With the latest versions of the frameworks, you can take advantage of features like AVX instructions for CPUs in TensorFlow, as well as MXNet’s Text API, which can be used to build vocabularies and load pre-trained word embeddings. As open source, the TensorFlow and MXNet Docker containers can be customized by adding libraries to the Dockerfile, pre-loading frequently used files and datasets, and changing internal configurations to meet your needs. You can also download the containers to your local environment and use the Amazon SageMaker Python SDK to test your scripts before deploying to Amazon SageMaker training or hosting environments, dramatically accelerating your debugging cycle. And now with support for larger notebook instance types, you can test locally on large datasets or compute intensive jobs.
These new features in Amazon SageMaker are available today in the US East (N. Virginia), US East (Ohio), EU (Ireland), and U.S. West (Oregon) AWS Regions. Visit the GitHub release notes for more information on TensorFlow 1.6 and Apache MXNet 1.1 features and the Amazon SageMaker Python SDK documentation for more details on how to use these containers locally. Visit the Amazon SageMaker pricing page for details on the added instance types for notebooks, training, and hosting.
