Upload SARM reward model weights, train config, preprocessor and readme

initial commit