JBHarris/dm-llm-tiny
JBHarris/dm-llm-tiny is a 1.1 billion parameter language model fine-tuned from TinyLlama-1.1B-Chat-v1.0. This model is specifically optimized for generating creative Dungeons & Dragons (D&D) content, including NPCs, quests, dialogue, locations, and encounters. It leverages QLoRA for efficient training on synthetic D&D instruction/response pairs. Its primary use case is as a quick idea generator for Dungeon Masters.
Loading preview...
DM-LLM-Tiny: A Specialized D&D Content Generator
JBHarris/dm-llm-tiny is a compact 1.1 billion parameter language model, built upon the TinyLlama-1.1B-Chat-v1.0 base. It has been meticulously fine-tuned using QLoRA (4-bit NF4 quantization with LoRA r=64) on approximately 5000 synthetic D&D instruction/response pairs generated by Claude.
Key Capabilities
This model excels at generating a variety of creative Dungeons & Dragons content, making it a valuable assistant for Dungeon Masters:
- NPCs: Creates memorable characters complete with backstories, motivations, and unique quirks.
- Quests: Develops quest hooks, outlines, and full quest arcs.
- Dialogue: Generates in-character conversations, monologues, and engaging banter.
- Locations: Provides vivid descriptions for dungeons, towns, and wilderness areas.
- Encounters: Designs combat, social, and puzzle-based encounters.
Good For
- Brainstorming: Ideal for quickly generating ideas and creative prompts for D&D campaigns.
- Rapid Prototyping: Useful for DMs who need quick content without extensive manual creation.
Limitations
As a 1.1 billion parameter model, dm-llm-tiny offers creative output but will not match the quality or depth of larger language models (7B+). It is best utilized as a supplementary tool for idea generation rather than a complete replacement for a human Dungeon Master's judgment and creativity.