The Amazon Deep Learning AMIs provide pre-configured environments for machine learning developers and data scientists to quickly start experimenting with deep learning models. The Conda-based Deep Learning AMI, which provides pre-installed pip packages of popular deep learning frameworks in separate virtual environments, now supports Caffe, includes TensorFlow version 1.4.1, Keras 2.1.2, Microsoft Cognitive Toolkit 2.3.1, and Theano 1.0.

The Conda-based AMI also provides TensorFlow Serving for developers to quickly create an inference end-point for their TensorFlow models, as well as the TensorBoard visualization tool. The AMIs come with easy-to-follow MNIST-based tutorials for both TensorFlow Serving and TensorBoard. Apache MXNet users can benefit from the MXNet Model Server to quickly deploy an HTTP-based inference endpoint for their models.

The CUDA 9 Source Code AMI, which provides pre-installed deep learning frameworks and their source code in a shared python environment, now includes TensorFlow 1.5.0-rc0 version. This version of TensorFlow works with CUDA 9 and cuDNN 7, making it the first version of TensorFlow that leverages the computing power of NVidia Volta GPUs powering Amazon EC2 P3 instances. Since the version is still pre-release, test before you use it in production.

All the Amazon Deep Learning AMIs for both Ubuntu and Amazon Linux platforms are updated with the latest NVidia GPU drivers and operating system versions that include security patches for the Spectre and Meltdown vulnerabilities discovered earlier this month.

You can find the Deep Learning AMI of your choice in the Quick Start section of the Step 1: Choose an Amazon Machine Image (AMI) in the EC2 instance launch wizard. Also visit our Documentation guide for help with selecting the right AMI for your project, simple tutorials and more resources.