MethodError: no method matching percival(::Val{:tron}, ...) when trying basic box-constrained optimization

[ForceBru]

Code

import Pkg
Pkg.activate(temp=true, io=devnull)
Pkg.add([
    (name="Percival", version="0.5.2"),
    (name="Nonconvex", version="2.1.1"),
    (name="NonconvexPercival", version="0.1.3"),
], io=devnull)
Pkg.status()


import Nonconvex as NC; using NonconvexPercival

component_pdf(x::Real, mu::Real, tau::Real) =
    tau * exp(-tau^2 * (x - mu)^2 / 2) / sqrt(2π)

data = randn(300)

loss(params::AbstractVector) = -sum(log,
    (1 - params[1]) * component_pdf(x, 0, 1)
    +    params[1]  * component_pdf(x, params[2], params[3])
    for x in data
)

model = NC.Model(loss)
NC.addvar!(model, [0, -Inf, 0], [1, Inf, Inf], init=[0.1, 0, 1])

@info "Begin optimization..."
res = optimize(
    model, AugLag(), [0.1, -3, 1],
    options=AugLagOptions(), convcriteria=NC.KKTCriteria()
)
@info "End optimization" res.minimizer

Error message

$ julia-1.8 bug_nonconvex.jl
Status `/private/var/folders/ys/3h0gnqns4b98zb66_vl_m35m0000gn/T/jl_4nXQJl/Project.toml`
  [01bcebdf] Nonconvex v2.1.1
  [4296f080] NonconvexPercival v0.1.3
  [01435c0c] Percival v0.5.2
[ Info: Begin optimization...
ERROR: LoadError: MethodError: no method matching percival(::Val{:tron}, ::ADNLPModels.ADNLPModel{Float64, Vector{Float64}, Vector{Int64}}; max_iter=1000, max_time=Inf, max_eval=100000, atol=1.0e-6, rtol=1.0e-6, subsolver_logger=Base.CoreLogging.NullLogger(), subproblem_modifier=NonconvexPercival.var"#4#5"{DataType, Int64}(Float64, 5), subsolver_max_eval=200, subsolver_kwargs=Dict(:max_cgiter => 100))
Closest candidates are:
  percival(::Val{:tron}, ::NLPModels.AbstractNLPModel; max_iter, max_time, max_eval, atol, rtol, subproblem_modifier, subsolver_logger, subsolver_kwargs) at ~/.julia/packages/Percival/RpJYY/src/method.jl:17 got unsupported keyword argument "subsolver_max_eval"
  percival(::Val{:ineq}, ::NLPModels.AbstractNLPModel; kwargs...) at ~/.julia/packages/Percival/RpJYY/src/method.jl:46
  percival(::Val{:equ}, ::NLPModels.AbstractNLPModel{T, V}; μ, max_iter, max_time, max_eval, atol, rtol, ctol, subsolver_logger, inity, subproblem_modifier, subsolver_max_eval, subsolver_kwargs) where {T, V} at ~/.julia/packages/Percival/RpJYY/src/method.jl:73
Stacktrace:
 [1] kwerr(::NamedTuple{(:max_iter, :max_time, :max_eval, :atol, :rtol, :subsolver_logger, :subproblem_modifier, :subsolver_max_eval, :subsolver_kwargs), Tuple{Int64, Float64, Int64, Float64, Float64, Base.CoreLogging.NullLogger, NonconvexPercival.var"#4#5"{DataType, Int64}, Int64, Dict{Symbol, Int64}}}, ::Function, ::Val{:tron}, ::ADNLPModels.ADNLPModel{Float64, Vector{Float64}, Vector{Int64}})
   @ Base ./error.jl:165
 [2] _percival(nlp::ADNLPModels.ADNLPModel{Float64, Vector{Float64}, Vector{Int64}}; T::Type, μ::Float64, max_iter::Int64, max_time::Float64, max_eval::Int64, atol::Float64, rtol::Float64, ctol::Float64, first_order::Bool, memory::Int64, subsolver_logger::Base.CoreLogging.NullLogger, inity::Vector{Float64}, max_cgiter::Int64, subsolver_max_eval::Int64, kwargs::Base.Pairs{Symbol, Union{}, Tuple{}, NamedTuple{(), Tuple{}}})
   @ NonconvexPercival ~/.julia/packages/NonconvexPercival/Y32FW/src/NonconvexPercival.jl:86
 [3] optimize!(workspace::NonconvexPercival.PercivalWorkspace{NonconvexCore.VecModel{NonconvexCore.Objective{NonconvexCore.var"#147#149"{typeof(loss), NonconvexCore.Unflatten{Tuple{Vector{Float64}}, NonconvexCore.var"#unflatten_to_Tuple#71"{Tuple{Int64}, Tuple{Int64}, Tuple{typeof(identity)}}}}, Base.RefValue{Float64}, Set{Symbol}}, NonconvexCore.VectorOfFunctions{Vector{NonconvexCore.EqConstraint}}, NonconvexCore.VectorOfFunctions{Vector{NonconvexCore.IneqConstraint}}, NonconvexCore.VectorOfFunctions{Vector{NonconvexCore.SDConstraint}}, Vector{Float64}, Vector{Float64}, Vector{Float64}}, ADNLPModels.ADNLPModel{Float64, Vector{Float64}, Vector{Int64}}, Vector{Float64}, PercivalOptions{NamedTuple{(:first_order, :memory, :inity), Tuple{Bool, Int64, typeof(ones)}}}, Base.RefValue{Int64}})
   @ NonconvexPercival ~/.julia/packages/NonconvexPercival/Y32FW/src/NonconvexPercival.jl:55
 [4] #optimize#144
   @ ~/.julia/packages/NonconvexCore/IPKCF/src/models/vec_model.jl:63 [inlined]
 [5] optimize(::Model{Vector{Any}}, ::PercivalAlg, ::Vector{Float64}; kwargs::Base.Pairs{Symbol, Any, Tuple{Symbol, Symbol}, NamedTuple{(:options, :convcriteria), Tuple{PercivalOptions{NamedTuple{(:first_order, :memory, :inity), Tuple{Bool, Int64, typeof(ones)}}}, KKTCriteria}}})
   @ NonconvexCore ~/.julia/packages/NonconvexCore/IPKCF/src/common.jl:233
 [6] top-level scope
   @ ~/test/bug_nonconvex.jl:28
in expression starting at /Users/forcebru/test/bug_nonconvex.jl:28
$

Apparently, percival(::Val{:tron}, ...) "got unsupported keyword argument subsolver_max_eval".

Versions

• Julia 1.8.1
• Percival 0.5.2
• Nonconvex 2.1.1
• NonconvexPercival 0.1.3

[mohamed82008]

Hi, sorry for the late reply. Just saw this. The issue is now fixed and tested in master. I will make a new release shortly.