BirdieByte1024/Qwen2.5-1.5B-dental-full

🦷 Qwen2.5-1.5B Dental Model (LoRA Adapter & Full Model)

This model is a fine-tuned version of Qwen/Qwen2.5-1.5B, trained with LoRA adapters on a dental procedure instruction dataset. It has been trained to explain ADA dental procedure codes in a way patients can understand.

Two versions are available:

• BirdieByte1024/Qwen2.5-1.5B-LoRA-dental: LoRA adapter only
• BirdieByte1024/Qwen2.5-1.5B-dental-full: Fully merged standalone model

---

🔍 Model Details

• Base model: Qwen/Qwen2.5-1.5B
• Training method: PEFT (LoRA)
• Tokenizer: Inherited from base model
• Model size: 1.5B parameters
• Precision: fp32
• Training hardware: GTX 1060 (6GB)

---

📚 Dataset

This model was trained on:

• TachyHealth/ADA_Dental_Code_to_SBS_V2

The dataset includes ADA codes and short/long descriptions useful for patient-friendly explanations.

---

💬 Prompt Format

This is an instruction-tuned model using a simple text format:

### Instruction:
Explain the following dental code.

### Code:
D7140 - Extraction, erupted tooth

### Response:

---

✅ How to Use (Transformers)

from transformers import AutoTokenizer, AutoModelForCausalLM

model_id = "BirdieByte1024/Qwen2.5-1.5B-dental-full"  # or LoRA version with PEFT if needed

tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(model_id, trust_remote_code=True).to("cuda")

prompt = """### Instruction:
Explain the following dental code.

### Code:
D7140 - Extraction, erupted tooth

### Response:"""

inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
outputs = model.generate(**inputs, max_new_tokens=100)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

---

🛠 Deployment Options

🖥 Local with Transformers (Python)

Use the example above with transformers for local inference.

🧠 Use with Ollama / llama.cpp (GGUF format)

To deploy via ollama, convert the merged model to GGUF and run:

ollama run qwen2.5-dental

Make sure to convert your model to GGUF first using transformers + transformers-gguf or use llama.cpp export tools.

---

⚠️ Limitations

• English-only dental domain coverage
• Not a diagnostic or real clinical system
• May hallucinate or oversimplify medical terms

---

✍️ Author

Created by BirdieByte1024 as part of a patient-education AI project using LoRA + Qwen models.

---

📜 License

Apache 2.0