Unable to generate any modality other than text

[jwzhi]

I was trying out the demo using the pretrained checkpoint from huggingface. And the exact same example as shown on the project webpage.
However, I found that it is unable to generate any modality other than text. After I read through a few issues, I found a lot of people have the exact same question as mine. Is there a bug in the code? Can you please look into it?
See the attached screenshots for more details.

![screencapture-localhost-24004-2024-01-10-23_26_50](https://github.com/NE

xT-GPT/NExT-GPT/assets/46155101/f6adc453-7396-4093-9d54-88df24a982d2)