Medusa decodage boppu quiz

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

1. Lan la boppu Medusa yu bari yi di wax?
   • Jeton yu bari ci ëlëg ci position yu wuute ci kanam
   • Perte de formation de model
   • ban GPU ngay jëfandikoo
   • Nataal biy wax ci token bi
2. Under which Medusa configuration does the paper describe lossless inference acceleration?
   • Dafay komprime diisaayu model bi ci anam wu jaar yoon
   • Medusa-1 with a frozen backbone and exact verification against the original model
   • Du musa def benn njuumte ci wàllu nàfar
   • Dina dindi soxla vocabulaire
3. Ci yan anam lañuy amalee wéyal kàndidaa yu Medusa ngir saytu?
   • Limu plat buñ jox poñ benn-benn
   • Tablo hash
   • Garab gi dugal benn poñ ci benn paas ak masku garab
   • Boucle buy dellusi ci kaw token bu nekk
4. Lan mooy njariñu Medusa ci dekodaasu spekulatif bi?
   • Dafay laaj ñaari model yu mag yu wuute
   • Ci nataal kese lay dox
   • Dafay efaase modelu base bi ginaaw biñu ko jëfandikoo
   • Dafay moytu soxla xeetu projet bu wuute ci yokk bopp ci benn
5. Lu tollu ci ban gaawaay la Medusa faral di xamle?
   • Lu tollu ci 2-3x
   • Lu tollu ci 0.5x (gëna ndànk)
   • 100x gën
   • Amul coppite buñ mëna natt
6. Luy boppu Medusa bu nekk, ci wàllu architecture?
   • Transformatër bu mat
   • ab couche feed-forward bu ndaw (MLP) ci kaw etaa bu nëbbu bi
   • Enkodeer nataal buy jëm kanam
   • Benn tokeniseer bu wuute
7. What does Medusa accept from a verified candidate tree?
   • Every speculative branch regardless of verification
   • Only the shortest candidate token
   • The longest prefix consistent with the base model’s decoding result
   • A random leaf from the tree
8. How do Medusa’s prediction-head offsets differ?
   • Every head predicts the same next token
   • Each head predicts a different vocabulary
   • The heads predict tokens only after decoding is complete
   • Each head predicts a token at a different position ahead in the sequence