Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Kubernetes와 SageMaker를 활용한
Machine Learning 워크로드...
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Agenda
Fully managed machine learning with SageM...
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Amazon SageMaker is a fully managed service that...
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Demo – ML Training with SageMaker console
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Demo – Managed ML development with SageMaker
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
SageMaker manages ML infrastructure
Build Train ...
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Kubernetes architecture
API Server
EtcdScheduler...
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Under the hood – Amazon SageMaker and
Kubernetes...
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Demo – XGBoost Kubernetes SageMaker Operator
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
SageMaker Operators for Kubernetes
https://githu...
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Kubeflow Components
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Pipelines – Machine Learning Job Orchestrator
• ...
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Creating Kubeflow Pipeline Components
@dsl.pipel...
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Demo – KMeans Classification with Kubeflow Pipel...
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
https://eksworkshop.com
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
https://github.com/aws-samples/eks-kubeflow-work...
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
© 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Demo – ML Workflow with Step functions
© 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
AWS Service integration with Step functions
http...
© 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Apache Airflow
• Built by Airbnb, open sourced
u...
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Apache Airflow SageMaker Operators
https://airfl...
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Demo – Recommendation batch job with Airflow
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
ML workflows with SageMaker and Apache Airflow
h...
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Machine learning cycle
Business
Problem
ML probl...
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Manage data on AWS
Business
Problem
ML problem
f...
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Build and train models using SageMaker
Business
...
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Deploy models using SageMaker
Business
Problem
M...
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
What about the lines between the steps?
Business...
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
AWS 머신러닝(ML) 교육 및 자격증
Amazon의 개발자와 데이터 과학자를 교육하는...
© 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
© 2019, Amazon Web Services, Inc. or its Affilia...
[AWS Innovate 온라인 컨퍼런스] Kubernetes와 SageMaker를 활용하여 Machine Learning 워크로드 관리하기 - 강성문, AWS 솔루션즈 아키텍트
Upcoming SlideShare
Loading in …5
×

[AWS Innovate 온라인 컨퍼런스] Kubernetes와 SageMaker를 활용하여 Machine Learning 워크로드 관리하기 - 강성문, AWS 솔루션즈 아키텍트

149 views

Published on

발표자료 다시보기: https://youtu.be/6sogVHw9jZ4

Machine Learning 워크로드를 실제 운영환경에서 사용하기 위하여 다양한 툴들과 방법들이 시도되고 있습니다. 본 세션에서는 ML 운영을 위해 어떤 툴들이 활용되고 있는지를 살펴보고, 그 중 엔터프라이즈 환경에서 많이 선택하고 았는 Kubernetes와 Kubeflow를 사용하여, 어떻게 Machine Learning 전처리와 Training 작업을 관리하고 운영환경에 배포할 수 있는지를 데모와 함께 알아봅니다.

Published in: Technology
  • Be the first to comment

[AWS Innovate 온라인 컨퍼런스] Kubernetes와 SageMaker를 활용하여 Machine Learning 워크로드 관리하기 - 강성문, AWS 솔루션즈 아키텍트

  1. 1. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved. Kubernetes와 SageMaker를 활용한 Machine Learning 워크로드 관리 Seongmoon Kang ML Special Solutions Architect AWS Korea
  2. 2. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved. Agenda Fully managed machine learning with SageMaker ML Workload management with Kubernetes - SageMaker Operator for Kubernetes - Kubeflow Other solutions (Step Functions, Apache Airflow) Machine Learning Lifecycle Management
  3. 3. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
  4. 4. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved. Amazon SageMaker is a fully managed service that covers the entire machine learning workflow Jupyter notebook instances High-performance algorithms Large-scale training Optimization One-click deployment Fully managed with auto-scaling
  5. 5. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved. Demo – ML Training with SageMaker console
  6. 6. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved. Demo – Managed ML development with SageMaker
  7. 7. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved. SageMaker manages ML infrastructure Build Train Deploy Pre-built notebook instances Highly optimized machine learning algorithms One-click training for ML, deep learning, and custom algorithms Automatic model tuning (hyperparameter optimization) Fully managed hosting at scale Deployment without engineering effort
  8. 8. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
  9. 9. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved. Kubernetes architecture API Server EtcdScheduler Controller Manager kubelet kubelet kubeletContainer Runtime Container Runtime Container Runtime Pod Pod Pod Pod Pod Pod Pod Pod Pod Kube- proxy Kube- proxy Kube- proxy kubectl YAML
  10. 10. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved. Under the hood – Amazon SageMaker and Kubernetes Kubectl apply YAML Key Features • Amazon SageMaker Operators for training, tuning, inference • Natively interact with Amazon SageMaker jobs using K8s tools (e.g., get pods, describe) • Stream and view logs from Amazon SageMaker in K8s • Helm Charts to assist with setup and spec creation
  11. 11. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved. Demo – XGBoost Kubernetes SageMaker Operator
  12. 12. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
  13. 13. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved. SageMaker Operators for Kubernetes https://github.com/aws/amazon-sagemaker-operator-for-k8s https://aws.amazon.com/blogs/machine-learning/introducing-amazon-sagemaker-operators-for-kubernetes/
  14. 14. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
  15. 15. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved. Kubeflow Components
  16. 16. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved. Pipelines – Machine Learning Job Orchestrator • Compose, deploy, and manage end-to-end ML workflows • End-to-end orchestration • Easy, rapid, and reliable experimentation • Easy re-use • Built using Pipelines SDK • kfp.compiler, kfp.components, kfp.Client • Uses Argo under the hood to orchestrate resources
  17. 17. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved. Creating Kubeflow Pipeline Components @dsl.pipeline( name='Sample Trainer', description=’’ ) def sample_train_pipeline(... ): create_cluster_op = CreateClusterOp('create-cluster', ...) analyze_op = AnalyzeOp('analyze', ...) transform_op = TransformOp('transform', ...) train_op = TrainerOp('train', ...) predict_op = PredictOp('predict', ...) confusion_matrix_op = ConfusionMatrixOp('confusion-matrix', ...) roc_op = RocOp('roc', ...) kfp.compiler.Compiler().compile(sample_train_pipeline , 'my- pipeline.zip’)
  18. 18. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved. Demo – KMeans Classification with Kubeflow Pipeline
  19. 19. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
  20. 20. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved. https://eksworkshop.com
  21. 21. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved. https://github.com/aws-samples/eks-kubeflow-workshop.git
  22. 22. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
  23. 23. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved. Demo – ML Workflow with Step functions
  24. 24. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved. AWS Service integration with Step functions https://docs.aws.amazon.com/ko_kr/step-functions/latest/dg/connect-sagemaker.html
  25. 25. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
  26. 26. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
  27. 27. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved. Apache Airflow • Built by Airbnb, open sourced under Apache • Platform to programmatically author, schedule and monitor workflows • Rich user interface to visualize job pipelines, monitor progress and troubleshoot issues • Need to be installed separately on Amazon EC2
  28. 28. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved. Apache Airflow SageMaker Operators https://airflow.apache.org/docs/1.10.3/_api/airflow/contrib/operators/index.html • sagemaker_training_operator • sagemaker_tuning_operator • sagemaker_model_operator • sagemaker_endpoint_config_operator • sagemaker_endpoint_operator • sagemaker_transform_operator • segment_track_event_operator
  29. 29. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved. Demo – Recommendation batch job with Airflow
  30. 30. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
  31. 31. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved. ML workflows with SageMaker and Apache Airflow https://aws.amazon.com/ko/blogs/machine-learning/build-end-to-end-machine-learning-workflows-with- amazon-sagemaker-and-apache-airflow/
  32. 32. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
  33. 33. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved. Machine learning cycle Business Problem ML problem framing Data collection Data integration Data preparation and cleaning Data visualization and analysis Feature engineering Model training and parameter tuning Model evaluation Monitoring and debugging Model deployment Predictions YESNO
  34. 34. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved. Manage data on AWS Business Problem ML problem framing Data collection Data integration Data preparation and cleaning Data visualization and analysis Feature engineering Model training and parameter tuning Model evaluation Monitoring and debugging Model deployment Predictions YESNO
  35. 35. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved. Build and train models using SageMaker Business Problem ML problem framing Data collection Data integration Data preparation and cleaning Data visualization and analysis Feature engineering Model training and parameter tuning Model evaluation Monitoring and debugging Model deployment Predictions YESNO
  36. 36. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved. Deploy models using SageMaker Business Problem ML problem framing Data collection Data integration Data preparation and cleaning Data visualization and analysis Feature engineering Model training and parameter tuning Model evaluation Monitoring and debugging Model deployment Predictions YESNO
  37. 37. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved. What about the lines between the steps? Business Problem ML problem framing Data collection Data integration Data preparation and cleaning Data visualization and analysis Feature engineering Model training and parameter tuning Model evaluation Monitoring and debugging Model deployment Predictions YESNO AWS Step Functions AWS Batch AWS CodePipeline
  38. 38. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved.
  39. 39. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved. AWS 머신러닝(ML) 교육 및 자격증 Amazon의 개발자와 데이터 과학자를 교육하는 데 직접 활용 되었던 커리큘럼을 기반으로 학습하세요! 전체 팀을 위한 머신러닝 교육 원하는 방법으로! 교육 유연성 제공 전문성에 대한 검증 비즈니스 의사 결정자, 데이터 과학자, 개발자, 데이터 플랫폼 엔지니어 등 역할에 따라 제공되는 맞춤형 학습 경로를 확인하세요. 약 65개 이상의 온라인 과정 및 AWS 전문 강사를 통해 실습과 실적용의 기회가 제공되는 강의실 교육이 준비되어 있습니다. 업계에서 인정받는 ‘AWS 공인 머신러닝 – 전문분야’ 자격증을 통해 머신러닝 모델을 구축, 학습, 튜닝 및 배포하는 데 필요한 전문 지식이 있음을 입증할 수 있습니다. https://aws.amazon.com/ko/traini ng/learning-paths/machine- learning/
  40. 40. © 2020, Amazon Web Services, Inc. or its affiliates. All rights reserved. © 2019, Amazon Web Services, Inc. or its Affiliates. All rights reserved. AWS Innovate 온라인 컨퍼런스에 참석해주셔서 대단히 감사합니다. aws-korea-marketing@amazon.com twitter.com/AWSKorea facebook.com/amazonwebservices.ko youtube.com/user/AWSKorea slideshare.net/awskorea twitch.tv/aws 저희가 준비한 내용, 어떻게 보셨나요? 더 나은 세미나를 위하여 설문을 꼭 작성해 주시기 바랍니다.

×