#5656th of 80 in Open modelsFor science, health and research, health and medicine
huatuo-llama-med-chinese
LoRA-tuned Chinese medical Q&A models on LLaMA, Alpaca, Bloom and Huozi
- Stars
- 5k
- License
- Apache-2.0
- Last commit
- Jul 2026
- Language
- Python
4,917 on 22 Apr 20265,000 on 11 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
BenTsao (formerly HuaTuo) releases LoRA adapters that instruction-tune 7B base models (LLaMA, Alpaca-Chinese, Bloom, Huozi) for Chinese medical question answering. Training data is about 8,000 Q&A pairs generated with the GPT-3.5 API from the cMeKG knowledge base, plus 1k multi-turn liver cancer samples from literature. The repo includes inference and fine-tuning scripts and a knowledge-tuning method.
Who it is for: Researchers fine-tuning LLMs for Chinese medical Q&A
Strengths
- Ships LoRA adapters for four base models, on Hugging Face and Baidu Netdisk
- Includes infer and finetune shell scripts plus the dataset format for custom data
- Documents training cost: one A100 80GB, 10 epochs, about 2h17m, around 40GB VRAM
- Knowledge-tuning method is described in an ACM TKDD paper
Weaknesses
- Training data is mostly GPT-generated and the authors admit it contains errors
- Literature model covers only liver cancer; other diseases are planned, not released
- Resources are for academic use only; the disclaimer prohibits commercial use
- Base models are 2023-era 7B; authors say LLaMA-based results are limited
What it needs
- GPU required
- Needs PyTorch, Hugging Face Transformers, PEFT, CUDA
- Models: LLaMA-7B, Alpaca-Chinese-7B, Bloom-7B, Huozi 1.0
Also in Open models
See all 80| Rank | Project | Score |
|---|---|---|
| 1 | ruviewWiFi CSI sensing on ESP32 for presence, breathing, heart rate and pose | 94 out of 100 |
| 2 | layaSingle-pass multilingual classifier returning typed decisions with calibrated odds | 88 out of 100 |
| 3 | rf-detrReal-time transformer models for object detection, segmentation and keypoints | 81 out of 100 |
| #4 | timesfmPretrained time-series foundation model for zero-shot forecasting from Google Research | 78 out of 100 |
| #5 | open_clipPyTorch library for training and running CLIP-style image-text models | 78 out of 100 |