Here is the prompt that achieved the best test accuracy across the optimization iterations.

def get_best_prompt(results):

return best_prompt, best_accuracy, best_iteration