This is a LoRA adapter trained on top of meta-llama/Llama-3.2-3B using Unsloth, for the purpose of aligning the model to doctor–patient conversations and dental implant-related Q&A.

The adapter improves the model's performance in instruction-following and medical dialogue within the dental implant domain (e.g. Straumann® surgical workflows).

🔧 Model Details

Base model: meta-llama/Llama-3.2-3B
Adapter type: LoRA via PEFT
Framework: Unsloth
Quantization for training: QLoRA (bnb 4-bit)
Training objective: Instruction-tuning on domain-specific dialogue
Dataset: BirdieByte1024/doctor-dental-llama-qa

🧠 Dataset

BirdieByte1024/doctor-dental-llama-qa
Includes synthetic doctor–patient chat covering:
- Straumann® dental implant systems
- Guided surgery workflows
- General clinical Q&A

💬 Expected Prompt Format

json

{
  "conversation": [
    { "from": "patient", "value": "What is the purpose of a healing abutment?" },
    { "from": "doctor", "value": "It helps shape the gum tissue and protect the implant site during healing." }
  ]
}

💻 How to Use the Adapter

python

from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel

# Load base model
base = AutoModelForCausalLM.from_pretrained("meta-llama/Llama-3.2-3B")
tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-3.2-3B")

# Load LoRA adapter
model = PeftModel.from_pretrained(base, "BirdieByte1024/doctor-dental-implant-LoRA-llama3.2-3B")

✅ Intended Use

Domain adaptation for dental and clinical chatbots
Offline inference for healthcare-specific assistants
Safe instruction-following aligned with patient communication

⚠️ Limitations

Not a diagnostic tool
May hallucinate or oversimplify
Based on non-clinical and synthetic data

🛠 Authors

Developed by (BirdieByte1024)
Fine-tuned using Unsloth and PEFT

📜 License

MIT