ML Frameworks
- What it is
- The bottom tier of the AWS AI/ML stack: run open-source machine learning frameworks yourself with complete control.
- Frameworks supported
- PyTorch, TensorFlow, JAX, Hugging Face Transformers, Apache MXNet, scikit-learn.
- How AWS supports them
- AWS Deep Learning AMIs — EC2 images with frameworks, GPU drivers, CUDA, and libraries preinstalled.
- AWS Deep Learning Containers — the same, packaged for Amazon Elastic Container Service (ECS), Amazon Elastic Kubernetes Service (EKS), and SageMaker.
- When this tier is the right choice
- Custom research, unusual architectures, or specific framework versions and distributed training strategies.
- The trade-off
- Maximum flexibility, maximum operational responsibility — you own environment setup, scaling, and orchestration.
- Most teams should start at Amazon SageMaker AI and drop to this tier only when they must.