Moderate Docker Question 90 of 221

How do you keep a GPU training image reproducible?

MLOps track · Speak this in 60–90 seconds · Faridabad & Delhi NCR

PICTURE THIS: DATA SPLIT

Train 70%Val 15%Test 15%

Fit on train, tune on val, report on test once.

Simple meaning

Pin the CUDA base image digest, driver-compatible tags, and Python lockfile, and record nvidia-smi output in the run.

WHY — Docker instead of guessing?

Why interviewers care about Docker:

This is a process

question about Docker.

Panels listen for order,

trade-offs, and what you would actually do on a MLOps project - not buzzwords.

Stay structured

Name the idea, why it exists, then one short example.

Close cleanly

End with when you use it and one common pitfall.

STEPS — What happens step by step?

Before you speak the answer, walk the interviewer through these steps:

1
Pin the CUDA base
image digest, driver-compatible tags, and Python lockfile, and record nvidia-smi output in the run.
2
Multi-stage builds can compile
wheels then copy them into a slimmer runtime.
3
You still document the
host driver range the image supports.
4
Give an example
One tiny concrete case you can say aloud.
5
Common mistake
What juniors usually get wrong.
6
Close
When you pick this over the alternative.

EXAMPLE — See it in action

Here's a short line you can speak, broken into clear beats:

Say this line

“Multi-stage builds can compile wheels then copy them into a slimmer runtime.”

Break into beats

Multistagebuildscancompilewheels

Speaking order

2987408337471632900

Note: Adapt this scaffold to your own project — keep it under 60–90 seconds.

Key takeaway

Pin the CUDA base image digest, driver-compatible tags, and Python lockfile, and record nvidia-smi output in the run. Multi-stage builds can compile wheels then copy them into a slimmer runtime.