precompile_funcs.jl

module PrecompileFuncs
    using FluxDash
    using CSV, DataFrames
    using FluxDash
    using FluxDash.Flux
    using Random
    function precompile_nn_train()
        df = CSV.read("FluxDash/datasets/real_estate.csv", DataFrame)

        n_row = nrow(df)
        df = df[shuffle(1:n_row)[:], :] #shuffle data

        labels = names(df)
        in_labels = labels[end]
        out_labels = labels[end]

        X_train, X_test, y_train, y_test = FluxDash.format_nn_data(
            df, in_labels, out_labels; training_percent=75
        ) 

        chn = Base.Channel{Tuple{Float64, Int, Chain}}(Inf)

        hidden_outs = [2]
        hidden_activations =["relu"]
        FluxDash.train_nn(
            X_train, y_train, in_labels, out_labels, hidden_outs,
            chn, hidden_activations; training_percent = 75,
            epoc = 4, opt = "adam", eta = 0.1
        )
        return "success"
    end
end