Error: SliceSampler ... shoud be initialized in the support #209

thorek1 · 2024-02-20T12:17:39Z

I am using a Turing model and it can happen especially at later stages/scans or when i use >1 chains that I get the below error (deleted parts to make it fit here).

how can I avoid this?

it appears he proposes a sample outside the prior bounds which returns infinite log probability -> error.

I dont know what the best solution is but from a user perspective it would be great if sampling continues. as a proposal: would it be possible to either a) ignore the error, or b) take draws until you get finite log probability?

SliceSampler supports contrained target, but the sampler should be initialized in the support: DynamicPPL.TypedVarInfo{@NamedTuple{z_ea::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_ea, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_ea, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, z_eb::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_eb, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_eb, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, z_eg::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_eg, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_eg, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, z_eqs::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_eqs, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_eqs, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, z_em::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_em, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_em, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, z_epinf::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_epinf, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_epinf, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, z_ew::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_ew, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_ew, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, crhoa::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:crhoa, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Beta{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:crhoa, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, crhob::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:crhob, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Beta{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:crhob, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, crhog::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:crhog, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Beta{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:crhog, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_ea, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}(Dict(z_ea => 1), [z_ea], UnitRange{Int64}[1:1], [-5.937935554636923], Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}[Truncated(Distributions.InverseGamma{Float64}(
invd: Distributions.Gamma{Float64}(α=2.0025, θ=9.97506234413965)
θ: 0.10025
)
; lower=0.01, upper=3.0)
], Set{DynamicPPL.Selector}[Set()], [0], Dict{String, BitVector}("del" => [0], "trans" => [1])), z_eb = DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_eb, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_eb, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}(Dict(z_eb => 1), [z_eb], UnitRange{Int64}[1:1], [-6.07248915333809], Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}[Truncated(Distributions.InverseGamma{Float64}(
invd: Distributions.Gamma{Float64}(α=2.0025, θ=9.97506234413965)
θ: 0.10025
)
; lower=0.025, upper=5.0)
], Set{DynamicPPL.Selector}[Set()], [0], Dict{String, BitVector}("del" => [0], "trans" => [1])), z_eg = DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_eg, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_eg, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}(Dict(z_eg => 1), [z_eg], UnitRange{Int64}[1:1], [-2.573719397373961], Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}[Truncated(Distributions.InverseGamma{Float64}(
invd: Distributions.Gamma{Float64}(α=2.0025, θ=9.97506234413965)
θ: 0.10025
)
; lower=0.01, upper=3.0)
], Set{DynamicPPL.Selector}[Set()], [0], Dict{String, BitVector}("del" => [0], "trans" => [1])), z_eqs = DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_eqs, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_eqs, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}(Dict(z_eqs => 1), [z_eqs], UnitRange{Int64}[1:1], [-3.540303267311355], Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}[Truncated(Distributions.InverseGamma{Float64}(
invd: Distributions.Gamma{Float64}(α=2.0025, θ=9.97506234413965)
θ: 0.10025
)
; lower=0.01, upper=3.0)
], Set{DynamicPPL.Selector}[Set()], [0], Dict{String, BitVector}("del" => [0], "trans" => [1])), z_em = DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_em, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_em, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}(Dict(z_em => 1), [z_em], UnitRange{Int64}[1:1], [-5.401526507060494], Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}[Truncated(Distributions.InverseGamma{Float64}(
invd: Distributions.Gamma{Float64}(α=2.0025, θ=9.97506234413965)
θ: 0.10025
)
; lower=0.01, upper=3.0)
], Set{DynamicPPL.Selector}[Set()], [0], Dict{String, BitVector}("del" => [0], "trans" => [1])), z_epinf = DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_epinf, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_epinf, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}(Dict(z_epinf => 1), [z_epinf], UnitRange{Int64}[1:1], [-4.364336756828878], Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}[Truncated(Distributions.InverseGamma{Float64}(
invd: Distributions.Gamma{Float64}(α=2.0025, θ=9.97506234413965)
θ: 0.10025
)
; lower=0.01, upper=3.0)
], Set{DynamicPPL.Selector}[Set()], [0], Dict{String, BitVector}("del" => [0], "trans" => [1])), z_ew = DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_ew, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_ew, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}(Dict(z_ew => 1), [z_ew], UnitRange{Int64}[1:1], [-4.976811514323299], Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}[Truncated(Distributions.InverseGamma{Float64}(
invd: Distributions.Gamma{Float64}(α=2.0025, θ=9.97506234413965)
θ: 0.10025
)
; lower=0.01, upper=3.0)
], Vector{Truncated{Distributions.Beta{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:crr, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}(Dict(crr => 1), [crr], UnitRange{Int64}[1:1], [-1.85321895906356], Truncated{Distributions.Beta{Float64}, Distributions.Continuous, Float64, Float64, Float64}[Truncated(Distributions.Beta{Float64}(α=13.312499999999998, β=4.437499999999998); lower=0.5, upper=0.975)], Set{DynamicPPL.Selector}[Set()], [0], Dict{String, BitVector}("del" => [0], "trans" => [1])), cry = DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:cry, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Normal{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:cry, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}(Dict(cry => 1), [cry], UnitRange{Int64}[1:1], [-0.4318612361537384], Truncated{Distributions.Normal{Float64}, Distributions.Continuous, Float64, Float64, Float64}[Truncated(Distributions.Normal{Float64}(μ=0.125, σ=0.05); lower=0.001, upper=0.5)], Set{DynamicPPL.Selector}[Set()], [0], Dict{String, BitVector}("del" => [0], "trans" => [1])), crdy = DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:crdy, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Normal{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:crdy, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}(Dict(crdy => 1), [crdy], UnitRange{Int64}[1:1], [-0.963199929104076], Truncated{Distributions.Normal{Float64}, Distributions.Continuous, Float64, Float64, Float64}[Truncated(Distributions.Normal{Float64}(μ=0.125, σ=0.05); lower=0.001, upper=0.5)], Set{DynamicPPL.Selector}[Set()], [0], Dict{String, BitVector}("del" => [0], "trans" => [1])), constepinf = DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:constepinf, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Gamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:constepinf, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}(Dict(constepinf => 1), [constepinf], UnitRange{Int64}[1:1], [-1.1202632540347432], Truncated{Distributions.Gamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}[Truncated(Distributions.Gamma{Float64}(α=39.06249999999999, θ=0.016000000000000004); lower=0.1, upper=2.0)], Set{DynamicPPL.Selector}[Set()], [0], Dict{String, BitVector}("del" => [0], "trans" => [1])), constebeta = DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:constebeta, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Gamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:constebeta, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}(Dict(constebeta => 1), [constebeta], UnitRange{Int64}[1:1], [-1.651054162980441], Truncated{Distributions.Gamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}[Truncated(Distributions.Gamma{Float64}(α=6.249999999999999, θ=0.04000000000000001); lower=0.01, upper=2.0)], Set{DynamicPPL.Selector}[Set()], [0], Dict{String, BitVector}("del" => [0], "trans" => [1])), constelab = DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:constelab, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Normal{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:constelab, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}(Dict(constelab => 1), [constelab], UnitRange{Int64}[1:1], [-0.32747896578836494], Truncated{Distributions.Normal{Float64}, Distributions.Continuous, Float64, Float64, Float64}[Truncated(Distributions.Normal{Float64}(μ=0.0, σ=2.0); lower=-10.0, upper=10.0)], Set{DynamicPPL.Selector}[Set()], [0], Dict{String, BitVector}("del" => [0], "trans" => [1])), ctrend = DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:ctrend, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Normal{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:ctrend, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}(Dict(ctrend => 1), [ctrend], UnitRange{Int64}[1:1], [-1.2120855104843935], Truncated{Distributions.Normal{Float64}, Distributions.Continuous, Float64, Float64, Float64}[Truncated(Distributions.Normal{Float64}(μ=0.4, σ=0.1); lower=0.1, upper=0.8)], Set{DynamicPPL.Selector}[Set()], [0], Dict{String, BitVector}("del" => [0], "trans" => [1])), cgy = DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:cgy, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Normal{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:cgy, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}(Dict(cgy => 1), [cgy], UnitRange{Int64}[1:1], [-1.1499676302234567], Truncated{Distributions.Normal{Float64}, Distributions.Continuous, Float64, Float64, Float64}[Truncated(Distributions.Normal{Float64}(μ=0.5, σ=0.25); lower=0.01, upper=2.0)], Set{DynamicPPL.Selector}[Set()], [0], Dict{String, BitVector}("del" => [0], "trans" => [1])), calfa = DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:calfa, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Normal{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:calfa, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}(Dict(calfa => 1), [calfa], UnitRange{Int64}[1:1], [-1.111465444391527], Truncated{Distributions.Normal{Float64}, Distributions.Continuous, Float64, Float64, Float64}[Truncated(Distributions.Normal{Float64}(μ=0.3, σ=0.05); lower=0.01, upper=1.0)], Set{DynamicPPL.Selector}[Set()], [0], Dict{String, BitVector}("del" => [0], "trans" => [1]))), Base.RefValue{Float64}(-Inf), Base.RefValue{Int64}(0))
Stacktrace:
  [1] error(s::String)
    @ Base ./error.jl:35
  [2] cached_log_potential(log_potential::InterpolatedLogPotential{...}, state::DynamicPPL.TypedVarInfo{@NamedTuple{z_ea::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_ea, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_ea, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, z_eb::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_eb, Setfield.IdentityLens}, Int64}, Vector{AbstractPPL.VarName{:chabb, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, cprobw::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:cprobw, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Beta{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:cprobw, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, csigl::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:csigl, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Normal{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:csigl, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, cprobp::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:cprobp, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Beta{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:cprobp, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, cindw::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:cindw, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Beta{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:cindw, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, cindp::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:cindp, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Beta{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:cindp, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, czcap::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:czcap, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Beta{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:czcap, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, cfc::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:cfc, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Normal{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:cfc, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, crpi::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:crpi, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Normal{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:crpi, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, crr::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:crr, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Beta{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:crr, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, cry::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:cry, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Normal{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:cry, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, crdy::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:crdy, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Normal{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:crdy, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, constepinf::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:constepinf, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Gamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:constepinf, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, constebeta::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:constebeta, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Gamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:constebeta, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, constelab::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:constelab, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Normal{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:constelab, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, ctrend::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:ctrend, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Normal{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:ctrend, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, cgy::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:cgy, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Normal{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:cgy, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, calfa::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:calfa, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Normal{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:calfa, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}}, Float64}, cached_lp::Float64)
    @ Pigeons ~/.julia/packages/Pigeons/UUYm9/src/explorers/SliceSampler.jl:36
  [3] SliceSampler supports contrained target, but the sampler should be initialized in the support: DynamicPPL.TypedVarInfo{@NamedTuple{z_ea::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_ea, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_ea, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, z_eb::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_eb, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_ea, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}(Dict(z_ea => 1), [z_ea], UnitRange{Int64}[1:1], [-5.937935554636923], Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}[Truncated(Distributions.InverseGamma{Float64}(
invd: Distributions.Gamma{Float64}(α=2.0025, θ=9.97506234413965)
θ: 0.10025
)
; lower=0.01, upper=3.0)
], Set{DynamicPPL.Selector}[Set()], [0], Dict{String, BitVector}("del" => [0], "trans" => [1])), z_eb = DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_eb, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_eb, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}(Dict(z_eb => 1), [z_eb], UnitRange{Int64}[1:1], [-6.07248915333809], Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}[Truncated(Distributions.InverseGamma{Float64}(
invd: Distributions.Gamma{Float64}(α=2.0025, θ=9.97506234413965)
θ: 0.10025
)
; lower=0.025, upper=5.0)
], Set{DynamicPPL.Selector}[Set()], [0], Dict{String, BitVector}("del" => [0], "trans" => [1])), z_eg = DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_eg, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_eg, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}(Dict(z_eg => 1), [z_eg], UnitRange{Int64}[1:1], [-2.573719397373961], Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}[Truncated(Distributions.InverseGamma{Float64}(
invd: Distributions.Gamma{Float64}(α=2.0025, θ=9.97506234413965)
θ: 0.10025
)
; lower=0.01, upper=3.0)
], Set{DynamicPPL.Selector}[Set()], [0], Dict{String, BitVector}("del" => [0], "trans" => [1])), z_eqs = DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_eqs, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_eqs, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}(Dict(z_eqs => 1), [z_eqs], UnitRange{Int64}[1:1], [-3.540303267311355], Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}[Truncated(Distributions.InverseGamma{Float64}(
invd: Distributions.Gamma{Float64}(α=2.0025, θ=9.97506234413965)
θ: 0.10025
)
; lower=0.01, upper=3.0)
], Set{DynamicPPL.Selector}[Set()], [0], Dict{String, BitVector}("del" => [0], "trans" => [1])), z_em = DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_em, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_em, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}(Dict(z_em => 1), [z_em], UnitRange{Int64}[1:1], [-5.401526507060494], Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}[Truncated(Distributions.InverseGamma{Float64}(
invd: Distributions.Gamma{Float64}(α=2.0025, θ=9.97506234413965)
θ: 0.10025
)
; lower=0.01, upper=3.0)
], Set{DynamicPPL.Selector}[Set()], [0], Dict{String, BitVector}("del" => [0], "trans" => [1])), z_epinf = DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_epinf, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_epinf, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}(Dict(z_epinf => 1), [z_epinf], UnitRange{Int64}[1:1], [-4.364336756828878], Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}[Truncated(Distributions.InverseGamma{Float64}(
invd: Distributions.Gamma{Float64}(α=2.0025, θ=9.97506234413965)
θ: 0.10025
)
; lower=0.01, upper=3.0)
], Set{DynamicPPL.Selector}[Set()], [0], Dict{String, BitVector}("del" => [0], "trans" => [1])), z_ew = DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_ew, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_ew, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}(Dict(z_ew => 1), [z_ew], UnitRange{Int64}[1:1], [-4.976811514323299], Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}[Truncated(Distributions.InverseGamma{Float64}(
invd: Distributions.Gamma{Float64}(α=2.0025, θ=9.97506234413965)
θ: 0.10025
)
; lower=0.01, upper=3.0)
], Set{DynamicPPL.Selector}[Set()], [0], Dict{String, BitVector}("del" => [0], "trans" => [1])), crhoa = DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:crhoa, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Beta{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:crhoa, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}(Dict(crhoa => 1), [crhoa], UnitRange{Int64}[1:1], [0.6316967241311127], Truncated{Distributions.Beta{Float64}, Distributions.Continuous, Float64, Float64, Float64}[Truncated(Distributions.Beta{Float64}(α=2.6249999999999996, β=2.6249999999999996); lower=0.01, upper=0.9999)], Set{DynamicPPL.Selector}[Set()], [0], Dict{String, BitVector}("del" => [0], "trans" => [1])), crhob = DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:crhob, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Beta{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:crhob, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}(Dict(crhob => 1), [crhob], UnitRange{Int64}[1:1], [-0.9168621366286988], Truncated{Distributions.Beta{Float64}, Distributions.Continuous, Float64, Float64, Float64}[Truncated(Distributions.Beta{Float64}(α=2.6249999999999996, β=2.6249999999999996); lower=0.01, upper=0.9999)], Set{DynamicPPL.Selector}[Set()], [0], Dict{String, BitVector}("del" => [0], "trans" => [1])), crhog = DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:crhog, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Beta{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:crhog, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}(Dict(crhog => 1), [crhog], UnitRange{Int64}[1:1], [0.7505066853531597], Truncated{Distributions.Beta{Float64}, Distributions.Continuous, Float64, Float64, Float64}[Truncated(Distributions.Beta{Float64}(α=2.6249999999999996, β=2.6249999999999996); lower=0.01, upper=0.9999)], Set{DynamicPPL.Selector}[Set()], [0], Dict{String, BitVector}("del" => [0], "trans" => [1])), crhoqs = DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:crhoqs, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Beta{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:crhoqs, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}(Dict(crhoqs => 1), [crhoqs], UnitRange{Int64}[1:1], [0.49319084611136077], Truncated{Distributions.Beta{Float64}, Distributions.Continuous, Float64, Float64, Float64}[Truncated(Distributions.Beta{Float64}(α=2.6249999999999996, β=2.6249999999999996); lower=0.01, upper=0.9999)], Set{DynamicPPL.Selector}[Set()], [0], Dict{String, BitVector}("del" => [0], "trans" => [1])), crhoms = DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:crhoms, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Beta{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:crhoms, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}(Dict(crhoms => 1), [crhoms], UnitRange{Int64}[1:1], [-0.8469899269176431], Truncated{Distributions.Beta{Float64}, Distributions.Continuous, Float64, Float64, Float64}[Truncated(Distributions.Beta{Float64}(α=2.6249999999999996, β=2.6249999999999996); lower=0.01, upper=0.9999)], Set{DynamicPPL.Selector}[Set()], [0], Dict{String, BitVector}("del" => [0], "trans" => [1])), crhopinf = DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:crhopinf, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Beta{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:crhopinf, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}(Dict(crhopinf => 1), [crhopinf], UnitRange{Int64}[1:1], [0.9274192514128416], Truncated{Distributions.Beta{Float64}, Distributions.Continuous, Float64, Float64, Float64}[Truncated(Distributions.Beta{Float64}(α=2.6249999999999996, β=2.6249999999999996); lower=0.01, upper=0.9999)], Set{DynamicPPL.Selector}[Set()], [0], Dict{String, BitVector}("del" => [0], "trans" => [1])), crhow = DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:crhow, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Beta{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:crhow, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}(Dict(crhow => 1), [crhow], UnitRange{Int64}[1:1], [1.6374616954636072], Truncated{Distributions.Beta{Float64}, Distributions.Continuous, Float64, Float64, Float64}[Truncated(Distributions.Beta{Float64}(α=2.6249999999999996, β=2.6249999999999996); lower=0.001, upper=0.9999)], Set{DynamicPPL.Selector}[Set()], [0], Dict{String, BitVector}("del" => [0], "trans" => [1])), cmap = DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:cmap, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Beta{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:cmap, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}(Dict(cmap => 1), [cmap], UnitRange{Int64}[1:1], [-0.34889744856776334], Truncated{Distributions.Beta{Float64}, Distributions.Continuous, Float64, Float64, Float64}Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Normal{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:calfa, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}(Dict(calfa => 1), [calfa], UnitRange{Int64}[1:1], [-1.111465444391527], Truncated{Distributions.Normal{Float64}, Distributions.Continuous, Float64, Float64, Float64}[Truncated(Distributions.Normal{Float64}(μ=0.3, σ=0.05); lower=0.01, upper=1.0)], Set{DynamicPPL.Selector}[Set()], [0], Dict{String, BitVector}("del" => [0], "trans" => [1]))), Base.RefValue{Float64}(-Inf), Base.RefValue{Int64}(0))
Stacktrace:
  [1] error(s::String)
    @ Base ./error.jl:35
  [2] cached_log_potential(log_potential::InterpolatedLogPotential{...}, state::DynamicPPL.TypedVarInfo{@NamedTuple{z_ea::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_ea, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_ea, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, z_eb::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_eb, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_eb, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, z_eg::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_eg, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_eg, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, z_eqs::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_eqs, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{Truncated{Distributions.Normal{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:calfa, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}}, Float64}, cached_lp::Float64)
    @ Pigeons ~/.julia/packages/Pigeons/UUYm9/src/explorers/SliceSampler.jl:36
  [3] slice_sample!(h::Pigeons.SliceSampler, state::Vector{Float64}, log_potential::InterpolatedLogPotential{...}, cached_lp::Float64, replica::Pigeons.Replica{DynamicPPL.TypedVarInfo{@NamedTuple{z_ea::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_ea, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_ea, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, z_eb::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_eb, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, constelab::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:constelab, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Normal{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:constelab, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, ctrend::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:ctrend, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Normal{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:ctrend, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, cgy::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:cgy, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Normal{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:cgy, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, calfa::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:calfa, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Normal{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:calfa, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}}, Float64}, @NamedTuple{explorer_acceptance_pr::OnlineStatsBase.GroupBy{Int64, Number, OnlineStatsBase.Mean{Float64, OnlineStatsBase.EqualWeight}}, explorer_n_steps::OnlineStatsBase.GroupBy{Int64, Number, OnlineStatsBase.Sum{Float64}}, swap_acceptance_pr::OnlineStatsBase.GroupBy{Tuple{Int64, Int64}, Number, OnlineStatsBase.Mean{Float64, OnlineStatsBase.EqualWeight}}, log_sum_ratio::OnlineStatsBase.GroupBy{Tuple{Int64, Int64}, Number, Pigeons.LogSum{Float64}}, traces::Dict{Pair{Int64, Int64}, Any}, round_trip::Pigeons.RoundTripRecorder, timing_extrema::NonReproducible{...}, allocation_extrema::NonReproducible{...}}})
    @ Pigeons ~/.julia/packages/Pigeons/UUYm9/src/explorers/SliceSampler.jl:44
  [4] slice_sample!
    @ ~/.julia/packages/Pigeons/UUYm9/ext/PigeonsDynamicPPLExt/state.jl:58 [inlined]
  [5] step!(explorer::Pigeons.SliceSampler, replica::Pigeons.Replica{DynamicPPL.TypedVarInfo{@NamedTuple{z_ea::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_ea, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_ea, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, z_eb::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_eb, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_eb, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, z_eg::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_eg, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_eg, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, z_eqs::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_eqs, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_eqs, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, z_em::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_em, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_em, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, z_epinf::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_epinf, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_epinf, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, z_ew::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_ew, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_ew, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, crhoa::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:crhoa, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Beta{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:crhoa, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, crhob::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:crhob, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Beta{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:crhob, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, crhog::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:crhog, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.Beta{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:crhog, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, OnlineStatsBase.EqualWeight}}, log_sum_ratio::OnlineStatsBase.GroupBy{Tuple{Int64, Int64}, Number, Pigeons.LogSum{Float64}}, traces::Dict{Pair{Int64, Int64}, Any}, round_trip::Pigeons.RoundTripRecorder, timing_extrema::NonReproducible{...}, allocation_extrema::NonReproducible{...}}}, shared::Shared{...})
    @ Pigeons ~/.julia/packages/Pigeons/UUYm9/src/explorers/SliceSampler.jl:28
  [6] slice_sample!(h::Pigeons.SliceSampler, state::Vector{Float64}, log_potential::InterpolatedLogPotential{...}, cached_lp::Float64, replica::Pigeons.Replica{DynamicPPL.TypedVarInfo{@NamedTuple{z_ea::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_ea, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_ea, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, OnlineStatsBase.EqualWeight}}, explorer_n_steps::OnlineStatsBase.GroupBy{Int64, Number, OnlineStatsBase.Sum{Float64}}, swap_acceptance_pr::OnlineStatsBase.GroupBy{Tuple{Int64, Int64}, Number, OnlineStatsBase.Mean{Float64, OnlineStatsBase.EqualWeight}}, log_sum_ratio::OnlineStatsBase.GroupBy{Tuple{Int64, Int64}, Number, Pigeons.LogSum{Float64}}, traces::Dict{Pair{Int64, Int64}, Any}, round_trip::Pigeons.RoundTripRecorder, timing_extrema::NonReproducible{...}, allocation_extrema::NonReproducible{...}}})
    @ Pigeons ~/.julia/packages/Pigeons/UUYm9/src/explorers/SliceSampler.jl:44
  [4] slice_sample!
    @ ~/.julia/packages/Pigeons/UUYm9/ext/PigeonsDynamicPPLExt/state.jl:58 [inlined]
  [5] step!(explorer::Pigeons.SliceSampler, replica::Pigeons.Replica{DynamicPPL.TypedVarInfo{@NamedTuple{z_ea::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_ea, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_ea, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, OnlineStatsBase.EqualWeight}}, explorer_n_steps::OnlineStatsBase.GroupBy{Int64, Number, OnlineStatsBase.Sum{Float64}}, swap_acceptance_pr::OnlineStatsBase.GroupBy{Tuple{Int64, Int64}, Number, OnlineStatsBase.Mean{Float64, OnlineStatsBase.EqualWeight}}, log_sum_ratio::OnlineStatsBase.GroupBy{Tuple{Int64, Int64}, Number, Pigeons.LogSum{Float64}}, traces::Dict{Pair{Int64, Int64}, Any}, round_trip::Pigeons.RoundTripRecorder, timing_extrema::NonReproducible{...}, allocation_extrema::NonReproducible{...}}}, shared::Shared{...})
    @ Pigeons ~/.julia/packages/Pigeons/UUYm9/src/explorers/SliceSampler.jl:28
  [6] explore!(pt::PT{...}, replica::Pigeons.Replica{DynamicPPL.TypedVarInfo{@NamedTuple{z_ea::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_ea, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_ea, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, z_eb::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_eb, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_eb, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, z_eg::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_eg, Setfield.IdentityLens}, Int64}, Vector{Truncated{Distributions.InverseGamma{Float64}, Distributions.Continuous, Float64, Float64, Float64}}, Vector{AbstractPPL.VarName{:z_eg, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}, z_eqs::DynamicPPL.Metadata{Dict{AbstractPPL.VarName{:z_eqs, Setfield.IdentityLens}, Int64}, Vector{AbstractPPL.VarName{:calfa, Setfield.IdentityLens}}, Vector{Float64}, Vector{Set{DynamicPPL.Selector}}}}, Float64}, @NamedTuple{explorer_acceptance_pr::OnlineStatsBase.GroupBy{Int64, Number, OnlineStatsBase.Mean{Float64, OnlineStatsBase.EqualWeight}}, explorer_n_steps::OnlineStatsBase.GroupBy{Int64, Number, OnlineStatsBase.Sum{Float64}}, swap_acceptance_pr::OnlineStatsBase.GroupBy{Tuple{Int64, Int64}, Number, OnlineStatsBase.Mean{Float64, OnlineStatsBase.EqualWeight}}, log_sum_ratio::OnlineStatsBase.GroupBy{Tuple{Int64, Int64}, Number, Pigeons.LogSum{Float64}}, traces::Dict{Pair{Int64, Int64}, Any}, round_trip::Pigeons.RoundTripRecorder, timing_extrema::NonReproducible{...}, allocation_extrema::NonReproducible{...}}}, explorer::Pigeons.SliceSampler)
    @ Pigeons ~/.julia/packages/Pigeons/UUYm9/src/pt/pigeons.jl:107
  [7] explore!
    @ ~/.julia/packages/Pigeons/UUYm9/src/pt/pigeons.jl:96 [inlined]
  [8] macro expansion
    @ ~/.julia/packages/Pigeons/UUYm9/src/pt/pigeons.jl:50 [inlined]
  [9] macro expansion
    @ ./timing.jl:503 [inlined]
 [10] run_one_round!(pt::PT{...})
    @ Pigeons ~/.julia/packages/Pigeons/UUYm9/src/pt/pigeons.jl:49
 [11] pigeons(pt::PT{...})
    @ Pigeons ~/.julia/packages/Pigeons/UUYm9/src/pt/pigeons.jl:18
 [12] pigeons(pt_arguments::Pigeons.Inputs{TuringLogPotential{...}, Nothing, Nothing, Nothing, Nothing}, ::Pigeons.ThisProcess)
    @ Pigeons ~/.julia/packages/Pigeons/UUYm9/src/api.jl:19
 [13] #pigeons#157
    @ ~/.julia/packages/Pigeons/UUYm9/src/api.jl:16 [inlined]

alexandrebouchard · 2024-02-20T17:18:13Z

Thanks for the report and suggestions. It might be easier to give a recommendation after reading the model in question, potentially trying it locally as well.

Conceptually, the slice sampler shrinks slides until they are within constraints, so we would expect that the once the sampler is in the support it should stay in it (hence the message talking about 'initialization' despite the problem occurring 'later' here).

One possibility is that the prior sampler might propose configurations that have zero probability under the likelihood. Would that be possible in your case? If so, the best fix might be to restrict the prior to the set of parameters allowed by the likelihood function.

PS: by default, the reason we did not do (a) and (b) by default is that they may void certain guarantees, e.g. correctness of log-normalization constant estimates. (b) might also lead to unbounded running time. However if we can identify a specific case where (a) would lead to correct result, I would be open to adding an option to facilitate it (potentially combined with warning messages).

thorek1 · 2024-02-20T19:18:47Z

You can find the model here (just switch the NUTS sampler for pigeons and thats what i work with locally)

if i understand correctly what you are saying, there might be an issue if the prior sampler makes a valid proposal but, in my case, the term in addlogprob is -Inf. so a fix (on my end) could be to catch infinite values and replace them with -1e300 and then hand them to addlogprob

let me try that

thorek1 · 2024-02-20T20:05:40Z

i guess the relevant part of my problem is this:

Turing.@model function loglikelihood_function(data, m, observables)
    z_ea    ~   Truncated(InverseGamma(0.1, 2.0, μσ = true),0.01,3)
    z_eb    ~   Truncated(InverseGamma(0.1, 2.0, μσ = true),0.025,5)
    ...
    kalman_prob = get_loglikelihood(m, data(observables), [z_ea, z_ea])

    Turing.@addlogprob! kalman_prob 
end

your proposed fix doesn't work in my case because i don't know the parameter combinations which return a finite kalman_prob

the fix I mentioned earlier (isfinite(kalman_prob) ? kalman_prob : -1e30) leads to:

─────────────────────────────────
  scans     restarts      Λ        time(s)    allc(B)  log(Z₁/Z₀)   min(α)     mean(α)    min(αₑ)   mean(αₑ) 
────────── ────────── ────────── ────────── ────────── ────────── ────────── ────────── ────────── ──────────
        2          0   2.27e-13        103   1.82e+11   1.14e-13          1          1          1          1 
        2          0   2.27e-13        103   1.82e+11   1.14e-13          1          1          1          1 
ERROR: LoadError: ERROR: LoadError: Maximum number of iterations 1024 reached. Dumping info:
        - Lbar   = -5.1729024534740455
        - Rbar   = -5.172902453474045
        - new_lp = -1.0e30
        - z      = -1.0e30

Stacktrace:
  [1] error(s::String)
    @ Base ./error.jl:35
  [2] Maximum number of iterations 1024 reached. Dumping info:
        - Lbar   = -5.1729024534740455
        - Rbar   = -5.172902453474045
        - new_lp = -1.0e30
        - z      = -1.0e30

Stacktrace:
  [1] error(s::String)
    @ Base ./error.jl:35
  [2] slice_shrink!(h::P...)

alexandrebouchard · 2024-02-20T20:07:21Z

Thanks for the additional info, I'll have a look..

thorek1 · 2024-02-20T21:10:21Z

just some additional info: the above error came with 4 chains. with one chain i managed to let it run through

miguelbiron · 2024-02-20T21:39:43Z

Hi @thorek1 -- here

https://github.com/thorek1/MacroModelling.jl/blob/a932f04fd08824a126fe7f655b5b3115efeda673/benchmark/SW07_estimation.jl#L228-L232

it seems that you are trying to manually add a log-likelihood term in a Turing model. The proper way to do this is by making sure the context is not prior evaluation; see this example

Pigeons.jl/examples/turing-galaxy.jl

Lines 21 to 33 in 764af10

    
           if DynamicPPL.leafcontext(__context__) !== DynamicPPL.PriorContext() 
        
               lη  = log.(η) 
        
               σ   = inv_σ2 .^ (-1//2) 
        
               lps = similar(η) 
        
               acc = zero(eltype(y)) 
        
               for yᵢ in y 
        
                   @inbounds for k in 1:K 
        
                       lps[k] = lη[k] + logpdf(Normal(μ[k], σ[k]), yᵢ) 
        
                   end 
        
                   acc += logsumexp(lps) 
        
               end 
        
               DynamicPPL.@addlogprob! acc 
        
           end

Otherwise, the term would appear in both prior and posterior evaluations, making it impossible to apply tempering.

thorek1 · 2024-02-20T21:48:37Z

so you suggest something like this!?

Turing.@model function loglikelihood_function(data, m, observables)
    z_ea    ~   Truncated(InverseGamma(0.1, 2.0, μσ = true),0.01,3)
    z_eb    ~   Truncated(InverseGamma(0.1, 2.0, μσ = true),0.025,5)
    ...
    if DynamicPPL.leafcontext(__context__) !== DynamicPPL.PriorContext() 
        kalman_prob = get_loglikelihood(m, data(observables), [z_ea, z_ea])

        Turing.@addlogprob! kalman_prob 
   end
end

let me give this a try

nikola-sur · 2024-02-20T21:52:35Z

Related to #206

thorek1 · 2024-02-20T22:17:50Z

now i remain stuck on a segfault error (single chain, no multithreading) but i cannot exclude this comes from my get_loglikelihood function

───────────────────────────────────────────────────────────────────────────────────────
  scans     restarts    time(s)    allc(B)  log(Z₁/Z₀)   mean(α)    min(αₑ)   mean(αₑ) 
────────── ────────── ────────── ────────── ────────── ────────── ────────── ──────────
        2          0       28.2   3.46e+10          0        NaN          1          1 
        2          0       28.2   3.46e+10          0        NaN          1          1 
        4          0       46.6   7.04e+10          0        NaN          1          1 
        4          0       46.6   7.04e+10          0        NaN          1          1 
        8          0        102    1.5e+11          0        NaN          1          1 
        8          0        102    1.5e+11          0        NaN          1          1 

[164] signal (11.1): Segmentation fault
in expression starting at /home/cdsw/sample_sw07.jl:185
Allocations: 265281838 (Pool: 234497717; Big: 30784121); GC: 1940

[164] signal (11.1): Segmentation fault
in expression starting at /home/cdsw/sample_sw07.jl:185
Allocations: 265281838 (Pool: 234497717; Big: 30784121); GC: 1940
Segmentation fault (core dumped)
35584

any hint as to how to debug those?

fix a seed and go step by step?

alexandrebouchard · 2024-02-20T22:51:13Z

Segfaults are the worst! Glad to hear there is progress though in terms of at least the preceding issue being resolved.

A few questions/ideas:

can the segfault be deterministically reproduced? (by default a seed of 1 is used, and in the Pigeons side randomness is carefully controlled).
is there anything that can be simplified (e.g. using a more naive AD system such as ForwardDiff, only a subset of the model, etc)?

To facilitate debugging, it might be helpful to use checkpoints; see https://pigeons.run/dev/checkpoints/

miguelbiron · 2024-02-20T23:29:58Z

After fixing the context issue, I was able to run this model using multithreading (see my code at the end). Of course, due to computational limitations, I'm using just 4 chains and 4 rounds, but this shows that the model does work with Pigeons.

In order to avoid the initialization in -Inf log-density region, I had to customize the initialization method. Basically, the idea is to sample from the prior until a non-infinite log-density state is found, and then use that as the initial point for all replicas.

=============================================
Code:

# install latest versions of Pigeons and MacroModelling
using Pkg
Pkg.add([
    PackageSpec(name="Pigeons", rev="main"),
    PackageSpec(name="MacroModelling", rev="main")
])

using MacroModelling
import Turing
import Turing: NUTS, sample, logpdf, Truncated#, Normal, Beta, Gamma, InverseGamma,
using Random, CSV, DataFrames, Zygote, AxisKeys, MCMCChains
# using ComponentArrays, Optimization, OptimizationNLopt, OptimizationOptimisers
import DynamicPPL: DynamicPPL, logjoint
using Pigeons

@model SW07 begin
    a[0] = calfa * rkf[0] + (1 - calfa) * (wf[0])

    zcapf[0] = (1 / (czcap / (1 - czcap))) * rkf[0]

    rkf[0] = wf[0] + labf[0] - kf[0]

    kf[0] = kpf[-1] + zcapf[0]

    invef[0] = (1 / (1 + cbetabar * cgamma)) * (invef[-1] + cbetabar * cgamma * invef[1] + (1 / (cgamma ^ 2 * csadjcost)) * pkf[0]) + qs[0]

    pkf[0] =  - rrf[0] + (1 / ((1 - chabb / cgamma) / (csigma * (1 + chabb / cgamma)))) * b[0] + (crk / (crk + (1 - ctou))) * rkf[1] + ((1 - ctou) / (crk + (1 - ctou))) * pkf[1]

    cf[0] = (chabb / cgamma) / (1 + chabb / cgamma) * cf[-1] + (1 / (1 + chabb / cgamma)) * cf[1] + ((csigma - 1) * cwhlc / (csigma * (1 + chabb / cgamma))) * (labf[0] - labf[1]) - (1 - chabb / cgamma) / (csigma * (1 + chabb / cgamma)) * (rrf[0]) + b[0]

    yf[0] = ccy * cf[0] + ciy * invef[0] + g[0] + crkky * zcapf[0]

    yf[0] = cfc * (calfa * kf[0] + (1 - calfa) * labf[0] + a[0])

    wf[0] = csigl * labf[0]	 + (1 / (1 - chabb / cgamma)) * cf[0] - (chabb / cgamma) / (1 - chabb / cgamma) * cf[-1]

    kpf[0] = (1 - cikbar) * kpf[-1] + (cikbar) * invef[0] + (cikbar) * (cgamma ^ 2 * csadjcost) * qs[0]

    mc[0] = calfa * rk[0] + (1 - calfa) * (w[0]) - a[0]

    zcap[0] = (1 / (czcap / (1 - czcap))) * rk[0]

    rk[0] = w[0] + lab[0] - k[0]

    k[0] = kp[-1] + zcap[0]

    inve[0] = (1 / (1 + cbetabar * cgamma)) * (inve[-1] + cbetabar * cgamma * inve[1] + (1 / (cgamma ^ 2 * csadjcost)) * pk[0]) + qs[0]

    pk[0] =  - r[0] + pinf[1] + (1 / ((1 - chabb / cgamma) / (csigma * (1 + chabb / cgamma)))) * b[0] + (crk / (crk + (1 - ctou))) * rk[1] + ((1 - ctou) / (crk + (1 - ctou))) * pk[1]

    c[0] = (chabb / cgamma) / (1 + chabb / cgamma) * c[-1] + (1 / (1 + chabb / cgamma)) * c[1] + ((csigma - 1) * cwhlc / (csigma * (1 + chabb / cgamma))) * (lab[0] - lab[1]) - (1 - chabb / cgamma) / (csigma * (1 + chabb / cgamma)) * (r[0] - pinf[1]) + b[0]

    y[0] = ccy * c[0] + ciy * inve[0] + g[0] + crkky * zcap[0]

    y[0] = cfc * (calfa * k[0] + (1 - calfa) * lab[0] + a[0])

    pinf[0] = (1 / (1 + cbetabar * cgamma * cindp)) * (cbetabar * cgamma * pinf[1] + cindp * pinf[-1] + ((1 - cprobp) * (1 - cbetabar * cgamma * cprobp) / cprobp) / ((cfc - 1) * curvp + 1) * (mc[0])) + spinf[0]

    w[0] = (1 / (1 + cbetabar * cgamma)) * w[-1] + (cbetabar * cgamma / (1 + cbetabar * cgamma)) * w[1] + (cindw / (1 + cbetabar * cgamma)) * pinf[-1] - (1 + cbetabar * cgamma * cindw) / (1 + cbetabar * cgamma) * pinf[0] + (cbetabar * cgamma) / (1 + cbetabar * cgamma) * pinf[1] + (1 - cprobw) * (1 - cbetabar * cgamma * cprobw) / ((1 + cbetabar * cgamma) * cprobw) * (1 / ((clandaw - 1) * curvw + 1)) * (csigl * lab[0] + (1 / (1 - chabb / cgamma)) * c[0] - ((chabb / cgamma) / (1 - chabb / cgamma)) * c[-1] - w[0]) + sw[0]

    r[0] = crpi * (1 - crr) * pinf[0] + cry * (1 - crr) * (y[0] - yf[0]) + crdy * (y[0] - yf[0] - y[-1] + yf[-1]) + crr * r[-1] + ms[0]

    a[0] = crhoa * a[-1] + z_ea * ea[x]

    b[0] = crhob * b[-1] + z_eb * eb[x]

    g[0] = crhog * g[-1] + z_eg * eg[x] + cgy * z_ea * ea[x]

    qs[0] = crhoqs * qs[-1] + z_eqs * eqs[x]

    ms[0] = crhoms * ms[-1] + z_em * em[x]

    spinf[0] = crhopinf * spinf[-1] + epinfma[0] - cmap * epinfma[-1]

    epinfma[0] = z_epinf * epinf[x]

    sw[0] = crhow * sw[-1] + ewma[0] - cmaw * ewma[-1]

    ewma[0] = z_ew * ew[x]

    kp[0] = (1 - cikbar) * kp[-1] + cikbar * inve[0] + cikbar * cgamma ^ 2 * csadjcost * qs[0]

    dy[0] = y[0] - y[-1] + ctrend

    dc[0] = c[0] - c[-1] + ctrend

    dinve[0] = inve[0] - inve[-1] + ctrend

    dw[0] = w[0] - w[-1] + ctrend

    pinfobs[0] = (pinf[0]) + constepinf

    robs[0] = (r[0]) + conster

    labobs[0] = lab[0] + constelab

end


@parameters SW07 begin  
    ctou=.025
    clandaw=1.5
    cg=0.18
    curvp=10
    curvw=10
    
    calfa=.24
    # cgamma=1.004
    # cbeta=.9995
    csigma=1.5
    # cpie=1.005
    cfc=1.5
    cgy=0.51
    
    csadjcost= 6.0144
    chabb=    0.6361    
    cprobw=   0.8087
    csigl=    1.9423
    cprobp=   0.6
    cindw=    0.3243
    cindp=    0.47
    czcap=    0.2696
    crpi=     1.488
    crr=      0.8762
    cry=      0.0593
    crdy=     0.2347
    
    crhoa=    0.9977
    crhob=    0.5799
    crhog=    0.9957
    crhols=   0.9928
    crhoqs=   0.7165
    crhoas=1 
    crhoms=0
    crhopinf=0
    crhow=0
    cmap = 0
    cmaw  = 0
    
    clandap=cfc
    cbetabar=cbeta*cgamma^(-csigma)
    cr=cpie/(cbeta*cgamma^(-csigma))
    crk=(cbeta^(-1))*(cgamma^csigma) - (1-ctou)
    cw = (calfa^calfa*(1-calfa)^(1-calfa)/(clandap*crk^calfa))^(1/(1-calfa))
    cikbar=(1-(1-ctou)/cgamma)
    cik=(1-(1-ctou)/cgamma)*cgamma
    clk=((1-calfa)/calfa)*(crk/cw)
    cky=cfc*(clk)^(calfa-1)
    ciy=cik*cky
    ccy=1-cg-cik*cky
    crkky=crk*cky
    cwhlc=(1/clandaw)*(1-calfa)/calfa*crk*cky/ccy
    cwly=1-crk*cky
    
    conster=(cr-1)*100
    # ctrend=(cgamma-1)*100
    ctrend=(1.004-1)*100
    # constepinf=(cpie-1)*100
    constepinf=(1.005-1)*100

    cpie=1+constepinf/100
    cgamma=1+ctrend/100 

    cbeta=1/(1+constebeta/100)
    constebeta = 100 / .9995 - 100

    constelab=0

    z_ea = 0.4618
    z_eb = 1.8513
    z_eg = 0.6090
    z_eqs = 0.6017
    z_em = 0.2397
    z_epinf = 0.1455
    z_ew = 0.2089
end

# load data
dat = CSV.read(joinpath(pkgdir(MacroModelling), "test", "data", "usmodel.csv"), DataFrame)
data = KeyedArray(Array(dat)',Variable = Symbol.(strip.(names(dat))), Time = 1:size(dat)[1])

# declare observables
observables = [:dy, :dc, :dinve, :labobs, :pinfobs, :dw, :robs]

# Subsample from 1966Q1 - 2004Q4
# subset observables in data
data = data(observables,75:230)

# functions to map mean and standard deviations to distribution parameters



DynamicPPL.@model function SW07_loglikelihood_function(data, m, observables,fixed_parameters)
    z_ea    ~   Truncated(InverseGamma(0.1, 2.0, μσ = true),0.01,3)
    z_eb    ~   Truncated(InverseGamma(0.1, 2.0, μσ = true),0.025,5)
    z_eg    ~   Truncated(InverseGamma(0.1, 2.0, μσ = true),0.01,3)
    z_eqs   ~   Truncated(InverseGamma(0.1, 2.0, μσ = true),0.01,3)
    z_em    ~   Truncated(InverseGamma(0.1, 2.0, μσ = true),0.01,3)
    z_epinf ~   Truncated(InverseGamma(0.1, 2.0, μσ = true),0.01,3)
    z_ew    ~   Truncated(InverseGamma(0.1, 2.0, μσ = true),0.01,3)
    crhoa   ~   Truncated(Beta(0.5, 0.20, μσ = true),.01,.9999)
    crhob   ~   Truncated(Beta(0.5, 0.20, μσ = true),.01,.9999)
    crhog   ~   Truncated(Beta(0.5, 0.20, μσ = true),.01,.9999)
    crhoqs  ~   Truncated(Beta(0.5, 0.20, μσ = true),.01,.9999)
    crhoms  ~   Truncated(Beta(0.5, 0.20, μσ = true),.01,.9999)
    crhopinf~   Truncated(Beta(0.5, 0.20, μσ = true),.01,.9999)
    crhow   ~   Truncated(Beta(0.5, 0.20, μσ = true),.001,.9999)
    cmap    ~   Truncated(Beta(0.5, 0.2, μσ = true),0.01,.9999)
    cmaw    ~   Truncated(Beta(0.5, 0.2, μσ = true),0.01,.9999)
    csadjcost~  Truncated(Normal(4,1.5),2,15)
    csigma  ~   Truncated(Normal(1.50,0.375),0.25,3)
    chabb   ~   Truncated(Beta(0.7, 0.1, μσ = true),0.001,0.99)
    cprobw  ~   Truncated(Beta(0.5, 0.1, μσ = true),0.3,0.95)
    csigl   ~   Truncated(Normal(2,0.75),0.25,10)
    cprobp  ~   Truncated(Beta(0.5, 0.10, μσ = true),0.5,0.95)
    cindw   ~   Truncated(Beta(0.5, 0.15, μσ = true),0.01,0.99)
    cindp   ~   Truncated(Beta(0.5, 0.15, μσ = true),0.01,0.99)
    czcap   ~   Truncated(Beta(0.5, 0.15, μσ = true),0.01,1)
    cfc     ~   Truncated(Normal(1.25,0.125),1.0,3)
    crpi    ~   Truncated(Normal(1.5,0.25),1.0,3)
    crr     ~   Truncated(Beta(0.75, 0.10, μσ = true),0.5,0.975)
    cry     ~   Truncated(Normal(0.125,0.05),0.001,0.5)
    crdy    ~   Truncated(Normal(0.125,0.05),0.001,0.5)
    constepinf~ Truncated(Gamma(0.625,0.1, μσ = true),0.1,2.0)
    constebeta~ Truncated(Gamma(0.25,0.1, μσ = true),0.01,2.0)
    constelab ~ Truncated(Normal(0.0,2.0),-10.0,10.0)
    ctrend  ~   Truncated(Normal(0.4,0.10),0.1,0.8)
    cgy     ~   Truncated(Normal(0.5,0.25),0.01,2.0)
    calfa   ~   Truncated(Normal(0.3,0.05),0.01,1.0)

    ctou, clandaw, cg, curvp, curvw, crhols, crhoas = fixed_parameters

    parameters_combined = [ctou,clandaw,cg,curvp,curvw,calfa,csigma,cfc,cgy,csadjcost,chabb,cprobw,csigl,cprobp,cindw,cindp,czcap,crpi,crr,cry,crdy,crhoa,crhob,crhog,crhols,crhoqs,crhoas,crhoms,crhopinf,crhow,cmap,cmaw,constelab,z_ea,z_eb,z_eg,z_eqs,z_em,z_epinf,z_ew,ctrend,constepinf,constebeta]

    if DynamicPPL.leafcontext(__context__) !== DynamicPPL.PriorContext() 
        kalman_prob = get_loglikelihood(m, data(observables), parameters_combined)

        # println(kalman_prob)
        
        DynamicPPL.@addlogprob! kalman_prob 
    end
end


SW07.parameter_values[indexin([:crhoms, :crhopinf, :crhow, :cmap, :cmaw],SW07.parameters)] .= 0.02

fixed_parameters = SW07.parameter_values[indexin([:ctou,:clandaw,:cg,:curvp,:curvw,:crhols,:crhoas],SW07.parameters)]

SW07_loglikelihood = SW07_loglikelihood_function(data, SW07, observables, fixed_parameters)

###############################################################################
# inference using Pigeons
###############################################################################

# generate a Pigeons log potential
sw07_lp = TuringLogPotential(SW07_loglikelihood)

# find a feasible starting point
pt = pigeons(target = sw07_lp, n_rounds = 0, n_chains = 1)
replica = pt.replicas[end]
xmax = deepcopy(replica.state)
lpmax = sw07_lp(xmax)
i = 0
while !isfinite(lpmax) && i < 100
    Pigeons.sample_iid!(sw07_lp, replica, pt.shared)
    new_lp = sw07_lp(replica.state)
    if new_lp > lpmax
        lpmax = new_lp
        xmax  = deepcopy(replica.state)
    end
    i += 1
end

# define a specific initialization for this model
Pigeons.initialization(::TuringLogPotential{typeof(SW07_loglikelihood)}, ::AbstractRNG, ::Int64) = deepcopy(xmax)

# run Pigeons
pt = pigeons(target = sw07_lp, n_chains = 4, n_rounds = 4, multithreaded=true)

thorek1 · 2024-02-20T23:34:24Z

neat. thanks a lot! i will let it run on a couple of use cases over night.

any chance of not having to go that route in the future? as in, not using the context code within the turing model and not needing to find the feasible point first?

miguelbiron · 2024-02-20T23:41:36Z

No worries, glad I could help.

Regarding your questions:

At the moment, checking context is required when manually adding logprob, as this is the only way we can automatically get both prior and joint log densities from the same Turing model. It is up to the DynamicPPL folks to improve/simplify this interface.
It would certainly be interesting to have a more robust initialization routine for Turing models, that can handle models where the support of the likelihood is strictly smaller than the one for the prior. We'll discuss this with the team to see if there are ways to improve.

alexandrebouchard · 2024-02-20T23:55:31Z

Nice!!

Creating a new issue for 2... (since it is a bit more general than slice sampling / Turing)

thorek1 · 2024-02-21T21:51:52Z

so far the proposed fixes work, rarely, as in most of the times i get segfaults. I am using checkpoints, am setting a seed, and will, time allowing, investigate why this still fails (on linux, with chains>1 and multithreading). any hints or best practices are most welcome

alexandrebouchard · 2024-02-21T22:26:40Z

Sorry to hear. Feel free to send us instructions to see if we can replicate these segfaults for investigation.

Based on an earlier message, I was under the impression that segfault was occurring even in the single thread context. Is this still true or they only occur in a multi-thread context?

thorek1 · 2024-02-21T22:38:37Z

collecting evidence...
setting a seed, using 1 chain and multithreading leads to non-reproducible segfaults

checking without multithreading now

thorek1 · 2024-02-21T22:44:54Z

here is what i am running (with this data file: usmodel.csv):

using MKL
# using LinearAlgebra
# BLAS.get_config()
using MacroModelling
using Serialization
using StatsPlots
import Turing
import Turing: NUTS, PG, IS, sample, logpdf, Truncated#, Normal, Beta, Gamma, InverseGamma,
using Random, CSV, DataFrames, Zygote, AxisKeys, MCMCChains
# using ComponentArrays, Optimization, OptimizationNLopt, OptimizationOptimisers
import DynamicPPL: logjoint
import DynamicPPL
import ChainRulesCore: @ignore_derivatives, ignore_derivatives
import Pigeons
import Random
Random.seed!(1)

println("Threads used: ", Threads.nthreads())

smpler = "pigeons" #
smple = "original" #
mdl = "linear" # 
chns = 1 # 

if smple == "original"
    smpl = "1966Q1-2004Q4"
    sample_idx = 75:230
    dat = CSV.read("usmodel.csv", DataFrame)
end
    
# define callback
# Define the path for the CSV file
csv_file_path = "sw07_$(mdl)_$(smpler)_$(smpl)_samples.csv"

# Initialize a DataFrame to store the data
df = DataFrame(iteration = Float64[])

function callback(rng, model, sampler, sample, state, i; kwargs...)
    # Prepare a row for the DataFrame
    row = Dict("iteration" => Float64(i))
    for (name, value) in sample.θ
        row[string(name)] = value
    end
    
    # If the DataFrame `df` does not have columns for these names, add them
    for name in keys(row)
        if !any(name .== names(df))
            df[!, name] = Union{Missing, Any}[missing for _ in 1:nrow(df)]
        end
    end
    
    # Append the new data to the DataFrame
    push!(df, row)
    
    # Write the updated DataFrame to the CSV file
    # Note: To avoid performance issues, consider writing periodically instead of on every callback
    CSV.write(csv_file_path, df, append=true)
end


@model SW07 begin
    a[0] = calfa * rkf[0] + (1 - calfa) * (wf[0])

    zcapf[0] = (1 / (czcap / (1 - czcap))) * rkf[0]

    rkf[0] = wf[0] + labf[0] - kf[0]

    kf[0] = kpf[-1] + zcapf[0]

    invef[0] = (1 / (1 + cbetabar * cgamma)) * (invef[-1] + cbetabar * cgamma * invef[1] + (1 / (cgamma ^ 2 * csadjcost)) * pkf[0]) + qs[0]

    pkf[0] =  - rrf[0] + (1 / ((1 - chabb / cgamma) / (csigma * (1 + chabb / cgamma)))) * b[0] + (crk / (crk + (1 - ctou))) * rkf[1] + ((1 - ctou) / (crk + (1 - ctou))) * pkf[1]

    cf[0] = (chabb / cgamma) / (1 + chabb / cgamma) * cf[-1] + (1 / (1 + chabb / cgamma)) * cf[1] + ((csigma - 1) * cwhlc / (csigma * (1 + chabb / cgamma))) * (labf[0] - labf[1]) - (1 - chabb / cgamma) / (csigma * (1 + chabb / cgamma)) * (rrf[0]) + b[0]

    yf[0] = ccy * cf[0] + ciy * invef[0] + g[0] + crkky * zcapf[0]

    yf[0] = cfc * (calfa * kf[0] + (1 - calfa) * labf[0] + a[0])

    wf[0] = csigl * labf[0]	 + (1 / (1 - chabb / cgamma)) * cf[0] - (chabb / cgamma) / (1 - chabb / cgamma) * cf[-1]

    kpf[0] = (1 - cikbar) * kpf[-1] + (cikbar) * invef[0] + (cikbar) * (cgamma ^ 2 * csadjcost) * qs[0]

    mc[0] = calfa * rk[0] + (1 - calfa) * (w[0]) - a[0]

    zcap[0] = (1 / (czcap / (1 - czcap))) * rk[0]

    rk[0] = w[0] + lab[0] - k[0]

    k[0] = kp[-1] + zcap[0]

    inve[0] = (1 / (1 + cbetabar * cgamma)) * (inve[-1] + cbetabar * cgamma * inve[1] + (1 / (cgamma ^ 2 * csadjcost)) * pk[0]) + qs[0]

    pk[0] =  - r[0] + pinf[1] + (1 / ((1 - chabb / cgamma) / (csigma * (1 + chabb / cgamma)))) * b[0] + (crk / (crk + (1 - ctou))) * rk[1] + ((1 - ctou) / (crk + (1 - ctou))) * pk[1]

    c[0] = (chabb / cgamma) / (1 + chabb / cgamma) * c[-1] + (1 / (1 + chabb / cgamma)) * c[1] + ((csigma - 1) * cwhlc / (csigma * (1 + chabb / cgamma))) * (lab[0] - lab[1]) - (1 - chabb / cgamma) / (csigma * (1 + chabb / cgamma)) * (r[0] - pinf[1]) + b[0]

    y[0] = ccy * c[0] + ciy * inve[0] + g[0] + crkky * zcap[0]

    y[0] = cfc * (calfa * k[0] + (1 - calfa) * lab[0] + a[0])

    pinf[0] = (1 / (1 + cbetabar * cgamma * cindp)) * (cbetabar * cgamma * pinf[1] + cindp * pinf[-1] + ((1 - cprobp) * (1 - cbetabar * cgamma * cprobp) / cprobp) / ((cfc - 1) * curvp + 1) * (mc[0])) + spinf[0]

    w[0] = (1 / (1 + cbetabar * cgamma)) * w[-1] + (cbetabar * cgamma / (1 + cbetabar * cgamma)) * w[1] + (cindw / (1 + cbetabar * cgamma)) * pinf[-1] - (1 + cbetabar * cgamma * cindw) / (1 + cbetabar * cgamma) * pinf[0] + (cbetabar * cgamma) / (1 + cbetabar * cgamma) * pinf[1] + (1 - cprobw) * (1 - cbetabar * cgamma * cprobw) / ((1 + cbetabar * cgamma) * cprobw) * (1 / ((clandaw - 1) * curvw + 1)) * (csigl * lab[0] + (1 / (1 - chabb / cgamma)) * c[0] - ((chabb / cgamma) / (1 - chabb / cgamma)) * c[-1] - w[0]) + sw[0]

    r[0] = crpi * (1 - crr) * pinf[0] + cry * (1 - crr) * (y[0] - yf[0]) + crdy * (y[0] - yf[0] - y[-1] + yf[-1]) + crr * r[-1] + ms[0]

    a[0] = crhoa * a[-1] + z_ea * ea[x]

    b[0] = crhob * b[-1] + z_eb * eb[x]

    g[0] = crhog * g[-1] + z_eg * eg[x] + cgy * z_ea * ea[x]

    qs[0] = crhoqs * qs[-1] + z_eqs * eqs[x]

    ms[0] = crhoms * ms[-1] + z_em * em[x]

    spinf[0] = crhopinf * spinf[-1] + epinfma[0] - cmap * epinfma[-1]

    epinfma[0] = z_epinf * epinf[x]

    sw[0] = crhow * sw[-1] + ewma[0] - cmaw * ewma[-1]

    ewma[0] = z_ew * ew[x]

    kp[0] = (1 - cikbar) * kp[-1] + cikbar * inve[0] + cikbar * cgamma ^ 2 * csadjcost * qs[0]

    dy[0] = y[0] - y[-1] + ctrend

    dc[0] = c[0] - c[-1] + ctrend

    dinve[0] = inve[0] - inve[-1] + ctrend

    dw[0] = w[0] - w[-1] + ctrend

    pinfobs[0] = (pinf[0]) + constepinf

    robs[0] = (r[0]) + conster

    labobs[0] = lab[0] + constelab

end


@parameters SW07 begin  
    ctou=.025
    clandaw=1.5
    cg=0.18
    curvp=10
    curvw=10
    
    calfa=.24
    # cgamma=1.004
    # cbeta=.9995
    csigma=1.5
    # cpie=1.005
    cfc=1.5
    cgy=0.51
    
    csadjcost= 6.0144
    chabb=    0.6361    
    cprobw=   0.8087
    csigl=    1.9423
    cprobp=   0.6
    cindw=    0.3243
    cindp=    0.47
    czcap=    0.2696
    crpi=     1.488
    crr=      0.8762
    cry=      0.0593
    crdy=     0.2347
    
    crhoa=    0.9977
    crhob=    0.5799
    crhog=    0.9957
    crhols=   0.9928
    crhoqs=   0.7165
    crhoas=1 
    crhoms=0
    crhopinf=0
    crhow=0
    cmap = 0
    cmaw  = 0
    
    clandap=cfc
    cbetabar=cbeta*cgamma^(-csigma)
    cr=cpie/(cbeta*cgamma^(-csigma))
    crk=(cbeta^(-1))*(cgamma^csigma) - (1-ctou)
    cw = (calfa^calfa*(1-calfa)^(1-calfa)/(clandap*crk^calfa))^(1/(1-calfa))
    cikbar=(1-(1-ctou)/cgamma)
    cik=(1-(1-ctou)/cgamma)*cgamma
    clk=((1-calfa)/calfa)*(crk/cw)
    cky=cfc*(clk)^(calfa-1)
    ciy=cik*cky
    ccy=1-cg-cik*cky
    crkky=crk*cky
    cwhlc=(1/clandaw)*(1-calfa)/calfa*crk*cky/ccy
    cwly=1-crk*cky
    
    conster=(cr-1)*100
    # ctrend=(cgamma-1)*100
    ctrend=(1.004-1)*100
    # constepinf=(cpie-1)*100
    constepinf=(1.005-1)*100

    cpie=1+constepinf/100
    cgamma=1+ctrend/100 

    cbeta=1/(1+constebeta/100)
    constebeta = 100 / .9995 - 100

    constelab=0

    z_ea = 0.4618
    z_eb = 1.8513
    z_eg = 0.6090
    z_eqs = 0.6017
    z_em = 0.2397
    z_epinf = 0.1455
    z_ew = 0.2089
end

# load data
data = KeyedArray(Array(dat)',Variable = Symbol.(strip.(names(dat))), Time = 1:size(dat)[1])

# declare observables
observables = [:dy, :dc, :dinve, :labobs, :pinfobs, :dw, :robs]

# Subsample
# subset observables in data
data = data(observables, sample_idx)

# functions to map mean and standard deviations to distribution parameters
Turing.@model function SW07_loglikelihood_function(data, m, observables,fixed_parameters)
    z_ea    ~   Truncated(InverseGamma(0.1, 2.0, μσ = true),0.01,3)
    z_eb    ~   Truncated(InverseGamma(0.1, 2.0, μσ = true),0.025,5)
    z_eg    ~   Truncated(InverseGamma(0.1, 2.0, μσ = true),0.01,3)
    z_eqs   ~   Truncated(InverseGamma(0.1, 2.0, μσ = true),0.01,3)
    z_em    ~   Truncated(InverseGamma(0.1, 2.0, μσ = true),0.01,3)
    z_epinf ~   Truncated(InverseGamma(0.1, 2.0, μσ = true),0.01,3)
    z_ew    ~   Truncated(InverseGamma(0.1, 2.0, μσ = true),0.01,3)
    crhoa   ~   Truncated(Beta(0.5, 0.20, μσ = true),.01,.9999)
    crhob   ~   Truncated(Beta(0.5, 0.20, μσ = true),.01,.9999)
    crhog   ~   Truncated(Beta(0.5, 0.20, μσ = true),.01,.9999)
    crhoqs  ~   Truncated(Beta(0.5, 0.20, μσ = true),.01,.9999)
    crhoms  ~   Truncated(Beta(0.5, 0.20, μσ = true),.01,.9999)
    crhopinf~   Truncated(Beta(0.5, 0.20, μσ = true),.01,.9999)
    crhow   ~   Truncated(Beta(0.5, 0.20, μσ = true),.001,.9999)
    cmap    ~   Truncated(Beta(0.5, 0.2, μσ = true),0.01,.9999)
    cmaw    ~   Truncated(Beta(0.5, 0.2, μσ = true),0.01,.9999)
    csadjcost~  Truncated(Normal(4,1.5),2,15)
    csigma  ~   Truncated(Normal(1.50,0.375),0.25,3)
    chabb   ~   Truncated(Beta(0.7, 0.1, μσ = true),0.001,0.99)
    cprobw  ~   Truncated(Beta(0.5, 0.1, μσ = true),0.3,0.95)
    csigl   ~   Truncated(Normal(2,0.75),0.25,10)
    cprobp  ~   Truncated(Beta(0.5, 0.10, μσ = true),0.5,0.95)
    cindw   ~   Truncated(Beta(0.5, 0.15, μσ = true),0.01,0.99)
    cindp   ~   Truncated(Beta(0.5, 0.15, μσ = true),0.01,0.99)
    czcap   ~   Truncated(Beta(0.5, 0.15, μσ = true),0.01,1)
    cfc     ~   Truncated(Normal(1.25,0.125),1.0,3)
    crpi    ~   Truncated(Normal(1.5,0.25),1.0,3)
    crr     ~   Truncated(Beta(0.75, 0.10, μσ = true),0.5,0.975)
    cry     ~   Truncated(Normal(0.125,0.05),0.001,0.5)
    crdy    ~   Truncated(Normal(0.125,0.05),0.001,0.5)
    constepinf~ Truncated(Gamma(0.625,0.1, μσ = true),0.1,2.0)
    constebeta~ Truncated(Gamma(0.25,0.1, μσ = true),0.01,2.0)
    constelab ~ Truncated(Normal(0.0,2.0),-10.0,10.0)
    ctrend  ~   Truncated(Normal(0.4,0.10),0.1,0.8)
    cgy     ~   Truncated(Normal(0.5,0.25),0.01,2.0)
    calfa   ~   Truncated(Normal(0.3,0.05),0.01,1.0)


    if DynamicPPL.leafcontext(__context__) !== DynamicPPL.PriorContext() 
        ctou, clandaw, cg, curvp, curvw, crhols, crhoas = fixed_parameters

        parameters_combined = [ctou,clandaw,cg,curvp,curvw,calfa,csigma,cfc,cgy,csadjcost,chabb,cprobw,csigl,cprobp,cindw,cindp,czcap,crpi,crr,cry,crdy,crhoa,crhob,crhog,crhols,crhoqs,crhoas,crhoms,crhopinf,crhow,cmap,cmaw,constelab,z_ea,z_eb,z_eg,z_eqs,z_em,z_epinf,z_ew,ctrend,constepinf,constebeta]

        kalman_prob = get_loglikelihood(m, data(observables), parameters_combined)

        Turing.@addlogprob! kalman_prob
    end
end

SW07.parameter_values[indexin([:crhoms, :crhopinf, :crhow, :cmap, :cmaw],SW07.parameters)] .= 0.02

fixed_parameters = SW07.parameter_values[indexin([:ctou,:clandaw,:cg,:curvp,:curvw,:crhols,:crhoas],SW07.parameters)]

dir_name = "sw07_$(mdl)_$(smpler)_$(smpl)_samples_$(chns)_chains"

if !isdir(dir_name) mkdir(dir_name) end

cd(dir_name)

println("Current working directory: ", pwd())

SW07_loglikelihood = SW07_loglikelihood_function(data, SW07, observables, fixed_parameters)

if smpler == "pigeons"
    # generate a Pigeons log potential
    sw07_lp = Pigeons.TuringLogPotential(SW07_loglikelihood)

    # find a feasible starting point
    pt = Pigeons.pigeons(target = sw07_lp, n_rounds = 0, n_chains = 1)
    
    replica = pt.replicas[end]
    XMAX = deepcopy(replica.state)
    LPmax = sw07_lp(XMAX)

    i = 0

    while !isfinite(LPmax) && i < 1000
        Pigeons.sample_iid!(sw07_lp, replica, pt.shared)
        new_LP = sw07_lp(replica.state)
        if new_LP > LPmax
            LPmax = new_LP
            XMAX  = deepcopy(replica.state)
        end
        i += 1
    end
    
    # define a specific initialization for this model
    Pigeons.initialization(::Pigeons.TuringLogPotential{typeof(SW07_loglikelihood)}, ::AbstractRNG, ::Int64) = deepcopy(XMAX)

    pt = Pigeons.pigeons(target = sw07_lp,
                checkpoint = true,
                record = [Pigeons.traces; Pigeons.round_trip; Pigeons.record_default(); Pigeons.disk],
                n_chains = chns,
                n_rounds = 10,
                multithreaded = false)

    samps = MCMCChains.Chains(Pigeons.get_sample(pt))
end

serialize("samples.jls", samps)

my_plot = StatsPlots.plot(samps)
StatsPlots.savefig(my_plot, "samples.png")

alexandrebouchard · 2024-02-21T22:48:10Z

Thank you! The manifest.toml file would be great too, to make sure we use same versions.

thorek1 · 2024-02-21T22:53:43Z

there you go
Manifest.txt
and the versioninfo()

Julia Version 1.10.1
Commit 7790d6f0641 (2024-02-13 20:41 UTC)
Build Info:
  Official https://julialang.org/ release
Platform Info:
  OS: Linux (x86_64-linux-gnu)
  CPU: 32 × AMD EPYC 7R13 Processor
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-15.0.7 (ORCJIT, znver3)

thorek1 · 2024-02-21T22:59:37Z

and i can now confirm single threaded runs also fail with segfaults and are not reproducible with the seed (they fail at different rounds)

alexandrebouchard mentioned this issue Feb 21, 2024

Better support for distributed initialization #210

Closed

thorek1 mentioned this issue Feb 21, 2024

change defaults: @addlogprob! not included in PriorContext() TuringLang/DynamicPPL.jl#580

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Error: SliceSampler ... shoud be initialized in the support #209

Error: SliceSampler ... shoud be initialized in the support #209

thorek1 commented Feb 20, 2024 •

edited

Loading

alexandrebouchard commented Feb 20, 2024

thorek1 commented Feb 20, 2024 •

edited

Loading

thorek1 commented Feb 20, 2024 •

edited

Loading

alexandrebouchard commented Feb 20, 2024

thorek1 commented Feb 20, 2024

miguelbiron commented Feb 20, 2024 •

edited

Loading

thorek1 commented Feb 20, 2024

nikola-sur commented Feb 20, 2024

thorek1 commented Feb 20, 2024

alexandrebouchard commented Feb 20, 2024

miguelbiron commented Feb 20, 2024

thorek1 commented Feb 20, 2024

miguelbiron commented Feb 20, 2024

alexandrebouchard commented Feb 20, 2024 •

edited

Loading

thorek1 commented Feb 21, 2024

alexandrebouchard commented Feb 21, 2024

thorek1 commented Feb 21, 2024

thorek1 commented Feb 21, 2024

alexandrebouchard commented Feb 21, 2024

thorek1 commented Feb 21, 2024 •

edited

Loading

thorek1 commented Feb 21, 2024 •

edited

Loading

Error: SliceSampler ... shoud be initialized in the support #209

Error: SliceSampler ... shoud be initialized in the support #209

Comments

thorek1 commented Feb 20, 2024 • edited Loading

alexandrebouchard commented Feb 20, 2024

thorek1 commented Feb 20, 2024 • edited Loading

thorek1 commented Feb 20, 2024 • edited Loading

alexandrebouchard commented Feb 20, 2024

thorek1 commented Feb 20, 2024

miguelbiron commented Feb 20, 2024 • edited Loading

thorek1 commented Feb 20, 2024

nikola-sur commented Feb 20, 2024

thorek1 commented Feb 20, 2024

alexandrebouchard commented Feb 20, 2024

miguelbiron commented Feb 20, 2024

thorek1 commented Feb 20, 2024

miguelbiron commented Feb 20, 2024

alexandrebouchard commented Feb 20, 2024 • edited Loading

thorek1 commented Feb 21, 2024

alexandrebouchard commented Feb 21, 2024

thorek1 commented Feb 21, 2024

thorek1 commented Feb 21, 2024

alexandrebouchard commented Feb 21, 2024

thorek1 commented Feb 21, 2024 • edited Loading

thorek1 commented Feb 21, 2024 • edited Loading

thorek1 commented Feb 20, 2024 •

edited

Loading

thorek1 commented Feb 20, 2024 •

edited

Loading

thorek1 commented Feb 20, 2024 •

edited

Loading

miguelbiron commented Feb 20, 2024 •

edited

Loading

alexandrebouchard commented Feb 20, 2024 •

edited

Loading

thorek1 commented Feb 21, 2024 •

edited

Loading

thorek1 commented Feb 21, 2024 •

edited

Loading