Llama 3.1 Prose Fine-Tune

I write a lot. Sometimes it’s fiction. I have a definite “style” to my prose writing.

I wanted to see if an LLM could be trained to imitate that writing style well.

So, I sat down, curated a data set from my writing and began the task of fine-tuning a small local model, Llama 3.1 8B by Meta.

The result is an LLM that when prompted to write in the style of my pen name, Astral FarKaster will generate prose that matches my writing style in rhythm and candor. Pretty amazing stuff.