2.4 Leveraging Gemini Enterprise Agent Platform for End-To-End ML Workflows PROFESSIONAL-CLOUD-ARCHITECT Practice Questions (Page 4)

Questions 16–20

1. 16foundation · easy

   How are GPUs and TPUs integrated into model serving on Agent Platform?

   • ABy configuring the serving infrastructure to use accelerators for low-latency inference.
   • BBy embedding the accelerator directly into the model file.
   • CBy requiring all serving requests to be processed in batch mode.
   • DBy using them only during training and not during serving.
   Reveal Solution Discussion Select an answer first
2. 17foundation · easy

   What is the role of GPUs and TPUs in ML model training on Agent Platform?

   • AThey are used to visualize training metrics.
   • BThey are used to store training data in a distributed manner.
   • CThey provide specialized hardware acceleration to speed up the training process.
   • DThey are required for all ML models, regardless of size.
   Reveal Solution Discussion Select an answer first
4. 18application · medium

   A company wants to run a training job that requires a specific version of a GPU driver and a custom networking setup. They want to use AI Hypercomputer. What is the best way to achieve this?

   • AUse a custom container with the required driver and configure the job to use a specific network.
   • BUse a TPU instead of a GPU to avoid driver issues.
   • CRequest a special quota for the GPU driver version.
   • DUse a standard pre-built container and hope the driver version is compatible.
   Reveal Solution Discussion Select an answer first
5. 19foundation · easy

   What is a best practice for running large-scale AI model training on Agent Platform?

   • AStoring all training data in memory on each VM.
   • BAvoiding the use of checkpoints to save time.
   • CRunning the entire training job on a single VM with a large CPU.
   • DUsing distributed training to parallelize the workload across multiple accelerators.
   Reveal Solution Discussion Select an answer first
6. 20foundation · easy

   Which step is part of preparing data for integration with Agent Platform?

   • AStoring data only in on-premises storage to ensure security.
   • BConverting all data to unstructured format for flexibility.
   • CDeleting all data that contains missing values.
   • DValidating the data schema and quality before using it in a pipeline.
   Reveal Solution Discussion Select an answer first