Idanwo ti o sopọ mọ itọsọna · Lile Ipele

Mamba ati Yiyan State alafo adanwo

Review Mamba’s input-dependent SSM parameters, fixed-size state, linear-time sequence processing, and hardware-aware parallel scan.

Jẹmọ itọsọna awọn ọnaMamba Ati Yiyan State alafo
Ibeere 1 ti 8

Kini anfani iṣiro akọkọ ti Mamba lori Amunawa boṣewa?

Awotẹlẹ gbogbo ibeere8 ibeere

Awọn idahun ati awọn alaye han bi o ṣe n mu ibeere naa.

  1. Kini anfani iṣiro akọkọ ti Mamba lori Amunawa boṣewa?
    • O ṣe iwọn laini pẹlu gigun ọkọọkan dipo imẹrin
    • Ko nilo data ikẹkọ
    • O ṣiṣẹ nikan lori awọn aworan
    • O yọ awọn nilo fun eyikeyi farasin ipinle
  2. Kini ọrọ 'aṣayan' tọka si ni Mamba?
    • O yan eyi ti GPU lati ṣiṣẹ lori
    • Awọn paramita SSM rẹ da lori ami titẹ sii, nitorinaa o le yan kini lati ranti tabi gbagbe
    • O yan awọn orukọ ati awọn ọrọ-ọrọ nikan
    • O yan ipin-aileto ti awọn ami lati foju
  3. Bawo ni awoṣe aaye aaye kan ṣe aṣoju itan ti ọkọọkan kan?
    • Nipa titoju gbogbo ti tẹlẹ àmi kedere
    • Nipa iyaworan maapu akiyesi lori gbogbo awọn orisii
    • Nipa titẹkuro itan sinu iwọn ti o wa titi ti o farapamọ ipo imudojuiwọn ni igbese nipa igbese
    • Ko tọju itan kankan
  4. Kilode ti Mamba ko le lo ọna abuja convolution iyara ti awọn SSM iṣaaju lo?
    • Nitori convolutions ni o wa arufin on GPUs
    • Nitoripe ko ni awọn paramita
    • Nitori ti o lakọkọ àmi ni ID ibere
    • Nitori ṣiṣe awọn ipilẹ-igbẹkẹle igbewọle npa awọn iyipada akoko-airapada nilo
  5. Ilana wo ni Mamba nlo lati duro ni iyara laibikita sisọnu ọna abuja convolution?
    • A hardware-mọ ni afiwe ọlọjẹ pa ni sare GPU iranti
    • Nṣiṣẹ patapata lori Sipiyu
    • Caching gbogbo àmi orisii
    • Foju pupọ julọ ti ọkọọkan
  6. Tani o ṣafihan Mamba ati ni aijọju nigbawo?
    • OpenAI ni ọdun 2018
    • Albert Gu ati Tri Dao ni ipari 2023
    • Google ni ọdun 2015
    • Ko si ẹnikan, o jẹ arosọ
  7. Which Mamba state-space parameters are made functions of the current input?
    • Only the fixed transition matrix A
    • The output vocabulary and tokenizer
    • B, C, and the step size delta
    • The attention query and key heads
  8. What do the variable B, C, and delta parameters let Mamba do?
    • Build a growing key-value cache for every token
    • Route each token to an expert network only
    • Replace the state update with full all-pairs attention
    • Choose how much information to let into the state and how much to retain based on the input