Practice
Learn
← Quit
0 / 0 correct
LLMs & AI
Hard
1 of 40
Prefill vs decode phases of LLM inference: which is accurate?
A
Decode processes the prompt in parallel
B
Prefill processes the whole prompt in parallel (compute-bound); decode generates one token at a time (memory-bandwidth-bound)
C
Prefill generates the answer; decode tokenizes the prompt
D
Both are compute-bound