[ilaksh]

I think it depends on the task and the result of the fine tune which is mainly based on the training dataset and ability of the base model, whether you will be able to maintain the quality.