Thanu10/Arogya_fine_tuned

Arogya Fine-Tuned (Gemma 3 - Medical Chatbot)

📌 Overview

Arogya is a fine-tuned version of Gemma 3 (4B) designed to act as a medical question-answering assistant.
It is trained on healthcare-related conversational data to provide general medical guidance.

---

🧠 Model Details

• Base Model: unsloth/gemma-3-4b-it-unsloth-bnb-4bit
• Fine-tuning Method: LoRA (Parameter-Efficient Fine-Tuning)
• Framework: Unsloth + TRL
• Quantization: 4-bit (for memory efficiency)

---

📊 Training Data

• Dataset: lavita/ChatDoctor-HealthCareMagic-100k
• Format: Instruction + Input → Response
• Converted into chat-style format for Gemma 3

---

⚙️ Training Configuration

• Max Steps: 30 (initial prototype)
• Learning Rate: 2e-4
• Batch Size: 2 (with gradient accumulation)
• Response-only training enabled

---

💡 Intended Use

• Educational purposes
• Demonstration of medical chatbot fine-tuning
• Research experiments in healthcare AI

---

⚠️ Limitations & Safety Notice

• This model is not a licensed medical professional
• Outputs may contain:
  • inaccurate or outdated medical advice
  • overconfident statements
• Do NOT use for real medical decision-making
• Always consult a qualified healthcare provider

---

🚀 Usage Example

from unsloth import FastModel

model, tokenizer = FastModel.from_pretrained(
    "your-username/Arogya_fine_tuned",
    load_in_4bit=True,
)

messages = [
    {"role": "user", "content": "I have a headache and fever. What should I do?"}
]

---

👨‍💻 Author

• Name: Thanujaya Tennekoon
• Affiliation: University of Ruhuna

---

📌 Future Improvements

• Increase training steps for better accuracy
• Add safety-aligned medical instructions
• Incorporate Sinhala language support
• Filter dataset for higher-quality responses