method: chosen_reward_ 

        num_inference_steps: 30

        {metric} : {mean} 
{metric} : {mean} 
{metric} : {mean} 
{metric} : {mean} 
{metric} : {mean}