Ikibazo gihuza ibibazo · Hagati Urwego
Apache Airflow ya ML Ikibazo Cyakazi
Koresha ibikorwa bya Airflow biterwa, gusubiramo, no gusubira mubikorwa bya ML, harimo amategeko ya trigger agena igihe imirimo yiteguye gukora.
Ikibazo 1 Bya 8
Muri Apache Airflow, niyihe abstraction yibanze ikoreshwa mugusobanura akazi?
Reba buri kibazo8 ibibazo
Ibisubizo nibisobanuro bigaragara mugihe ufata ikibazo.
- Muri Apache Airflow, niyihe abstraction yibanze ikoreshwa mugusobanura akazi?
- YAML igaragara
- SQL Reba
- DAG (Igishushanyo cya Acyclic Graph) cyanditswe muri Python
- Ikaye ya Jupyter
- Ni uruhe ruhare Airflow ikunze kugira mu miyoboro ya ML?
- Irategura kandi ikanateganya imirimo ikora ku zindi sisitemu
- Itoza uburyo bwimbitse bwo kwiga kuri GPUs
- Irabika ibiranga ibice byo kwibuka
- Ikora ibyitegererezo kuri HTTP
- Niyihe Airflow yubaka ikoreshwa mugutegereza imiterere yo hanze, nka dosiye igera mububiko?
- Impinduka
- Ikidendezi
- Makro
- Rukuruzi
- Niki Airflow 'backfill' ikwemerera gukora?
- Siba DAG ishaje ikora burundu
- Koresha DAG hejuru yitariki yashize kugirango usubiremo amateka yamateka
- Shishoza amakuru yububiko
- Hagarika gahunda byigihe gito
- Which Airflow component determines when eligible task instances should be scheduled?
- Gahunda
- The graph view in the web UI
- A model’s feature encoder
- The artifact file itself
- Niki 'gahunda-yamenyesheje gahunda' yatangijwe muburyo bushya bwa Airflow?
- Gushishoza amakuru yose yatambutse hagati yimirimo
- Mu buryo bwikora kuranga amakuru yimyitozo
- Kugabanya umubare wa GPU kuri buri gikorwa
- Gukurura DAGs iyo dataset ivuguruwe aho kuba kumasaha yagenwe
- An Airflow cleanup task must run after upstream tasks finish even if one fails. Which setting controls this behavior?
- Umubare wicyitegererezo
- The cleanup task’s trigger rule
- The feature normalization method
- The UI theme selected for the DAG
- A team backfills three months of ML feature data in Airflow. What makes repeating an interrupted task safer?
- Appending duplicate outputs each time it retries
- Ignoring the run’s historical date
- An idempotent operation with the correct historical inputs
- Changing the training labels on every retry