Unexpected Double Processing in model.predict() on Hugging Face Spaces

Hey everyone,

I’ve been working with a model that’s been behaving a bit oddly on Hugging Face Spaces and I was hoping to get some insights from you all. The model works fine when I run it locally and on Google Colab. Link to my space

However, when I use the same gradio app.py on Hugging Face Spaces, it seems like model.predict() is running twice. This results in a prediction full of NaNs, which is not the case when I run the model locally or on Colab.

I’ve attached screenshot of the logs from Hugging Face Spaces. Has anyone else experienced something similar or have any ideas about what might be causing this?