Physically training the models isn't the hard part. It knowing why you using x method, how to interpret the result, and what dials to turn that's the hard part. And there's no automating that.