Catalog

Running Kubernetes pods in parallel across multiple nodes to efficiently train large models is difficult. Scaling compute resources, setting up drivers and networking, and scheduling pods to maximize instance utilization is not only challenging, it’s undifferentiated heavy lifting. AWS Batch is a managed service that works with your existing Amazon EKS clusters to manage the full lifecycle of nodes and pods for ML training and other compute-intensive workloads. In this chalk talk, dive deep on how to offload the complexities of large ML model training on Amazon EKS using AWS Batch multi-node processing jobs, reducing your cost and operational overhead.