easy5 pts · first answer countsgpu-computingcheckpointingreliability

Where to write checkpoints

A training job saves checkpoints only to the GPU node's local disk. The node is replaced after a hardware fault. What is the problem, and the fix?

A.Nothing, local disks are copied to new nodesB.Checkpoints are fine but will load slowlyC.The GPU memory keeps a copy of the checkpointD.Checkpoints are lost; write them to durable shared storage

Discussion 0

Join free to post in the discussion.

Where to write checkpoints

Discussion 0

Related questions