Scalar indexing when using GPU

[Moelf]

Warning: Performing scalar indexing on task Task (runnable, started) @0x00007fe8fe4968b0.
│ Invocation of getindex resulted in scalar indexing of a GPU array.
│ This is typically caused by calling an iterating implementation of a method.
│ Such implementations *do not* execute on the GPU, but very slowly on the CPU,
│ and therefore should be avoided.
│ 
│ If you want to allow scalar iteration, use `allowscalar` or `@allowscalar`
│ to enable scalar iteration globally or for the operations in question.
└ @ GPUArraysCore ~/.julia/packages/GPUArraysCore/aNaXo/src/GPUArraysCore.jl:145

The training setup looks something like:

evotree_config = EvoTreeClassifier(;
        nrounds=num_round,
        eta=η,
        max_depth,
        min_weight=3.0,
        nbins=128,
    )

fit_evotree(evotree_config, dtrain;
            target_name="proc",
            fnames=string.(input_names),
            w_name="wgts",
            deval,
            metric=:logloss,
            print_every_n=5,
            verbosity=1,
            early_stopping_rounds, device="gpu")

[Moelf]

metric=:logloss,

this is the problem, for classification, it needs to be :mlogloss -- perhaps this should throw an error?

[jeremiedb]

Yes I agree that it should have thrown an error.
Upcomnig release for v0.17 should fix this by having both the loss and metric baked right into in EvoTreeClassifier:

EvoTrees.jl/src/learners.jl

Lines 238 to 239 in 835d8a4

	loss = :mlogloss
	metric = :mlogloss

I'm hoping to have it release by next week.