Customize your Amazon SageMaker model deployment software and driver versions

The article summarizes a new feature introduced by Amazon SageMaker that allows customers to customize the software and driver versions when deploying machine learning models for inference.

Specifically, the article covers:

• Previously, customers had to use preset software and driver versions defined by SageMaker on the managed instances behind an endpoint.
• Now, customers can specify the "InferenceAmiVersion" parameter when configuring endpoints to select the combination of software and driver versions (such as Nvidia driver and CUDA version) that best meets their requirements.
• This allows customers to tailor their hosting environment to meet the performance, compatibility, scalability, and operational requirements of their ML applications.
• Customers can now downgrade and upgrade driver versions for their endpoints on their own schedule.
• The feature is available in all regions where SageMaker is available.