Modified Yolov10 with CARAFE integrating to streamlit error

[Lukite404]

Hello everyone, I need help with something, it has something to do with integrating a custom trained YOLOv10 model to streamlit.

Me and my group are using YOLOv10 to classify different classes of cocoa beans (after fermentation), this classes are A, B, and C. Right now, we are entering the integration phase and we are encountering a problem that's giving us a hard time to fix: After modifying YOLOv10, we tried to run an inference in streamlit, but we got an error.

The reason why we got an error is because we modify YOLOv10 with CARAFE. CARAFE is not a library included in YOLOv10 but we wanted to try to change the upsampling block in the head with CARAFE to see if it produces good results. But because of this, we don't know how to fix this problem. We are planning add an API to make things easier to run the inference in streamlit but we wanted to look for other solutions that are much more time efficient and effective. The file we used to run the inference to streamlit is a .pt file. We need your help in finding a way to integrate a custom trained YOLOv10 model