Saving models to disk

JLD.jl is the preferred way of saving ScikitLearn.jl models. If you also use Python models (via @sk_import), you will have to import PyCallJLD as well.

using PyCall, JLD, PyCallJLD
using ScikitLearn
using ScikitLearn.Pipelines
@sk_import decomposition: PCA
@sk_import linear_model: LinearRegression

pca = PCA()
lm = LinearRegression()

X=rand(10, 3); y=rand(10);

pip = Pipeline([("PCA", pca), ("LinearRegression", lm)])
fit!(pip, X, y)   # fit to some dataset

JLD.save("pipeline.jld", "pip", pip)
pip = JLD.load("pipeline.jld", "pip") # Load back the pipeline