Gabriel Birnbaum

https://arxiv.org/abs/2006.09661 this is also relevant.

What is the status of multi-GPU support for PINNs?

Agree. But is there a simple way (meaning, just using Plots) to visualize a PINN with 3 space dimensions I am missing though? We usually reserve one dimension for `u`...

It took some researching, but I figured out how to plot this with Makie ```julia using Makie ts, xs, ys, zs = [infimum(d.domain):0.1:supremum(d.domain) for d in domains] u_predict = [collect(phi([t,x,y,z],...

> We just shouldn't use GridTraining It might be worth pointing that out in the docs. We could add a little caveat wherever GridTraining is being used (e.g. https://neuralpde.sciml.ai/stable/pinn/wave/) The...

I got that same error message when trying to implement the Integral with Interval(x, Inf). ```julia if a isa Num a = NeuralPDE.build_symbolic_loss_function(nothing, indvars,depvars, dict_indvars,dict_depvars, dict_depvar_input, phi, derivative_, nothing, chain,...

I'll try to tackle this. Probably involves a change to _transform_expression and build_symbolic_loss_function. Will also think about how we can support compound expressions with two variables ```julia @parameters t @variables...

Specially since IDEs are much slower than PDEs in NeuralPDE (training sometimes goes from 1 minute to 1 hour by adding one integral). This is to be expected as we...

Just tried optimizing with ADAM + BFGS and that didn't work well either just fyi :)

@ChrisRackauckas I'm not sure I know how to. How does one go about doing that?