Here is my specific deployment process, with only the GPU of the head node working in the end
《ray文生图部署代码,(只有head节点的gpu运行)》