Amazon Web Services has announced several more useful updates to Amazon SageMaker, a platform for building, training, and deploying machine self-learning models during AWS re:Invent.
New features have been introduced to the platform that are designed to improve model handling, including the introduction of new classes in the SageMaker Python SDK: ModelBuilder and SchemaBuilder.
ModelBuilder, selects a compatible SageMaker container to deploy to, and captures the necessary dependencies. SchemaBuilder manages the tasks of serializing and deserializing model inputs and outputs.
“You can use the tools to deploy the model in your local development environment to experiment with it, fix any runtime errors, and when ready, transition from local testing to deploy the model on SageMaker with a single line of code”, Antje Barth, principal developer advocate at AWS, wrote in a blog post.
SageMaker Studio has been enhanced with updated deployment workflows, offering guidance to assist in selecting the most optimal endpoint configuration.
Furthermore, SageMaker has received improvements in its inference capabilities, contributing to reduced deployment costs and latency. These enhancements enable the deployment of one or more foundation models on a single endpoint, with control over memory allocation and the number of accelerators assigned to each model.
The system also features automatic monitoring of inference requests, intelligently routing them based on the availability of instances. According to Amazon, this advanced capability has the potential to slash deployment costs by up to 50% and decrease latency by up to 20%.
Additionally, Amazon SageMaker Canvas, a no-code interface designed for constructing machine learning models, has some updates. Users can now leverage natural language prompts during the data preparation process.
Within the chat interface, the application furnishes a variety of guided prompts tailored to the specific database being utilized. Alternatively, users have the flexibility to create their own prompts. For instance, they can instruct the system to generate a data quality report, filter out rows based on specific criteria, and perform various other tasks.
Moreover, users now have the ability to incorporate foundation models from Amazon Bedrock and Amazon SageMaker Jumpstart. This added capability empowers companies to deploy models specifically tailored to their unique business requirements.
SageMaker Canvas takes charge of the entire training process, facilitating fine-tuning of the model post-creation. Additionally, it offers an in-depth analysis of the generated model, presenting metrics such as perplexity and loss curves, training loss, and validation loss.
