Guide-yakabatana mibvunzo · Pakati Nhanho

Cross-Attention Quiz

Understand how cross-attention lets one sequence or modality query representations from another sequence or modality in transformer models.

Nzira dzekutungamira dzinoenderanaCross Attention
Mubvunzo 1 ye 8

Ndeupi mutsauko mukuru pakati pekuzvitarisa uye kutarisisa?

Tarisa mubvunzo wese8 mibvunzo

Mhinduro uye tsananguro zvinoonekwa paunenge uchitora bvunzo.

  1. Ndeupi mutsauko mukuru pakati pekuzvitarisa uye kutarisisa?
    • Mukuyambuka-kutarisisa, mibvunzo inobva kune imwe kutevedzana uye makiyi / kukosha kubva kune imwe
    • Cross-attention haishandisi softmax
    • Kuzvitarisa kunoshanda chete pamifananidzo
    • Muchinjikwa-kutarisisa haugone kudzidziswa ne gradient descent
  2. Mune encoder-decoder Transformer, mibvunzo yekuyambuka-yakatarisa inobvepi?
    • Encoder zvabuda
    • Decoder mamiriro azvino
    • Iwo akaiswa embeddings chete
    • Yakasiyana isina matrix
  3. Sei iwo maviri akatevedzana mukuchinjika-kutarisa achigona kuve nehurefu hwakasiyana kana modalities?
    • Nekuti kutarisa kunoda zvakaenzana-kureba zvinopinza
    • Nekuti inofuratira imwe yeanoteedzana zvachose
    • Nekuti mibvunzo nemakiyi zvinobva kune akasiyana, akazvimirira masosi
    • Nekuti inoisa zvese kune yakagadziriswa saizi kutanga
  4. Ko Stable Diffusion inoshandisa sei kuyambuka-kutarisisa?
    • Kumanikidza mifananidzo kuita madiki mafaira
    • Kushandura mavara pakati pemitauro
    • Kubvisa ruzha kubva pakuteerera
    • Kugadzirisa nzvimbo dzakagadzirwa dzemifananidzo pane iyo text prompt
  5. Ndeupi mashandiro esvomhu anodyidzana nekutarisa wega?
    • Yakayerwa dot-chigadzirwa kutarisisa nesoftmax
    • Convolution pamusoro pe pixels
    • Recurrent state updates
    • Hard-coded yekutarisa matafura
  6. What gate does Flamingo use to control information flow through its cross-attention layers?
    • A hard threshold initialized to one
    • A tanh gate initialized to zero
    • A fixed residual weight of 0.5
    • A learned vocabulary lookup table
  7. What input sequence structure does Flamingo’s paper say its visual-language model can handle?
    • Only one image followed by a fixed-length label
    • Only text tokens with no visual inputs
    • Visual and textual data interleaved in arbitrary sequences
    • Only video frames with no text
  8. What is one purpose of cross-attention in Flamingo’s visual-language architecture?
    • It replaces all text tokens with pixels
    • It prevents any information from passing between modalities
    • It performs only image resizing
    • It lets language-model layers incorporate visual features from image/video inputs