Module 3 Quiz

Test your understanding of vLLM, PagedAttention, continuous batching, the CPU track and its NUMA patch, and quantization from the lesson and lab.

Which mechanisms let vLLM achieve roughly 3× the throughput of a naive server under load? (Select all that apply)

(select all that apply)

The lesson compares PagedAttention to a familiar operating-system technique. Which one, and why?

You built a client against Ollama in M2. Now you want it to talk to the vLLM server from this lab instead. What has to change?

The CPU Dockerfile patches cpu_worker.py with a one-line sed. What problem does that patch fix?

A colleague wants to fit a 7B model onto a smaller GPU and asks about quantization trade-offs. Which statement is accurate?