Challenges and Opportunities in Making AI Easy and Efficient with...
Talkat Kubernetes AI Day Europe 2022 · on Monday, 2022-05-16· at 09:25· Pavilion 3, Room D | Level 2 | Central Forum·
summary
Kubernetes (noted as Kuberenetes in the event listing) is framed as a strong platform for ML/HPC because it brings portability from laptop to cloud, autoscaling and hardware provisioning, and frees data scientists from infra work. Concrete problems include low GPU utilization driven by no fractional GPU allocation (one GPU per container causes overprovisioning), brittle distributed training that needs framework-agnostic elastic training and transparent checkpoint/restore for preemptions and maintenance, and observability gaps for long-range metrics like precision, drift and fairness.
AI-written summary · about
chapters (4)
- 0:00Start
- 3:37Challenge: Failure Resilient Training
- 5:23Opportunity Framework Agnostic Elastic Training
- 7:09Challenge: Kubernetes observability primitives are inadequate for model performance monitoring
related talks
12mKeynote: Building Enterprise AI/ML Platform on Kubernetes
27mHarnessing Kubernetes for Scalable AI/ML Workloads
25mEmpowering AI-Driven Drug Discovery: Overcoming Challenges in Building a ML Platform on Kubernetes
54mEconomics and Best Practices of Running AI/ML Workloads on Kubernetes
26mMachine Learning Meets K8s
06mSponsored Keynote: Building Your AI Factory: Wh...
25mExtending Kubernetes for AI | Lessons Learned From Platform Engineering
17mChallenges and takeaways of managing AI workloads on cloud