How do you train yolox models?

Thank you for sharing your excellent work.

In the previous issues, it is mentioned that the SSOD loss is not supported by yolox model. I guess this is because `YOLOAnchorAssigner` does not support anchor-free heads. I will appreciate if you can provide more information on how you trained the yolox model with this framework.

Is it only trained for the burn in period with the domain adaptation?