quiz bu lëkkale ak tegtal · Diggante Niveau

Quiz buy gëna baaxal cache KV

Saytu màggug cache KV, xaaj xët yi, séddoo laajte yuñ boole, anam yi ñuy jëfandikoowaat prefix yi, kantite, ak eskalaasioŋu mémoire ak contexte ak concurrence.

Yooni njiit yu jëm ciKv Cache Optimisation
Laaj 1 de 8

Luy denc ci cache KV te lu tax?

Xoolaatal laaj bu nekk8 laaj

Sooy def quiz bi, tontu yi ak leeral yi dañuy feeñ.

  1. Luy denc ci cache KV te lu tax?
    • Caabi ak valeur yiñ njëkka xayma ngir ñu baña xaymaat jéego bu nekk
    • Done yiñ tàggat ci model bi ngir tàggataat
    • Baatu-jàll jëfandikukat ngir xamle
    • Texte bi mujjee génn kese
  2. Lu tax cache KV mën nekk jafe-jafe mémoire?
    • Dafay yàq GPU bi ba fàww
    • Dafay màgg ak guddaayu toppalante bi, dayo lote bi, ak diisaay yi, yenn saa muy weesu diisaayu model bi
    • Du dox ludul ci mbind yu gàtt 10 baat
    • Dafay ñaari yoon sistem biy doxal
  3. Ban konsepti sistem jëfandikoo la PageAttention lebal?
    • Utilite yiy dindi disk yi
    • Scan antivirus
    • Paging mémoire virtuel ak ay blok yu am dayo bu takku
    • Dayo ekraŋ bi
  4. naka lay def ba Laajte yuñ boole ak Laajte yu bari di wàññi dayo cache KV?
    • Soo efaasee jumtukaayu bàyyi xel
    • Soo denc cache bi ni ay nataal
    • Ni ñuy tàggatee ci done yu néew
    • Soo bàyyee boppu laajte yu bari ñu séddoo boppu caabi/valeur yu néew
  5. Lan mooy benn ci njariñu séddoo prefix ci cache KV?
    • Laajte yi am benn laajtu ndoorte mën nañu jëfandikoowaat benn blok yiñ denc ci cache bi
    • Dafay encrypter model bi
    • Dafay yeexal bepp laaj ci anam wu tolloo
    • Dina dindi soxla bàyyi xel
  6. Compared with storing the same KV entries in 16-bit, what does 8-bit storage do to raw cache-value bytes?
    • It doubles the raw bytes
    • It roughly halves the raw bytes, before metadata and other runtime memory
    • It removes the key and value tensors entirely
    • It guarantees twice the inference speed
  7. With full-context attention, no prefix sharing, and a fixed model and cache dtype, what happens if sequence length and batch size both double?
    • It stays constant because model weights are fixed
    • It roughly doubles
    • Raw KV storage grows to roughly four times its previous size
    • It becomes one quarter as large
  8. What must generally match for two requests to reuse the same cached prefix blocks?
    • Only the user account name
    • The requests must have different model weights
    • The prompts must have the same character count but different tokens
    • Their tokenized prefix and relevant model/cache configuration must be compatible