Model Overview

netdur/gemma-3-4b-radiology is a specialized 4.3 billion parameter Gemma 3 model, fine-tuned to generate detailed radiology reports from ultrasound images. Developed by netdur, this model utilizes the Unsloth library for efficient fine-tuning and is provided in GGUF format, compatible with llama.cpp for local inference.

Key Capabilities

• Radiology Report Generation: Specifically trained to interpret ultrasound images and produce structured textual reports.
• Medical Imaging Analysis: Designed for tasks requiring descriptive analysis of medical visuals.
• Gemma Architecture: Built upon the Gemma 3 4B base model, leveraging its foundational language understanding.
• Optimized for Inference: Provided in GGUF format, enabling efficient execution with llama.cpp and GPU acceleration (e.g., Metal on macOS).

Good For

• Automated Radiology Reporting: Assisting radiologists by generating initial drafts or summaries of ultrasound findings.
• Medical AI Development: Serving as a base model for further research and development in medical image-to-text tasks.
• Educational Tools: Demonstrating AI capabilities in medical diagnostics and report generation.

This model was fine-tuned using the unsloth/Radiology_mini dataset, focusing on accurate and descriptive output for medical professionals.