herrji

Seems to be the visual-question-answering only works for single image. My hunch is that GPT is always the best choice for question answering task, and its dependent input should be...

Since the error complains missing config files, I just do `!git clone https://github.com/open-mmlab/mmagic.git` and `!cp -r /content/mmagic/configs /usr/local/lib/python3.10/dist-packages/`, and it seems to work fine. I use Colab. Good luck.

Something is wrong in installation script IMO.