Modelpublic

ashish1866/lemma-rewriter-qwen3-4b-gguf

sourceHugging Faceupdated 2mo agoView on Hugging Face

0likes17downloads

Model Card

Lemma Qwen3 4B GGUF

A GGUF model artifact used by the Lemma academic text-rewriting project. It is intended for local CPU/GPU inference through Ollama or llama.cpp.

Artifact details

—File: lemma-rewriter-Qwen3-4B-Q4_K_M.gguf
—Format: GGUF
—Architecture: Qwen3
—Parameter size: 4B
—Quantization: Q4KM
—Quantized by: Unsloth (as recorded in the GGUF metadata)
—Native context length: 40,960 tokens
—SHA-256: DCB876A4440A507D66038C6CFE3160FB2E161E3D9E264E296B474D670868616F

The embedded GGUF metadata identifies this artifact as Lemma Qwen3 4b Gguf Recovered and does not declare a license. The artifact's complete training and redistribution provenance has not been independently verified.

License and provenance

This repository intentionally does not declare a license for this exact GGUF. Do not assume that the license of a related base model or another Lemma model automatically applies to this artifact. Confirm the redistribution rights and applicable license before commercial use, modification, or further redistribution.

The official Qwen3-4B model information is available at Qwen/Qwen3-4B. This link is provided for provenance reference only and is not a license claim for this GGUF.

Ollama usage

Download this repository, then create an Ollama model using the included Modelfile:

powershell

hf download ashish1866/lemma-rewriter-qwen3-4b-gguf `
  lemma-rewriter-Qwen3-4B-Q4_K_M.gguf `
  --local-dir .

ollama create lemma-model -f Modelfile
ollama run lemma-model

The included template explicitly uses the Qwen3 ChatML turn markers and closes the thinking block so that rewriting responses do not include reasoning text.

Lemma prompt behavior

The model card does not prescribe a system prompt. The included Ollama Modelfile configures it as an academic editor that preserves meaning, facts, numbers, and citations and returns only rewritten text.

Intended use

Use this model for local experimentation and academic/technical paraphrasing workflows. Review generated text for factual accuracy, citations, and appropriate attribution before using it in a publication or submitted work.