Poisson Example not working on Julia 1.8

[bernaloo]

Julia 1.8 Running on macOS Catalina 10.15.7 (Intel)
After installing all needed dependencies, trying to run the Poisson Example from "Introduction to NeuralPDE. "
Shows the following Problem:

ERROR: LoadError: TypeError: in NonAdaptiveLoss, in T, expected T<:Real, got Type{Any} Stacktrace: [1] symbolic_discretize(pde_system::PDESystem, discretization::PhysicsInformedNN{QuadratureTraining{IntegralsCubature.CubatureJLh, Float64}, Nothing, NeuralPDE.Phi{Lux.Chain{NamedTuple{(:layer_1, :layer_2, :layer_3), Tuple{Dense{typeof(σ), Matrix{Float32}, Vector{Float32}}, Dense{typeof(σ), Matrix{Float32}, Vector{Float32}}, Dense{typeof(identity), Matrix{Float32}, Vector{Float32}}}}}, NamedTuple{(:layer_1, :layer_2, :layer_3), Tuple{NamedTuple{(), Tuple{}}, NamedTuple{(), Tuple{}}, NamedTuple{(), Tuple{}}}}}, typeof(NeuralPDE.numeric_derivative), Bool, Nothing, Nothing, Nothing, Base.Pairs{Symbol, Union{}, Tuple{}, NamedTuple{(), Tuple{}}}}) @ NeuralPDE ~/.julia/packages/NeuralPDE/Qjcw1/src/discretize.jl:491 [2] discretize(pde_system::PDESystem, discretization::PhysicsInformedNN{QuadratureTraining{IntegralsCubature.CubatureJLh, Float64}, Nothing, NeuralPDE.Phi{Lux.Chain{NamedTuple{(:layer_1, :layer_2, :layer_3), Tuple{Dense{typeof(σ), Matrix{Float32}, Vector{Float32}}, Dense{typeof(σ), Matrix{Float32}, Vector{Float32}}, Dense{typeof(identity), Matrix{Float32}, Vector{Float32}}}}}, NamedTuple{(:layer_1, :layer_2, :layer_3), Tuple{NamedTuple{(), Tuple{}}, NamedTuple{(), Tuple{}}, NamedTuple{(), Tuple{}}}}}, typeof(NeuralPDE.numeric_derivative), Bool, Nothing, Nothing, Nothing, Base.Pairs{Symbol, Union{}, Tuple{}, NamedTuple{(), Tuple{}}}}) @ NeuralPDE ~/.julia/packages/NeuralPDE/Qjcw1/src/discretize.jl:678 [3] top-level scope @ ~/Projects/julia/NeuralPDEs/PoissonNPDE.jl:28 [4] include(fname::String) @ Base.MainInclude ./client.jl:476 [5] top-level scope @ REPL[18]:1 [6] top-level scope @ ~/.julia/packages/CUDA/DfvRa/src/initialization.jl:52`

[ChrisRackauckas]

Is this with Julia v1.8.1? What version of NeuralPDE?

[bernaloo]

@ChrisRackauckas
This is with Julia v1.8.0 and NeuralPDE v5.2.0
Now that you say there is 1.8.1
I can give that a try.

[bernaloo]

solved with Julia 1.8.1

[YichengDWu]

What's changed in v1.8.1 so that it worked?

[ChrisRackauckas]

ERROR: LoadError: TypeError: in NonAdaptiveLoss, in T, expected T<:Real, got Type{Any} Stacktrace: [1]

Sounds like it was just an older version of NeuralPDE from before the refactor, so not a Julia update but the package update from a previous major version. Just a guess from the information there though.

[YichengDWu]

Oh I was wondering about the new release 1.8.1 cuz the result of the ks equation is quite bad https://neuralpde.sciml.ai/stable/examples/ks/
but always good on my laptop.