Job Verification

Description

Check the number of AscendAI processors required by the job. For Atlas training product, the number can only be 1, 2, 4, or 8.

Implementation Details

For details about code implementation, see ValidNPUJob in the open source code. ValidNPUJob is used to check whether the configuration delivered by the user is proper. The system does not check whether the actual resources in the cluster environment are sufficient. Instead, the system only checks whether the key fields of the job are complete, whether the field values are correct, and whether the fields match.

Parent topic: Scheduling Algorithm of Ascend AI Processors