Maswali yanayounganishwa na mwongozo · Kati Kiwango

Maswali ya Miundo ya Kiwango cha Byte Isiyo na Tokenizer

Elewa miundo ya lugha ya biti ghafi, viraka vya BLT kulingana na entropy, na ufanisi na utendakazi wa uondoaji wa msamiati wa tokeni zisizobadilika.

Njia za mwongozo zinazohusianaVielelezo vya Kiwango cha Byte cha Tokenizer Bila Malipo
Swali 1 ya 8

Mtindo wa kiwango cha baiti bila tokeniza husoma nini kama vitengo vyake vya ingizo?

Hakiki kila swali8 maswali

Majibu na maelezo huonekana unapojibu maswali.

  1. Mtindo wa kiwango cha baiti bila tokeniza husoma nini kama vitengo vyake vya ingizo?
    • Baiti ghafi (thamani za UTF-8 0-255)
    • Maneno madogo kutoka kwa msamiati thabiti
    • Maneno yote yaliyotenganishwa na nafasi
    • Upachikaji uliokokotolewa mapema pekee
  2. Ni nini kikwazo kikuu cha vitendo cha kulisha ka mbichi kwa kibadilishaji cha kawaida?
    • Byte haziwezi kuwakilisha maandishi yasiyo ya Kiingereza
    • Mlolongo unakuwa mrefu zaidi, na hivyo kuongeza gharama ya kuhesabu
    • Kuna thamani nyingi sana za baiti za kupachikwa
    • Byte hupoteza maana ya neno kabisa
  3. Je, Kibadilishaji cha Byte Latent (BLT) huamua vipi mahali pa kuweka baiti katika viraka?
    • Kwa kugawanyika kwa kila mhusika wa nafasi
    • Kwa kutumia kiraka cha saizi isiyobadilika ya ka 4
    • Kulingana na utabiri wa kiwango cha baiti (kutokuwa na uhakika)
    • Nasibu wakati wa kila pasi ya mbele
  4. Why can a fixed subword vocabulary use more tokens for some languages than others?
    • Subword tokenizers only accept ASCII
    • All languages have identical token boundaries
    • They cannot represent numerals
    • Its token inventory may reflect the language distribution in the data used to build it
  5. What does removing a fixed tokenizer guarantee about byte-level input?
    • The same byte alphabet can represent text from different scripts, while learned language ability still depends on data
    • The model understands every language without training data
    • It guarantees equal performance for all scripts
    • The model no longer needs a vocabulary or learned embeddings
  6. Ni modeli gani ya awali iliyoonyesha usindikaji wa kiwango kidogo lakini ilikosolewa kwa kuwa polepole?
    • BERT
    • KwaT5
    • Neno2Vec
    • GPT-2
  7. What representations does BLT’s large global Transformer process?
    • Individual words from a fixed subword vocabulary only
    • Pikseli za picha
    • Dynamically sized byte patches
    • Robot joint-angle commands
  8. Which components surround the global Transformer in BLT?
    • A reward model and a value network
    • A speech vocoder and a spectrogram
    • A camera and a lidar sensor
    • A local byte encoder and a local byte decoder