Save/load generated surrogate model

[jacktang]

Hi,

I want to save/load the surrogate model using JLD2 and BSON, and find the packages can only serialize objects rather than function. So if I load the saved model, it throws error:

UndefVarError: XGBoost not defined

Stacktrace:
  [1] (::BSON.var"#31#32")(m::Module, f::String)
    @ BSON ~/.julia/packages/BSON/rOaki/src/extensions.jl:21
  [2] BottomRF
    @ ./reduce.jl:81 [inlined]
  [3] _foldl_impl(op::Base.BottomRF{BSON.var"#31#32"}, init::Module, itr::Vector{Any})
    @ Base ./reduce.jl:58
  [4] foldl_impl
    @ ./reduce.jl:48 [inlined]
  [5] mapfoldl_impl
    @ ./reduce.jl:44 [inlined]
  [6] _mapreduce_dim
    @ ./reducedim.jl:327 [inlined]
  [7] #mapreduce#725
    @ ./reducedim.jl:322 [inlined]
  [8] #reduce#727
    @ ./reducedim.jl:371 [inlined]
  [9] resolve(fs::Vector{Any}, init::Module)
    @ BSON ~/.julia/packages/BSON/rOaki/src/extensions.jl:21
 [10] (::BSON.var"#35#36")(d::Dict{Symbol, Any}, init::Module)
    @ BSON ~/.julia/packages/BSON/rOaki/src/extensions.jl:64
 [11] _raise_recursive(d::Dict{Symbol, Any}, cache::IdDict{Any, Any}, init::Module)
    @ BSON ~/.julia/packages/BSON/rOaki/src/read.jl:80

BTW: the question on julialang discourse: https://discourse.julialang.org/t/save-and-load-the-model-generated-by-surrogates-jl/83302

[vikram-s-narayan]

Can you please share your full minimum working example so that we can examine it more carefully? Also, can you try with some other surrogate like Radial Basis?

[jacktang]

@vikram-s-narayan For example the random forrest surrogate model: https://surrogates.sciml.ai/dev/randomforest/

using Surrogates
using SurrogatesRandomForest
using Plots
using JLD2
default()

f(x) = sin(x) + sin(10/3 * x)
n_samples = 80
lower_bound = 2.7
upper_bound = 7.5
x = sample(n_samples, lower_bound, upper_bound, SobolSample())
y = f.(x)
scatter(x, y, label="Sampled points", xlims=(lower_bound, upper_bound))
plot!(f, label="True function", xlims=(lower_bound, upper_bound), legend=:top)

num_round = 100
randomforest_surrogate = RandomForestSurrogate(x ,y ,lower_bound, upper_bound, num_round = num_round)
plot(x, y, seriestype=:scatter, label="Sampled points", xlims=(lower_bound, upper_bound), legend=:top)
plot!(f, label="True function",  xlims=(lower_bound, upper_bound), legend=:top)
plot!(randomforest_surrogate, label="Surrogate function",  xlims=(lower_bound, upper_bound), legend=:top)

save_object("/tmp/rf_model.jld", randomforest_surrogate)
sm = load_object("/tmp/rf_model.jld")
sm.(3.5)

PolynomialChaos model works fine, so I think Radial basis should be ok too.

[vikram-s-narayan]

Thanks. With the code you provided the error reads - "ERROR: Call to XGBoost C function XGBoosterPredict failed: ". This appears to be a known issue with XGBoost. One of the package dependencies in SurrogatesRandomForest is XGBoost.

[jacktang]

@vikram-s-narayan If I understand it correctly, the solution is replacing XGBoost by EvoTrees.jl in Surrogates.jl?

[vikram-s-narayan]

@jacktang - I'm not sure if there are any other dependencies involved and whether the function calls need to be changed.