using Statistics, Flux loss(X, y) = Flux.Losses.logitcrossentropy(model(X), y) accuracy(X, y) = Statistics.mean(Flux.onecold(model(X)) .== Flux.onecold(y))