Monitor and Visualize Training Metrics of your Machine Learning Models with Amazon SageMaker and Amazon CloudWatch

Amazon SageMaker can now publish training metrics to Amazon CloudWatch in real-time. You can then use CloudWatch to query, monitor, and visualize your Amazon SageMaker training jobs.

Model training is an important process aimed at enabling the model to predict the required outcomes of your business requirements. As part of the training process, machine learning algorithms produce metrics such as training loss and validation accuracy. These metrics help you understand if the model is learning well and where additional tuning is needed. With this new enhancement, you can publish these metrics to AWS CloudWatch. Once published, you can then visualize the metrics in the ClouldWatch console and query them using both SageMaker APIs and CloudWatch APIs.

Training metrics can now be monitored, queried, and visualized in all AWS regions where SageMaker is currently available, and is supported both on the built-in algorithms as well as custom algorithms. For more information, please visit the related blog here.