Maswali yanayounganishwa na mwongozo · Kati Kiwango

Maswali ya CLIP na Vielelezo vya Lugha ya Maono

Kagua uwakilishi ulioshirikiwa wa CLIP na uoanishaji tofauti, na lebo zisizo na picha, familia za programu za kusimba, hesabu za alama za kundi na unCLIP.

Njia za mwongozo zinazohusianaClip Na Vision Lugha Models
Swali 1 ya 8

Ni nini wazo kuu nyuma ya CLIP?

Hakiki kila swali8 maswali

Majibu na maelezo huonekana unapojibu maswali.

  1. Ni nini wazo kuu nyuma ya CLIP?
    • Inabana picha kuwa faili ndogo
    • Kuweka picha na maandishi kwenye nafasi sawa ya upachikaji iliyoshirikiwa
    • Inazalisha miundo ya 3D kutoka kwa picha
    • Kutafsiri hotuba kuwa maandishi
  2. CLIP hutumia mbinu gani ya mafunzo?
    • Mafunzo ya kuimarisha kutoka kwa maoni ya kibinadamu
    • Kuweka lebo kwa mkono kila kitu katika kila picha
    • Mafunzo pinzani ambayo huleta kulinganisha jozi za maandishi ya picha pamoja
    • Kunakili uzito kutoka kwa muundo wa hotuba
  3. What does zero-shot classification mean in the original CLIP transfer setup?
    • Using text-defined candidate categories without new task-specific supervised classifier fitting
    • Using a model that never received any pretraining data
    • Running without any computation or electricity
    • A guarantee that every category is classified correctly
  4. Je, CLIP hupima vipi ikiwa picha na manukuu yanalingana?
    • Kwa kulinganisha saizi ya faili
    • Kwa kuhesabu saizi za kila rangi
    • Kwa kuangalia muhuri wa muda wa picha
    • Kwa kufanana kwa cosine kati ya vekta zao
  5. Je, CLIP ilifunzwa kuhusu data ngapi?
    • Picha elfu chache zenye lebo ya mkono
    • Maandishi pekee, hakuna picha
    • Takriban jozi milioni 400 za maandishi ya picha kutoka kwa wavuti
    • Mkusanyiko mmoja wa makumbusho ulioratibiwa
  6. How does the cited unCLIP image-generation paper use CLIP representations?
    • The original standalone CLIP encoder directly outputs every image pixel
    • A prior supplies CLIP image embeddings to a generative decoder
    • It removes all image representations and uses only filenames
    • It uses CLIP only as a network-encryption protocol
  7. Which image-encoder families were evaluated in the original CLIP paper?
    • Only recurrent text encoders
    • Only a mandatory linear pixel average
    • Only a raw audio codec
    • Modified ResNets and Vision Transformers
  8. A CLIP batch contains 32 paired images and texts. How many entries are in its full cross-modal score matrix?
    • 1024
    • 32
    • 64
    • 992