**kwargs for custom forward functions

[ILbr077]

why won't it use kwargs to adress cases like def forward(x, steps) or stuff like that when the model does not get two tensors but a variable that affect the model flow.
I think it can be done in this way:

def summary(model,shape,**kwargs):
#create dummy tensor in shape
#check the flow with model.forward(dummy, **kwargs)