Laajteb posisioŋ ALiBi

Sooy def quiz bi, tontu yi ak leeral yi dañuy feeñ.

1. Luy ALiBi?
   • Fexe ak Bias Lineer
   • Bias lineer buy méngoo
   • Kodage ñaari yoon yu méngoo
   • Indexation ligneaire absolu
2. naka la ALiBi di yar ay token yu sori?
   • Ni ñuy zeroe embeddings yi ci bitti palanteer bu takku
   • Soo yokke benn penalti bu méngoo ak diggante laaj bi ak position bu am solo bi
   • Soo jàngee benn embedding bu wuute ngir position bu nekk
   • Sooy wëlbati vecteur laaj yi ak caabi yi
3. Lan mooy njariñu jëfandikoo gi gëna mag ci ALiBi?
   • Dafay dindi soxla softmax
   • Dafay tax bàyyi xel di daw ci fàttaliku bu sax
   • Dafay extrapoler bu baax ci toppalante yu gëna gudd yi ñuy gis ci tàggat yaram
   • Dina dindi soxla bopp yu bari yu ñuy bàyyi xel ci
4. Lan moo wuutale ligneer bias ci boppu attention yi?
   • Bopp bu nekk dafa bokk benn pente
   • piste yi ñu ngi koy jàng ci diiru tàggat yaram
   • Bopp bu njëkk bi kese mooy am benn par
   • Bopp bu nekk amna pente boppam, ñu defar benn xeetu geometrik
5. Buñu ko méngale ak position yiñ gëna xam, ALiBi dafay yokk ñaata paramet yuñ jàng?
   • Amul - pente bias yi dañu fixe, jàngu ñu
   • Benn ci posisioŋ bu nekk
   • Benn token bu nekk
   • Ñaar ci boppu
6. Ban model bu mag bu siiw moo jëfandikoo ALiBi ngir am xibaar ci position?
   • GPT-2
   • BLOOM
   • Bert
   • Word2Vec
7. At what stage does ALiBi apply its distance-based bias?
   • To token IDs before they enter the embedding table
   • To the final vocabulary probabilities after decoding
   • To query-key attention scores before the softmax
   • To the training labels after loss calculation
8. Why can ALiBi support inference on sequences longer than those used in training?
   • It removes the need for attention at longer lengths
   • It automatically retrains the model while generating
   • It stores a complete answer for every possible sequence length
   • Its distance-based bias does not require a separately learned embedding for every absolute position