README.md

---
license: mit
---

# ChiMed-GPT

ChiMed-GPT is a Chinese medical large language model (LLM) that is built by continually training [Ziya-v2](https://arxiv.org/abs/2311.03301) on Chinese medical data, where pre-training, supervised fine-tuning (SFT), and reinforcement learning from human feedback (RLHF) are performed.

More information about the model is coming soon.

## Citation

If you use or extend our work, please cite the following [paper](https://arxiv.org/abs/2311.06025):
```
@article{USTC-ChiMed-GPT,
  title="{ChiMed-GPT: A Chinese Medical Large Language Model with Full Training Regime and Better Alignment to Human Preferences}",
  author={Yuanhe Tian, Ruyi Gan, Yan Song, Jiaxing Zhang, Yongdong Zhang},
  journal={arXiv preprint arXiv:2311.06025},
  year={2023},
}
```

## Usage
```python
from transformers import AutoTokenizer
from transformers import LlamaForCausalLM
import torch

query="[human]:感冒怎么处理？\n[bot]:"
model = LlamaForCausalLM.from_pretrained('SYNLP/ChiMed-GPT-1.0', torch_dtype=torch.float16, device_map="auto").eval()
tokenizer = AutoTokenizer.from_pretrained(ckpt)
input_ids = tokenizer(query, return_tensors="pt").input_ids.to('cuda:0')
generate_ids = model.generate(
            input_ids,
            max_new_tokens=512, 
            do_sample = True, 
            top_p = 0.9)
output = tokenizer.batch_decode(generate_ids)[0]
print(output)
```