Guide-yakabatana mibvunzo · Zvakaoma Nhanho

PagedAttention uye vLLM Quiz

Yedza PagedAttention KV mabhuroko, zvine musoro-kune-muviri mepu, uye kugovera, kusanganisira mutsauko pakati pecache memory savings uye yakazara modhi yekuchengetedza.

Nzira dzekutungamira dzinoenderanaPagedattention Uye Vllm
Mubvunzo 1 ye 8

Chii chinochengeterwa 'KV cache' panguva yekugadzira zvinyorwa?

Tarisa mubvunzo wese8 mibvunzo

Mhinduro uye tsananguro zvinoonekwa paunenge uchitora bvunzo.

  1. Chii chinochengeterwa 'KV cache' panguva yekugadzira zvinyorwa?
    • Data yekudzidzisa yemuenzaniso
    • Mvumo yekupinda yemushandisi
    • Makiyi uye kukosha mavheji ematokeni apfuura kuti tokens nyowani dzigone kuvatarisa
    • Ikopi yakamanikidzwa yezviyero zvemuenzaniso
  2. Ndeipi inoshanda-system pfungwa yakafuridzira PagedAttention?
    • Disk defragmentation zvishandiso
    • Virtual memory paging
    • File compression
    • CPU bazi kufanotaura
  3. Ndeipi dambudziko neyakagoverwa yeKV cache inogadziriswa PagedAttention?
    • Inononoka kubatanidza internet
    • Kushaikwa kwedata rekudzidzisa
    • Pasiwedhi chengetedzo isina simba
    • Memory tsvina kubva pakuchengetedza mabhuroko makuru akabatana pachikumbiro chega chega
  4. Sei PagedAttention inorega akawanda anobuda achigovera ndangariro kune yakajairwa kukurumidza?
    • Kutevedzana kwakasiyana kunonongedzera kune mamwe mapeji emuviri kuburikidza nekopi-pa-kunyora
    • Inodzima kukurumidza mushure mekutanga kushandiswa
    • Inochengeta kukurumidza pa diski
    • Inodzokorora zvakare modhi kune imwe neimwe inobuda
  5. VLLM nePagedAttention zvakagadziridzwa kupi?
    • A 1990s academic lab
    • UC Berkeley (yakabudiswa 2023)
    • Chikwata chekushambadzira chemutengesi wehardware
    • Iyo isingazivikanwe online foramu
  6. Ndeipi yakakosha bhenefiti yevLLM kune mumwe munhu anoshanda LLM?
    • Yakakura modhi yechokwadi pamabhenji
    • Kunyora data otomatiki
    • Yemahara GPU Hardware
    • Kupinda kwepamusoro uye zvimwe zvikumbiro zvinowirirana paGPU
  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