uaritm/gemmamed_cardio

Hugging Face
VISIONConcurrency Cost:1Model Size:4.3BQuant:BF16Ctx Length:32kPublished:Dec 7, 2025License:apache-2.0Architecture:Transformer0.0K Open Weights Warm

uaritm/gemmamed_cardio is a specialized instruction-following Gemma-4B-Instruct model, fine-tuned by Uaritm for cardiology-related information and medical queries in Ukrainian. This model underwent a two-stage LoRA fine-tuning process, including linguistic adaptation on a large Ukrainian corpus and domain specialization on cardiovascular health data. It is optimized for efficient inference on consumer hardware, provided as a GGUF Q4KM quantized file (~2.4 GB) with a context length of 4096 tokens, excelling as a high-quality cardiology assistant in Ukrainian.

Loading preview...

Overview

uaritm/gemmamed_cardio is a highly specialized, instruction-following language model based on the Gemma-4B-Instruct architecture, developed by Uaritm. It is meticulously fine-tuned to provide cardiology-related information and answer medical queries exclusively in Ukrainian. The model is distributed as a highly optimized GGUF Q4KM quantized file, making it suitable for efficient inference on consumer hardware.

Key Capabilities

  • Specialized Cardiology Assistant: Designed to act as a clinical cardiologist, providing recommendations based on patient data and epicrises.
  • Ukrainian Language Proficiency: Adapted and specialized for high-quality interaction in Ukrainian.
  • Two-Stage Fine-Tuning: Utilizes a LoRA fine-tuning approach, first for general Ukrainian linguistic adaptation (on 14500 cardiological epicrises) and then for deep domain specialization in cardiovascular health.
  • Efficient Inference: Provided in GGUF Q4KM format (~2.4 GB) for use with llama.cpp, Ollama, or LM Studio.
  • Context Length: Supports a context length of 4096 tokens.

Good For

  • Ukrainian medical professionals seeking an AI assistant for cardiology-specific queries.
  • Applications requiring specialized medical information in the cardiovascular domain in Ukrainian.
  • Research and development in Ukrainian clinical natural language processing.

Limitations

  • Optimized primarily for cardiology and cardiac surgery; accuracy is reduced outside these domains.
  • Text-only model with no vision capabilities.
  • May generate incomplete or generalized recommendations.

Finetunes
1 models