Web Services Japan Startup ML Solutions Architect Yoshitaka Haribara, Ph.D. 2020-05-22 Amazon SageMaker Build, train, and deploy machine learning models
AI は、Amazon SageMaker マネージドスポットトレーニングで ML モデルのトレーニングコストを 70% 節約」 https://aws.amazon.com/jp/blogs/news/cinnamon-ai-saves-70-on-ml-model- training-costs-with-amazon-sagemaker-managed-spot-training/
Amazon SageMaker and TensorFlow Work for You — Mobileye guest post” https://medium.com/@julsimon/making-amazon-sagemaker-and-tensorflow- work-for-you-893365184233
2x のパフォーマンス Amazon SageMaker Neo Neo Broad framework support Broad hardware support Open-source Neo-AI device runtime and compiler 1/10th the size of original frameworks https://github.com/neo-ai/
Airflow にも SageMaker Operator が用意されている • Python で記述した DAG (有向非巡回グラフ) でワークフロー管理 • Amazon SageMaker とのインテグレーションも • EC2 + RDS は別途必要 (マネージドサービスではない) Raw data Cleaned data Train data Test data Amazon SageMaker Training / HPO Model artifact Amazon SageMaker Batch transform Airflow DAG Filter long-tailed data sparse data format → RecordIO protobuf Analyze model performance based on test data Operator PythonOperator PythonOperator SageMakerTrainOperator/ SageMakerTransformOperator PythonOperator SageMakerTuningOperator Blog: https://aws.amazon.com/jp/blogs/news/build-end-to-end-machine-learning-workflows-with-amazon-sagemaker-and-apache-airflow/ Prediction results