How to setup conversation with vision chat?

Open kylekam opened this issue 1 year ago • 1 comments

I'm currently able to use run_vision_chat.sh with a limited number of video frames being passed in for a single text query. The text result is output from the model and then the process ends. However, the paper shows examples of a continuous dialogue about a video and I was wondering if it's possible to set this up.

Mar 07 '24 02:03 kylekam

@kylekam what tpu version are using?

Mar 12 '24 12:03 infocodiste