NVIDIA 开源 GPU 编排平台 Run:ai
NVIDIA 把 Run:ai 开源了，管 GPU 集群的可以看看，Lambda 上还能配 Slurm 或 SkyPilot 一起用。
NVIDIA 将 GPU 编排平台 Run:ai 开源，用户不再被锁定在某一种 GPU 调度方式上。云服务商 Lambda 自 2022 年起就支持 Run:ai，平台上还可选 Slurm、Kubernetes 和 SkyPilot 等编排器。开源后，Lambda 表示更容易把 Run:ai 打包、测试并作为托管选项保持更新。Lambda 同时上线了节点级自动修复功能，用于保持 GPU 节点健康。
NVIDIA just open-sourced Run:ai. Good news if you'd rather not be locked into one way of scheduling GPUs.
Run:ai is one of several orchestrators you can run on Lambda, alongside Slurm, Kubernetes, and SkyPilot. You pick how to allocate and share your GPUs. We keep them healthy underneath, node-level auto-remediation ships in Lambda Kubernetes Service today.
We've supported Run:ai on Lambda since 2022. Open source makes it easier to package, test, and keep it current as a managed option.
Read more: https://t.co/QvBsSa2ZAV
