Where to start

AI Engineering & Accelerated Computing

A short check to suggest a starting chapter. It's advice only: your answers aren't saved, nothing gets unlocked, and it doesn't count as evidence of a skill. It hasn't been tested on learners yet, and the questions are in English.

Why is a vectorised NumPy sum usually faster than a Python loop?

Python loops are limited to one million steps

The work runs in compiled code over contiguous data

NumPy always uses a GPU

Why can a GPU timing be wrong if you only read the clock around a model call?

GPU work is asynchronous, so the call may return before it finishes

Clocks don't work while a GPU is running

GPU timings are always identical

What does dynamic batching usually trade?

Accuracy for security

Memory for accuracy

Some added waiting per request for higher throughput

A training job leaves the GPU idle 60 % of the time. Where do you look first?

A larger model

The data-loading pipeline and transfers

The learning rate

In CUDA, what does a kernel launch <<<N, 256>>> request?

N blocks of 256 threads each

N bytes of memory per 256 threads

N GPUs with 256 cores each

Why add a deadline to a call to a model server?

It makes the model more accurate

It stops slow calls from tying up resources without limit

It replaces the need for monitoring

Understand

Learn by doing. Understand why.

Learn

Explore the programmes Higher education High school Programming Data Science & AI Search

Understand

The method Create an account Log in

Help and privacy

Privacy Contact