AWS 机器学习· Shantanu Tripathi·· 8 hours agoSelectedAI score72
Amazon SageMaker Inference Supports NVIDIA MIG Hardware-Level GPU Partitioning
SageMaker Inference now supports Multi-Instance GPU to maximize GPU utilization for model endpoints
AI guide
Amazon SageMaker Inference announced support for hardware-level GPU partitioning powered by NVIDIA MIG, enabling users to divide a single physical GPU into up to seven isolated instances on managed endpoints to deploy different models.
Why it matters
Through managed hardware partitioning, developers can achieve isolated co-location of multiple models on a single GPU without maintaining their own clusters, providing a practical path toward fine-grained compute cost control.
Source:AWS 机器学习 · aws.amazon.com