I am developing a machine learning pipeline with kubeflow on the GCP Vertex AI platform. At one step, I scale my training data using the MinMaxScaler from the scikit learn package. I need to out that scaler in order to import it into a custom component further down the pipeline. Does that scaler object need to be output as an artifact or a model? Or something else?
When I output a data set from a component I use the command df.to_pickle(output_dataset_name) based on the output name parameter in the function of the custom component. How would I do that for a sk-learn object?