search

OpenMOSS / AnyGPT

Code for "AnyGPT: Unified Multimodal LLM with Discrete Sequence Modeling"
629 stars 43 forks source link

text chat is not so good. #17

Closed mixiazhiyang closed 2 months ago

mixiazhiyang commented 2 months ago

image it's like no intelligence. Am I wrongly prompting?

mixiazhiyang commented 2 months ago

besides, do you provide any interface, like gradio, or something showed in your demo page? Thank you!

JunZhan2000 commented 2 months ago

First, this model is primarily trained on English data, so we recommend communicating in English. Secondly, it focuses more on multimodal dialogue capabilities, and we did not include pure text instruction data in its training.