issues
search
shell-nlp
/
gpt_server
gpt_server是一个用于生产级部署LLMs或Embedding的开源框架。
Apache License 2.0
101
stars
11
forks
source link
纯CPU无法启动
#9
Closed
firstjack
closed
2 months ago
firstjack
commented
3 months ago
最新版本,这是配置文件
qwen: #自定义的模型名称 alias: gpt-4,gpt-3.5-turbo,gpt-3.5-turbo-16k # 别名 例如 gpt4,gpt3 enable: true # false true model_name_or_path: /home/k/Qwen1.5-7B-Chat-GPTQ-Int4/ model_type: qwen # qwen chatglm3 yi internlm work_mode: vllm # vllm hf device: cpu # gpu / cpu workers:
gpus:
0 启动后提示 ImportError: libcuda.so.1: cannot open shared object file: No such file or directory
shell-nlp
commented
3 months ago
cpu,确实有问题,我研究一下
shell-nlp
commented
3 months ago
我看了vllm的文档,pip 安装的 vllm 不支持 cpu,
最新版本,这是配置文件