SageMaker AI launches model customization without GPU management
The story so far
Amazon has released the no-GPU managed model customization feature for SageMaker AI, allowing users to control post-training logic through training sessions without managing underlying infrastructure. This feature supports technologies such as SFT, DPO, RLVR, RLAIF, and multi-round RL, and offers options for LoRA or full fine-tuning. Users pay for tokens processed by SageMaker AI, with metering stopping when the reward function scoring batch ends, eliminating the need for reserved capacity. User data is retained in Amazon S3 buckets in selected AWS regions, and checkpoints are also stored there. Amazon commits not to use user data or custom weights to train other models.
AI-written from the reports · Updated 27 days ago
Coverage timeline
Follow the coverage to see every side of the story.
- AWS Machine Learning BlogAnnouncing new ways to customize models with SageMaker AI, without managing GPUs
SageMaker AI introduces model customization features without GPU management, allowing users to control post-training logic through training sessions without managing underlying infrastructure. Supports technologies such as SFT, DPO, RLVR, RLAIF, and multi-round RL, with options for LoRA or full fine-tuning.