Describe Solutions to Maximize Performance in AI Workloads CERTIFIED-AI-EXPERT Practice Questions (Page 6)

Questions 26–30

1. 26foundation · medium

   What is the primary purpose of performance monitoring in an AI workload?

   • ATo identify inefficiencies and guide optimization efforts.
   • BTo reduce the cost of training.
   • CTo increase the accuracy of the model.
   • DTo automatically fix all performance issues.
   Reveal Solution Discussion Select an answer first
2. 27foundation · medium

   In which scenario is model parallelism more appropriate than data parallelism?

   • AWhen the training is I/O bound.
   • BWhen the model is too large to fit into the memory of a single GPU.
   • CWhen the dataset is too large to fit into memory.
   • DWhen the model has a small number of parameters.
   Reveal Solution Discussion Select an answer first
4. 28foundation · medium

   Which technique directly increases GPU utilization by processing multiple samples simultaneously?

   • ABatching
   • BLearning rate scheduling
   • CGradient clipping
   • DMixed precision training
   Reveal Solution Discussion Select an answer first
5. 29foundation · medium

   Which technique reduces communication overhead in distributed training by sending only the gradients that changed significantly?

   • AIncreasing the number of GPUs
   • BUsing a larger batch size
   • CGradient compression
   • DEnabling checkpointing
   Reveal Solution Discussion Select an answer first
6. 30foundation · medium

   What is the primary purpose of caching in an AI workload?

   • ATo reduce the model's memory footprint.
   • BTo increase the size of the training dataset.
   • CTo improve model accuracy.
   • DTo reduce data access latency by storing frequently used data closer to the compute.
   Reveal Solution Discussion Select an answer first