jujbob commited on
Commit
b6548e7
ยท
verified ยท
1 Parent(s): 805d96c

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +20 -13
README.md CHANGED
@@ -11,30 +11,37 @@ license: apache-2.0
11
  </p>
12
 
13
 
14
-
15
-
16
  ## ๐Ÿš€ Update News
17
  - **2026-03-05**: Official release of KORMo-Diffusion.
18
  - **2026-03-02**: Official release of KORMo-VL.
19
  - **2025-10-13**: Official release of KORMo-10B-sft.
20
  ---
21
- ## ๐Ÿ’ก About KORMo-VL
22
- KORMo-Diffusion์— ๋Œ€ํ•ด์„œ...
23
 
24
- <video width="640" height="360" controls>
25
- <source src="https://huggingface.co/KORMo-VL/KORMo-Diffusion/resolve/main/kormo_diffusion_assets/kormo_t2i.mp4" type="video/mp4">
26
- </video>
27
- ---
28
 
29
- ## ๐Ÿ”— Links (About KORMo-10B)
 
30
 
31
- - ๐Ÿ“– **Technical Report**: [๐Ÿ‘‰ Paper](https://huggingface.co/papers/2510.09426) , [๐Ÿ‘‰ ํ•œ๊ตญ์–ด ์š”์•ฝppt](https://github.com/MLP-Lab/KORMo-tutorial/blob/main/20251009_MLP_KORMo(Korean).pdf)
32
- - ๐Ÿค— **Hugging Face**: [๐Ÿ‘‰ Model Download](https://huggingface.co/KORMo-Team)
33
- - ๐Ÿ’ป **GitHub Repository**: [๐Ÿ‘‰ Training and Inference Code](https://github.com/MLP-Lab/KORMo-tutorial)
34
- - ๐Ÿ”‰ **Tutorial**: [๐Ÿ‘‰ Instruction Tuning over google colab](https://colab.research.google.com/github/MLP-Lab/KORMo-tutorial/blob/main/tutorial/02.sft_qlora.ipynb) [๐Ÿ‘‰ Youtube Tutorial](https://www.youtube.com/@MLPLab)
35
 
36
  ---
37
 
 
 
 
 
 
 
 
 
 
 
 
 
 
38
 
39
  ## ๐Ÿ“ˆ T2I Performance
40
 
 
11
  </p>
12
 
13
 
 
 
14
  ## ๐Ÿš€ Update News
15
  - **2026-03-05**: Official release of KORMo-Diffusion.
16
  - **2026-03-02**: Official release of KORMo-VL.
17
  - **2025-10-13**: Official release of KORMo-10B-sft.
18
  ---
19
+ ## ๐Ÿ’ก About KORMo-VL-Diffusion
 
20
 
21
+ **KORMo-VL** is a vision-language model developed **from scratch by the KAIST MLP Lab (https://sites.google.com/view/aailab)**, built on top of **KORMo-10B**.
22
+ The system consists of two components:
 
 
23
 
24
+ * **Vision-Language Model (VLM)**
25
+ * **Image Generation Model**
26
 
27
+ The KORMo-VL-Diffusion model, designed for image generation, was trained from scratch with a high proportion of images reflecting Korean daily environments and culture.
28
+ <span style="color:red">Unfortunately, due to limited GPU resources during the research process, we are sharing the intermediate results of the model at this stage.</span>
 
 
29
 
30
  ---
31
 
32
+ KORMo-VL์€ KAIST MLP ์—ฐ๊ตฌ์‹ค์—์„œ **from scratch๋กœ ๊ฐœ๋ฐœํ•œ ์‹œ๊ฐ-์–ธ์–ด ๋ชจ๋ธ**๋กœ, KORMo-10B๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ (1) ์‹œ๊ฐ-์–ธ์–ด ๋ชจ๋ธ๊ณผ (2) ์ด๋ฏธ์ง€ ์ƒ์„ฑ ๋ชจ๋ธ๋กœ ๊ตฌ์„ฑ๋˜์–ด ์žˆ์Šต๋‹ˆ๋‹ค.
33
+
34
+ ์ด ์ค‘ **์ด๋ฏธ์ง€ ์ƒ์„ฑ์„ ์œ„ํ•œ KORMo-VL-Diffusion** ๋ชจ๋ธ์€ ํ•œ๊ตญ์˜ ์ƒํ™œ ํ™˜๊ฒฝ๊ณผ ๋ฌธํ™”๋ฅผ ๋ฐ˜์˜ํ•˜๊ธฐ ์œ„ํ•ด ๊ตญ๋‚ด ํ™˜๊ฒฝ ์ด๋ฏธ์ง€๋ฅผ ๊ฐ€๋Šฅํ•œ ๋†’์€ ๋น„์œจ๋กœ ์‚ฌ์šฉํ•˜์—ฌ **from scratch๋ถ€ํ„ฐ ํ•™์Šต๋œ ๋ชจ๋ธ**์ž…๋‹ˆ๋‹ค.
35
+ <span style="color:red">๋‹ค๋งŒ ์—ฐ๊ตฌ ์ง„ํ–‰ ์ค‘ GPU ์ž์›์„ ์ถ”๊ฐ€๋กœ ํ™•๋ณดํ•˜์ง€ ๋ชปํ•ด **ํ˜„์žฌ๋Š” ์ค‘๊ฐ„ ๊ฒฐ๊ณผ๋ฌผ์„ ๊ณต์œ ํ•˜๊ฒŒ ๋˜์—ˆ์Šต๋‹ˆ๋‹ค.**</span>
36
+
37
+ * **LLM:** KORMo-VL
38
+ * **Languages:** Korean / English
39
+ * **Training Data:** Synthetic data + public datasets (e.g., AI Hub, details to be released)
40
+
41
+ ํ–ฅํ›„ ํ•ด๋‹น ๋ชจ๋ธ์„ ์ถฉ๋ถ„ํžˆ ํ•™์Šตํ•  ์ˆ˜ ์žˆ๋Š” ํ™˜๊ฒฝ์ด ๋งˆ๋ จ๋œ๋‹ค๋ฉด **์™„์„ฑ๋œ ๋ชจ๋ธ๋กœ ๋ฐœ์ „์‹œํ‚ค๋Š” ๊ฒƒ์„ ๋ชฉํ‘œ๋กœ ํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค.**
42
+ ์ค‘๊ฐ„ ๊ฒฐ๊ณผ๋ฌผ ์œ„์—์„œ ์ถ”๊ฐ€ ํŠœ๋‹์ด๋‚˜ ์—ฐ๊ตฌ๋ฅผ ์ง„ํ–‰ํ•˜๊ณ  ์‹ถ์€ ๋ถ„๋“ค์€ **์ž์œ ๋กญ๊ฒŒ ํ™œ์šฉํ•ด ๋ณด์‹œ๊ธฐ ๋ฐ”๋ž๋‹ˆ๋‹ค.**
43
+
44
+
45
 
46
  ## ๐Ÿ“ˆ T2I Performance
47