MLCOST-12: Select an optimal ML framework

Organize, track, compare and evaluate machine learning (ML) experiments and model versions. Identify the most cost-effective and optimal combination of instance types and ML frameworks. Examples of ML frameworks include TensorFlow, PyTorch, and Scikit-learn.

Implementation plan

• Use Amazon SageMaker Experiments - Amazon SageMaker Experiments lets you organize, track, compare, and evaluate your machine learning experiments. Using this service, you can experiment with various ML frameworks and see which one gives you the most cost-effective performance. AWS Deep Learning AMIs and AWS Deep Learning Containers enable you to use several open-source ML frameworks for training on your infrastructure. AWS Deep Learning AMIs have popular deep learning frameworks and interfaces preinstalled including TensorFlow, PyTorch, Apache MXNet, Chainer, Gluon, Horovod, and Keras. The AMI or container can be launched on powerful infrastructure that has been optimized for ML performance. SageMaker also allows you to bring your own container where you can use any framework you choose.

Documents

• Use Machine Learning Frameworks, Python, and R with Amazon SageMaker

Blogs

• Right-sizing resources and avoiding unnecessary costs in Amazon SageMaker

• Amazon SageMaker Experiments – Organize, Track and Compare your Machine Learning Trainings