Mini-Gemini facilitates a range of dense and Mixture of Experts (MoE) Large Language Models (LLMs), spanning from 2B to 34B parameters, capable of simultaneousl

Mini-Gemini facilitates a range of dense and Mixture of Experts (MoE) Large Language Models (LLMs), spanning from 2B to 34B parameters, capable of simultaneously understanding, reasoning about, and generating images. This repository is developed on the foundation provided by LLaVA.