Posted On: Feb 1, 2023
Amazon SageMaker Training reduces the time and cost to train and tune machine learning (ML) models at scale without the need to manage infrastructure. In addition to providing built-in libraries and tools, SageMaker works with popular open-source foundation models such as GPT, BERT, and DALL·E and ML frameworks, such as PyTorch and TensorFlow. We are excited to announce that SageMaker Training now supports using images with pre-installed frameworks or algorithms stored in your private Docker registry to build ML models.
Typically, machine learning practitioners working in enterprises want to use a registry for their container image since it is an organization-wide practice to maintain a central location for their images and artifacts. Amazon ECR is a standard example of such a centralized registry used by enterprise teams. For some teams, there is a need to run training jobs using different third party registries that they have built and maintained outside of AWS. With this new feature, data scientists have the flexibility to train customized machine learning/deep learning (ML/DL) models, using any private Docker registry of their choice. SageMaker model training can now authenticate with your private Docker registry so that you can have an additional layer of security and the peace of mind that requests to your container images are serviced only for authorized entities. For a step by step instructions, please read our documentation.