Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Explore advanced AI model training techniques in this 55-minute conference talk from AWS re:Invent 2025 focusing on Amazon SageMaker HyperPod's revolutionary checkpointless and elastic training capabilities. Discover how checkpointless training eliminates costly downtime by automatically recovering from infrastructure faults in minutes rather than hours through peer-to-peer state transfer technology that bypasses traditional checkpoint restart methods. Learn about elastic training's ability to dynamically expand capacity by claiming idle accelerators or gracefully contract when higher-priority tasks require resources, all without manual intervention. Understand how these innovations maintain continuous training momentum despite infrastructure failures or resource availability fluctuations, enabling you to scale and accelerate generative AI model development across hundreds to thousands of AI accelerators for more efficient and resilient machine learning workflows.
