The model needs to use two losses_ FN, how to operate?

Reappear MTCNN A network model , On the output class、box Separate calculation loss

【 Operation steps & Problem phenomenon 】

1、 Can I use WithLossCell&TrainOneStepCell Methods? ？ If you can, how should you operate ？

2、 If not , Is there any other way to achieve , Can reference samples be provided ？

reason ： As shown in the figure ,num_epochs The parameter is set before 1, Change to max_epoch Just fine