After correctly configuring the NExT-GPT-Legacy demo (using pre-trained weights and following the instructions in NExT-GPT-Legacy/README.md, completing the bash app.sh script), the Gradio web demo interface does not respond. After entering a prompt, the demo interface does not display the corresponding inference results. Additionally, there are errors in the input for audio, video, etc., as shown in the image:

input:

output:

console:

As can be seen, my backend is displaying the prompt and the corresponding text modality inference results. What should I do to ensure that normal inference works on the web interface?