Docker containerization in ML Python - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

Output of Dockerfile RUN command

What will be the output when building this Dockerfile snippet?

ML Python

FROM python:3.8-slim
RUN echo "Hello from Docker" > /message.txt
RUN cat /message.txt

A Hello from Docker

B Hello from Docker\nHello from Docker

C No output

D Error: /message.txt not found

💡 Hint

RUN commands execute during build and output their results in the build logs.

The RUN command executes shell commands during image build. The first RUN writes the message, the second RUN reads and outputs it.

❓ Model Choice

intermediate

Choosing the best Docker base image for ML model deployment

You want to deploy a TensorFlow model with GPU support in a Docker container. Which base image is the best choice?

A ubuntu:20.04

B python:3.9-slim

C tensorflow/tensorflow:latest-gpu

D node:16-alpine

💡 Hint

Look for images that include TensorFlow and GPU support.

The tensorflow/tensorflow:latest-gpu image includes TensorFlow with GPU support pre-installed, making it ideal for deploying GPU models.

❓ Hyperparameter

advanced

Optimizing Docker container for faster ML training

Which Dockerfile instruction helps reduce image size and speeds up ML training container startup?

A Combining RUN commands with && to reduce layers

B Installing all packages in separate RUN commands

C Using multiple RUN commands separately

D Using COPY after RUN commands

💡 Hint

Fewer layers in Docker images reduce size and improve startup.

Combining RUN commands with && creates fewer layers, reducing image size and improving container startup time.

🔧 Debug

advanced

Debugging Docker container failing to access GPU

You built a Docker container for ML training with GPU support, but inside the container, the GPU is not detected. What is the most likely cause?

A The host machine does not have a GPU

B The container was not run with the --gpus flag

C The Docker image does not have CUDA libraries installed

D The Dockerfile uses python:3.8-slim base image

💡 Hint

GPU access requires special runtime flags when running containers.

To access GPUs inside a container, you must run it with the --gpus flag; otherwise, the container cannot see the GPU hardware.

❓ Metrics

expert

Measuring Docker container startup time impact on ML inference latency

You want to measure how Docker container startup time affects ML model inference latency in production. Which approach is best?

A Ignore container startup time as it does not affect inference latency

B Include container startup time in latency measurement for each inference request

C Measure container startup time separately and add average to inference latency

D Measure inference latency only after container is fully started

💡 Hint

Inference latency should reflect only the model prediction time during normal operation.

Inference latency metrics typically exclude container startup time because startup happens once, not per request.