[MLLM] MiniGPT-4
Todo list.
[Submitted on 20 Apr 2023]
MiniGPT-4: Enhancing Vision-Language Understanding with Advanced Large Language Models
Saleforce的AI Projects:
MiniGPT-4 仅使用一个投影层将一个冻结的视觉编码器(BLIP-2)与一个冻结的 LLM(Vicuna)对齐。
多模态对齐是指找到两种或多种模态的instances中sub-components之间的对应关系,例如:给定一张图片和一个描述,找到词或者短语对应图片中的区域。
Ref: CV多模态和AIGC的原理解析:从CLIP、BLIP到Stable Diffusion、Midjourney
https://instruction-tuning-with-gpt-4.github.io/
[Submitted on 6 Apr 2023]
Instruction Tuning with GPT-4 [看上去不错的实践报告]
Vicuna: An Open-Source Chatbot Impressing GPT-4 with 90%* ChatGPT Quality
by: The Vicuna Team, Mar 30, 2023