This workshop provides a set of labs designed to give you hands-on experience building, training, and deploying machine learning models using Amazon SageMaker. The target audience includes data scientists, machine learning engineers, solutions architects, and software developers.
The workshop can be used in self-paced fashion, or delivered by a solutions architect in a 1-day or 2-day format.
The following labs are provided:
- create your first notebook instance ("setup" lab), which is a prerequisite for the other labs
- explore logistic regression using SageMaker's built-in XGBoost algorithm
- classify images using SageMaker's built-in Image Classification algorithm with a domain of 256 classes of objects (horse, kayak, teapot, ...)
- use DeepAR, one of SageMaker's built-in algorithms, to perform forecasting of electricity demand
- try out text classification using SageMaker's built-in BlazingText algorithm
- bring your own neural network script to a container provided by Amazon SageMaker
- perform hyperparameter optimization
- perform batch inference to get predictions on a large number of observations in bulk
- perform A/B testing when deploying a new version of an existing model hosted by Amazon SageMaker
- use auto-scaling to improve scalability of an endpoint hosted by Amazon SageMaker
- use inference pipelines to build and deploy feature preprocessing pipelines and reuse them for training and inference
- bring your own Docker container to Amazon SageMaker
- try the Amazon Textract service, demonstrating how it can be used to identify headers and footers