Hugging Face Overview
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.