navneetsatyamkumar/Medalpaca-distillation

MedAlpaca Distilled Model Demo

This Hugging Face Space demonstrates a distilled Qwen model fine-tuned for medical question answering using knowledge distillation techniques.

Model Details

• —Base Model: Qwen/Qwen2.5-0.5B
• —LoRA Adapter: navneetsatyamkumar/Medalpaca-distillation
• —Training Method: Knowledge Distillation + LoRA Fine-tuning
• —Domain: Medical Question Answering

Features

• —Interactive chat interface for medical Q&A
• —Adjustable generation parameters (temperature, top-p, max length)
• —Real-time model loading and inference
• —Medical-focused prompting template

Usage

1. 1.Click "Load Model" to initialize the model
2. 2.Ask medical questions in the chat interface
3. 3.Adjust generation parameters as needed
4. 4.View responses from the distilled model

Technical Implementation

The app uses:

• —Gradio for the web interface
• —Transformers for model loading
• —PEFT for LoRA adapter integration
• —PyTorch for inference

Framework Versions

• —PEFT 0.16.0
• —Transformers >= 4.35.0
• —Gradio >= 4.0.0