From CPU to GPU - Progressive Delivery for Complex ML Deployments

Explore a 34-minute conference talk that delves into the challenges and solutions of implementing progressive delivery strategies for complex Machine Learning deployments in GPU environments. Learn how to overcome obstacles like extended image pull times and resource constraints while maintaining cost efficiency in GPU-based workloads. Discover the evolution from traditional deployment approaches to advanced scenarios, including cost-efficient progressive delivery for high CPU-consuming workloads on GPU nodes and orchestrating sequential multi-resource deployments. Gain insights from real-world experience in designing CI/CD for GenAI applications, and understand the cost implications and best practices for GenAI model deployments. Presented by Sumit Jain and Divyansh Saxena from Adobe at a CNCF event, examine practical solutions for managing ML architectures in modern deployment strategies.