How to use from the
Use from the
llama-cpp-python library
# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="lerugray/tailor-king-7b",
	filename="tailor-king-7b-Q5_K_M.gguf",
)
output = llm(
	"Once upon a time,",
	max_tokens=512,
	echo=True
)
print(output)

tailor-king: a Jan van Leiden / Münster register model

A 7B voice tune that writes in the register of the Anabaptist kingdom of Münster (1534–35): apocalyptic, theocratic, scriptural. In 1534 a Dutch journeyman tailor, Jan van Leiden, was crowned king of a city that abolished money, held all goods in common, and waited for the end of the world. This is its voice.

It is a register, not a biography or a source. Read it for the voice, not the record.

What it does

Ask it about the New Jerusalem, the holding of goods in common, the godless princes, or the present day, and it answers from the throne of the last days. It translates the modern subject into the kingdom's sixteenth-century apocalyptic frame: a bank becomes usury forbidden twice in Scripture; private wealth becomes the oppression of the poor against the witness of Acts that all who believed held everything in common.

How it was built

  • Base: Qwen2.5-7B-Instruct.
  • Method: completion-style causal-LM fine-tuning; the adapter merged onto the fp16 base before conversion to GGUF (Q5_K_M).
  • Source: the voice it learns is mostly the kingdom's own theologian, Bernhard Rothmann, whose Restitution survives in public-domain German. The English here is our own fresh translation, so the spine is unencumbered. Roughly 1,200 completion records (authentic chunks oversampled), plus a small modern-bridge layer that lets the king-prophet's lens fall on the present. The corpus is not distributed.
  • Inference: a lead-in frame elicits the first-person voice; plain chat narrates about the kingdom instead of as it. Recommended temperature ~0.65.

Intended use

Creative writing, Reformation-history pedagogy in a register, interactive fiction, voice prototyping. The output is a historical and artistic register. It is not an endorsement, a call to action, or advice.

Limitations and honest notes

  • It invents. It will produce names, dates, and scripture that never existed and present them as real. Read it for the voice, not the record.
  • A regurgitation audit found no long verbatim reproduction of its source texts. The model writes new prose in the register rather than reciting source passages. The training corpus text is not distributed.
  • Authored theology, not one man's words. Jan van Leiden's surviving "confessions" are captor-mediated interrogation records, so the model leans on the kingdom's authored theology (chiefly Rothmann) rather than any one man's verified speech.
  • Period worldview. It speaks from a sixteenth-century apocalyptic, theocratic framework, absolutism and all. That is the artifact, not a recommendation.

Run it

ollama run hf.co/lerugray/tailor-king-7b

License

CC-BY-NC-4.0. Non-commercial research, educational, and creative use. The source work (Rothmann's Restitution) is public domain and the English translation is our own, so the weights could ship permissively; the non-commercial clause is a deliberately conservative choice given the synthetic modern-bridge component and the persona framing. Attribution: Ray Weiss / The Elect. No warranty.

Part of The Elect — a small fleet of public-domain historical-voice models. https://lerugray.github.io/the-elect/

Downloads last month
19
GGUF
Model size
8B params
Architecture
qwen2
Hardware compatibility
Log In to add your hardware

5-bit

Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for lerugray/tailor-king-7b

Base model

Qwen/Qwen2.5-7B
Quantized
(341)
this model