nvidia and h2o accelerate ml on gpus · pdf fileblue line infotech blueconnect ... digi...

40
Joshua Patterson — NVIDIA Arno Candel — H2O.ai NVIDIA and H2O Accelerate ML on GPUs

Upload: leliem

Post on 25-Mar-2018

214 views

Category:

Documents


1 download

TRANSCRIPT

Joshua Patterson — NVIDIA Arno Candel — H2O.ai

NVIDIA and H2OAccelerate ML on GPUs

2

NVIDIA Leader in AI Computing

GPU Computing

Gaming Pro Visualization Data Center Self-Driving Cars

3

AMAZING ACHIEVEMENTS IN AI

Play Go Play Doom Learn Paint Style Synthesize Voice

Write Captions Learn Motor Skills Learn to Walk Drive

4

LIFE AFTER MOORE’S LAW

1980 1990 2000 2010 2020

102

103

104

105

106

107

40 Years of Microprocessor Trend Data

Original data up to the year 2010 collected and plotted by M. Horowitz, F. Labonte, O. Shacham, K. Olukotun, L. Hammond, and C. Batten New plot and data collected for 2010-2015 by K. Rupp

Single-threaded perf

1.5X per year

1.1X per yearTransistors(thousands)

5

RISE OF GPU COMPUTING

1980 1990 2000 2010 2020

GPU-Computing perf 1.5X per year

1000X by 2025

Original data up to the year 2010 collected and plotted by M. Horowitz, F. Labonte, O. Shacham, K. Olukotun, L. Hammond, and C. Batten New plot and data collected for 2010-2015 by K. Rupp

102

103

104

105

106

107

Single-threaded perf

1.5X per year

1.1X per year

APPLICATIONS

SYSTEMS

ALGORITHMS

CUDA

ARCHITECTURE

6

DGX SYSTEMS CLOUD

Servers in Every Shape and Size

ALL GPU

The Essential AI Tools for Instant Productivity Everywhere

NVIDIA GPU COMPUTING MODEL EVERYWHERE, ANYWHERE

7

END-TO-END SOLUTIONS FOR DATA SCIENCE

DGX-1

Fully integrated deep learning solution

EMBEDDED

Inference at the Edge

DESKTOP DATA CENTER

Accelerators for PCs

Tesla V100

Most advanced data center GPU

Jetson TX1, Drive PX2 DGX Station, Titan Xp

ENTERPRISE

8

Go

NVIDIA DGX STATION

SPECIFICATIONS

At a GlanceGPUs 4x NVIDIA® Tesla® V100

TFLOPS (GPU FP16) 480

GPU Memory 16 GB per GPU

NVIDIA Tensor Cores 2,560 (total)

NVIDIA CUDA Cores 20,480 (total)

CPUIntel Xeon E5-2698 v4 2.2 GHz (20-core)

System Memory 256 GB LRDIMM DDR4

StorageData: 3 x 1.92 TB SSD RAID 0 OS: 1 x 1.92 TB SSD

Network Dual 10 Gb LAN

Display 3x DisplayPort, 4K Resolution

Acoustics < 35 dB

Maximum Power Requirements 1500 W

Operating Temperature Range 10 - 30 oC

SoftwareUbuntu Desktop Linux OS DGX Recommended GPU Driver CUDA Toolkit

Learn more: www.nvidia.com/station

9

DGX STATIONThe Personal AI Supercomputer

400 x86 CPU’s – in a workstation

Desk-friendly Whisper-quiet

Experiment on Station Scale on DGX-1 / Cloud

EFFORTLESS PRODUCTIVITYDESIGNED FOR

THE OFFICEVOLTA-POWERED PERFORMANCE

10

OUR DATA CENTER STRATEGY: NVIDIA DGX-1

Highest Performance, Fully Integrated System

960 TFLOPS

300 GB/s NVLink Hybrid Cube Mesh

8x Tesla V100 16GB

2x Xeon | 8 TB RAID 0

Quad IB 100Gbps, Dual 10GbE

3U — 3200W

8 TB SSD 8 x Tesla V100 16GB

At a Glance

Learn more: www.nvidia.com/station

11

Registry of Containers, Datasets, and Pre-trained models

NVIDIA GPU CLOUD

CSPs

NVIDIA GPU CLOUD

Containerized in NVDocker | Optimization across the full stack Always up-to-date | Fully tested and maintained by NVIDIA | Beta in July

GPU-accelerated Cloud Platform Optimized for Deep Learning

12

How GPU Acceleration WorksApplication Code

+

GPU CPU

Compute-Intensive Functions

Rest of Sequential CPU Code

13

GPU Acceleration In Action

Deep learning researcher & educator. Founder: fast.ai; Faculty: USF & Singularity University; // Previously - CEO: Enlitic; President: Kaggle; CEO Fastmail

Rewrote @scikit_learn PolynomialFeatures in @ContinuumIO Numba. Got a 40x speedup (would be bigger with more data!) 12 lines of code

14

GPU Acceleration In Action

15

What’s machine learning?

18

This is a team effort!

19

20

Who is H2O.ai?

21

10,000 Companies using H2O - World Wide Community Adoption

A.C. Nielsen A1 Telekom AustriaAAPTAbovenet CommunicationsAcademic Administrative and Research NetworkAcademic Computer Centre Cyfronet HAccelerated Data WorksAccentureAccenture ServicesAce Ina HoldingsAce InternationalAce Telecom Acton Acxiom Oration Adamo Telecom Iberia Administracion Nacional De TelecomunicacionesAdmiral Objekt Waesche & ArbeitskleidungAdobe SystemsAdobe Systems India Adsl Maroc TelecomAdvanced Cable CommunicationsAdvanced Computer SolutionsAffectoAfrihost-DynamicAinet Telekommunikations-Netzwerk BetriebsAir Bank A.S. Air Liquide Sa Airess CeskoAkamai TechnologiesAktia Saastopankki OyAktiv-I SzolgaltatoAl-Shahad Information Technology Albert Einstein College of Medicine of Yeshiva University Albert-Ludwigs-Universitaet FreiburgAlexander & Alexander Information Technology Algar Telecom Aliyun ComputingAllbusiness.Com Allianz Maned Operations & Services Se

Bell CanadaBeltelecomBeyond The Network AmericaBezeq International-Bh - Tec Bharti AirtelBibliotheque Nationale De FranceBig Fish GamesBigleaf NetworksBiglobe Bilink Bimeh Dormitory Sharif University of TechnologyBio-Rad LaboratoriesBiocontrolBisiness Network JvBite CommunicationsBiznetBiznet MetronetBlekinge Institute of TechnologyBlue Line InfotechBlueconnectBoingo WirelessBol.Com Bv Boots UK RetailBoranetBorlange Energi Boston Scientific Oration Bouygues Telecom Division Mobile Bouygues Telecom Sa Brain TelecommunicationBright House NetworksBrighthouse Networks Cfl Division Brighthouse Networks IndianapolisBristish PetroleumBritish Sky BroadcastingBroadriver CommunicationBroadstripeBrutele Sc Bryant University

Case Western Reserve University Catalina Marketing Catalina Marketing Oration Cect-Chinacomm Communications Cedars-Sinai Health SystemsCelgene Oration Center For Governmental Research Centerbeam Central Telegraph Public Joint-Stock Centre De Calcul El-Khawarizmi - Cck Centre For Advanced Computing Centro De Tecnologia Da Informa O Renato ArcherCeom Israel Cerfnet Cerner Oration Certara USACeu Cgi GroupChampaign Telephone Charles University Charlesbrauer Charter Communications CheggChengdu West Dimension Digital TechnologyCheonanjeonhwakukjang Chico Board of Trade China Digital Kingdom TechnologyChina Education and Research Network Chinatelecom Group Beijing CoChongqing Times Newper Office Chs - Bna Lan Chunghwa Telecom Data Communication Business GroupCik Telecom CiscoCisco SystemsCisco Systems Ironport Division Citadel Investment Group L.L.C. Citrix SystemsCity University

Delft University of Technology Network Delhi Technical University(Dce) Deloitte Deloitte ServicesDeloitte Touche Tohmatsu ServicesDeloitte and Touch Regional Consulting ServicesDelphon Industries Delta Dental Plan of Michigan Delta Leasedline Network Deluxe Oration Den Networks Dena Deutsche Telekom Deutsches Reisebuero Develon Dhirubhai Ambani Institute of Information Dialog Axiata Digi Tavkozlesi Es SzolgaltatoDigia Digital Entertainment Digital Hosting TechnologyDigital Network Associates - Franchisee Digital Ocean Digital RealmDigital RiverDigital-Entertainment-Industry-Development-Co--Zhongshan ZhoDigitalocean Cloud Direct Supply Discoveries In Sight Dishnet Wireless Distributel Communications Disy Informationssysteme Diverge ConsultingDna Oy Doclernet Dongbeicaijingdaxue-Dl-Ln Doorway As Dotomi Drivetime

Enbridge PipelinesEncy For Science Technology and Research End-User NumericableEnergy Sciences Network Enom OrporatedEnsync Business Solutions PtyEntanet InternationalEnterprise TeamingEnzu Eotvos Lorand University of SciencesEpam SystemsEpm Telecomunicaciones E.S.P. Epsilon Data Manement DbaEquantEquinox ConsultingErasmus McErasmus University RotterdamEricsson Business CommunicationsEricsson Network SystemsEscout ConsultingEspn Estate Valuations and Pricing SystemsEtapa EpEtex CommunicationsEtheric NetworksEthio TelecomEthz Swiss Federal Institute of Technology Zurich Etisalat Lanka (Private)European Bioinformatics InstituteEvergy Excell Media Exe2 Newton Abbot Exetel Act DslExponential-E FPL FibernetFacebookFachhochschule DortmundFachhochschule NordwestschweizFaculty of Sciences University of Lisbon

Companies Using H2O.ai

2015

2016 Now

2017

Goal

14,000

10,281

6,427

3,810

H2O.ai Users

2015

2016 Now

2017

Goal

140,000

97,620

54,163

38,257

10,000+ Companies use H2O — World Wide Community Adoption

22

H2O.ai Select Paying Customers

Financial InsuranceMarketing TelecomHealthcareRetail

“Overall customer satisfaction is very high.” - Gartner

Advisory & Accounting

23

AI in Financial Services

Wholesale / Commercial Banking• Know Your Customers (KYC) • Anti-Money Laundering

(AML)

Card/Payments Business• Transaction Frauds • Real-time Targeting • Credit Risk Scoring • In-Context Promotion

Retail Banking• Deposit Fraud • Customer Churn Prediction • Auto-Loan

IT Infrastructure• Security Cyberlake • DoS Detection and Protection • Master Data Management

24

AI in Healthcare

Flu Season Prediction

Personalized Drug Matching

Medical Claim Fraud Detection

Emergency Room and Hospital Management

Drug Discovery

Remote Patient Monitoring

Early Cancer Detection / Oncology

Medical Imaging and Diagnostics

Product Recommendation

25

H2O.ai is a Visionary in the Gartner Magic Quadrant

for Data Science Platforms

“Overall customer satisfaction is very high.”

“H2O is especially suited to IoT edge and device scenarios.”

“H2O had the highest reference customer analytics support score of all the vendors.”

“H2O.ai has significant adoption by large enterprises such as Macy’s, Comcast, and Capital One.”

“H2O.ai is best known for developing open source, cluster-distributed ML algorithms at a time (2011) when big data demanded them, but no one else had them.”

H2O.ai is a Strong Performer in the Forrester Predictive Analytics & Machine Learning

H2O.ai Deep Water Included in Gartner Deep Learning Report

Publish: January 2017

H2O.ai named alongside Caffe, Facebook Torch, Google TensorFlow, and Intel Nervana, as a platform that assists users in creating their own deep-learning and AI solutions.

H2O.ai Strongly Positioned in Key Analyst Reports

26

The Road Ahead

H2O AI Platform Timeline

Users

Advanced Data Scientists

Developers/Engineers

Dev Ops

App Developers

H2O Core

Data.table

Sparkling Water

Roadmap

Steam

Visual Interpretation

Deep Learning

Analysts

H2O AI EditionQ3 2017

2012 2014 2016 2017 2018

H2O GPU Edition

GPU2019

ASIC

Auto ML

27

28

Accuracy, Speed and Interpretability

30

https://www.youtube.com/watch?v=4RKSXNfreLE

31

https://www.youtube.com/watch?v=NkeSDrifJdg

171 with latest solver

87

51

32

33

This performance based on NVIDIA’s technology will lead to…

34

Driverless AI for the Digital Brain — Enabled by Fast Model Training

Data

Visual Model Interpretation

Driverless AIPipeline

Distributed Multi-CPU Multi-GPU

Model Repository

Feature Engine Deploy

Model Fitness

Auto ML Deep Learning Algorithms Data Prep

H2O Kaggle Grandmasters

H2O Systems Engineers

H2O Customers Business Leaders

H2O PhDs & Professors

Accuracy Speed Interpretability

36

Driverless AI — Competitive with Kagglers!

Top 8 position in Kaggle with zero manual labor!(ranked above multiple Kaggle Grandmasters)

https://www.kaggle.com/c/mercedes-benz-greener-manufacturing/leaderboard

37

Model Interpretability — Insights Through Computing

38

39

GPU OPEN ANALYTICS INITIATIVEgithub.com/gpuopenanalytics

GPU Data Frame (GDF)

Ingest/ Parse

Exploratory Analysis

Feature Engineering

ML/DL Algorithms

Grid Search

Scoring

ModelExport

Thank You