Idanwo ti o sopọ mọ itọsọna · Alabọde Ipele

MaskGIT Parallel Tokini Idanwo Iyipada koodu

Check MaskGIT's masked-token training and iterative parallel decoding, including its confidence-based remasking schedule and bidirectional context.

Jẹmọ itọsọna awọn ọnaMaskgit Parallel Tokini Iyipada koodu
Ibeere 1 ti 8

Bawo ni MaskGIT ṣe n ṣe iyipada awọn ami aworan yatọ si oluyipada VQGAN?

Awotẹlẹ gbogbo ibeere8 ibeere

Awọn idahun ati awọn alaye han bi o ṣe n mu ibeere naa.

  1. Bawo ni MaskGIT ṣe n ṣe iyipada awọn ami aworan yatọ si oluyipada VQGAN?
    • O ṣe asọtẹlẹ awọn ami ni afiwe ati ṣe atunṣe lori awọn igbesẹ diẹ
    • O ṣe asọtẹlẹ ami-ami kan ni akoko kan ni aṣẹ raster
    • O nlo ko si transformer
    • O ṣe ipilẹṣẹ ọrọ nikan
  2. Idi ikẹkọ wo ni MaskGIT yawo lati awọn awoṣe ede?
    • Asọtẹlẹ ọrọ-nikan
    • Asọtẹlẹ ami ti o boju, bii BERT
    • Pipa Pipa Pipa
    • Ẹkọ imudara lati awọn esi eniyan
  3. Lakoko iran, awọn ami wo ni MaskGIT ṣe ni aṣetunṣe kọọkan?
    • Awọn ti o kere julọ-igbekele
    • ID àmi
    • Awọn asọtẹlẹ igbẹkẹle ti o ga julọ
    • Gbogbo àmi ni ẹẹkan lori igbese kan
  4. In the original MaskGIT generation examples, roughly how many refinement iterations were often used?
    • Exactly one iteration for every image and task
    • Hundreds of sequential token predictions
    • A fixed number that cannot be changed by the user
    • About 8 to 12 in those reported configurations
  5. Iṣeto wo ni o ṣakoso melo ni awọn ami-ami MaskGIT ṣe afihan igbesẹ kọọkan?
    • A cosine masking iṣeto
    • Ofin laini kan ti n ṣafihan gbogbo ni ẹẹkan
    • Ko si iṣeto; laileto ni
    • A ti o wa titi ọkan àmi fun igbese
  6. Kini idi ti akiyesi bidirectional ṣe iranlọwọ MaskGIT?
    • O tọju aworan naa lati awoṣe
    • O jẹ ki ami kọọkan rii gbogbo aworan apa kan fun ipo to dara julọ
    • O nikan wulẹ ni sẹyìn àmi
    • O yọ awọn nilo fun eyikeyi àmi
  7. What token grid does MaskGIT begin with when generating an image?
    • A fully decoded image that only needs color correction
    • A left-to-right stream with no masked positions
    • A grid with every image token masked
    • A grid of text embeddings from the prompt only
  8. How can MaskGIT fill a masked region during image inpainting?
    • Discard the observed image and generate an unrelated full image
    • Use only a left-to-right text decoder with no image tokens
    • Replace every visible token with a random codebook entry
    • Keep observed image tokens as context and iteratively predict the masked positions