AWS
Amazon Web Services releases and Terraform AWS provider.
- AWS What's New mlawsgaengineer ·
AWS SageMaker Notebooks Expand P4de Instance Support to Tokyo Region
Amazon EC2 P4de instances, featuring 8 NVIDIA A100 GPUs with 80GB HBM2e memory each, are now generally available on SageMaker notebook instances in the Asia Pacific (Tokyo) region. These instances offer up to 60% better ML training performance and 20% lower cost compared to P4d instances, benefiting workloads with large, high-resolution datasets and accelerating time to market for ML models. This expansion is relevant for ML engineers and data scientists training large models.
feature - AWS What's New mlinfraawspreviewengineer ·
AWS Neuron 2.30.0 Enhances Trainium3 Capabilities and Developer Tools
AWS Neuron 2.30.0 is now generally available, featuring NKI 0.4.0 with new AWS Trainium3 hardware support and 22 new NKI Library kernels. This release benefits ML developers by improving model porting and validation with expanded Neuron Agentic Development skills and introduces the Neuron DRA Driver for Kubernetes. Key updates include hardware-specific instructions, FP8 support, and performance enhancements for custom kernel development and deployment on Trainium and Inferentia instances.
feature patch - AWS What's New mlawsengineer ·
SageMaker Studio adds interactive interface for Feature Store management
Amazon SageMaker Unified Studio now offers an interactive interface for managing SageMaker Feature Store, reducing the need for coding common tasks. This feature benefits data scientists, ML engineers, and business analysts by simplifying feature discovery, creation, modification, and monitoring within a unified environment.
feature - AWS What's New mlawsengineer ·
SageMaker Studio supports GPU capacity reservation via Flexible Training Plans
Amazon SageMaker Studio now supports GPU capacity reservations for its IDEs using Flexible Training Plans (FTP). This feature provides predictable access to high-performance compute resources and can reduce costs by up to 65% compared to on-demand instances. Affected users are developers and ML engineers working within SageMaker Studio who require consistent access to GPUs for their workloads.
feature