AdaBoostStumpClassifier MethodError: zero(::Type{Symbol})

The function fit! fails with number of iterations > 5.

bdt = let
    _model = AdaBoostStumpClassifier(; n_iterations = 10) 
    fit!(_model, X_train, y_train)
end

fails with an error,

MethodError: no method matching zero(::Type{Symbol})

The function `zero` exists, but no method is defined for this combination of argument types.

Closest candidates are:
  zero(::Type{Union{}}, Any...)
   @ Base number.jl:310
  zero(::Type{Dates.DateTime})
   @ Dates ~/.julia/juliaup/julia-1.11.5+0.aarch64.apple.darwin14/share/julia/stdlib/v1.11/Dates/src/types.jl:458
  zero(::Type{Pkg.Resolve.VersionWeight})
   @ Pkg ~/.julia/juliaup/julia-1.11.5+0.aarch64.apple.darwin14/share/julia/stdlib/v1.11/Pkg/src/Resolve/versionweights.jl:15
  ...

It depends on dataset to train, see MWE, it works on one set, fails on the other

MWE

begin
    using Random
    using DataFrames
    using DecisionTree
    Random.seed!(1234)
end

function classify_signal_background(x, y)
    # Sinusoidal boundary
    # if sin(2.5π * (x - 0.55)) / 5 + 0.3 + 0.4x < y < 0.7 + 0.4x # note: this one has no problem
    if (x-0.25)^2 + (y-0.25)^2 < 0.05 || (x-0.65)^2 + (y-0.65)^2 < 0.05
        return :signal
    else
        return :background
    end
end

const features = [:f1, :f2];

df = let
    _df = DataFrame(rand(500, 2), features)
    transform!(_df, features => ByRow(classify_signal_background) => :y)
end

bdt = let
    _model = AdaBoostStumpClassifier(; n_iterations = 40)
	X_train = df[:,features] |> Matrix
    y_train = df[:, :y]
    fit!(_model,X_train, y_train)
end

Start at the AdaBoostStumpClassifier implementation and its fit! method, then run the supplied Julia MWE with both datasets and iteration counts. Trace the failing path that produces zero(::Type{Symbol}); done means fitting succeeds for the reproduced labeled data with n_iterations greater than 5.