Maswali yanayounganishwa na mwongozo · Kati Kiwango

Maswali ya Kurekebisha Haraka

Pitia vidokezo laini (soft prompts) kwenye safu ya ingizo, usanidi wa muundo uliogandishwa, na matokeo yanayotegemea kiwango yaliyoripotiwa kwa ajili ya prompt tuning.

Njia za mwongozo zinazohusianaUrekebishaji wa haraka
Swali 1 ya 8

Je, 'uchochezi laini' katika urekebishaji wa haraka ni nini?

Hakiki kila swali8 maswali

Majibu na maelezo huonekana unapojibu maswali.

  1. Je, 'uchochezi laini' katika urekebishaji wa haraka ni nini?
    • Umejifunza vekta za kupachika ambazo hazihitaji ramani ya maneno halisi
    • Maagizo ya maandishi yenye adabu
    • Nakala ndogo iliyosafishwa ya mfano
    • Seti ya kumbukumbu za matokeo yaliyogandishwa
  2. Je, urekebishaji wa haraka huongeza wapi vekta zake zinazoweza kufunzwa?
    • Katika kila safu ya transformer
    • Kwenye safu ya upachikaji wa ingizo pekee
    • Tu kwenye safu ya pato
    • Ndani ya hali ya optimizer
  3. Je, karatasi asili ilionyesha athari gani ya kushangaza kuhusu ukubwa wa mfano?
    • Urekebishaji wa haraka unazidi kuwa mbaya kadiri miundo inavyokua
    • Mizani haina athari kwa urekebishaji wa papo hapo
    • Kadiri miundo inavyokuwa kubwa, urekebishaji wa papo hapo unakaribia ubora kamili wa usanifu
    • Miundo mikubwa haiwezi tena kusasishwa kwa haraka
  4. Nani alianzisha urekebishaji wa haraka?
    • DeepMind mnamo 2016
    • Anthropic mwaka wa 2023
    • Microsoft mwaka wa 2019
    • Watafiti katika Google (Lester, Al-Rfou, Constant) mnamo 2021
  5. Je, urekebishaji wa haraka huwezeshaje kuhudumia kazi nyingi kutoka kwa modeli moja?
    • Kila kazi hutumia kidokezo chake kidogo huku kielelezo kikigandishwa
    • Inarudisha muundo mzima kwa kila ombi
    • Inafuta kazi ambazo hazijatumiwa kutoka kwa uzani
    • Inahitaji GPU tofauti kwa kila kazi
  6. How does Lester and colleagues’ paper describe prompt tuning in relation to prefix tuning?
    • As a replacement for frozen language models
    • As a simpler related method that the paper compares with prefix tuning
    • As a tokenizer with no trainable parameters
    • As a method that updates every model weight
  7. Which model family did Lester and colleagues use to study prompt tuning across scales?
    • WaveNet
    • AlphaFold2
    • T5
    • Msitu wa Kutengwa
  8. How are soft prompts optimized in prompt tuning?
    • By changing all model weights for each task
    • By replacing embeddings with hard-coded words only
    • By training a new tokenizer for every example
    • By backpropagation through a frozen language model to update the prompt vectors