README.md

---
license: apache-2.0
datasets:
- tabtoyou/KoLLaVA-CC3M-Pretrain-595K
---
# EEVE-LLaVA-2.8B-CLIP_KO-pretrain

![이미지](https://cdn.donmai.us/sample/d8/91/__nagase_kotono_idoly_pride_drawn_by_sakuranoron__sample-d8915e20211fe88aba61f1c215cc32d1.jpg)  

**This model is a pretrained version of the llava multimodal projector.**

**You can use this model to finetune [EEVE-Korean-Instruct-2.8B-v1.0](https://huggingface.co/yanolja/EEVE-Korean-Instruct-2.8B-v1.0) model.**  
  
**Use [Bingsu/clip-vit-large-patch14-ko](https://huggingface.co/Bingsu/clip-vit-large-patch14-ko)**  

## Hardware
***3x NVIDIA A100 40GB***  
## Software
***Using [MoE-LLaVA](https://github.com/PKU-YuanGroup/MoE-LLaVA)***