Hello. I am trying to fine-tune GPT-3 from a blog. My goal is to provide a writing assistent that can write in the same style as the author of the blog. There are hundreds of thousands of words combined in all of the blog posts. How would I go about finetuning on this? what would be my prompt and completion in this case?
#Fine-tune GPT-3 from blog
midnight ore
Should I split up each blog post into many different prompts and completions?
I was thinking maybe like the first sentence of each paragraph can be a prompt, and the entire paragraph can be the conclusion
Or should I have the article title as the prompt, and the entire article as the completion?