/ko-news-corpus-2
|
22 |
+
- Saxo/ko-news-corpus-3
|
23 |
+
- Saxo/ko-news-corpus-4
|
24 |
+
- Saxo/ko-news-corpus-5
|
25 |
+
- Saxo/ko-news-corpus-6
|
26 |
+
- Saxo/ko-news-corpus-7
|
27 |
+
- Saxo/ko-news-corpus-8
|
28 |
+
- Saxo/ko-news-corpus-9
|
29 |
+
- maywell/ko_Ultrafeedback_binarized
|
30 |
+
- youjunhyeok/ko-orca-pair-and-ultrafeedback-dpo
|
31 |
+
- lilacai/glaive-function-calling-v2-sharegpt
|
32 |
+
- kuotient/gsm8k-ko
|
33 |
+
language:
|
34 |
+
- ko
|
35 |
+
- en
|
36 |
+
- jp
|
37 |
+
- cn
|
38 |
+
pipeline_tag: text-generation
|
39 |
+
---
|
40 |
+
|
41 |
+
# Model Card for Model ID
|
42 |
+
|
43 |
+
<div align="center">
|
44 |
+
<img src="https://www.linkbricks.com/wp-content/uploads/2022/03/%E1%84%85%E1%85%B5%E1%86%BC%E1%84%8F%E1%85%B3%E1%84%87%E1%85%B3%E1%84%85%E1%85%B5%E1%86%A8%E1%84%89%E1%85%B3%E1%84%85%E1%85%A9%E1%84%80%E1%85%A9-2-1024x804.png" />
|
45 |
+
</div>
|
46 |
+
|
47 |
+
|
48 |
+
AI ์ ๋น
๋ฐ์ดํฐ ๋ถ์ ์ ๋ฌธ ๊ธฐ์
์ธ Linkbricks์ ๋ฐ์ดํฐ์ฌ์ด์ธํฐ์คํธ์ธ ์ง์ค์ฑ(Saxo) ์ด์ฌ๊ฐ <br>
|
49 |
+
gemma-2-27b-it ๋ฒ ์ด์ค๋ชจ๋ธ์ ์ฌ์ฉํด์ H100-80G 8๊ฐ๋ฅผ ํตํด ์ฝ 20%์ ๋์ ํ๋ผ๋ฏธํฐ๋ฅผ ํ๊ตญ์ด CPT(Continued-Pretraining)->SFT->DPO ํ <br>
|
50 |
+
ํ๊ธ ์ธ์ด ๋ชจ๋ธ์ธ Saxo/Linkbricks-Horizon-AI-Korean-Advanced-27B์ ์ถ๊ฐ์ ์ธ ํ๊ธ๋ฐ์ดํฐ๋ค์ ํ์ฉํด์ ๋ค์ํ ํ
์คํฌ๋ณ ํ๊ตญ์ด-์ค๊ตญ์ด-์์ด-์ผ๋ณธ์ด ๊ต์ฐจ ํ์ต ๋ฐ์ดํฐ์ ์ํ ๋ฐ <br>
|
51 |
+
๋
ผ๋ฆฌํ๋จ ๋ฐ์ดํฐ๋ฅผ ํตํ์ฌ ํ์ค์ผ์ ์ธ์ด ๊ต์ฐจ ์ฆ๊ฐ ์ฒ๋ฆฌ์ ๋ณต์กํ ๋
ผ๋ฆฌ ๋ฌธ์ ์ญ์ ๋์ ๊ฐ๋ฅํ๋๋ก ํ๋ จํ ๋ชจ๋ธ์ด๋ค.<br>
|
52 |
+
-์์ด์ฒ๋ฆฌ ๊ฐํ ๋ฒ์ <br>
|
53 |
+
-ํ ํฌ๋์ด์ ๋ ๋จ์ด ํ์ฅ ์์ด ๋ฒ ์ด์ค ๋ชจ๋ธ ๊ทธ๋๋ก ์ฌ์ฉ<br>
|
54 |
+
-๊ณ ๊ฐ ๋ฆฌ๋ทฐ๋ ๋ณต์กํ ํ๊ธ ์ถ๋ก ๋ฐ ์์
ํฌ์คํ
๊ณ ์ฐจ์ ๋ถ์ ๋ฐ ์ฝ๋ฉ๊ณผ ์๋ฌธ, ์ํ, ๋
ผ๋ฆฌํ๋จ ๋ฑ์ด ๊ฐํ๋ ๋ชจ๋ธ<br>
|
55 |
+
-128k-Context Window<br>
|
56 |
+
<br><br>
|
57 |
+
|
58 |
+
Dr. Yunsung Ji (Saxo), a data scientist at Linkbricks, a company specializing in AI and big data analytics, has developed a Korean language model<br>
|
59 |
+
using the Saxo/Linkbricks-Horizon-AI-Korean-Advanced-27B, a Korean language model that uses the gemma-2-27b-it basemodel to train about 20% of the parameters through 8 H100-80Gs <br>
|
60 |
+
using Korean CPT (Continued-Pretraining)->SFT->DPO.
|
61 |
+
It is a model trained to handle cross-lingual augmentation and complex logic problems by utilizing additional Korean data, cross-training data of Korean, Chinese, English, and Japanese by various tasks, and math and logic judgment data.
|
62 |
+
|
63 |
+
Translated with DeepL.com (free version) <br>
|
64 |
+
-Reinforced English Language processing<br>
|
65 |
+
-Tokenizer uses the base model without word expansion<br>
|
66 |
+
-Models enhanced with high-dimensional analysis of customer reviews and social posts, as well as coding, writing, math, decision making and complex inference<br>
|
67 |
+
-128k-Context Window<br>
|
68 |
+
-Deepspeed Stage=3, use rslora and BAdam Layer Mode<br>
|
69 |
+
<br><br>
|
70 |
+
|
71 |
+
<a href="www.linkbricks.com">www.linkbricks.com</a>, <a href="www.linkbricks.vc">www.linkbricks.vc</a>
|