2 篇文档带有标签[hami]

Install HAMi on a GPU cluster and schedule SGLang inference services with GPU partitioning.

用 nvml-mock 模拟 8 张 A100 GPU 验证 HAMi 调度能力，无需真实 GPU。