Java – SBT: how to package instances of classes as jars?

My code is basically like this:

class FoodTrainer(images: S3Path) { // data is >100GB file living in S3
  def train(): FoodClassifier       // Very expensive - takes ~5 hours!
}

class FoodClassifier {          // Light-weight API class
  def isHotDog(input: Image): Boolean
}

I want to call Val classifier = new foodtrainer (s3dir) when jar assembly (SBT assembly) Train () and publish a jar with a classifier instance that is immediately available to downstream library users

What is the easiest way? What are the established examples of this? I know it is a fairly common idiom in ML projects and can publish well-trained models http://nlp.stanford.edu/software/stanford-corenlp-models-current.jar

How can I use an SBT assembly to do this without having to check large model classes or data files into my version control?

Solution

You should serialize the training data into your own file You can then package this data file into a jar Your production code opens the file and reads it instead of running the training algorithm