[jacquesm]

You definitely want to tee the process before you start doing any ML artifact correction to make sure that it actually is an improvement.