Ikibazo gihuza ibibazo · Hagati Urwego

Kubernetes ya ML Ikibazo Cyakazi

Koresha Kubernetes ML pod gushyira, ibikoresho bya GPU ibikoresho, hamwe no kwihanganira umwanda, harimo n'impamvu amahugurwa yatanzwe akeneye ubushobozi bukwiye.

Inzira zijyanye no kuyoboraKubernetes Kubikorwa bya ML
Ikibazo 1 Bya 8

Nuwuhe murimo wibanze wa gahunda ya Kubernetes kumurimo wa ML?

Reba buri kibazo8 ibibazo

Ibisubizo nibisobanuro bigaragara mugihe ufata ikibazo.

  1. Nuwuhe murimo wibanze wa gahunda ya Kubernetes kumurimo wa ML?
    • Kwandika kode y'amahugurwa kubashakashatsi mu makuru
    • Guhitamo urwego muri cluster ikora buri kintu cyabitswe
    • Kuranga amakuru yamahugurwa mu buryo bwikora
    • Guhitamo imiyoboro yububiko
  2. Nigute Kubernetes ituma GPU iboneka kuri pod?
    • Binyuze mumacomeka yibikoresho yamamaza ibikoresho nka nvidia.com/gpu
    • Kohereza imeri umucuruzi wa GPU kugirango abone uruhushya
    • GPU ntishobora gukoreshwa muri Kubernetes
    • Muguhindura CPU yibanze muri GPU mugihe cyo gukora
  3. With a NoSchedule taint on a GPU node, what does a matching pod toleration do?
    • Forces the pod onto that node regardless of capacity
    • Changes the model’s optimizer
    • Allows the pod to be considered for that node without guaranteeing placement
    • Encrypts every GPU tensor
  4. Niki gikoresho cyongeramo ubushobozi bwa ML nkukwirakwiza-imyitozo hejuru ya Kubernetes?
    • Microsoft Ijambo
    • Apache yo gutanga paji y'urubuga ruhamye
    • Urupapuro rwerekana urupapuro
    • Kubeflow
  5. Kuki Kubernetes ikwiranye no guturika ML akazi?
    • Iteganya imirimo kubuntu kandi ikarekura ibyuma iyo imirimo irangiye
    • Irabika burundu GPU zose kumukoresha umwe
    • Irinda akazi ako ari ko kose kugabanuka
    • Ikoresha ikintu kimwe gusa kuri cluster
  6. Ni iki 'gupima umwanzuro wa nyuma kuri zeru' bisobanura kuri Kubernetes?
    • Gusiba icyitegererezo burundu
    • Gushiraho ukuri kuri zeru ku ijana
    • Gukoresha zeru zeru mugihe nta traffic, hanyuma uzunguruka kubisabwa bikurikira
    • Kugabanya ibipimo by'icyitegererezo kuri zeru
  7. One ML pod requests four GPUs, while two nodes each have only two available. Can ordinary scheduling combine those nodes for that one pod?
    • Yes; every pod automatically spans all cluster nodes
    • Yes; the scheduler converts CPU cores into the missing GPUs
    • Only if the pod name contains the word distributed
    • No; the pod must fit on one eligible node
  8. An inference service scales to zero while idle. Which user-facing delay should its team test when traffic resumes?
    • The training-set labeling speed
    • Cold-start delay while a serving replica becomes ready
    • The file size of an unrelated notebook
    • The number of discarded vocabulary tokens