---
license: apache-2.0
language:
- zh
base_model:
- THUDM/glm-4-9b
pipeline_tag: text-generation
---
# MentalGLM is a series of large language models designed for mental health analysis tasks in Chinese.
We have developed the MentalGLM series, the first open-source LLMs designed for explainable mental health analysis targeting Chinese social media, based on GLM-4-9b and GLM-4-9b-chat.

## How to use

```bash
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

device = "cuda"

tokenizer = AutoTokenizer.from_pretrained("zwzzz/MentalGLM", trust_remote_code=True)

query = "考虑以下这个帖子，帖子体现了什么认知路径？这已经够糟糕的了。不过在那一周我将完全失去我的支持。我没有什么可期待的。"

inputs = tokenizer.apply_chat_template([{"role": "user", "content": query}],
                                       add_generation_prompt=True,
                                       tokenize=True,
                                       return_tensors="pt",
                                       return_dict=True
                                       )

inputs = inputs.to(device)
model = AutoModelForCausalLM.from_pretrained(
    "zwzzz/MentalGLM",
    torch_dtype=torch.bfloat16,
    low_cpu_mem_usage=True,
    trust_remote_code=True
).to(device).eval()

gen_kwargs = {"max_length": 1000, "do_sample": True, "top_k": 1}
with torch.no_grad():
    outputs = model.generate(**inputs, **gen_kwargs)
    outputs = outputs[:, inputs['input_ids'].shape[1]:]
    print(tokenizer.decode(outputs[0], skip_special_tokens=True))
```

## Citation

Article address：[https://arxiv.org/pdf/2410.10323.pdf](https://arxiv.org/pdf/2410.10323.pdf)
```bash
@article{zhai2024mentalglm,
  title={MentalGLM Series: Explainable Large Language Models for Mental Health Analysis on Chinese Social Media},
  author={Zhai, Wei and Bai, Nan and Zhao, Qing and Li, Jianqiang and Wang, Fan and Qi, Hongzhi and Jiang, Meng and Wang, Xiaoqin and Yang, Bing Xiang and Fu, Guanghui},
  journal={arXiv preprint arXiv:2410.10323},
  year={2024}
}
```