MultitargetNeuralNetworkRegressor doc example doesn't work as intended

[EssamWisam]

I was trying to run the following example available in the documentation

import Pkg; Pkg.activate("./scratchpad")

using MLJ
import MLJFlux
using Flux
import Optimisers

X, y = make_regression(100, 9; n_targets = 2) # both tables
schema(y)
schema(X)

(X, Xtest), (y, ytest) = partition((X, y), 0.7, multi=true);

builder = MLJFlux.@builder begin
    init=Flux.glorot_uniform(rng)
    Chain(
        Dense(n_in, 64, relu, init=init),
        Dense(64, 32, relu, init=init),
        Dense(32, n_out, init=init),
    )
end

MultitargetNeuralNetworkRegressor = @load MultitargetNeuralNetworkRegressor pkg = MLJFlux
model = MultitargetNeuralNetworkRegressor(builder=builder, rng=123, epochs=20)

pipe = Standardizer |> TransformedTargetModel(model, transformer=Standardizer)

mach = machine(pipe, X, y)
fit!(mach, verbosity=2)

# first element initial loss, 2:end per epoch training losses
report(mach).transformed_target_model_deterministic.model.training_losses

# custom MLJ loss:
multi_loss(yhat, y) = l2(MLJ.matrix(yhat), MLJ.matrix(y))

# CV estimate, based on `(X, y)`:
evaluate!(mach, resampling=CV(nfolds=5), measure=multi_loss)

# loss for `(Xtest, test)`:
fit!(mach) # trains on all data `(X, y)`
yhat = predict(mach, Xtest)
multi_loss(yhat, ytest)

Error log is so long but here is a snippet:

ERROR: MethodError: no method matching abs(::Vector{Float64})

Closest candidates are:
  abs(::Missing)
   @ Base missing.jl:101
  abs(::Bool)
   @ Base bool.jl:153
  abs(::Pkg.Resolve.VersionWeight)
   @ Pkg ~/.julia/juliaup/julia-1.10.4+0.aarch64.apple.darwin14/share/julia/stdlib/v1.10/Pkg/src/Resolve/versionweights.jl:32
  ...

Stacktrace:
  [1] pth_power_of_absolute_difference(yhat::Vector{Float64}, y::Vector{Float64}, p::Int64)
    @ StatisticalMeasures.Functions ~/.julia/packages/StatisticalMeasures/hPDX2/src/functions.jl:12
  [2] (::StatisticalMeasures.LPLossOnScalars{Int64})(yhat::Vector{Float64}, y::Vector{Float64})
    @ StatisticalMeasures ~/.julia/packages/StatisticalMeasuresBase/7euvM/src/tools_for_implementers.jl:353
  [3] call(::StatisticalMeasuresBase.SupportsMissingsMeasure{…}, ::Vector{…}, ::Vector{…})
    @ StatisticalMeasuresBase ~/.julia/packages/StatisticalMeasuresBase/7euvM/src/supports_missings_measure.jl:21
  [4] (::StatisticalMeasuresBase.SupportsMissingsMeasure{…})(arg::Vector{…}, args::Vector{…})
    @ StatisticalMeasuresBase ~/.julia/packages/StatisticalMeasuresBase/7euvM/src/wrappers.jl:26
  [5] (::StatisticalMeasuresBase.var"#28#29"{StatisticalMeasuresBase.Multimeasure{…}, Nothing})(::Tuple{Vector{…}, Vector{…}})
    @ StatisticalMeasuresBase ~/.julia/packages/StatisticalMeasuresBase/7euvM/src/multimeasure.jl:158
  [6] _broadcast_getindex_evalf
    @ ./broadcast.jl:709 [inlined]
  [7] _broadcast_getindex
    @ ./broadcast.jl:682 [inlined]
  [8] _getindex

Which seems to stem from the following:

Possibly some missing dots?

[ablaom]

Thanks @EssamWisam for reporting this.

The docstring you quote is indeed incorrect. Since MLJBase 1.0, multitarget_l2 should be used here instead of l2. Also, there is no longer any need to convert y and yhat to matrices, as multi target measures in StatisticalMeasures support most tables.

A fix to the docstring is cross-referenced above if you wish to see the corrected example now.