Hi team,
I want to set up my own fine-tuned model to mimic my writing style(blogs, articles). But I’m not sure what are the detailed steps to do.
Based on my understanding and the reference: Fine tuning gpt-3 to write in the writing style of a news outlet
One of the possible way is to extract the “news/fact” to prompt and the “high frequent words/sentence in my past articles” to completion to train the model, so that each time i ask the model to say something, it would try to follow my speaking habit.
But actually there is no general guidance yet, do we have other recommendation for this scenario?
Should we add some pre-notes like “from now on, you should act like xxx for the following response” to allow the model to output comments/sentences that imitating my writing style after feeding the prompt and completion? Is that a good approach? not sure if we can hide that first-prompt for the new conversation.
Thank you