Amazon SageMaker Deep Dive

Slide 1

Slide 1 text

Slide 2

Slide 2 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T Amazon SageMaker Deep Dive Yoshitaka Haribara Solutions Architect Amazon Web Services Japan K.K. A 2 - 0 8

Slide 3

Slide 3 text

Slide 4

Slide 4 text

Slide 5

Slide 5 text

Slide 6

Slide 6 text

Slide 7

Slide 7 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T Customer-focused 90%=- ML (" JG D Multi-framework 9:'#) ! Pace of innovation H+200=- ML 25 /79:8BF> Breadth and depth IA AI/ML (%*C0 Security and analytics KE$&QL4 28BS;.P?N Embedded R&D OJ,@ ( state-of-the-art 3< AWS 8R6M 1 (

Slide 8

Slide 8 text

Slide 9

Slide 9 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T AI ML AMAZON SAGEMAKER A M A Z O N E C 2 C 5 I n s t a n c e s A M A Z O N E C 2 P 3 I n s t a n c e s F P G A s Frameworks AWS & A m a z o n R e k o g n i t i o n I m a g e / V i d e o A m a z o n P o l l y A m a z o n T r a n s c r i b e A m a z o n T r a n s l a t e A m a z o n C o m p r e h e n d A m a z o n L e x Chatbots A m a z o n F o r e c a s t Forecasting A m a z o n T e x t r a c t A m a z o n P e r s o n a l i z e Recommendations Vision Speech Language E l a s t i c I n f e r e n c e Infrastructure Interfaces

Slide 10

Slide 10 text

Slide 11

Slide 11 text

Slide 12

Slide 12 text

Slide 13

Slide 13 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T 07)=(96:(9 1;!#" !# 53 2/+' $453 " (9 ,.-/ ML %<*8& 1 2 3 $453 +'# Amazon SageMaker Ground Truth !# # Amazon SageMaker

Slide 14

Slide 14 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T @J3W2OIP2O!%BS&(' &( FD C?:1 )EFD ' 2O <>=? ML %$ /U4N0 1 2 3 )EFD :1$( Amazon SageMaker Ground Truth AWS Marketplace for Machine Learning &( # "( • k-means #$( • Factorization Machines (& () • DeepAR (,KR7Q) • BlazingText (Word2Vec) • XGBoost • ;T+L9.8* • Seq2Seq • LDA / Neural Topic Modelling (!%) • 56++V • AG2OH (-M / +L) Amazon SageMaker

Slide 15

Slide 15 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T 07)=(96:(9 1;!#" !# 53 2/+' $453 " (9 ,.-/ ML %<*8& 1 2 3 $453 +'# Amazon EC2 P3 Instances Amazon SageMaker RL Amazon SageMaker Ground Truth AWS Marketplace for Machine Learning !# # Amazon SageMaker

Slide 16

Slide 16 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T !# # 07)=(96:(9 1;!#" !# 53 2/+' $453 " (9 ,.-/ ML %<*8& 1 2 3 $453 +'# Amazon EC2 P3 Instances Amazon SageMaker RL Amazon SageMaker Ground Truth AWS Marketplace for Machine Learning Amazon SageMaker

Slide 17

Slide 17 text

Slide 18

Slide 18 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T !# # 07)=(96:(9 1;!#" !# 53 2/+' $453 " (9 ,.-/ ML %<*8& 1 2 3 $453 +'# Amazon EC2 P3 Instances Amazon SageMaker RL Amazon SageMaker Ground Truth Amazon Elastic Inference AWS Marketplace for Machine Learning Amazon SageMaker Neo Amazon SageMaker

Slide 19

Slide 19 text

Slide 20

Slide 20 text

Slide 21

Slide 21 text

Slide 22

Slide 22 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T S U M M I T © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. SageMaker Python SDK import sagemaker from sagemaker.mxnet import MXNet # Estimator estimator = MXNet("train.py", # role=sagemaker.get_execution_role(), train_instance_count=1, train_instance_type="ml.p3.2xlarge", framework_version="1.4.0")

Slide 23

Slide 23 text

Slide 24

Slide 24 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T S U M M I T © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. SageMaker Python SDK import sagemaker from sagemaker.mxnet import MXNet # #$ Estimator estimator = MXNet("train.py", # ! role=sagemaker.get_execution_role(), train_instance_count=1, train_instance_type="ml.p3.2xlarge", framework_version="1.4.0") estimator.fit("s3://mybucket/data/train") # fit predictor = estimator.deploy(initial_instance_count=1, instance_type="ml.m4.xlarge") # deploy "

Slide 25

Slide 25 text

Slide 26

Slide 26 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T S U M M I T © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. (Script Mode) train.py import argparse if __name__ == '__main__’: parser = argparse.ArgumentParser() # hyperparameters parser.add_argument('--epochs', type=int, default=10) # input data and model directories parser.add_argument('--train', type=str, default=os.environ['SM_CHANNEL_TRAIN']) parser.add_argument('--test', type=str, default=os.environ['SM_CHANNEL_TEST']) parser.add_argument('--model-dir', type=str, default=os.environ['SM_MODEL_DIR']) args, _ = parser.parse_known_args() /opt/ml/input/data/train /opt/ml/input/data/test /opt/ml/model

Slide 27

Slide 27 text

Slide 28

Slide 28 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T Amazon SageMaker Jupyter Notebook/Lab Amazon S3 The Jupyter Trademark is registered with the U.S. Patent & Trademark Office.

Slide 29

Slide 29 text

Slide 30

Slide 30 text

Slide 31

Slide 31 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T Amazon SageMaker Jupyter Notebook/Lab Amazon S3 Amazon EC2 P3 Instances Amazon ECR The Jupyter Trademark is registered with the U.S. Patent & Trademark Office. /opt/ml/input/data/train /opt/ml/input/data/test /opt/ml/model

Slide 32

Slide 32 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T Amazon SageMaker Amazon EC2 P3 Instances Jupyter Notebook/Lab Amazon S3 The Jupyter Trademark is registered with the U.S. Patent & Trademark Office.

Slide 33

Slide 33 text

Slide 34

Slide 34 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T Amazon SageMaker Amazon EC2 P3 Instances Jupyter Notebook/Lab Endpoint/ Batch transform Amazon S3 Amazon ECR The Jupyter Trademark is registered with the U.S. Patent & Trademark Office.

Slide 35

Slide 35 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T Amazon SageMaker Amazon EC2 P3 Instances Endpoint/ Batch transform Amazon S3 Amazon API Gateway AWS Lambda User The Jupyter Trademark is registered with the U.S. Patent & Trademark Office.

Slide 36

Slide 36 text

Slide 37

Slide 37 text

Slide 38

Slide 38 text

Slide 39

Slide 39 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T AWS Step Functions • JSON L A • A • CloudWatch Event Start End Train Deploy Fetch data AWS Lambda Amazon SageMaker AWS Lambda (Amazon SageMaker) Amazon CloudWatch Events (Schedule / event trigger)

Slide 40

Slide 40 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T AWS Step Functions workflow Test data Train data Data Scientists/ Developers Git webhook docker push AWS Glue Amazon S3 (data) Amazon SageMaker Training Job / Batch Transform AWS CodeCommit or 3rd party Git repository Amazon S3 (raw data) Amazon Elastic Container Registry (ECR) AWS CodeBuild Endpoint https://aws.amazon.com/blogs/machine-learning/automated-and-continuous-deployment-of-amazon-sagemaker-models-with-aws-step-functions/ AWS Lambda SageMaker Endpoint deploy Amazon S3 (trained model) git push

Slide 41

Slide 41 text

Slide 42

Slide 42 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T AWS Step Functions workflow Test data Train data Data Scientists/ Developers Git webhook docker push AWS Glue Amazon S3 (data) Amazon SageMaker Training Job / Batch Transform AWS CodeCommit or 3rd party Git repository Amazon S3 (raw data) AWS CodeBuild Endpoint https://aws.amazon.com/blogs/machine-learning/automated-and-continuous-deployment-of-amazon-sagemaker-models-with-aws-step-functions/ AWS Lambda SageMaker Endpoint deploy Amazon S3 (trained model) git push Amazon Elastic Container Registry (ECR)

Slide 43

Slide 43 text

Slide 44

Slide 44 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T Apache Airflow • ) 2 a DPM S CMR • 2 + ( D A • + G E 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

Slide 45

Slide 45 text

Slide 46

Slide 46 text

Slide 47

Slide 47 text

Slide 48

Slide 48 text

Slide 49

Slide 49 text

Slide 50

Slide 50 text

Slide 51

Slide 51 text

Slide 52

Slide 52 text

Slide 53

Slide 53 text

Slide 54

Slide 54 text

Slide 55

Slide 55 text

Slide 56

Slide 56 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T Model aws sagemaker create-model --model-name model1 --primary-container ‘{“Image”: “123.dkr.ecr.amazonaws.com/algo”, “ModelDataUrl”: “s3://bkt/model1.tar.gz”}’ --execution-role-arn arn:aws:iam::123:role/me

Slide 57

Slide 57 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T Model Endpoint configuration aws sagemaker create-model --model-name model1 --primary-container ‘{“Image”: “123.dkr.ecr.amazonaws.com/algo”, “ModelDataUrl”: “s3://bkt/model1.tar.gz”}’ --execution-role-arn arn:aws:iam::123:role/me aws sagemaker create-endpoint-config --endpoint-config-name model1-config --production-variants ‘{“InitialInstanceCount”: 2, “InstanceType”: “ml.m4.xlarge”, “InitialVariantWeight”: 1, “ModelName”: “model1”, “VariantName”: “AllTraffic”}’

Slide 58

Slide 58 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T Model Endpoint configuration Endpoint aws sagemaker create-model --model-name model1 --primary-container ‘{“Image”: “123.dkr.ecr.amazonaws.com/algo”, “ModelDataUrl”: “s3://bkt/model1.tar.gz”}’ --execution-role-arn arn:aws:iam::123:role/me aws sagemaker create-endpoint-config --endpoint-config-name model1-config --production-variants ‘{“InitialInstanceCount”: 2, “InstanceType”: “ml.m4.xlarge”, “InitialVariantWeight”: 1, “ModelName”: “model1”, “VariantName”: “AllTraffic”}’ aws sagemaker create-endpoint --endpoint-name my-endpoint --endpoint-config-name model1-config

Slide 59

Slide 59 text

Slide 60

Slide 60 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T Model v2 aws sagemaker create-model --model-name model2 --primary-container ‘{“Image”: “123.dkr.ecr.amazonaws.com/algo”, “ModelDataUrl”: “s3://bkt/model2.tar.gz”}’ --execution-role-arn arn:aws:iam::123:role/me

Slide 61

Slide 61 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T Model v2 endpoint configuration aws sagemaker create-model --model-name model2 --primary-container ‘{“Image”: “123.dkr.ecr.amazonaws.com/algo”, “ModelDataUrl”: “s3://bkt/model2.tar.gz”}’ --execution-role-arn arn:aws:iam::123:role/me aws sagemaker create-endpoint-config --endpoint-config-name model2-config --production-variants ‘{“InitialInstanceCount”: 2, “InstanceType”: “ml.m4.xlarge”, “InitialVariantWeight”: 1, “ModelName”: “model2”, “VariantName”: “AllTraffic”}’

Slide 62

Slide 62 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T Model v2 endpoint configuration endpoint aws sagemaker create-model --model-name model2 --primary-container ‘{“Image”: “123.dkr.ecr.amazonaws.com/algo”, “ModelDataUrl”: “s3://bkt/model2.tar.gz”} --execution-role-arn arn:aws:iam::123:role/me aws sagemaker create-endpoint-config --endpoint-config-name model2-config --production-variants ‘{“InitialInstanceCount”: 2, “InstanceType”: “ml.m4.xlarge”, “InitialVariantWeight”: 1, “ModelName”: “model2”, “VariantName”: “AllTraffic”}’ aws sagemaker update-endpoint --endpoint-name my-endpoint --endpoint-config-name model2-config

Slide 63

Slide 63 text

Slide 64

Slide 64 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T endpoint configuration aws sagemaker create-endpoint-config --endpoint-config-name both-models-config --production-variants ‘[{“InitialInstanceCount”: 2, “InstanceType”: “ml.m4.xlarge”, “InitialVariantWeight”: 95, “ModelName”: “model1”, “VariantName”: “model1-traffic”}, {“InitialInstanceCount”: 2, “InstanceType”: “ml.m4.xlarge”, “InitialVariantWeight”: 5, “ModelName”: “model2”, “VariantName”: “model2-traffic”}]’

Slide 65

Slide 65 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T endpoint configuration endpoint aws sagemaker create-endpoint-config --endpoint-config-name both-models-config --production-variants ‘[{“InitialInstanceCount”: 2, “InstanceType”: “ml.m4.xlarge”, “InitialVariantWeight”: 95, “ModelName”: “model1”, “VariantName”: “model1-traffic”}, {“InitialInstanceCount”: 2, “InstanceType”: “ml.m4.xlarge”, “InitialVariantWeight”: 5, “ModelName”: “model2”, “VariantName”: “model2-traffic”}]’ aws sagemaker update-endpoint --endpoint-name my-endpoint --endpoint-config-name both-models-config

Slide 66

Slide 66 text

Slide 67

Slide 67 text

Slide 68

Slide 68 text

Slide 69

Slide 69 text

Slide 70

Slide 70 text

Slide 71

Slide 71 text

Slide 72

Slide 72 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T CPU automatic scaling policy aws application-autoscaling register-scalable-target --service-namespace sagemaker --resource-id endpoint/my-endpoint/variant/model2 --scalable-dimension sagemaker:variant:DesiredInstanceCount --min-capacity 2 --max-capacity 5 aws application-autoscaling put-scaling-policy --policy-name model2-scaling --service-namespace sagemaker --resource-id endpoint/my-endpoint/variant/model2 --scalable-dimension sagemaker:variant:DesiredInstanceCount --policy-type TargetTrackingScaling --target-tracking-scaling-policy-configuration ‘{"TargetValue": 50, "CustomizedMetricSpecification": {"MetricName": "CPUUtilization", "Namespace": "/aws/sagemaker/Endpoints", "Dimensions": [{"Name": "EndpointName", "Value": "my-endpoint"}, {"Name": "VariantName","Value": ”model2"}], "Statistic": "Average", "Unit": "Percent”}}’

Slide 73

Slide 73 text

Slide 74

Slide 74 text

Slide 75

Slide 75 text

Slide 76

Slide 76 text

Slide 77

Slide 77 text

Slide 78

Slide 78 text

Slide 79

Slide 79 text

Slide 80

Slide 80 text

Slide 81

Slide 81 text

Slide 82

Slide 82 text

Slide 83

Slide 83 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T Parse Model Optimize Tensors Generate Code Optimize Graph TensorFlow, MXNet, PyTorch, XGBoost ")C %05 ML " (NN) & %$ @B1?+: 8D 6E/ (-% shape %$4* .;>!,2 %= '#"$ =97A<3 Neo (TVM / treelite) Pruning Operator fusion Nested loop tiling Vectorization / Tensorization Data layout transform

Slide 84

Slide 84 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T WP • D , T GcrzG • FmyGo dFlszaT • , , P e kn • GcrzG ul P T • , , A B T GcrzG G • D , A A S pkGf • T T fnwbS M • w gatI ihIE K J IS MN D D ,

Slide 85

Slide 85 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T Related breakouts | l k hu WML G 32 Ot dg a 32 2 C : A 5 : 5 ry e ,4D 4E 248 04 mik SW cens 2D :5. A 4 -1 ,4D 4E 248 04 ho z O c k 32 2 C : A 5 : 5 2C : 04 D4 ,4D

Slide 86

Slide 86 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T ! S U M M I T © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. Yoshitaka Haribara [email protected] @_hariby

Slide 87

Slide 87 text

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved. S U M M I T References • ML@Loft [Blog#1] • Amazon SageMaker [Web#1, Blog#2, #3, #4, #5] • AWS Black Belt Online Seminar • Basic [Movie, Slides] • Advanced [Movie, Slides] • AWS : Keras [Blog], Apache Airflow [Blog], Kubeflow [Blog], • HPO: SageMaker default [Blog], Optuna [Blog] • GPU [Blog] • SageMaker Containers [GitHub] • Jupyter /IDE [SageMaker Python SDK] API