3 ms·It cheapens the cost of fine tuning, it doesn’t make the model itself smaller at inference time.by stu2b50 4y agoIt cheapens the cost of fine tuning, it doesn’t make the model itself smaller at inference time.