sagemaker-python-sdk

Why we included this project

For teams already on AWS, this is the most direct way to drive SageMaker from Python. It wraps training, real-time and batch inference, hyperparameter tuning, the model registry, and processing jobs in one object-oriented API, so a training script can become a live endpoint in a few method calls. The 3.x release split the library into modular packages for core, training, serving, and MLOps, and folded framework-specific classes into a single ModelTrainer and ModelBuilder, which helps when you juggle PyTorch, TensorFlow, or custom containers. If you want managed infrastructure and GPU provisioning without building your own orchestration layer, and you want training and deployment logic versioned in code, this SDK is worth a look. For anyone committed to SageMaker, it is effectively the control plane for ML operations.