Maswali yanayounganishwa na mwongozo · Ngumu Kiwango

Maswali ya Mbinu za Kuzingatia

Jaribio la umakini wa hoja-ufunguo-thamani, kuongeza kiwango cha upeo laini, vichwa sambamba, gharama mnene ya umakini, hesabu za thamani iliyopimwa, na ufichaji wa kisababishi kwa ajili ya uzalishaji.

Njia za mwongozo zinazohusianaTaratibu za Kuzingatia
Swali 1 ya 8

What does token attention combine when forming an output representation?

Hakiki kila swali8 maswali

Majibu na maelezo huonekana unapojibu maswali.

  1. What does token attention combine when forming an output representation?
    • Permanent copies of the entire conversation on disk
    • Value information from allowed positions using computed weights
    • Only manually chosen keywords with no learned scores
    • A translation dictionary that removes every other token
  2. Ni karatasi gani ya kihistoria ya 2017 iliyoanzisha usanifu wa kibadilishaji?
    • 'Kujifunza kwa Mabaki ya Kina kwa Utambuzi wa Picha'
    • 'Kucheza Atari na Mafunzo ya Kuimarisha Kina'
    • 'Tahadhari Ndio Wote Unaohitaji'
    • 'Generative Adversarial Networks'
  3. Je, ni vekta gani tatu zilizokokotwa kwa kila ishara inayozingatiwa?
    • Hoja, ufunguo na thamani
    • Ingizo, pato, na upendeleo
    • Wastani, wastani, na hali
    • Kisimbaji, avkodare, na hasara
  4. Katika fomula softmax(QK^T / √d_k) V, kwa nini ugawanye kwa √d_k?
    • Ili kubadilisha matokeo kuwa nambari kamili
    • Ili kuondoa nambari hasi
    • Kufanya modeli kukimbia kwenye GPU
    • Ili kuzuia alama za bidhaa za nukta zisiwe kubwa sana kabla ya softmax
  5. Why does dense self-attention involve growing pairwise work for very long token sequences?
    • Kwa sababu ni lazima kupakua data mpya kutoka kwa mtandao
    • Kwa sababu maandishi marefu huwa na maneno adimu
    • Kwa sababu gharama hukua na mraba wa urefu wa mfuatano (O(n²))
    • Kwa sababu inabadilika hadi lugha ya programu polepole
  6. What capacity does multi-head attention provide through separate learned projections?
    • A guarantee that every head has one unique named meaning
    • Several attention representations that can capture different relationships in parallel
    • Removal of every token embedding
    • A reduction of vocabulary size by definition
  7. In a scalar attention example, values two and six receive weights three-quarters and one-quarter. What is their weighted output?
    • Nane
    • Nne
    • Mbili
    • Tatu
  8. What does a causal attention mask prevent in an autoregressive text decoder?
    • Using later token positions when forming the current position's representation
    • Using any previous token
    • Computing any query or key vectors
    • Learning all model parameters