Maswali yanayounganishwa na mwongozo · Kati Kiwango

Maswali ya LAION na Fungua Hifadhidata

Jaribu jinsi LAION-5B inavyooanisha, kuchuja na kupanga picha na manukuu kutoka kwa wavuti, ikijumuisha upachikaji wake wa CLIP na vikwazo vya uzalishaji tena.

Njia za mwongozo zinazohusianaLaion na Fungua Hifadhidata
Swali 1 ya 8

LAION inasambaza nini kimsingi katika hifadhidata zake za maandishi ya picha?

Hakiki kila swali8 maswali

Majibu na maelezo huonekana unapojibu maswali.

  1. LAION inasambaza nini kimsingi katika hifadhidata zake za maandishi ya picha?
    • URL na metadata zinazoelekeza kwenye picha kwenye wavuti
    • Faili za picha halisi
    • Manukuu ya maandishi pekee bila picha
    • Uzito wa mfano uliofundishwa mapema
  2. Takriban ni jozi ngapi za maandishi ya picha katika LAION-5B?
    • milioni 5
    • bilioni 5.85
    • milioni 500
    • 50 bilioni
  3. LAION ilitumia muundo gani kuchuja jozi za maandishi-picha kwa ubora wa upatanishi?
    • GPT-3
    • Stable Diffusion
    • CLIP
    • ResNet-50
  4. Ni modeli gani maarufu ya uwazi iliyofunzwa kwa kutumia data ya LAION?
    • BERT
    • AlphaFold
    • Whisper
    • Stable Diffusion
  5. What issue led LAION to withdraw LAION-5B in December 2023?
    • Researchers found URLs to images suspected to be child sexual abuse material
    • The dataset contained no English captions
    • The CLIP model could not read any image URLs
    • Every image in the dataset had been duplicated
  6. Je, ni udhaifu gani unaoendelea wa kiufundi unaotokana na LAION kuhifadhi URL pekee badala ya picha?
    • Uchujaji wa polepole wa CLIP
    • Uozo wa kiungo unaoharibu uwezo wa kuzaliana kwa wakati
    • Gharama za juu za kuhifadhi kwa LAION
    • Kutokuwa na uwezo wa kuauni lugha nyingi
  7. Which precomputed representation can support fast similarity search across LAION-5B?
    • Full copies of every source image
    • Audio transcripts aligned to each image
    • CLIP embeddings for the image-text data
    • The complete weights of every trained text-to-image model
  8. Which three language groupings does the LAION-5B paper describe?
    • English, Spanish, and French only
    • Text-only, video, and speech datasets
    • Training, validation, and test images
    • English, multilingual, and language-undetermined data