Amazon SageMaker allows you to build, train, and deploy machine learning models without worrying about maintaining multiple environments and workflows. It provides the flexibility to use the same models, frameworks, and algorithms you already use today, but with the freedom to focus all of your time on your models rather than the complexities of scaling and application integration.
Benefits
Automatic Model Tuning
Amazon SageMaker comes with automated hyperparameter optimization (HPO), adjusting thousands of different combinations of algorithm parameters, to arrive at the most accurate predictions the model is capable of producing.
Choose Any Framework
Amazon SageMaker supports all frameworks and comes pre-configured with TensorFlow and Apache MXNet. If you want to train with an alternative framework, you can bring your own in a Docker container.
Integrate into Existing Workflows
Amazon SageMaker provides discrete modules for authoring, training, and hosting models so that you can choose to keep the parts of your existing workflows that you like, and insert Amazon SageMaker into the parts you don’t.
Powerful ML Algorithms
Amazon SageMaker provides built-in algorithms like K-Means Clustering, ResNet, and Factorization Machines. But, they’ve been optimized for large datasets to perform two- to ten-times faster than what can be achieved anywhere else.
Sample Notebooks provided
Amazon SageMaker provides a library of Juptyer notebooks with sample code for a wide variety of machine learning and deep learning projects to help you get new projects started quickly.
Fast Training without High Cost
Training on Amazon SageMaker can automatically distribute processing across an infinite number of nodes - including P3 instances built on NVIDIA V100 GPUs – and since training environments live only as long as the job, you pay only for what you use.
Stay Agile
With Amazon SageMaker, you can deploy your model into production without making application code changes. It also allows for automatic A/B testing of models and handles all of the routine maintenance and monitoring of your production model servers.
Train with Any Deep Learning Framework
With Amazon SageMaker, you can use the deep learning framework of your choice for model training. Amazon SageMaker is pre-configured to run TensorFlow and Apache MXNet; two popular deep learning frameworks. You can also bring your own Docker container with any framework you like - such as Caffe2, Chainer, PyTorch, Microsoft Cognitive Toolkit (CNTK), or Torch.
Learn how to bring your own TensorFlow model and deploy on Amazon SageMaker >>
Train your next model with Amazon SageMaker