Posted On: Apr 17, 2020

Amazon SageMaker customers can now select ml.g4dn and ml.c5n instances for training machine learning models. Amazon SageMaker is a modular, fully-managed platform that enables developers and data scientists to quickly and easily build, train, and deploy machine learning models at any scale.

Amazon ml.g4dn instances deliver the most cost-effective and versatile GPU instances designed to accelerate computationally demanding machine learning training and model evaluation workloads. ml.g4dn instances are equipped with NVIDIA T4 Tensor Core GPUs, AWS-custom Second Generation Intel® Xeon® Scalable (Cascade Lake) processors, and the AWS Nitro system. Nitro’s local NVMe-based SSD storage provides direct access to up to 900 GB of fast, local NVMe storage. ml.g4dn instances deliver deliver up to 65 TFLOPs of FP16 performance for a lower price and can be cost-effective when used for small-scale machine learning training jobs that are less sensitive to time-to-train.

Amazon ml.c5n instances are ideal for running advanced compute-intensive workloads such as batch data processing and distributed deep learning inference. ml.c5n instances are network-optimized variants of ml.c5 instances, powered by Intel® Xeon® Scalable processors (Skylake) and the fourth generation of custom Nitro card and Elastic Network Adapter (ENA) device, to deliver up to 100 Gbps of network bandwidth per instance. They offer significantly higher network performance across all instance sizes, ranging from 25 Gbps of peak bandwidth on smaller instance sizes to 100 Gbps of network bandwidth on the largest instance size. In addition, ml.c5n instances also feature 33% more available memory compared to ml.c5 instances making them ideal for applications that can take advantage of improved network throughput and packet rate performance.

To get started, visit the Amazon SageMaker product page.