NeuralPDE debuging: is it posible to get the loss of a point in the domain?

Hi.

I am solving some differential equations with NeuralPDE and the solutions don’t look like expected. So, I really want to plot the loss at every point in the domain to see what is going on. Is there any way to get this loss information out for each point, maybe from symbolic_discretize or something else ?

When I look inside the structure from symbolic_discretize
I see
symbolic_bc_loss_functions
symbolic_pde_loss_functions
And insider .loss_functions I find.
additional_loss_function
bc_loss_functions
datafree_bc_loss_functions
datafree_pde_loss_functions
full_loss_function
pde_loss_functions

Is there any documentation on how to use those loss functions ?
I can find examples with bc_loss_functions and pde_loss_functions
but not the others.

We are redoing the lowering process to give the symbolic loss function

When that’s done this should be better documented and easier to do.