🦷 doctor-dental-implant-llama3.2-3B-full-model

This model is a fine-tuned version of meta-llama/Llama-3.2-3B, trained using the Unsloth framework on a domain-specific instruction dataset focused on medical and dental implant conversations.

The model has been optimized for chat-style reasoning in doctor–patient scenarios, particularly within the domain of Straumann® dental implant systems, as well as general medical question answering.

---

🔍 Model Details

• Base model: meta-llama/Llama-3.2-3B
• Training framework: Unsloth with LoRA + QLoRA support
• Training format: Conversational JSON with {"from": "patient"/"doctor", "value": ...} messages
• Checkpoint format: Full model merged, usable as standard HF or GGUF (Ollama / llama.cpp)
• Tokenizer: Inherited from base model
• Model size: 3B parameters (efficient for consumer-grade inference)

---

📚 Dataset

This model was trained on:

• BirdieByte1024/doctor-dental-llama-qa

The dataset contains synthetic and handbook-derived doctor-patient conversations focused on:

• Dental implant systems (e.g. surgical kits, guided procedures)
• General medical Q&A relevant to clinics and telemedicine
• Clinical assistant-style instruction-following

---

💬 Prompt Format

The model expects a chat-style format:

{
  "conversation": [
    { "from": "patient", "value": "What are the advantages of guided implant surgery?" },
    { "from": "doctor", "value": "Guided surgery improves accuracy, safety, and esthetic outcomes." }
  ]
}

---

✅ Intended Use

• Virtual assistants in dental or medical Q&A
• Instruction-tuned experimentation on health topics
• Local chatbot agents (Ollama / llama.cpp compatible)

---

⚠️ Limitations

• Model is not a medical device or diagnostic tool
• Hallucinations and factual errors may occur
• Content was fine-tuned using synthetic and handbook-based sources (not real EMR)

---

🧪 Example Prompt

{
  "conversation": [
    { "from": "human", "value": "What should I expect after a Straumann implant surgery?" },
    { "from": "assistant", "value": "[MODEL RESPONSE HERE]" }
  ]
}

---

🛠 Deployment

Local Use with Hugging Face Transformers

from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("BirdieByte1024/doctor-dental-implant-llama3.2-3B-full-model")
model = AutoModelForCausalLM.from_pretrained("BirdieByte1024/doctor-dental-implant-llama3.2-3B-full-model")

GGUF / Ollama / llama.cpp

ollama run doctor-dental-llama3.2

If using a local Modelfile, ensure the prompt template matches chat formatting (no Alpaca-style).

---

✍️ Author

Created by (BirdieByte1024) as part of a medical AI research project using Unsloth and LLaMA 3.2.

---

📜 License

MIT