Amazon SageMaker
Build, Train, and Deploy Machine Learning Models Quickly & Easily, at scale
Cobus Bernard
Senior Developer Advocate
Amazon Web Services
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Agenda
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
• Machine Learning onAWS
• Amazon Machine Learningstack
• The Machine Learning Lifecycle
• SageMaker Overview
• Demo
• SageMaker features
Definitions
• Supervised, Unsupervised, Reinforcement Learning – Learning
methods
• Model – The function produced by training
• Inference/Predictions – Requests to a model
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Our mission at AWS
Put machine learning in the hands
of every developer
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Amazon SageMaker: Launch Customer
“ With Amazon SageMaker, we can accelerate our Artificial
Intelligence initiatives at scale by building and deploying our
algorithms on the platform. We will create novel large-scale
machine learning and AI algorithms and deploy them on this
platform to solve complex problems that can power prosperity for
our customers.
”
–Ashok Srivastava, Chief Data Officer, Intuit
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Model hosting
(Amazon SageMaker)
Near real-time fraud detection in AWS using Amazon SageMaker
Calculate
features
Reader
Cleanser
Processor
Data
Lookup
Training
Feature store Model training
(Amazon SageMaker)
Model
Client service
Amazon
EMR
Amazon
SageMaker
Amazon
SageMaker
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
ML F R A M E W O R K S &
IN F R A S T R U C T U R E
A I SE R V IC E S
ML SE R V IC E S Data labeling |Pre-built algorithms & notebooks |One-click training and deployment
Build, train, and deploy machine learning models fast
Easily add intelligence to applications without machine
learning skills
Vision |Documents |Speech |Language |Chatbots |Forecasting |Recommendations
Flexibility & choice, highest-performing infrastructure
Support for ML frameworks |Compute options purpose-built for ML
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
M L F R A M E W O R K S &
IN F R A ST R U C T U R E
A I SE R V IC E S
I M A G E V I D E O
Vision S p eech Chatbots
M L SE R V IC E S
Fr a m e w o r k s In t e r f a ce s In f r a s t r u c t u r e
E C 2 P 3
& P 3 d n
E C 2 C 5 F P G A s G R E E N G R A S S E L A S T I C
I N F E R E N C E
Language Forecasting Recommendations
R E K O G N I T I O N R E K O G N I T I O N
T E X T R A C T
P O L L Y T R A N S C R I B E
New
T R A N S L A T E C O M P R E H E N D L E X
C O M P R E H E N D
M E D I C A L New
F O R E C A S T
New
P E R S O N A L I Z E
New
A M A Z O N
S A G E M A K E R
New
G R O U N D T R U T H
N O T E B O O K S New
A W S M A R K E T P L A C E
A L G O R I T H M S
R E I N F O R C E M E N T
L E A R N I N G
New
T R A I N I N G
O P T I M I Z A T I O N
( N E O ) New
D E P L O Y M E N T
H O S T I N G
New
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Whatdoes the Machine Learning
Lifecycle look like?
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Data Visualization &
Analysis
Business Problem –
ML problem framing Data Collection
Data Integration
Data Preparation &
Cleaning
Feature Engineering
Model Training &
Parameter Tuning
Model Evaluation
Are Business
Goals met?
Model Deployment
Monitoring &
Debugging
YesNo
DataAugmentation
Feature
Augmentation
The Machine Learning Process
Re-training
Predictions
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Let’s Dive Deeper into eachphase
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Data Visualization &
Analysis
Business Problem –
ML problem framing
Data Integration
Data Preparation &
Cleaning
Feature Engineering
Model Training &
Parameter Tuning
Model Evaluation
Are Business
Goals met?
Model Deployment
Monitoring &
Debugging
YesNo
DataAugmentation
Feature
Augmentation
Problem discovery
Re-training
Data Collection
• Help formulate theright
questions
• Domain Knowledge
Predictions
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Data Visualization &
Analysis
Business Problem –
ML problem framing
Data Integration
Data Preparation &
Cleaning
Feature Engineering
Model Training &
Parameter Tuning
Model Evaluation
Are Business
Goals met?
Model Deployment
Monitoring &
Debugging
YesNo
DataAugmentation
Feature
Augmentation
Retraining
Data Collection
• Need a dataplatform?
• Amazon S3
• AWS Glue
• Amazon Athena
• Amazon EMR
• Amazon Redshift
Spectrum
Integration
Predictions
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Data Visualization &
Analysis
Business Problem –
ML problem framing
Data Integration
Data Preparation &
Cleaning
Feature Engineering
Model Training &
Parameter Tuning
Model Evaluation
Are Business
Goals met?
Model Deployment
Monitoring &
Debugging
YesNo
DataAugmentation
Feature
Augmentation
Retraining
Data Collection
Model Training
Predictions
• Setup and manage
Notebook Environments
• Setup and manage
Training Clusters
• Write Data Connectors
• Scale ML algorithms to
large datasets
• Distribute ML training
algorithm to multiple
machines
• Secure Model artifacts
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Data Visualization &
Analysis
Business Problem –
ML problem framing
Data Integration
Data Preparation &
Cleaning
Feature Engineering
Model Training &
Parameter Tuning
Model Evaluation
Are Business
Goals met?
Model Deployment
Monitoring &
Debugging
YesNo
DataAugmentation
Feature
Augmentation
Retraining
Data Collection
Model Deployment
Predictions
• Setup and manageModel
Inference Clusters
• Manage and ScaleModel
Inference APIs
• Monitor and DebugModel
Predictions
• Models versioningand
performance tracking
• Automate New Model
version promotion to
production (A/B testing)
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Pre-built
notebooks for
common
problems
Built-in,
high-
performance
algorithms
DE P LO Y
BU I LD
Amazon SageMaker
Build, train, tune, and host your own models
TR A I N
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Authoring &
notebooks
ETL Access toAWS
database services
Access to AmazonS3
data lake
VPC
Zero setup notebook instance
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Pre-built
notebooks for
common
problems
Built-in,
high-
performance
algorithms
One-click
training
Hyperparameter
Tuning
DE P LO Y
BU I LD TR A I N & TU N E
Amazon SageMaker
Build, train, tune, and host your own models
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Amazon ECR
Trainingdata
Modelartifacts
Training code Helper code
Model Training (on EC2)
Training code
Amazon SageMaker
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Amazon SageMaker
Built-in algorithms
Deep learning
frameworks
MXNet & Gluon
Tensorflow
PyTorch
Chainer
Custom Image
Training (single machine
or distributed cluster)
Data
Flexible and scalable model training
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Pre-built
notebooks for
common
problems
Built-in,
high-
performance
algorithms
One-click
training
Hyperparameter
Tuning
BU I LD TR A I N & TU N E
Amazon SageMaker
Build, train, tune, and host your own models
Fully managed
hosting with
Auto Scaling
One-click
deployment
DE PLO Y
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Amazon ECR
Trainingdata
Modelartifacts
Training code Helper code
Model Training (on EC2)
Inference code Helpercode
Model Hosting (on EC2)
GroundTruth
Client application
Inferencecode
Training code
Inferenceresponse
InferenceEndpoint
Inference request
Amazon SageMaker
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Amazon ECR
30 50
10 10
Model Artifacts
InferenceImage
Model versions
Endpoint configuration
InferenceEndpoint
Amazon SageMaker
 Auto Scaling
 A/B Testing
 Low latency &high
throughput
 Bring your ownmodel
One-click model deployment
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Fully managed
hosting with
Auto Scaling
One-click
deployment
Pre-built
notebooks for
common
problems
Built-in, high
performance
algorithms
One-click
training
BU I LD TR A I N & TU N E DE PLO Y
End-to-end encryption with AWS
KMS
End-to-end VPC support
Compliance and auditcapabilities
Metadata and experiment management capabilities
Pay as yougo
Amazon SageMaker
Build, train, tune, and host your own models
Hyperparameter
Tuning
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Amazon SageMaker is the best place
to runTensorFlow in the cloud
AW S - O P T I M I Z E D T E N S O R F L O W
• Fully-managed training and hosting
• Near-linear scaling across 100s of GPU
• 7 5 % lower inference costs with Amazon
Elastic Inference
• 3x faster network throughput with EC2 P3
ST O C K T E N S O R F L O W
6 5 %
9 0 %
Sc a l i n g e f f i c ie n c y w i t h 2 5 6 G P U s
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Amazon SageMaker Demo
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Additional features of AmazonSageMaker
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Training
9 0 %
Training drives research
and development
Inference (Prediction)
1 0 %
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Predictions drive
complexity and costin
production
Inference (Prediction)
9 0 %
Training
1 0 %
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Lower inference costs Match capacityto
demand
Available between
1 to 32 TFLOPS per
accelerator
Integrated with
Amazon EC2 and
Amazon SageMaker
KEY FEATURES
Support for TensorFlow,
Apache MXNet, andONNX
with PyTorch coming soon
Single and
mixed-precision
operations
Amazon Elastic Inference
Reduce deep learning inference costs up to 7 5 %
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Neo
Amazon SageMaker Neo
Train once, run anywhere
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
K E Y F E A T U R E S
Open-source device runtime and compiler,
1/10th the size of originalframeworks
Amazon SageMaker Neo
Train once, run anywhere with 2xperformance
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Automatic model tuning – improve model quality
Neural networks
Number of layers
Hidden layer width
Learning rate
Embedding
dimensions
Dropout
…
Decision trees
Tree depth
Max leaf nodes
Gamma
Eta
AWS Lambda
Alpha
…
…
“Hyperparameters”
(algorithm parameters that significantly affect model quality)
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Successful models require high-quality data
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Amazon SageMaker Ground Truth
Label machine learning training data easily and accurately
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Subscribe in a
single click
Available in
Amazon SageMaker
KEY FEATURES
Automatic labeling via machine learning
IP protection
Automated billing and metering
Browse or search
AWS Marketplace
S E L L ER S
Broad selection of paid,free and
open-source algorithms and models
Data protection
Discoverable on your AWS bill
BU Y E R S
AWS Marketplace for Machine Learning
ML algorithms and models available instantly
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
A fully autonomous 1/18th-scale race car designed to help you learn about
reinforcement learning through autonomousdriving
• Build machine learning models in Amazon
SageMaker
• Train, test, and iterate on the track using the
AWS DeepRacer 3D racing simulator
• Compete in the world’s first global autonomous
racing league, to race for prizes and a chance to
advance to win the coveted AWS DeepRacer Cup
Amazon SageMaker RL
Get started with Reinforcement learning using AWS DeepRacer
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Amazon SageMaker RL
Get started with Reinforcement learning using AWS DeepRacer
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
© 2019, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
2D & 3D physics
environments and
OpenGym support
Support Amazon Sumerian,
AWS RoboMaker and the open
source Robotics Operating
System (ROS)project
Fully
managed
Example notebooks
and tutorials
K E Y F E A T U R E S
Amazon SageMaker RL
Reinforcement learning for every developer and data scientist
Key Advantages of Amazon SageMaker
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
One-click
model training
& deployment
10x
better algorithm
performance
Trainonce
run anywhere
A M A Z O N SAG E M A K E R
70%
cost reduction for data
labeling using GroundTruth
2x
performance increases from
model optimization withNeo
75%
cost reduction for inference
with Elastic Inference
RE D UC E CO S TS INC RE A S E P ER FO R MA N CE EA S E- OF - US E
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Useful Amazon SageMaker resources
Questions
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

Deep Dive Amazon SageMaker

  • 1.
    Amazon SageMaker Build, Train,and Deploy Machine Learning Models Quickly & Easily, at scale Cobus Bernard Senior Developer Advocate Amazon Web Services © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 2.
    Agenda © 2020, AmazonWeb Services, Inc. or its Affiliates. All rights reserved. • Machine Learning onAWS • Amazon Machine Learningstack • The Machine Learning Lifecycle • SageMaker Overview • Demo • SageMaker features
  • 3.
    Definitions • Supervised, Unsupervised,Reinforcement Learning – Learning methods • Model – The function produced by training • Inference/Predictions – Requests to a model © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 4.
    Our mission atAWS Put machine learning in the hands of every developer © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 5.
    © 2020, AmazonWeb Services, Inc. or its Affiliates. All rights reserved.
  • 6.
    Amazon SageMaker: LaunchCustomer “ With Amazon SageMaker, we can accelerate our Artificial Intelligence initiatives at scale by building and deploying our algorithms on the platform. We will create novel large-scale machine learning and AI algorithms and deploy them on this platform to solve complex problems that can power prosperity for our customers. ” –Ashok Srivastava, Chief Data Officer, Intuit © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 7.
    Model hosting (Amazon SageMaker) Nearreal-time fraud detection in AWS using Amazon SageMaker Calculate features Reader Cleanser Processor Data Lookup Training Feature store Model training (Amazon SageMaker) Model Client service Amazon EMR Amazon SageMaker Amazon SageMaker © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 8.
    ML F RA M E W O R K S & IN F R A S T R U C T U R E A I SE R V IC E S ML SE R V IC E S Data labeling |Pre-built algorithms & notebooks |One-click training and deployment Build, train, and deploy machine learning models fast Easily add intelligence to applications without machine learning skills Vision |Documents |Speech |Language |Chatbots |Forecasting |Recommendations Flexibility & choice, highest-performing infrastructure Support for ML frameworks |Compute options purpose-built for ML © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 9.
    M L FR A M E W O R K S & IN F R A ST R U C T U R E A I SE R V IC E S I M A G E V I D E O Vision S p eech Chatbots M L SE R V IC E S Fr a m e w o r k s In t e r f a ce s In f r a s t r u c t u r e E C 2 P 3 & P 3 d n E C 2 C 5 F P G A s G R E E N G R A S S E L A S T I C I N F E R E N C E Language Forecasting Recommendations R E K O G N I T I O N R E K O G N I T I O N T E X T R A C T P O L L Y T R A N S C R I B E New T R A N S L A T E C O M P R E H E N D L E X C O M P R E H E N D M E D I C A L New F O R E C A S T New P E R S O N A L I Z E New A M A Z O N S A G E M A K E R New G R O U N D T R U T H N O T E B O O K S New A W S M A R K E T P L A C E A L G O R I T H M S R E I N F O R C E M E N T L E A R N I N G New T R A I N I N G O P T I M I Z A T I O N ( N E O ) New D E P L O Y M E N T H O S T I N G New © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 10.
    Whatdoes the MachineLearning Lifecycle look like? © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 11.
    Data Visualization & Analysis BusinessProblem – ML problem framing Data Collection Data Integration Data Preparation & Cleaning Feature Engineering Model Training & Parameter Tuning Model Evaluation Are Business Goals met? Model Deployment Monitoring & Debugging YesNo DataAugmentation Feature Augmentation The Machine Learning Process Re-training Predictions © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 12.
    Let’s Dive Deeperinto eachphase © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 13.
    Data Visualization & Analysis BusinessProblem – ML problem framing Data Integration Data Preparation & Cleaning Feature Engineering Model Training & Parameter Tuning Model Evaluation Are Business Goals met? Model Deployment Monitoring & Debugging YesNo DataAugmentation Feature Augmentation Problem discovery Re-training Data Collection • Help formulate theright questions • Domain Knowledge Predictions © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 14.
    Data Visualization & Analysis BusinessProblem – ML problem framing Data Integration Data Preparation & Cleaning Feature Engineering Model Training & Parameter Tuning Model Evaluation Are Business Goals met? Model Deployment Monitoring & Debugging YesNo DataAugmentation Feature Augmentation Retraining Data Collection • Need a dataplatform? • Amazon S3 • AWS Glue • Amazon Athena • Amazon EMR • Amazon Redshift Spectrum Integration Predictions © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 15.
    Data Visualization & Analysis BusinessProblem – ML problem framing Data Integration Data Preparation & Cleaning Feature Engineering Model Training & Parameter Tuning Model Evaluation Are Business Goals met? Model Deployment Monitoring & Debugging YesNo DataAugmentation Feature Augmentation Retraining Data Collection Model Training Predictions • Setup and manage Notebook Environments • Setup and manage Training Clusters • Write Data Connectors • Scale ML algorithms to large datasets • Distribute ML training algorithm to multiple machines • Secure Model artifacts © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 16.
    Data Visualization & Analysis BusinessProblem – ML problem framing Data Integration Data Preparation & Cleaning Feature Engineering Model Training & Parameter Tuning Model Evaluation Are Business Goals met? Model Deployment Monitoring & Debugging YesNo DataAugmentation Feature Augmentation Retraining Data Collection Model Deployment Predictions • Setup and manageModel Inference Clusters • Manage and ScaleModel Inference APIs • Monitor and DebugModel Predictions • Models versioningand performance tracking • Automate New Model version promotion to production (A/B testing) © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 17.
    Pre-built notebooks for common problems Built-in, high- performance algorithms DE PLO Y BU I LD Amazon SageMaker Build, train, tune, and host your own models TR A I N © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 18.
    Authoring & notebooks ETL AccesstoAWS database services Access to AmazonS3 data lake VPC Zero setup notebook instance © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 19.
    Pre-built notebooks for common problems Built-in, high- performance algorithms One-click training Hyperparameter Tuning DE PLO Y BU I LD TR A I N & TU N E Amazon SageMaker Build, train, tune, and host your own models © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 20.
    Amazon ECR Trainingdata Modelartifacts Training codeHelper code Model Training (on EC2) Training code Amazon SageMaker © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 21.
    Amazon SageMaker Built-in algorithms Deeplearning frameworks MXNet & Gluon Tensorflow PyTorch Chainer Custom Image Training (single machine or distributed cluster) Data Flexible and scalable model training © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 22.
    Pre-built notebooks for common problems Built-in, high- performance algorithms One-click training Hyperparameter Tuning BU ILD TR A I N & TU N E Amazon SageMaker Build, train, tune, and host your own models Fully managed hosting with Auto Scaling One-click deployment DE PLO Y © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 23.
    Amazon ECR Trainingdata Modelartifacts Training codeHelper code Model Training (on EC2) Inference code Helpercode Model Hosting (on EC2) GroundTruth Client application Inferencecode Training code Inferenceresponse InferenceEndpoint Inference request Amazon SageMaker © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 24.
    Amazon ECR 30 50 1010 Model Artifacts InferenceImage Model versions Endpoint configuration InferenceEndpoint Amazon SageMaker  Auto Scaling  A/B Testing  Low latency &high throughput  Bring your ownmodel One-click model deployment © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 25.
    Fully managed hosting with AutoScaling One-click deployment Pre-built notebooks for common problems Built-in, high performance algorithms One-click training BU I LD TR A I N & TU N E DE PLO Y End-to-end encryption with AWS KMS End-to-end VPC support Compliance and auditcapabilities Metadata and experiment management capabilities Pay as yougo Amazon SageMaker Build, train, tune, and host your own models Hyperparameter Tuning © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 26.
    Amazon SageMaker isthe best place to runTensorFlow in the cloud AW S - O P T I M I Z E D T E N S O R F L O W • Fully-managed training and hosting • Near-linear scaling across 100s of GPU • 7 5 % lower inference costs with Amazon Elastic Inference • 3x faster network throughput with EC2 P3 ST O C K T E N S O R F L O W 6 5 % 9 0 % Sc a l i n g e f f i c ie n c y w i t h 2 5 6 G P U s © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 27.
    Amazon SageMaker Demo ©2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 28.
    Additional features ofAmazonSageMaker © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 29.
    Training 9 0 % Trainingdrives research and development Inference (Prediction) 1 0 % © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 30.
    Predictions drive complexity andcostin production Inference (Prediction) 9 0 % Training 1 0 % © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 31.
    Lower inference costsMatch capacityto demand Available between 1 to 32 TFLOPS per accelerator Integrated with Amazon EC2 and Amazon SageMaker KEY FEATURES Support for TensorFlow, Apache MXNet, andONNX with PyTorch coming soon Single and mixed-precision operations Amazon Elastic Inference Reduce deep learning inference costs up to 7 5 % © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 32.
    Neo Amazon SageMaker Neo Trainonce, run anywhere © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 33.
    K E YF E A T U R E S Open-source device runtime and compiler, 1/10th the size of originalframeworks Amazon SageMaker Neo Train once, run anywhere with 2xperformance © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 34.
    Automatic model tuning– improve model quality Neural networks Number of layers Hidden layer width Learning rate Embedding dimensions Dropout … Decision trees Tree depth Max leaf nodes Gamma Eta AWS Lambda Alpha … … “Hyperparameters” (algorithm parameters that significantly affect model quality) © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 35.
    Successful models requirehigh-quality data © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 36.
    Amazon SageMaker GroundTruth Label machine learning training data easily and accurately © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 37.
    Subscribe in a singleclick Available in Amazon SageMaker KEY FEATURES Automatic labeling via machine learning IP protection Automated billing and metering Browse or search AWS Marketplace S E L L ER S Broad selection of paid,free and open-source algorithms and models Data protection Discoverable on your AWS bill BU Y E R S AWS Marketplace for Machine Learning ML algorithms and models available instantly © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 38.
    A fully autonomous1/18th-scale race car designed to help you learn about reinforcement learning through autonomousdriving • Build machine learning models in Amazon SageMaker • Train, test, and iterate on the track using the AWS DeepRacer 3D racing simulator • Compete in the world’s first global autonomous racing league, to race for prizes and a chance to advance to win the coveted AWS DeepRacer Cup Amazon SageMaker RL Get started with Reinforcement learning using AWS DeepRacer © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 39.
    Amazon SageMaker RL Getstarted with Reinforcement learning using AWS DeepRacer © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 40.
    © 2019, AmazonWeb Services, Inc. or its Affiliates. All rights reserved. 2D & 3D physics environments and OpenGym support Support Amazon Sumerian, AWS RoboMaker and the open source Robotics Operating System (ROS)project Fully managed Example notebooks and tutorials K E Y F E A T U R E S Amazon SageMaker RL Reinforcement learning for every developer and data scientist
  • 41.
    Key Advantages ofAmazon SageMaker © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 42.
    One-click model training & deployment 10x betteralgorithm performance Trainonce run anywhere A M A Z O N SAG E M A K E R 70% cost reduction for data labeling using GroundTruth 2x performance increases from model optimization withNeo 75% cost reduction for inference with Elastic Inference RE D UC E CO S TS INC RE A S E P ER FO R MA N CE EA S E- OF - US E © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
  • 43.
    Useful Amazon SageMakerresources Questions © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.