Kubernetes for ML workloads in MLOps - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to create a Kubernetes pod that runs a machine learning container.

MLOps

apiVersion: v1
kind: Pod
metadata:
  name: ml-pod
spec:
  containers:
  - name: ml-container
    image: [1]

Drag options to blanks, or click blank then click option'

Atensorflow/tensorflow:latest

Bubuntu:latest

Cnginx:alpine

Dmysql:5.7

Attempts:

3 left

💡 Hint

Common Mistakes

Using a web server image like nginx instead of an ML image.

Using a database image like mysql which is unrelated to ML workloads.

The correct image for running ML workloads is the TensorFlow image, which contains ML libraries.

2fill in blank

medium

Complete the code to specify resource limits for the ML container in the pod.

MLOps

spec:
  containers:
  - name: ml-container
    resources:
      limits:
        cpu: [1]

Drag options to blanks, or click blank then click option'

A2Gi

B2

C2MB

D500m

Attempts:

3 left

💡 Hint

Common Mistakes

Using memory units like '2Gi' for CPU limits.

Using plain numbers without units for CPU limits.

CPU limits are specified in cores or millicores; '500m' means 0.5 CPU cores.

3fill in blank

hard

Fix the error in the YAML to mount a volume for ML data inside the container.

MLOps

spec:
  containers:
  - name: ml-container
    volumeMounts:
    - name: data-volume
      mountPath: [1]
  volumes:
  - name: data-volume
    persistentVolumeClaim:
      claimName: ml-data-pvc

Drag options to blanks, or click blank then click option'

A/var/lib/data

B/data-volume

C/mnt/data

D/etc/data

Attempts:

3 left

💡 Hint

Common Mistakes

Using mount paths that are not directories or reserved system paths.

Using the volume name as the mount path.

The mountPath should be a directory inside the container where data is accessed, commonly '/mnt/data'.

4fill in blank

hard

Fill both blanks to create a Kubernetes Job that runs an ML training script once.

MLOps

apiVersion: batch/v1
kind: Job
metadata:
  name: ml-training-job
spec:
  template:
    spec:
      containers:
      - name: trainer
        image: [1]
        command: ["python", [2]]
      restartPolicy: Never

Drag options to blanks, or click blank then click option'

Aml-training-image:latest

B"train.py"

Capp.py

Dtensorflow/tensorflow:latest

Attempts:

3 left

💡 Hint

Common Mistakes

Using a generic TensorFlow image without the training script.

Using the wrong script name in the command.

The job uses a custom ML training image and runs the 'train.py' script with python.

5fill in blank

hard

Fill all three blanks to define a Kubernetes Deployment for an ML model server with 3 replicas and environment variable.

MLOps

apiVersion: apps/v1
kind: Deployment
metadata:
  name: ml-model-server
spec:
  replicas: [1]
  selector:
    matchLabels:
      app: ml-server
  template:
    metadata:
      labels:
        app: ml-server
    spec:
      containers:
      - name: model-server
        image: [2]
        env:
        - name: MODEL_NAME
          value: [3]

Drag options to blanks, or click blank then click option'

A3

Bml-model-server:latest

C"image-classifier"

D1

Attempts:

3 left

💡 Hint

Common Mistakes

Setting replicas to 1 instead of 3.

Not quoting the environment variable value.

Using wrong image names.

The deployment runs 3 replicas of the model server container with the environment variable MODEL_NAME set to 'image-classifier'.