---

原文链接：http://wangguo.site/posts/9d8c1768.html

ChatGLM2-6B 是开源中英双语对话模型 ChatGLM-6B 的第二代版本
GitHub地址：https://github.com/THUDM/ChatGLM2-6B

1、先看效果

2、本地部署

部署环境

wsl2-ubuntu22.04 LTS

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 525.104      Driver Version: 528.79       CUDA Version: 12.0     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  NVIDIA GeForce ...  On   | 00000000:01:00.0  On |                  N/A |
| N/A   45C    P8     5W /  80W |    928MiB /  6144MiB |      3%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|    0   N/A  N/A        23      G   /Xwayland                       N/A      |
+-----------------------------------------------------------------------------+

下载

git clone https://github.com/THUDM/ChatGLM2-6B
cd ChatGLM2-6B

创建虚拟环境，安装库

virtualenv venv
source venv/bin/activate

pip install -r requirements.txt

本地模型下载

git clone https://huggingface.co/THUDM/chatglm2-6b-int4

然后在清华大学云盘下载相应的模型参数文件,并将文件拷贝到chatglm2-6b-int4文件夹下

int-4推理

需要先修改web_demo.py,修改内容

//第7行,修改为本地模型参数地址

# model = AutoModel.from_pretrained("THUDM/chatglm2-6b-int4", trust_remote_code=True).cuda()

model = AutoModel.from_pretrained("./chatglm2-6b-int4", trust_remote_code=True).cuda()

web_demo.py

执行

python web_demo.py

遇到的问题

问题1

OSError: model/chatglm2-6b is not a local folder and is not a valid model identifier listed on 'https://huggingface.co/models'
If this is a private repository, make sure to pass a token having permission to this repo with `use_auth_token` or log in with `huggingface-cli login` and pass `use_auth_token=True`.

– 解决，输入

huggingface-cli login 

并点击链接生成new token,拷贝到shell中输入即可

问题2

RuntimeError: Internal: src/sentencepiece_processor.cc(1101) [model_proto->ParseFromArray(serialized.data(), serialized.size())]

– 解决，输入

sudo apt install libcudart11.0 libcublaslt11