Downloads · 30 days
6
4% of all-time downloads
LutzMoran/pi-of-ai-rules-baker
pi-of-ai-rules-baker is a machine learning model from LutzMoran. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for gguf. The card lists the license as mit.
Part of Pi-of-AI — the "Raspberry Pi of AI." This is the Rules-Baker experiment: a small model with a project's house-style coding rules baked into its weights, so it obeys them with an empty system prompt (no rule-bl…
Downloads · 30 days
6
4% of all-time downloads
All-time downloads
141
Public
Repo size
531 MB
Likes
0
Public
Click a slice to open those files.
.gguf531 MB · 100%
From the Hugging Face model README
Part of Pi-of-AI — the "Raspberry Pi of AI." This is the Rules-Baker experiment: a small model with a project's house-style coding rules baked into its weights, so it obeys them with an empty system prompt (no rule-bloat in the context window).
smoke_housestyle.gguf is a pipeline smoke-test, not a finished model. It was
LoRA-trained on a tiny mock dataset purely to prove the train → merge → GGUF →
run-in-browser loop end to end. It overfits the mock template and does not yet
follow arbitrary tasks. The real, task-following bake comes from Stage 3+ teacher
data (see the repo roadmap). Published here for transparency and to demonstrate
the closing-the-loop test — treat it as a demo, not a tool.
q8_0, ~506 MB)The teacher model sees your coding rules and writes compliant code; the stored student training prompt has the rules stripped out. The student learns to obey silently — so at inference time your IDE context stays clean and fast.
Open rules_baker/web/index.html from the repo, then drag-drop this .gguf in.
Same tiny prompt, rules obeyed — that's the closing-the-loop test.
MIT. See the project repo.