Kediska Hagaajinta Kaydka KV

Jawaabaha iyo sharraxaadaha waxay soo baxayaan markaad qaadanayso kediska.

1. Waa maxay kaydka KV iyo sababta?
   • Furayaasha dareenka iyo qiyamka horay loo xisaabiyay si aan dib loo xisaabin tillaabo kasta
   • Xogta tababbarka moodeelka ee dib u tababbarka
   • Erayada sirta ah ee isticmaalaha si loo xaqiijiyo
   • Soo saarida qoraalka ugu dambeeya kaliya
2. Muxuu kaydka KV u noqon karaa dhibaato xusuusta?
   • Waxay si joogto ah u kharribtaa GPU-da
   • Waxay ku kortaa dhererka isku xigxiga, cabbirka dufcadda, iyo lakabyada, mararka qaarkoodna waxay dhaaftaa miisaanka moodeelka
   • Waxa kaliya oo ku shaqeeya qoraal ka gaaban 10 kelmadood
   • Waxay nuqul ka sameysaa nidaamka hawlgalka
3. Waa maxay fikradda nidaamka hawlgalka ee PagedAttention amaahdo?
   • Utility defragmentation Disk
   • Iskaanka fayraska
   • Xusuus-qurux badan oo leh baloogyo cabbir go'an
   • Miisaanka xallinta shaashadda
4. Sidee Su'aalo-kooxeed iyo Feejignaan Su'aalo-badan u dhimaan cabbirka kaydka KV?
   • Adigoo tirtiraya habka dareenka
   • Adoo kaydinta kaydka sawir ahaan ahaan
   • By tababar ku saabsan xog yar
   • Adigoo u ogolaanaya madax badan oo su'aalo waydiin ah inay wadaagaan madax furaha/qiimaha ka yar
5. Waa maxay hal faa'iido oo wadaaga horgalaha ee kaydka KV?
   • Codsiyada leh isla markiiba bilawga waxay dib u isticmaali karaan baloogyadii la kaydiyay
   • Waxay siraysaa qaabka
   • Si isku mid ah ayay u dhimaysaa codsi kasta
   • Waxay meesha ka saartaa baahida dareenka
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