Chinese-LLaVA

Yu Shu2, Siwei Dong2, Wenhao Huang3, Jialei Wang2, Yemin Shi1*
1LinkSoul.AI, 2Beijing Academy of Artificial Intelligence, China, 301.ai
*Corresponding author: ymshi@linksoul.ai
Model Code Data

Abstract

To contribute to the Chinese open source community, we adapt LLaVA to support visual instruction following in Chinese.

Our model makes the following contributions:

  • We adapt LLaVA to support visual instruction following in Chinese.
  • We release a Chinese and English visual instruction following dataset Chinese-LLaVA-Vision-Instructions.
  • We release the code in https://github.com/LinkSoul-AI/Chinese-LLaVA.
  • We release the models in Chinese-LLaVA-Chinese-Llama-2-7B and Chinese-LLaVA-Baichuan-7B