saul-goodman-persona : GGUF

This model was finetuned and converted to GGUF format using Unsloth.

Example usage:

• For text only LLMs: llama-cli -hf giangdvdotdev/saul-goodman-persona --jinja
• For multimodal models: llama-mtmd-cli -hf giangdvdotdev/saul-goodman-persona --jinja

Available Model files:

• llama-3.1-8b-instruct.Q5_K_M.gguf

Ollama

An Ollama Modelfile is included for easy deployment.

Training data

The full fine-tuning dataset ships in this repo:

Path	What it is
data/saul_dataset.jsonl	500 single-turn chat examples used for the LoRA
data/README.md	Dataset card — format, composition, boundary
persona_spec.md	Character bible used as the generation-time system prompt
scripts/seed_tasks.jsonl	~50 hand-picked seed tasks across 10 categories
scripts/generate_dataset.py	Generator that expands seeds into the dataset

Rows carry no system message — that is what makes the persona unconditional. You do not need a persona prompt at inference; just talk to it.

500 rows across 10 categories (negotiation, persuasion, moral-gray advice, banter, crime flavor, everyday mundane, legal flavor, life advice, one-liner reactions, NPC scene prompts), 50 each. Responses are synthetic. See the dataset card for the full breakdown and the content boundary.

Note

Saul Goodman is a character from Breaking Bad / Better Call Saul, owned by their rights holders. This is an unofficial fan/parody project made for learning purposes, not affiliated with or endorsed by them. Output is styled fiction — not legal, financial, or any other kind of real advice.

This was trained 2x faster with Unsloth