PagedAttention and vLLM Quiz

Answers and explanations appear as you take the quiz.

1. What does the 'KV cache' store during text generation?
   • The model's training data
   • The user's login credentials
   • The key and value vectors for past tokens so new tokens can attend to them
   • A compressed copy of the model weights
2. What operating-system concept inspired PagedAttention?
   • Disk defragmentation tools
   • Virtual memory paging
   • File compression
   • CPU branch prediction
3. What problem with traditional KV cache allocation does PagedAttention solve?
   • Slow internet connections
   • Lack of training data
   • Weak password security
   • Memory waste from reserving large contiguous blocks for each request
4. How does PagedAttention let multiple outputs share memory for a common prompt?
   • Different sequences point to the same physical pages via copy-on-write
   • It deletes the prompt after the first use
   • It stores the prompt on disk
   • It retrains the model for each output
5. Where was vLLM and PagedAttention originally developed?
   • A 1990s academic lab
   • UC Berkeley (published 2023)
   • A hardware vendor's marketing team
   • An anonymous online forum
6. What is the main practical benefit of vLLM for someone serving an LLM?
   • Larger model accuracy on benchmarks
   • Automatic data labeling
   • Free GPU hardware
   • Higher throughput and more concurrent requests per GPU
7. A sequence’s logical KV block 3 resides in a different physical location. What tells PagedAttention where to read it?
   • The sequence’s block table mapping logical blocks to physical blocks
   • The model’s class-label list
   • The learning-rate scheduler
   • A requirement that every block follow block 2 contiguously
8. A PagedAttention cache ends with a half-filled block. Which statement about memory waste is accurate?
   • Paging guarantees every allocated byte contains a token
   • The entire model must be copied for each unused slot
   • Unused slots can remain in that final block despite reduced allocation fragmentation
   • The block table cannot represent a partial sequence