NPU Computing Power Allocation

Ascend NPU hardware resources include AI Cores (used for AI model computing), AI CPUs, and memory. The function of NPU computing power allocation divides the NPU into vNPUs based on the specifications specified by users. Each vNPU corresponds to several AI Cores, AI CPUs, and memory resources. If the NPU is not divided, it is identified as an entire card in edge scenarios. For details about computing power allocation, see "Instructions" in the Ascend Virtualization Instance (AVI) User Guide.

Currently, the AtlasEdge supports only the computing power allocation of the Atlas 300I Pro inference card deployed on the Atlas 500 Pro AI edge server (model 3000), and containers can be deployed only through FusionDirector.