I am reproducing your model result, and I have a question about the code #18
Description
In the "YOLOv10WorldHead", you use ".detach()" to cut the grad from "one2one" supervise loss. I don't know why
{{ message }}
In the "YOLOv10WorldHead", you use ".detach()" to cut the grad from "one2one" supervise loss. I don't know why