introduction to machine learning on ibm power systems
TRANSCRIPT
© IBM Corporation, 2016
Introduction to Machine Learning on IBM Power Systems
20th July 2017
Presented by David Spurway
IBM Power Systems Product Manager
IBM Systems, UK and Ireland
2 © IBM Corporation, 2016
Augmented intelligence, Artificial Intelligence, Cognitive
driving innovation Faster
3 © IBM Corporation, 2016
My friends at Uni…
4 © IBM Corporation, 2016
MNIST dataset
5 © IBM Corporation, 2016
6 © IBM Corporation, 2016
7 © IBM Corporation, 2016
GPUs are like minions
The individual cores in a GPU
are not very powerful
But gather loads together, and
remarkable things can happen!
8 © IBM Corporation, 2016
Power Systems and NVIDIA GPU Roadmap
NVIDIA GPU NVIDIA GPU with NVLink
2015
Power Chip Power Chip
with NVLink
2016
80 GB/s
Peak*PCIe x1632 GB/s
https://devblogs.nvidia.com/parallelforall/inside-pascal/
9 © IBM Corporation, 2016
10 © IBM Corporation, 2016
From Big Data to AI client journey
11 © IBM Corporation, 2016
OBSERVATION DECISIONINTERPRETATION EVALUATION
010101010101010111100010011001010111 0000000000010101010100000000000 11110101111000 000000000000 111111 010101 101010 10101010100
PrescriptiveBest Outcomes?
DescriptiveWhat Has Happened?
CognitiveLearn Dynamically
PredictiveWhat Could Happen?
12 © IBM Corporation, 2016
010101010101010111100010011001010111 0000000000010101010100000000000 11110101111000 000000000000 111111 010101 101010 10101010100
OBSERVATION DECISIONINTERPRETATION EVALUATION
PrescriptiveBest Outcomes?
DescriptiveWhat Has Happened?
CognitiveLearn Dynamically
PredictiveWhat Could Happen?
ACTIONDATA
How many fraudsduring last month? Per Country ?
Which Transactions will be fraudulent ?
What is the best action in light of potential fraud ? In Natural Language
: « Explain me whythis transaction is
fraudulent ?
13 © IBM Corporation, 2016
PrescriptiveBest Outcomes?
DescriptiveWhat Has Happened?
CognitiveLearn Dynamically
PredictiveWhat Could Happen?
- Artificial -Intelligence
- Big Data -
NLP
Robot
KnowledgeBase
Deep LearningMachine Learning010101010101010111100010011001010111
1000101
1000101
1000101
111010111010
00000000000010101010100000000000 111101011
14 © IBM Corporation, 2016
Prepare the data
ALL DATA
Input VAR
Training Data
Test Data
Machine Learning Algorithms
Predictive Model PREDICTION ? Test Data
ACCURACY ?
Machine Learning Algorithms use training data to create a
predictive model: its accuracy is tested on holdback data
Machine Learning
TensorFlow Caffee Torch Theano Chainer Spark ML ……
Prepare Data Build/Train Model Deploy/Score Monitor/Refine
Iterative Development
15 © IBM Corporation, 2016
Prepare the data
ALL DATA
Input VAR
Training Data
Test Data
Machine Learning Algorithms
Predictive Model PREDICTION ? Test Data
ACCURACY ?
Machine Learning Algorithms use training data to create a
predictive model: its accuracy is tested on holdback data
Machine Learning
TensorFlow Caffee Torch Theano Chainer Spark ML ……
Recognising Patterns
Face Detection
Spoken Words
Transportation Security
Extracting Insight
From Text/Video
Avoiding spam
Healthcare improved diag
Discovering Anomalies
Financial Frauds
Sensor Readings
Manufacturing optimisation
Making Predictions
Stock Trade Client Behavior
Retail shopping / promotion
Use Cases / Industries
16 © IBM Corporation, 2016
Deep Learning Goes to the Dogs
• https://openpowerfoundation.org/blogs/deep-learning-goes-to-the-dogs/
• http://vision.stanford.edu/aditya86/ImageNetDogs/
• The Stanford Dogs dataset contains images of 120 breeds of dogs from
around the world. This dataset has been built using images and annotation
from ImageNet for the task of fine-grained image categorization.
17 © IBM Corporation, 2016
My recent buyer’s journey…
18 © IBM Corporation, 2016
Some challenges…
https://uk.pinterest.com/pin/197595502370835426/sent/?sender=5279
06524979412201&invite_code=7fc292525ac44aafa6736bdec95dd1b5
Speziale Floral Lace Fit & Flare Dress
Items in this section are
temporarily out of stock
19 © IBM Corporation, 2016
Where I ended up going…
Petite Clothing
Update your wardrobe with Wallis'
stunning must have petite range.
Designed for women who are 5'3" and
under
20 © IBM Corporation, 2016
Example of Datasets available
http://mmlab.ie.cuhk.edu.hk/projects/DeepFashion.html
21 © IBM Corporation, 2016
DeepFashion: In-shop Clothes Retrieval
Details
In-shop Clothes Retrieval
Benchmark evaluates the performance of in-
shop Clothes Retrievel. This is a large subset of
DeepFashion, containing large pose and scale
variations. It also has large diversities, large
quantities, and rich annotations, including
• 7,982 number of clothing items;
• 52,712 number of in-shop clothes images,
and ~200,000 cross-pose/scale pairs;
• Each image is annotated by bounding
box, clothing type and pose type.
22 © IBM Corporation, 2016
ML Use Cases By Industry
23 © IBM Corporation, 2016
Manufacturing
Retail
Healthcare & Life SciencesFinancial Services
HospitalityUtilities
• Predictive
Maintenance
• Process Optimisation
• Demand Forecasting
• Risk Analysis
• Cross/Up Selling
• Credit Checks
• Customer
Segmentation
• Patient Triage
• Proactive Health
Management
• Real Time Alerts and
Diagnostics
• Disease Identification
• Inventory Planning
• Cross-Channel
marketing
• Customer ROI and
Lifetime Value
• Smart Grid
Management
• Carbon Emissions
• Customer Specific
Pricing
• Scheduling
• Pricing
• Social Media
Analytics and
Customer Sentiment
24 © IBM Corporation, 2016
TensorFlow
Theano
DIGITSCaffe
More..Torch
• Becoming De Facto
standard
• Easy to learn
• Designed to support
GPU execution
• Awesome for Image
Recognition (original
use case)
• Poor support for
language modelling –
due to legacy
limitations
• Web GUI
• Easy to use
• Abstracts features of
other frameworks
• Used @ Twitter and
• Better debugging –
Automatic
Differentiation
(reverse-mode)
• No Compile Time
https://github.com/zer0n/deepframeworks• Supported and
Developed by
“Worlds Largest”
Academic DL Lab
• Faster than TF and
with a wider range of
functions
25 © IBM Corporation, 2016
Introducing PowerAI:
Get Started Fast with Deep Learning
Enabled by High Performance Computing Infrastructure
Package of Pre-Compiled Major Deep Learning
Frameworks
Easy to install & get started with Deep Learning with Enterprise-Class Support
Optimized for Performance To Take Advantage of
NVLink
26 © IBM Corporation, 2016
| 26
Introducing IBM Power System S822LC for HPCFirst Custom-Built GPU Accelerator Server with NVLink
2.5x Faster CPU-GPU Data Communication via NVLink
NVLink80 GB/s
GPU
P8
GPU GPU
P8
GPU
PCIe32 GB/s
GPU
x86
GPU GPU
x86
GPU
No NVLink between CPU & GPU for x86 Servers: PCIe Bottleneck
NVIDIA P100 Pascal GPU
POWER8 NVLink Server x86 Servers with PCIe
• Custom-built GPU Accelerator Server• High-Speed NVLink Connections between
CPUs & GPUs and among GPUs• Features novel NVIDIA P100 Pascal GPU
accelerator
27 © IBM Corporation, 2016
90%Reduction in
inspection times
Significant
Decreasein inspection times
Significant
Increasein checkable quantities/ day
Significantly
Decreasedrate of Safety Risks
• The utility provider inspects its vast
transmission network via hand, with
skilled workers placed into high-risk
environments. This method is costly,
occasionally dangerous, and difficult to
scale.
• To address this and augment worker
productivity, the provider is seeking to
deploy drones to make visual inspections
of transmission towers.
• To automate the image processing, the
provider is using PowerAI to train a
deep learning network to ID potential
maintenance issues captured by the
drones.
• IBM is the only vendor who can provide
the unique supremacy of NVIDIA Tesla
P100 GPUs connected to POWER8
CPUs with NVIDIA NVLink technology
for deep learning.
• IBM’s integrated portfolio of solutions also
allows the provider to not only apply deep
learning but also in-memory DBMS and
high speed storage to store and analyze
various data using Power Systems and
IBM ESS and Spectrum Scale.
Asian Electric
Utility Provider
Maintenance
Inspection
28 © IBM Corporation, 2016
90%Reduction in
inspection times
ImprovedAccuracy of Risk analysis in
credit application process
IncreasedCapital available for
Investment and other
revenue-generating
opportunities
Increased Responsiveness to clients.
Accelerated
Time to respond to clients.
Applying inference in real
time shortens the time to
answer for all clients:
improving the customer
experience.
• A major bank in Oceania is seeking to
apply deep learning to the credit risk
analysis for credit card applications. Their
main goal from this undertaking is to
explore options to add self-learning
capabilities to the current credit risk
marking process.
• By using deep learning to improve the
accuracy of risk analysis, the bank can
determine how much capital needs to be
held to cover that risk.
• Even an improvement of just 1%
accuracy in marking credit risk would
reduce their capital holding requirements,
allowing the freed up capital to be
invested, generating more income for the
bank and it’s account holders.
• The solution uses the IBM S822LCs for
HPC systems, each with four NVIDIA Tesla
P100 GPUs and 1TB of memory, and the
IBM Data Engine for NoSQL CAPI-
attached flash system.
• In addition, IBM is providing Apache Spark
via IBM Spectrum Conductor for SQL and
Spectrum Scale.
Large BankCredit Risk
Analysis
© IBM Corporation, 2016
Questions?David Spurway – IBM Power Systems Product Manager
Email: [email protected]
Phone: 07717 892 896
Twitter, LinkedIn, YouTube
30 © IBM Corporation, 2016
PowerAI Installation and OptimizationOverview
Designed to optimize deployments of PowerAI on Power System servers. This
offering is appropriate both for those clients who are interested in exploring the
capabilities and benefits of implementing PowerAI, as well as those clients
who may currently have Deep Learning and HPC solutions implemented.
Target Audience
• Clients wanting to purse Deep Learning using PowerAI on Power Systems
• Clients who have a 8335-GTB (Minsky) as part of the insertion program
• Clients who have existing applications built on CAFFE, Torch, TensorFlow,
Theano
Benefits
• The customer will have a working PowerAI proof of concept that they can
use for further evaluation of PowerAI and Deep Learning based solutions
• Skills transfer from our experts helps you fully exploit the capabilities
PowerAI
Qualifying Questions
• Does the customer have an interest in Deep Learning/ Machine Learning?
• Does the customer have a research team that needs assistance with setting
up the infrastructure to take advantage of the NVIDIA Deep Learning
capabilities that PowerAI enables
• Has the account team offered a 8335-GTB (Minsky) as part of a larger
Power transaction that could justify Pre-Sales or Post Sales funding for
implementation
Key Features
• Design, Install and Configure PowerAI Ubuntu infrastructure (servers,
network, file systems )
• Install CUDA Libraries & NVIDIA Drivers
• Verify access to GPU and apply best practices
• Install and validate PowerAI packages and sample applications utilizing such
technologies such as Jupyter notebooks, and Spectrum Conductor
• Install Spectrum Computer (LSF) job scheduling
• BlueMind sofware installation and configuration (China only)
• Provide Power infrastructure skills transfer to customer personnel
• Assist in migration of existing customer ML-DL workload onto PowerAI
• Optional Remote On-Demand Assistance as follow-up
Duration
The service varies depending on the size and complexity of the implementation,
but can be customized to specific client requirements.
Resources
Learn more about PowerAI on IBM Power Systems at:
https://www.ibm.com/us-en/marketplace/deep-learning-platform/
Team Contacts
Owner Americas: Fred Robinson [email protected]
Owner China: Yong Bao Tan [email protected]
Owner Europe: David Uttley [email protected]
Owner India: Subramaniam Meenakshisundaram [email protected]
Owner Japan: Shinichi Niimi [email protected]
Find a Lab Services Opportunity Manager in your area ->
http://ibm.biz/LabServicesOM
IBM Systems Lab Services — Power Systems
© IBM Corporation, 2016
Thank you!David Spurway – IBM Power Systems Product Manager
Email: [email protected]
Phone: 07717 892 896
Twitter, LinkedIn, YouTube
32 © IBM Corporation, 2016
Trademarks and notes
IBM Corporation 2015
• IBM, the IBM logo and ibm.com are registered trademarks, and other company, product, or service names may be trademarks or service marks of International Business Machines Corporation in the United States, other countries, or both. A current list of IBM trademarks is available on the web at “Copyright and trademark information” at www.ibm.com/legal/copytrade.shtml
• Other company, product, and service names may be trademarks or service marks of others.
• References in this publication to IBM products or services do not imply that IBM intends to make them available in all countries in which IBM operates.
• IBM and IBM Credit LLC do not, nor intend to, offer or provide accounting, tax or legal advice to clients. Clients should consult with their own financial, tax and legal advisors. Any tax or accounting treatment decisions made by or on behalf of the client are the sole responsibility of the customer.
• IBM Global Financing offerings are provided through IBM Credit LLC in the United States, IBM Canada Ltd. in Canada, and other IBM subsidiaries and divisions worldwide to qualified commercial and government clients. Rates and availability are based on a client’s credit rating, financing terms, offering type, equipment type and options, and may vary by country. Some offerings are not available in certain countries. Other restrictions may apply. Rates and offerings are subject to change, extension or withdrawal without notice.
33 © IBM Corporation, 2016
Special notices
This document was developed for IBM offerings in the United States as of the date of publication. IBM may not make these offerings available in
other countries, and the information is subject to change without notice. Consult your local IBM business contact for information on the IBM offerings
available in your area.
Information in this document concerning non-IBM products was obtained from the suppliers of these products or other public sources. Questions on
the capabilities of non-IBM products should be addressed to the suppliers of those products.
IBM may have patents or pending patent applications covering subject matter in this document. The furnishing of this document does not give you
any license to these patents. Send license inquires, in writing, to IBM Director of Licensing, IBM Corporation, New Castle Drive, Armonk, NY 10504-
1785 USA.
All statements regarding IBM future direction and intent are subject to change or withdrawal without notice, and represent goals and objectives only.
The information contained in this document has not been submitted to any formal IBM test and is provided "AS IS" with no warranties or guarantees
either expressed or implied.
All examples cited or described in this document are presented as illustrations of the manner in which some IBM products can be used and the
results that may be achieved. Actual environmental costs and performance characteristics will vary depending on individual client configurations and
conditions.
IBM Global Financing offerings are provided through IBM Credit Corporation in the United States and other IBM subsidiaries and divisions worldwide
to qualified commercial and government clients. Rates are based on a client's credit rating, financing terms, offering type, equipment type and
options, and may vary by country. Other restrictions may apply. Rates and offerings are subject to change, extension or withdrawal without notice.
IBM is not responsible for printing errors in this document that result in pricing or information inaccuracies.
All prices shown are IBM's United States suggested list prices and are subject to change without notice; reseller prices may vary.
IBM hardware products are manufactured from new parts, or new and serviceable used parts. Regardless, our warranty terms apply.
Any performance data contained in this document was determined in a controlled environment. Actual results may vary significantly and are
dependent on many factors including system hardware configuration and software design and configuration. Some measurements quoted in this
document may have been made on development-level systems. There is no guarantee these measurements will be the same on generally-available
systems. Some measurements quoted in this document may have been estimated through extrapolation. Users of this document should verify the
applicable data for their specific environment.