Guide-yakabatana mibvunzo · Zvakaoma Nhanho

State Space Models uye Mamba Quiz

Test state-space sequence complexity, Mamba selectivity, hardware-aware scan, recurrent inference state, and hybrid models.

Nzira dzekutungamira dzinoenderanaState Space Models Uye Mamba
Mubvunzo 1 ye 8

Ko chimiro chemuchadenga chinoyera sei nehurefu hwekutevedzana, zvichienzaniswa nekuzvitarisira wega?

Tarisa mubvunzo wese8 mibvunzo

Mhinduro uye tsananguro zvinoonekwa paunenge uchitora bvunzo.

  1. Ko chimiro chemuchadenga chinoyera sei nehurefu hwekutevedzana, zvichienzaniswa nekuzvitarisira wega?
    • Linearly, maringe neattention's quadratic scaling
    • Quadratically, maringe netarisiro ye linear scaling
    • Ose anoyera zvakafanana
    • Exponentially kune ese ari maviri
  2. Ndeipi iyo yepakati innovation iyo Mamba yakawedzera kune yakange yakarongeka SSMs seS4?
    • Kushandisa softmax kutarisa mukati
    • Kugadzira iyo mamiriro-yekuvandudza ma parameter zvinoenderana nekuisa (selectivity)
    • Kubvisa iyo yakavanzika mamiriro zvachose
    • Kuchinjira ku quadratic-time computation
  3. Sei Mamba yaida tsika yehardware-aware parallel scan?
    • Nekuti maGPU haakwanise kuita matrix kuwanza
    • Kuderedza kuwanda kwezvikamu
    • Nekuti kugadzira ma paramita anotsamira kunotyora iri nyore convolution trick
    • Nekuti haishandise yakadzidza paramita
  4. Munguva yekufungidzira, yakakosha kushanda zvakanaka kweSSM seMamba ndeyekuti inoshandisa:
    • Memory inokura nehurefu hwakazara hwekutevedzana, senge KV cache
    • Hapana ndangariro zvachose
    • Yakawanda ndangariro kupfuura Transformer yehukuru hwakaenzana
    • Inongogara ndangariro pa tokeni yakagadzirwa
  5. Mukirasi yakarongeka maSSM (asati asarudza), iyo inodzokororwa inogona kuverengerwa nemazvo nekuti inogona kunyorwazve se:
    • Long convolution
    • Softmax kushanda
    • Muti wesarudzo
    • Random sango
  6. Which language-modeling corpus did the Mamba paper use in its standard autoregressive scaling experiments?
    • ImageNet-1K
    • Murwi
    • LibriSpeech only
    • D4RL
  7. In the Mamba block, what operation combines the selective-SSM output with the other projected branch?
    • A nearest-neighbor lookup over audio frames
    • A cross-attention layer over image patches
    • An elementwise multiplicative interaction with a gated branch
    • A graph-degree normalization step
  8. What language-modeling result does the Mamba paper report for its 3B model?
    • It requires a Transformer cache twice as large
    • It replaces the training corpus with audio only
    • It can be trained without any data
    • It matches Transformers twice its size in pretraining and downstream evaluation