LLaMA-3 8B Medical Vietnamese Chatbot (V4)

Fine-tuned version of LLaMA-3 8B for Vietnamese medical Q&A. Developed as part of graduation thesis at An Giang University.

Model Details

  • Base model: LLaMA-3 8B
  • Fine-tuning method: QLoRA (4-bit NF4) + LoRA rank=16
  • Training data: ~1,440 Vietnamese medical Q&A samples
  • Language: Vietnamese (primary), English
  • Format: GGUF q4_k_m

Benchmark Results

Metric Base Model Fine-tuned V4
Perplexity ↓ 7.058 6.526
ROUGE-1 ↑ 0.4918 0.5602
ROUGE-2 ↑ 0.1486 0.2091
ROUGE-L ↑ 0.2841 0.2997

Usage with LM Studio

  1. Search your-username/llama-3-8b-medical-vi in LM Studio
  2. Download q4_k_m variant
  3. Use system prompt:
Bạn là bác sĩ AI chuyên tư vấn sức khỏe. Hãy trả lời câu hỏi của bệnh nhân một cách chính xác, rõ ràng và có trách nhiệm. Luôn khuyên bệnh nhân gặp bác sĩ trực tiếp khi cần thiết.

Ethan2004/llama-3-8b-medical-vi ├── 📄 Model Card: Vietnamese Medical Chatbot ├── 🏷️ Tags: medical, vietnamese, llama-3, gguf ├── 📥 llama-3-8b-medical-vi.Q4_K_M.gguf (4.8GB) └── ⭐ License: Apache 2.0

Downloads last month
34
GGUF
Model size
8B params
Architecture
llama
Hardware compatibility
Log In to add your hardware

4-bit

Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for Ethan2004/llama-3-8b-medical-vi

Adapter
(709)
this model