An AI training job is running slower than expected. The infrastructure team suspects a compute bottleneck. The job involves processing high-resolution images with a complex convolutional neural network (CNN). They review the logs from the training script.
Epoch 1/100 - 3600s - loss: 1.253 - acc: 0.54
...
CPU_Utilization_During_Epoch: 99% (all cores)
GPU_0_Utilization_During_Epoch: 8%
GPU_1_Utilization_During_Epoch: 7%
...
Epoch 2/100 - 3610s - loss: 1.102 - acc: 0.61
What is the most likely cause of the poor training performance?

• A. The model is too simple and does not effectively utilize the GPU's parallel processing capabilities.
• B. The network is saturated, preventing the GPUs from receiving data.
• C. The training script is CPU-bound, likely performing data augmentation or preprocessing on the CPU instead of offloading it to the GPUs.
• D. The storage system cannot deliver data fast enough to the compute node.

The pod running the vector database on the Kubernetes cluster fails to start. An MLOps engineer runs 'kubectl describe pod vector-db-pod-0' and sees the following event message:
Events:
Type Reason Age From Message
- - - -
Warning FailedScheduling 30s default-scheduler 0/8 nodes are available: 8 node(s) did not match pod anti-affinity rules.
The pod's manifest contains the following 'affinity' definition:
affinity:
podAntiAffinity:
requiredDuringSchedulingIgnoredDuringExecution:
- labelSelector:
matchExpressions:
- key: app
operator: In
values:
- vector-db
topologyKey: "kubernetes.io/hostname"
What is the most likely reason the pod cannot be scheduled?

• A. The 'vector-db-pod-0' is requesting more GPU resources than are available on any node.
• B. The pod's anti-affinity rule prevents it from being scheduled on any node that is already running another pod with the label 'app=vector-db'.
• C. The 'topologyKey' is invalid; it should be 'failure-domain.beta.kubernetes.io/zone'.
• D. The PersistentVolumeClaim for the pod is in a 'Pending' state.

Which of the following platforms can be used to manage containerized AI workloads on Kubernetes? (Choose two)

• A. Google VertexAI
• B. KubeFlow
• C. TensorFlow Extended
• D. RunAI

An MLOps engineer is troubleshooting a failed Kubeflow pipeline step. The step was designed to create a clone of a dataset for a training job using the NetApp DataOps Toolkit. The pod logs for the failed pipeline step show the following:
Traceback (most recent call last):
File "create_clone.py", line 15, in <module>
clone_pvc(source_pvc_name="training-data-v2", new_pvc_name="train-job-34a-data") NameError: name 'clone_pvc' is not defined The engineer reviews the Python script for the pipeline step:
# create_clone.py
import os
from netapp_dataops.k8s import create_pvc
# Other code
print("Cloning source dataset for training run...")
clone_pvc(
source_pvc_name="training-data-v2",
new_pvc_name="train-job-34a-data"
)
print("Clone created successfully.")
What is the cause of the error?

• A. The Python script is attempting to use the 'clone_pvc' function, but it was not imported from the
  'netapp_dataops.k8s' library.
• B. The Kubernetes cluster is not running NetApp Trident.
• C. The source PVC 'training-data-v2' does not exist.
• D. The NetApp DataOps Toolkit is not installed in the container image used for this pipeline step.