ERROR: using Cholesky decomposition when using Turing.jl Bayesien inference

New to Julia
1

Hi everyone, i’am sharing my code snippet where i try to do some bayesien inference like here

@model IPPModel() = begin

#assign priors for the SPDE layer
μ₁ ~ L₁priors.μprior
l₁ ~ L₁priors.lprior
σ₁ ~ L₁priors.σprior
w₁ ~ wprior
μ = μ₁ .* exp.(spatialSPDE(smoothclamp(l₁, lmin, lmax), σ₁, w₁))
μi = μ
println(integrate_on_mesh(μ, mesh.point, mesh.cell))
Turing.@addlogprob! logsum(μi) - integrate_on_mesh(μ, mesh.point, mesh.cell)
return μ
end

function (m::spatialSPDE)(κ::Real, σ::Real, w::AbstractVector)
L = Symmetric(κ * C_inv * κ) # for alpha=2.0
F = cholesky(L).PtL’
return F \ (σ * w) # solve SPDE
end

the problem when using the cholesky decomposition i get this error:
ERROR: LoadError: TaskFailedException

nested task error: TaskFailedException

    nested task error: TypeError: in Sparse, in Tv, expected Tv<:Union{Float64, ComplexF64}, got Type{ForwardDiff.Dual{ForwardDiff.Tag{DynamicPPL.DynamicPPLTag, Float64}, Float64, 3}}
    Stacktrace:
      [1] SparseArrays.CHOLMOD.Sparse(A::SparseMatrixCSC{ForwardDiff.Dual{ForwardDiff.Tag{DynamicPPL.DynamicPPLTag, Float64}, Float64, 3}, Int64}, stype::Int64)
        @ SparseArrays.CHOLMOD C:\Users\asus\AppData\Local\Programs\Julia-1.10.2\share\julia\stdlib\v1.10\SparseArrays\src\solvers\cholmod.jl:917
      [2] SparseArrays.CHOLMOD.Sparse(A::Symmetric{ForwardDiff.Dual{ForwardDiff.Tag{DynamicPPL.DynamicPPLTag, Float64}, Float64, 3}, SparseMatrixCSC{ForwardDiff.Dual{ForwardDiff.Tag{DynamicPPL.DynamicPPLTag, Float64}, Float64, 3}, Int64}})
        @ SparseArrays.CHOLMOD C:\Users\asus\AppData\Local\Programs\Julia-1.10.2\share\julia\stdlib\v1.10\SparseArrays\src\solvers\cholmod.jl:931
      [3] cholesky(A::Symmetric{ForwardDiff.Dual{ForwardDiff.Tag{DynamicPPL.DynamicPPLTag, Float64}, Float64, 3}, SparseMatrixCSC{ForwardDiff.Dual{ForwardDiff.Tag{DynamicPPL.DynamicPPLTag, Float64}, Float64, 3}, Int64}}; kws::@Kwargs{})
        @ SparseArrays.CHOLMOD C:\Users\asus\AppData\Local\Programs\Julia-1.10.2\share\julia\stdlib\v1.10\SparseArrays\src\solvers\cholmod.jl:1437
      [4] cholesky(A::Symmetric{ForwardDiff.Dual{ForwardDiff.Tag{DynamicPPL.DynamicPPLTag, Float64}, Float64, 3}, SparseMatrixCSC{ForwardDiff.Dual{ForwardDiff.Tag{DynamicPPL.DynamicPPLTag, Float64}, Float64, 3}, Int64}})
        @ SparseArrays.CHOLMOD C:\Users\asus\AppData\Local\Programs\Julia-1.10.2\share\julia\stdlib\v1.10\SparseArrays\src\solvers\cholmod.jl:1437
      [5] safe_cholesky(L::Symmetric{ForwardDiff.Dual{ForwardDiff.Tag{DynamicPPL.DynamicPPLTag, Float64}, Float64, 3}, SparseMatrixCSC{ForwardDiff.Dual{ForwardDiff.Tag{DynamicPPL.DynamicPPLTag, Float64}, Float64, 3}, Int64}})
        @ Main d:\Statistical_tools\test_delaunay_triangulation_refined\src\spatialSPDE.j

what can i do to make it without deleting the cholesky decomposition since it’s needed here

2

Hello!

This code you’ve provided isn’t runnable, because we don’t have access to the variables and types you’re using. For example – what is L₁priors?

Could you please provide an example that can be run as a single script? If you need specific packages to be installed, please list those too.

Thank you :slight_smile: