prose-generation

Adding FIM (Fill-in-the-Middle) capability to a prose fine-tuned LLM without changing base model