|
|
||
|---|---|---|
| .gitignore | ||
| .python-version | ||
| README.md | ||
| requirements.txt | ||
| start.sh | ||
| stop.sh | ||
| vllm-embedding.pid | ||
| vllm-rerank.pid | ||
| vllm.pid | ||
README.md
简介
本来想直接在docker里面跑 vllm ,但是 mac m2 没有对应的镜像,所以改为直接系统上用python 跑
准备
下载需要运行的模型,记得修改 start.sh 中的模型配置
例: 使用 Qwen2.5-0.5B-Instruct 模型调试
# mac 系统
# 1. 安装 git-lfs,Git LFS(Large File Storage)是 Git 的一个扩展,专门用于高效地版本控制大型文件,如音频、视频、数据集、图形资产等
brew install git-lfs
git lfs install
# 2. 下载模型
mkdir models
cd models
git clone https://www.modelscope.cn/Qwen/Qwen2.5-0.5B-Instruct.git
命令
# 启动
sh start.sh
# 关闭
sh stop.sh
# 查看是否成功
curl -H "Authorization: Bearer sk-local-827ccb0eea8a706c4c34a16891f84e7b" http://localhost:8022/v1/models
# embedding
curl http://localhost:8023/v1/embeddings \
-X POST \
-H "Authorization: Bearer sk-local-827ccb0eea8a706c4c34a16891f84e7b" \
-H "Content-Type: application/json" \
-d '{"input": ["Instruct: Given a web search query, retrieve relevant passages that answer the query\nQuery: What is the capital of China?", "Instruct: Given a web search query, retrieve relevant passages that answer the query\nQuery: Explain gravity"]}'