UCSB-AI/MiniGPT-5
A multimodal LLM that jointly generates coherent text and images using generative vokens as bridging tokens.
Not currently ranked — collecting fresh signals.
star history
MiniGPT-5 is an interleaved vision-and-language generation model that uses a novel concept of generative vokens to harmonize image and text output. It employs a two-staged training strategy that does not require comprehensive image descriptions, making training more efficient. The approach incorporates classifier-free guidance to improve voken effectiveness for image generation.
Frequently asked
- What is UCSB-AI/MiniGPT-5?
- A multimodal LLM that jointly generates coherent text and images using generative vokens as bridging tokens.
- Is MiniGPT-5 open source?
- Yes — UCSB-AI/MiniGPT-5 is open source, released under the Apache-2.0 license.
- What language is MiniGPT-5 written in?
- UCSB-AI/MiniGPT-5 is primarily written in Python.
- How popular is MiniGPT-5?
- UCSB-AI/MiniGPT-5 has 866 stars on GitHub.
- Where can I find MiniGPT-5?
- UCSB-AI/MiniGPT-5 is on GitHub at https://github.com/UCSB-AI/MiniGPT-5.