Blazingly Fast Machine Learning Inference

Slides:



Advertisements
Similar presentations
Copyright © 2014 Oracle and/or its affiliates. All rights reserved. | Your customer as a segment of one That changes every second! Hein Van Der Merwe Chief.
Advertisements

A Java Architecture for the Internet of Things Noel Poore, Architect Pete St. Pierre, Product Manager Java Platform Group, Internet of Things September.
Oracle Cloud Marketplace Neelesh Gurnani Director, Product Development Arif Khan Director, Product Management September 29, 2014 Copyright © 2014, Oracle.
Copyright © 2014, Oracle and/or its affiliates. All rights reserved. | Advanced Metadata Modeling Modeling for the Oracle Business Intelligence Cloud.
The Safe Harbor The following is intended to outline our general product direction. It is intended for information purposes only, and may not be incorporated.
American Airlines AAdvantage®:
QAD Solutions for Configurable Products Frank Feustel, Director, Product Management, QAD Inc. Jim Smith, QAD Improvement Leader, R. W. Beckett Corporation.
Copyright © 2015, Oracle and/or its affiliates. All rights reserved. JD Edwards Summit The Newest JDE Module – Rental Management Joel Sandberg Sales Consultant.
“This presentation is for informational purposes only and may not be incorporated into a contract or agreement.”
Copyright © 2015, Oracle and/or its affiliates. All rights reserved. JD Edwards Summit PaaS from an Applications Perspective Charles McGuinness Director,
QAD's Customer Engagement Dan Blake Consultancy Development Director, QAD QAD Explore 2012.
1Copyright © 2012, Oracle and/or its affiliates. All rights reserved. Insert Information Protection Policy Classification from Slide 8 Reporting from Contract.
A New IMS-Like Architecture for Enterprise Applications Reid Stidolph Master Principle Solutions Architect Communications Global Business Unit October.
Oracle Contact Center Anywhere: Go To Market - Positioning James Owens – BDM, Specialist Sales APAC.
RMB Billing UX Design Concepts / Proposals Peter Picone.
Hadoop 2 cluster with Oracle Solaris Zones, ZFS and unified archives Orgad Kimchi - Principal Software Engineer September 29, 2014 Oracle Confidential.
1Copyright © 2012, Oracle and/or its affiliates. All rights reserved. Insert Information Protection Policy Classification from Slide 8 Contract Management.
| Building the Effective Enterprise The Power of.Net Upgrade Jeffrey Steele – Solution Consultant, QAD.
CON Software-Defined Networking in a Hybrid, Open Data Center Krishna Srinivasan Senior Principal Product Strategy Manager Oracle Virtual Networking.
1. S318417: OAUG SysAdmin SIG Angelo Rosado, Oracle Senior Product Manager Kenneth Baxter, Oracle Strategy Product Manager Biju Mohan, Oracle Principal.
Workforce Scheduling Release 5.0 for Windows Implementation Overview OWS Development Team.
RESTful Microservices In Java With Jersey Jakub Podlešák Software Engineer Oracle, Application Server Group September 29, 2014 Copyright © 2014, Oracle.
Copyright © 2015, Oracle and/or its affiliates. All rights reserved. JD Edwards Summit Recruit and Learn Solutions Extend HR into the Cloud Marie Scott.
Oracle SIS and Sakai Integration Linda Feng, Architect Oracle Academic Enterprise Initiative.
For Oracle employees and authorized partners only. Do not distribute to third parties. © 2008 Oracle Corporation – Proprietary and Confidential.
Travel and Transportation General Session and Industry Excellence Awards Vijay Anand, Oracle Sundar Swaminathan, Oracle September 30, 2014 Copyright ©
-1- For Oracle employees and authorized partners only. Do not distribute to third parties. © 2009 Oracle Corporation – Proprietary and Confidential Oracle.
Copyright © 2013, Oracle and/or its affiliates. All rights reserved. Insert Information Protection Policy Classification from Slide 12 1.
Slide 1. What's New in NetBeans IDE 7.1 Name Title.
Copyright © 2014, Oracle and/or its affiliates. All rights reserved. | Daddy, what's a middle wear? An incredibly oversimplified explanation of what Middleware.
JD Edwards EnterpriseOne In-Memory Sales Advisor
Copyright © 2014 Oracle and/or its affiliates. All rights reserved. |
Wallpaper only – on screen during welcome and chat
<Insert Picture Here>
My Oracle Support (The next generation Metalink experience) lynn
OpenWorld 2018 Accelerate Your Success on Oracle Cloud Infrastructure with Applications Unlimited Support Viviana Tilita Senior Technical Support Engineer.
Saudi Telecom: Better Together
OpenWorld 2018 How to Create Chatbots with OMCe
Build a Web App with Oracle REST Data Services &
Project Helidon Introduction
Assessing the Security of the Cloud
#.
OpenWorld 2018 How to Combine Data from Source Sites
Managing Development Projects Across Oracle Cloud Services: A Guide
Charles Phillips screen
JD Edwards Support and Oracle Cloud Infrastructure: A Successful Path to Oracle Cloud
Confidential – Oracle Internal/Restricted/Highly Restricted
QAD Reporting Framework
Key Considerations in User Lifecycle Management
Project Helidon Deep Dive
Confidential – Oracle Internal/Restricted/Highly Restricted
How to Cure Those Digital Adoption Blues: Oracle Guided Learning
Confidential – Oracle Internal/Restricted/Highly Restricted
OpenWorld How to Prepare Data from Business Intelligence Cloud Service
Confidential – Oracle Internal/Restricted/Highly Restricted
Confidential – Oracle Internal/Restricted/Highly Restricted
Oracle Database Multilingual Engine
OpenWorld 2018 HCM Oracle
OpenWorld 2018 Oracle API Platform: How to Manage Typical Workflows
Your Next Chatbot From Idea to Execution Donna Scheele
Your Finance Cloud End User Adoption and Enablement Starts Here
Leverage the virtual team using Oracle Support Communities Phil Dixon Director Service Automation Product Management Global Customer Support October,
Confidential – Oracle Internal/Restricted/Highly Restricted
Oracle Essbase: Scenario and What-If Modeling Made Simple
Principal Product Manager Oracle Data Science Platform
R11 Product Release Update March 2008
1.
“This presentation is for informational purposes only and may not be incorporated into a contract or agreement.”
Safe Harbor Statement The following is intended to outline our general product direction. It is intended for information purposes only, and may not be.
Oracle Java SE Subscriptions; Protect Your Investment in Java SE
Presentation transcript:

Blazingly Fast Machine Learning Inference Vish Abrams Architect, Cloud Development Machine Learning Team, Oracle Cloud Infrastructure October 22, 2018

Safe Harbor Statement The following is intended to outline our general product direction. It is intended for information purposes only, and may not be incorporated into any contract. It is not a commitment to deliver any material, code, or functionality, and should not be relied upon in making purchasing decisions. The development, release, timing, and pricing of any features or functionality described for Oracle’s products may change and remains at the sole discretion of Oracle Corporation.

Program Agenda 1 Machine Learning Model Serving What is GraphPipe? Advantages Protocol Deep Dive Real World Demo More Info 2 3 4 5 6

Program Agenda 1 Machine Learning Inference What is GraphPipe? Performance Protocol Deep Dive Real World Demo More Info 2 3 4 5 6

Machine Learning Inference (Model Serving) Building machine learning models has become much easier due to open source frameworks like TensorFlow and Pytorch Serving machine learning models means putting your trained model onto a server so that it can be accessed by client applications This involves two components: the ML client and the ML server. The client talks to the server using some kind of communication protocol: often JSON over HTTP.

ML Client

ML Server

Program Agenda 1 Machine Learning Model Serving What is GraphPipe? Advantages Protocol Deep Dive Real World Demo More Info 2 3 4 5 6

What is GraphPipe? GraphPipe is an open source protocol and collection of software designed to simplify machine learning model deployment and decouple it from framework-specific model implementations.

In other words, it turns this: What is GraphPipe? In other words, it turns this: mxnet server tensorflow serving custom server standard json custom protocol protocol buffers custom client autogenerated client

What is GraphPipe? Into this: graphpipe-onnx graphpipe-tf

GraphPipe Features A minimalist machine learning transport specification based on flatbuffers Simple reference model servers for Tensorflow, Caffe2, and ONNX. Efficient client implementations in Go, Python, and Java.

Why Did we Make It? Production deployments of AI agents are around the corner Model Serving is an important part of production solutions Existing solutions suffer from various problems: Inconsistent Inefficient Custom Clients A standard along with simple implementations moves the industry forward

Program Agenda 1 Machine Learning Model Serving What is GraphPipe? Advantages Protocol Deep Dive Real World Demo More Info 2 3 4 5 6

Ease of Development Model Servers are written in Go – a very accessible language Flatbuffer code generation makes it easy to produce new clients Open spec makes it possible to integrate with existing servers

Protocol Performance

Serving Performance

Program Agenda 1 Machine Learning Model Serving What is GraphPipe? Advantages Protocol Deep Dive Real World Demo More Info 2 3 4 5 6

Flatbuffers Extensible protocol Small code footprint Near-zero deserialization overhead

Protocol Summary

Program Agenda 1 Machine Learning Model Serving What is GraphPipe? Advantages Protocol Deep Dive Real World Demo More Info 2 3 4 5 6

AlphaZero Timeline AlphaGo Beats Fan Hui Oct-16 Paper published in Nature Jan-16 AlphaGo Beats Lee Sedol Mar-16 AlphaGo Beats Ke Jie May-17 AlphaGoZero published Oct-17 AlphaZero published Dec-17

AlphaZero Algorithm for training a machine to play any game* Any game that can be represented with a Markov Process Trained without human information through self play Needs a structured representation of the game state Needs rules for transitioning from one state to the next

The Game Playing Black Box Neural Network Position Move

Training the Network Training Labeled Data Neural Network

Generating Data Self-Play (MCTS) Neural Network Labeled Data

AlphaZero In a nutshell Neural Network Self-Play Training Labeled Data

AlphaZero for Connnect Four We trained a network to play Connect Four using 150 cycles of this process (and playing about 1,000,000 games during self-play) The network finds the correct move in about 99% of positions We used GraphPipe as part of the training process because we were generating games across a cluster of 5 machines with GPUs But GraphPipe is even more useful for deploying this model so that people can use it How do we deploy our model for use in an application? GraphPIpe!

Serving the AlphaZero Trained Network Position Web Frontend Neural Network GraphPipe GraphPipe Move

Live Demo!

Live Demo This is an event branded Section Header with Graphic slide ideal for including a picture with a brief title and optional subtitle. This slide can also be used as a Q and A slide. Do not customize this slide with your own background. Subtitle

Actual Architecture

Program Agenda 1 Machine Learning Model Serving What is GraphPipe? Advantages Protocol Deep Dive Real World Demo More Info 2 3 4 5 6

GraphPipe https://oracle.github.io/graphpipe/ https://github.com/oracle/graphpipe https://github.com/oracle/graphpipe-go https://github.com/oracle/graphpipe-py https://github.com/oracle/graphpipe-tf-py https://hub.docker.com/r/sleepsonthefloor/ https://hackernoon.com/machine-learning-model-pipelines-part-i- e138b7a7c1ef

AlphaZero https://azfour.com/ https://medium.com/oracledevs/lessons-from-implementing-alphazero- 7e36e9054191 https://medium.com/@sleepsonthefloor/azfour-a-connect-four-webapp- powered-by-the-alphazero-algorithm-d0c82d6f3ae9 https://medium.com/applied-data-science/alphago-zero-explained-in-one- diagram-365f5abf67e0 https://deepmind.com/documents/119/agz_unformatted_nature.pdf https://arxiv.org/abs/1712.01815

Questions and Answers Subtitle This is an event branded Section Header with Graphic slide ideal for including a picture with a brief title and optional subtitle. This slide can also be used as a Q and A slide. Do not customize this slide with your own background. Subtitle