Practice
Learn
← Quit
0 / 0 correct
LLMs & AI
Hard
1 of 10
What problem does PagedAttention (vLLM) solve?
A
Model loading time
B
Fragmentation of KV cache memory: it stores KV in fixed-size blocks like virtual-memory pages, allocated on demand
C
Attention numerical overflow
D
Slow tokenization