---
tags:
- gguf
- llama.cpp
- unsloth
- vision-language-model
- qwen3-vl
library_name: llama.cpp
pipeline_tag: visual-question-answering
license: apache-2.0
datasets:
- prapaa/eastrus-vl
language:
- en
base_model:
- unsloth/Qwen3-VL-8B-Instruct-unsloth-bnb-4bit
---

# eastrus-vl-qwen3-8b-gguf

A **GGUF-exported multimodal (vision-language) model** based on **Qwen3-VL-8B-Instruct**, fine-tuned for **cattle estrus–related vulval image assessment**. The model is intended to produce **symptom-by-symptom observations** and a **single confidence-style score**, rather than a hard binary decision.

This repo contains:
- A **quantized GGUF** model for inference (`Q4_K_M`)
- A matching **multimodal projection** file (`mmproj`) required by `llama.cpp` for vision inputs

## Model details
- **Base model**: Qwen3-VL-8B-Instruct (VLM)
- **Fine-tuning**: Unsloth (LoRA / efficient finetuning workflow)
- **Export**: `llama.cpp` GGUF conversion via Unsloth
- **Quantization**: `Q4_K_M` (balanced quality/speed)

## Intended use
- **Primary**: Assistive analysis of cattle vulval imagery for estrus-related visual signs (educational/research workflow support).
- **Not a medical device**: Outputs should not be used as the sole basis for veterinary diagnosis, treatment, or critical farm management decisions.

## Output format (what you should expect)
The model is trained to generate:
- A structured, human-readable **symptom list** (mucus color, swelling severity, redness severity, moisture level, mucus viscosity, tissue turgidity)
- A single **confidence-style score** (0–100%)
- A **JSON structured summary** of all observed symptoms

## How to run (llama.cpp)

### 1) Text-only prompt (sanity check)
```bash
llama-cli -hf prapaa/eastrus-vl-qwen3-8b-gguf --jinja -p "Is there eastrus?"
```