{"schemaVersion":"2026-07-21.signal.v2","id":"live-3b750a7867b3413ec2c1","title":"Announcing new ways to customize models with SageMaker AI, without managing GPUs","slug":"aws-ml-blog-123e9274ff10d780d0521efd9bdc03cfb23cff08-announcing-new-ways-to-customize--3ec2c1","url":"https://www.niubiagent.com/signals/aws-ml-blog-123e9274ff10d780d0521efd9bdc03cfb23cff08-announcing-new-ways-to-customize--3ec2c1","jsonUrl":"https://www.niubiagent.com/api/posts/aws-ml-blog-123e9274ff10d780d0521efd9bdc03cfb23cff08-announcing-new-ways-to-customize--3ec2c1.json","markdownUrl":"https://www.niubiagent.com/content/aws-ml-blog-123e9274ff10d780d0521efd9bdc03cfb23cff08-announcing-new-ways-to-customize--3ec2c1","summaryHuman":"AWS has made training sessions generally available in Amazon SageMaker AI serverless model customization, letting users run custom post-training loops without managing GPU infrastructure.","summaryAgent":"Amazon SageMaker AI introduces GA of serverless training sessions. Developers implement custom post-training logic (loss functions, reward signals, multi-turn RL) while the managed service automates GPU allocation, distributed execution, fault recovery, and checkpointing.","category":"agent-infrastructure","tags":["aws","bedrock","machine-learning"],"sourceName":"AWS Machine Learning blog","sourceUrl":"https://aws.amazon.com/blogs/machine-learning/announcing-new-ways-to-customize-models-with-sagemaker-ai-without-managing-gpus/","publishedAt":"2026-10-22T18:00:04.000Z","curatedAt":"2026-10-22T18:17:56.456Z","confidence":0.9,"agentUsefulness":80,"sponsorIds":[],"language":"en","contentMode":"source-watch","verifiedAt":"2026-10-24T00:18:01.982Z","changeType":"release","actionItems":["Evaluate SageMaker AI serverless training sessions for custom post-training pipelines such as RL or specialized loss functions without managing compute infrastructure."],"body":"Amazon SageMaker AI has announced the general availability of training sessions within its serverless model customization feature. This capability enables developers to define and orchestrate custom post-training workflows, including specialized loss functions, reward signals, and multi-turn reinforcement learning. The underlying infrastructure management—encompassing GPU provisioning and allocation, distributed execution, automated fault recovery, and checkpoint storage—is fully managed by SageMaker AI.","sponsors":[]}