Query about Uni Planner training model

[Diwakar-7674]

Normally any deep learning network having only forward method and same number of inputs during training and inference also but your uni planner network having 7 inputs during training, but only 4 input during inference
And why uni planner network having infer method also and how it trained with 4 inputs without loss function
could you please explain it?