Guide-yakabatana mibvunzo · Pakati Nhanho

Diffusion Transformers Quiz

Ongorora DiT chigamba tokeni, mamiriro, kuyera, vhidhiyo nzvimbo-nguva inomiririra, uye mutengo wekutarisisa wematokeni mazhinji.

Nzira dzekutungamira dzinoenderanaDiffusion Transformers
Mubvunzo 1 ye 8

Ndechipi chikamu chinotsiviwa neDiffusion Transformer ichienzaniswa nemhando dzechinyakare dzekuparadzira?

Tarisa mubvunzo wese8 mibvunzo

Mhinduro uye tsananguro zvinoonekwa paunenge uchitora bvunzo.

  1. Ndechipi chikamu chinotsiviwa neDiffusion Transformer ichienzaniswa nemhando dzechinyakare dzekuparadzira?
    • Iyo convolutional U-Net denoiser
    • The text encoder
    • Mufananidzo decoder
    • Hurongwa hweruzha
  2. Ko DiT inoshandura sei chifananidzo kuita chimwe chinhu chinoshandurwa neTransformer?
    • Inoshandura pixels zvakananga kumashoko
    • Iyo inotsemura iyo yakavanzika chifananidzo kuita zvigamba zvinobatwa setokens
    • Inoshandisa chiratidzo chimwe chemufananidzo wose
    • Inoshandura mufananidzo kuti uve audio kutanga
  3. Pepa rekutanga reDiT rakawanei nezve kuyera?
    • Performance plateaus nekukurumidza nehukuru
    • Mamodheru makuru anogara achiita zvakanyanya
    • Kuita kunonatsiridza kufanofungidzira sezvo saizi yemhando inokura uye saizi yechigamba ichidzikira
    • Saizi yechigamba haina mhedzisiro
  4. MaDiTs anowanzo pinza sei majekiseni senge nhanho yenguva uye mameseji ekukurumidza?
    • Nekuiisa semamwe mapixels
    • Nokudzidzisazve modhi nhanho imwe neimwe
    • Via rakasiyana convolutional bazi
    • Kuburikidza neadaptive layer normalization (adaLN)
  5. Chii chinonzi 'adaLN-zero' kutanga chinogonesa?
    • Inoita kuti yega yega Transformer block itange sechitupa basa rekudzidziswa kwakagadzikana
    • Inobvisa iyo normalization layers
    • Inoomesa modhi yese
    • Inobvisa mufananidzo wekuisa
  6. Ndeipi inozivikanwa sisitimu inoshandisa Transformer musana pane spacetime zvigamba zvekugadzira vhidhiyo?
    • GAN yechinyakare
    • OpenAI's Sora
    • A recurrent neural network
    • Muenzaniso wemuti wesarudzo
  7. Why can full self-attention become costly when a DiT processes many patches?
    • Its cost stays constant for any token count
    • It must retrain the U-Net at each denoising step
    • Its pairwise token comparisons grow quadratically with the number of tokens
    • It requires a separate language model for every patch
  8. After a DiT processes latent image patches with its Transformer blocks, what happens to the patch outputs?
    • As one text token for an entire video
    • As only a sequence of audio waveforms
    • As a separate convolutional U-Net for each frame
    • They are reassembled and decoded back into pixels