Downloads · 30 days
0
vervelo/medical-qwen-0.5b-mlx
medical-qwen-0.5b-mlx is a machine learning model from vervelo. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for mlx. The card lists the license as other.
This repository contains a LoRA Adapter (34 MB) for Qwen2.5-0.5B-Instruct, fine-tuned natively on Apple Silicon for clinical dialogue-to-SOAP-note summarization.
Downloads · 30 days
0
Access
Public
Updated Apr 16, 2026
Repo size
35.2 MB
Likes
0
Public
Click a slice to open those files.
.safetensors35.2 MB · 100%
From the Hugging Face model README
This repository contains a LoRA Adapter (34 MB) for Qwen2.5-0.5B-Instruct, fine-tuned natively on Apple Silicon for clinical dialogue-to-SOAP-note summarization.
Unlike a standard 1GB+ foundation model, this Adapter-only release (34 MB) provides a highly portable "medical brain patch." It requires the base Qwen2.5-0.5B-Instruct model to function, allowing for ultra-fast updates and deployments in clinical environments.
Benchmarked against the 16x larger Llama 3 8B (Base) to evaluate the performance of domain-specific fine-tuning.
| Metric | Specialized (0.5B) | General Baseline (8B) |
|---|---|---|
| Mean ROUGE-L | 0.1131 | 0.1217 |
| Avg Inference Time | 2.02s | 4.54s |
| Relative Performance | 93.0% | Baseline |
While trained with MLX, these adapters are architecture-agnostic:
transformers + PEFT.llama.cpp).To test the model immediately, run the included inference script:
python3 run_inference.py "Patient: I have a persistent dry cough. Doctor: Any fever?"
This model is for research purposes only and is not a clinical diagnostic tool.