GPU support in containers in MLOps - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to specify the GPU runtime when running a Docker container.

MLOps

docker run --gpus [1] nvidia/cuda:11.0-base nvidia-smi

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

Acpu

Bnone

Call

Ddefault

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'none' disables GPU access.

Using 'cpu' is invalid for GPU runtime.

Using 'default' does not specify GPU usage.

Using --gpus all allows the container to access all GPUs on the host.

2fill in blank

medium

Complete the Dockerfile line to install NVIDIA CUDA toolkit inside the container.

MLOps

RUN apt-get update && apt-get install -y [1]

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

Apython3

Bgit

Cnginx

Dcuda-toolkit-11-0

Attempts:

3 left

💡 Hint

Common Mistakes

Installing unrelated packages like python3 or nginx.

Forgetting to update apt-get before installing.

Installing cuda-toolkit-11-0 provides CUDA support inside the container.

3fill in blank

hard

Fix the error in the Docker run command to enable GPU support.

MLOps

docker run --runtime=[1] nvidia/cuda:11.0-base nvidia-smi

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

Adocker

Bnvidia

Cdefault

Dgpu

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'docker' or 'default' runtimes disables GPU support.

Using 'gpu' is not a valid runtime name.

The correct runtime for NVIDIA GPU support is nvidia.

4fill in blank

hard

Fill both blanks to create a Docker Compose service with GPU support.

MLOps

services:
  gpu-service:
    image: nvidia/cuda:11.0-base
    deploy:
      resources:
        reservations:
          devices:
            - driver: [1]
              count: [2]
              capabilities: [gpu]

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

Anvidia

Ball

C2

Ddefault

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'default' as driver disables GPU support.

Using 'all' as count is invalid; count expects a number.

The device driver should be nvidia and count specifies how many GPUs to reserve, e.g., 2.

5fill in blank

hard

Fill all three blanks to write a Dockerfile snippet that sets environment variables for CUDA and runs a GPU test.

MLOps

ENV CUDA_VERSION=[1]
ENV PATH=/usr/local/cuda-[2]/bin:${PATH}
RUN nvidia-smi --query-gpu=name,memory.total --format=csv > [3]

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

A11.0

C/tmp/gpu_info.csv

D/var/log/gpu.log

Attempts:

3 left

💡 Hint

Common Mistakes

Mismatching CUDA versions in ENV variables.

Saving output to a non-writable or unrelated file path.

Set CUDA version to 11.0, update PATH accordingly, and save GPU info to /tmp/gpu_info.csv.