Alison B. Lowndes, profile picture
Uploaded byAlison B. Lowndes
PDF, PPTX632 views

NVIDIA @ Infinite Conference, London

This document provides an overview of Alison B Lowndes' work in artificial intelligence and how AI is being applied across various industries. It discusses using sensors and AI for applications in automotive, communications, consumer goods, financial services, education, manufacturing, media, online services, healthcare, oil and gas, retail, transportation, and utilities. It also briefly outlines Alison's role at a frontier development lab focusing on using AI to help "spaceship earth."

Technology◦
Related topics:
Neural Networks•

Embed presentation

Download as PDF, PPTX
1 / 101
2 / 101
3 / 101
4 / 101
5 / 101
6 / 101
7 / 101
8 / 101
9 / 101
10 / 101
11 / 101
12 / 101
13 / 101
14 / 101
15 / 101
16 / 101
17 / 101
18 / 101
19 / 101
20 / 101
21 / 101
22 / 101
23 / 101
24 / 101
25 / 101
26 / 101
27 / 101
28 / 101
29 / 101
30 / 101
31 / 101
32 / 101
33 / 101
34 / 101
35 / 101
36 / 101
37 / 101
38 / 101
39 / 101
40 / 101
41 / 101
42 / 101
43 / 101
44 / 101
45 / 101
46 / 101
47 / 101
48 / 101
49 / 101
50 / 101
51 / 101
52 / 101
53 / 101
54 / 101
55 / 101
56 / 101
57 / 101
58 / 101
59 / 101
60 / 101
61 / 101
62 / 101
63 / 101
64 / 101
65 / 101
66 / 101
67 / 101
68 / 101
69 / 101
70 / 101
71 / 101
72 / 101
73 / 101
74 / 101
75 / 101
76 / 101
77 / 101
78 / 101
79 / 101
80 / 101
81 / 101
82 / 101
83 / 101
84 / 101
85 / 101
86 / 101
87 / 101
88 / 101
89 / 101
90 / 101
91 / 101
92 / 101
93 / 101
94 / 101
95 / 101
96 / 101
97 / 101
98 / 101
99 / 101
100 / 101
101 / 101
ALISON B LOWNDES AI DevRel | EMEA Fuelling the AI Revolution with Gaming @alisonblowndes
2 The day job AUTOMOTIVE Auto sensors reporting location, problems COMMUNICATIONS Location-based advertising CONSUMER PACKAGED GOODS Sentiment analysis of what’s hot, problems $ FINANCIAL SERVICES Risk & portfolio analysis New products EDUCATION & RESEARCH Experiment sensor analysis HIGH TECHNOLOGY / INDUSTRIAL MFG. Mfg. quality Warranty analysis LIFE SCIENCES MEDIA/ENTERTAINMENT Viewers / advertising effectiveness ON-LINE SERVICES / SOCIAL MEDIA People & career matching HEALTH CARE Patient sensors, monitoring, EHRs OIL & GAS Drilling exploration sensor analysis RETAIL Consumer sentiment TRAVEL & TRANSPORTATION Sensor analysis for optimal traffic flows UTILITIES Smart Meter analysis for network capacity, LAW ENFORCEMENT & DEFENSE Threat analysis - social media monitoring, photo analysis
www.FrontierDevelopmentLab.org
A MISSION CONTROL FOR SPACESHIP EARTH KICK OFF WORKSHOP ESA ESRIN, ROME 25th - 29th June ‘18 RESEARCH SPRINT 30th June - 17th August
5
6 EMBED TORNADO VIDEO
7
8 NVIDIA RESEARCH
9 TITAN V THE MOST POWERFUL PC GPU EVER 5,120 CUDA cores 640 NEW Tensor cores 12GB HBM2 Memory 1.7Gbps 110 Tensor TFLOPS 13.8 TFLOPS fp16 | 6.9 fp32
10 THE RISE OF GPU COMPUTING Big Data Needs Algorithms and Compute That Scales 1980 1990 2000 2010 2020 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 103 105 107 1.5X per year END OF MOORE’S LAW Single-threaded perf GPU-Computing perf 1.5X per year 1.1X per year CPU vs. GPU
11 “the machine equivalent of experience”
Definitions
14 2012
15 CONVOLUTION 0 0 0 0 0 0 0 0 1 1 1 0 0 0 0 1 2 2 1 1 1 0 1 2 2 2 1 1 0 1 2 2 2 1 1 0 0 1 1 1 1 1 0 0 0 0 0 0 0 4 0 0 0 0 0 0 0 -4 1 0 -8 Source Pixel Convolution kernel New pixel value (destination pixel) Centre element of the kernel is placed over the source pixel. The source pixel is then replaced with a weighted sum of itself and nearby pixels. http://www.biomachina.org/
FAST FOURIER TRANSFORMATION (FFT) = + +
DEEP LEARNING NEW PROGRAMMING MODEL DEVICE EDGE CLOUD
19 Describe the differences between these 2 bikes……..
20 Image “Volvo XC90” Image source: “Unsupervised Learning of Hierarchical Representations with Convolutional Deep Belief Networks” ICML 2009 & Comm. ACM 2011. Honglak Lee, Roger Grosse, Rajesh Ranganath, and Andrew Ng. CONVOLUTIONAL NEURAL NETWORKS
21 http://distill.pub/2017/momentum/
22 Li et al, University of Maryland & US Naval Academy https://arxiv.org/pdf/1712.09913.pdf
23 Residual Networks
24 ResNets vs Highway Nets (IDSIA) https://arxiv.org/pdf/1612.07771.pdf Klaus Greff, Rupesh K. Srivastava Really great explanation of “representation” Compares the two.. shows for language modelling, translation HN >> RN. Not quite as simple as each layer building a new level of representation from the previous - since removing any layer doesn’t critically disrupt.
25 Capsules & routing with EM, Hinton et al https://arxiv.org/pdf/1710.09829.pdf [NIPS2017] SegCaps: capsules for object segmentation https://arxiv.org/pdf/1804.04241.pdf University of Central Florida
26 PRUNING
27 TRAINING VS INFERENCE
28 Long short-term memory (LSTM) Hochreiter (1991) analysed vanishing gradient “LSTM falls out of this almost naturally” Gates control importance of the corresponding activations Training via backprop unfolded in time LSTM: input gate output gate Long time dependencies are preserved until input gate is closed (-) and forget gate is open (O) forget gate Fig from Vinyals et al, Google April 2015 NIC Generator Fig from Graves, Schmidhuber et al, Supervised Sequence Labelling with RNNs
AlexNet
CAMBRIAN EXPLOSION Convolutional Networks ReLuEncoder/Decoder Dropout PoolingConcat BatchNorm Recurrent Networks GRULSTM CTC Beam Search WaveNet Attention Generative Adversarial Networks 3D-GAN Speech Enhancement GANCoupled GAN Conditional GANMedGAN Reinforcement Learning DQN Simulation DDPG New Species Neural Collaborative FilteringMixture of Experts Block Sparse LSTM Capsule Nets
31 Shakir Mohamed and Danilo Rezende, DeepMind, UAI 2017
32 IMAGINE NO LIMITATIONS TO DRUG DISCOVERY It takes 12 years and $2.6B to bring a new drug to market. Insilico Medicine takes an innovative approach to speed drug discovery and uses GPU-accelerated Generative Adversarial Networks (GANs) to "invent" new molecular structures on demand. With GANs, in only a few months’ time, Insilico did what would have taken years with conventional research methods – identified 69 new molecules. To date, Insilico has generated 5,000+ new molecules, integrated GANs into a comprehensive drug discovery framework, and established working collaborations with large pharmaceutical companies. This unique approach could dramatically reduce the time and cost of developing novel substances with medicinal properties.
33 Dreamcatcher https://autodeskresearch.com/projects/dreamcatcher
34
35 “After finishing this book, you will have a deep understanding of how to set technical direction for a machine learning project.” www.mlyearning.org
36
37
38 Brain Computer Interfaces Focused on treatment for disease and dysfunction eg epilepsy, depression, Parkinsons but ultimately to advance human intelligence by restoring and extending cognitive vibrancy. “We’re either going to have to merge with AI or be left behind”; Elon Musk
REINFORCEMENT LEARNING “from our own mistakes”
40 RL TIMELINE •1984: Rich Suttons thesis •1989: Q-learning •1991: REINFORCE •2006: MCTS (Szepesvari) •2009: David Silver’s thesis •2013: DQN •2015: DQN in Nature •2016: A3C & DRL •2017: Soft Q-learning/ AlphaGo Zero •2018: GA3C
41 Tabish Rashid et al, Oxford/DeepMind QMIX: https://arxiv.org/pdf/1803.11485.pdf
42 DEEPMIND ALPHA* 1.0 => 2.0 => 0.0 * ..a deeply structured hybrid (Gary Marcus, Jan 2018)
43 ROBOTS THINK / REASON SENSE / PERCEIVE ACT
44 ATLAS, BOSTON DYNAMICS Inserting video: Insert/Video/Video from File. Insert video by browsing your directory and selecting OK. File types that works best in PowerPoint are mp4 or wmv
45
46 Holodeck: photorealistic, collaborative VR environment Real-world experience through sight, sound, and haptics Explore high-fidelity, full-resolution models THE DESIGN LAB OF THE FUTURE IMMERSIVE AI - HOLODECK
47
48
49 NVIDIA ISAAC ROBOTICS PLATFORM SIMULATION TRAINING DEPLOYMENT SDK https://developer.nvidia.com/isaac-sdk
50
51 INSERT TELEPORTATION VIDEO OF TIM DRIVING A VR VERSION OF A REAL SDC
A massive Deep Learning challenge
53
54 BRIDGING THE REALITY GAP https://arxiv.org/pdf/1804.06516.pdf NVR+University of Toronto “leveraging the power of synthetic data improves upon results obtained using real data alone”
55 END TO END SYSTEM FOR AV Sources: NVIDIA, RAND Corporation SIMULATETRAIN MODELSCOLLECT DATA RE-SIMULATE Lanes Lights Path Signs PedestriansCars Up to 2 petabytes per car per year 20 DNNs 1+ million images per DNN 10B+ miles to ensure safe driving Validate and verify for self-driving cars by 2020 MAPPING 5 + DNNs Create and update HD maps
56 LEVEL 5 DEMANDS EXTREME COMPUTING + 10X camera resolution + Surround LIDAR point-cloud processing + Camera & LIDAR localization to HD map + Tracking all surrounding objects + New map generation + Sophisticated path planning & control + Algorithm diversity + Sensor & computing fail-operate + ASIL-D Functional Safety + Excess computing capacity Level 2 Level 5
57 COMPUTATIONAL SCALE REQUIRED 3 million labeled images 1 DGX-1 trains 300k labeled images on 1 DNN in 1 day 10 DNNs required for self-driving 10 parallel experiments at all times 100 DGX-1 per car
58 Xavier (ASIL-D rated) “Our in-depth technical assessment confirms the Xavier SoC architecture is suitable for use in autonomous driving applications and highlights NVIDIA’s commitment to enable safe autonomous driving”; TÜV SÜD, Germany
59 XAVIER DLA NOW OPEN SOURCE Command Interface Tensor Execution Micro-controller Memory Interface Input DMA (Activations and Weights) Unified 512KB Input Buffer Activations and Weights Sparse Weight Decompression Native Winograd Input Transform MAC Array 2048 Int8 or 1024 Int16 or 1024 FP16 Output Accumulators Output Postprocess or (Activation Function, Pooling etc.) Output DMA WWW.NVDLA.ORG
60 Jetson Xavier developer.nvidia.com/ jetson-xavier
61 AI SAFETY: • Understanding • Trust
62 Dynamic Memory Networks MetaMind now Salesforce (2yrs ago) https://arxiv.org/pdf/1603.01417v1.pdf
63 “The Building Blocks of Interpretability” https://distill.pub/2018/building-blocks/ Chris Olah, Google et al, March 2018
64
66 A PLETHORA OF HEALTHCARE STORIES Molecular Energetics For Drug Discovery AI for Drug Discovery Medical Decision Making Treatment Outcomes Reducing Cancer Diagnosis Errors by 85% Predicting Toxicology Predicting Growth Problems Image Processing Gene Mutations Detect Colon Polyps Predicting Disease from Medical Records Enabling Detection of Fatty Acid Liver Disease
67
68 Radiation Therapy with AI ViewRay MRIdian (imaging & radiation dose)
V-NET CINEMATIC RENDERINGAUTOMAP GPU DEEP LEARNING REVOLUTIONIZING MEDICAL IMAGING CUDA Source: https://arxiv.org/pdf/1704.08841.pdf Source: https://arxiv.org/pdf/1606.04797.pdf Source: Elliot Fishman, M.D. & Siemens
70 https://arxiv.org/abs/1804.07723 INPAINTING
71 COMPUTATIONAL & DATA SCIENCE AI SUPERCOMPUTING WILL TRANSFORM HPC Extending Reach of HPC By Combining Computational & Data Science DATA SCIENCECOMPUTATIONAL SCIENCE Turbulent Flow Molecular Dynamics Structural Analysis N-body Simulation “Next move?” “Is there cancer?”“What’s happening?” “What does she mean?” Understanding Universe Clean EnergyDrug Discovery Monitoring Climate Change
72 GPU READY APPS AWARENESS : ADOPTION : UTILIZATION ✓ Easy setup ✓ Best app practices ✓ Optimal results ✓ Higher productivity ✓ Faster discoveries App Users Life Science Developers Deep Learning IT Managers HPC/Data Centers Simple, ISV-approved, step-by-step instructions 1. 2. 3. 4. 5. Target Customers Quick Start Guide User Benefits www.nvidia.com/gpu-ready-apps
73
74
75 POWERING THE DEEP LEARNING ECOSYSTEM NVIDIA SDK accelerates every major framework COMPUTER VISION OBJECT DETECTION IMAGE CLASSIFICATION SPEECH & AUDIO VOICE RECOGNITION LANGUAGE TRANSLATION NATURAL LANGUAGE PROCESSING RECOMMENDATION ENGINES SENTIMENT ANALYSIS DEEP LEARNING FRAMEWORKS NVIDIA DEEP LEARNING SDK developer.nvidia.com/deep-learning-software
76 developer.nvidia.com
May 8-11, 2017 | Silicon Valley CUDA 9 https://devblogs.nvidia.com/parallelforall/cuda-9-features-revealed/
78NVIDIA CONFIDENTIAL. DO NOT DISTRIBUTE. NVIDIA DEEP LEARNING SDK UPDATE GPU-accelerated DL Primitives Faster training Optimizations for RNNs Leading frameworks support cuDNN 7 Multi-node distributed training (multiple machines) Leading frameworks support Multi-GPU & Multi-node NCCL 2 TensorFlow model reader Object detection INT8 RNNs support High-performance Inference Engine TensorRT 4
79 Circa 2000 - Torch7 - 4th (using odd numbers only 1,3,5,7) Web-scale learning in speech, image and video applications Maintained by top researchers including Soumith Chintala - Research Engineer @ Facebook All the goodness of Torch7 with an intuitive Python frontend that focuses on rapid prototyping, readable code & support for a wide variety of deep learning models. https://pytorch.org/2018/05/02/road-to-1.0.html
80 Benchmarks Jan 2018https://github.com/u39kun/deep-learning-benchmark
81 Apex - A PyTorch Extension ● Goal: Raise PyTorch customer awareness and increase adoption of NVIDIA Tensor Cores ● Content: Provide an easy to use set of utility functions in PyTorch for mixed-precision optimizations ● Benefit: Few lines of code to achieve improved training speed while maintaining accuracy and stability of single precision (Tensor Cores) ● Target audience: Deep learning researchers and developers of PyTorch with NVIDIA Volta ● Key Features: AMP (Auditor for mixed-precision) and Optimizer Wrapper (Dynamic loss scaling and master parameters) ● Teams Involved: Leading NVIDIA PyTorch team and collaboration with external FB PyTorch team Overview
82 DALI Full input pipeline acceleration including data loading and augmentation Drop-in integration with direct plugins to frameworks – MxNet, TensorFlow, PyTorch Portable workflows through multiple input formats and configurable graphs Unblock CPU with GPU-accelerated DL pre-processing library Version 0.1 supports: • Resnet-50 image classification & segmentation training • Input formats – JPEG, LMDB, RecordIO, TFRecord • Python APIs to define, build and run an input pipelineAvailable as open source - https://github.com/NVIDIA/DALI • DALI Samples & Tutorial: https://github.com/NVIDIA/DALI/blob/master/examples/Getting%20started.ipynb • nvJPEG (Webpage, Documentation) - https://developer.nvidia.com/nvjpeg
NEW NVIDIA TENSORRT 4 DRIVE PX 2 JETSON TX2 NVIDIA DLA TESLA P4 TensorRT TESLA V100 Programmable Inference Accelerator Compile and Optimize Neural Networks | Support for Every Framework Optimize for Each Target Platform
84 TensorRT INTEGRATED WITH TensorFlow 8x faster Inference Than TensorFlow Only * 14 86 705 0 100 200 300 400 500 600 700 800 CPU Only FP32 P4 FP32 P4 INT8 TensorRT Images/sec Throughput at < 7ms latency (TensorFlow ResNet-50) * Available in TensorFlow 1.7 https://github.com/tensorflow/tensorflow * Min CPU latency measured was 70 ms. It is not < 7 ms. CPU: Skylake Gold 6140, 2.5GHz, Ubuntu 16.04; 18 CPU threads. Pascal P4; CUDA (384.111; v9.0.176); Batch size: CPU=1, TF_GPU=1 (latency 12 ms) , TF-TRT=4 w/ latency=6ms
85 NVIDIA TensorRT 4 Recommender, Speech & Machine Translation RNN and MLP Layers ⚫ ONNX Import ⚫ NVIDIA DRIVE Support Free download to members of NVIDIA Developer Program soon at developer.nvidia.com/tensorrt 50x Faster Inference for ONNX Models Easily import and accelerate inference for ONNX frameworks with native ONNX parser in TensorRT Accelerate inference of recommenders, speech and machine translation apps with new layers and optimizations Deploy optimized deep learning inference models NVIDIA DRIVE Xavier Support for NVIDIA DRIVE Xavier 1 45x 0X 10X 20X 30X 40X 50X CPU TensorRT
86 TESLA V100 32GB THE MOST ADVANCED DATA CENTER GPU EVER BUILT 5,120 CUDA cores 640 NEW Tensor cores 7.5 FP64 TFLOPS | 15 FP32 TFLOPS 120 Tensor TFLOP 20MB SM RF | 16MB Cache | 32GB HBM2 @ 900 GB/s 300 GB/s NVLink https://devblogs.nvidia.com/parallelforall/inside-volta/
87 TENSORCORES D = A * B + C (4x4)
88 Tensor Cores: More resources Further information Mixed-precision blog: https://devblogs.nvidia.com/mixed-precision-training-deep-neural-networks/ Mixed-precision best practices: https://docs.nvidia.com/deeplearning/sdk/mixed-precision-training/index.html Mixed-precision arVix paper: https://arxiv.org/abs/1710.03740 GTC 2018 Sessions: Training with Mixed Precision: Theory and Practice and Training with Mixed Precision: Real Examples Call to action: 1) Share these resources with your customers and partners 2) Provide feedback to help us prioritize new examples and improve examples DGX/NGC deep learning framework containers: Latest versions of software stack and Tensor Core optimized examples - DGX/NGC Registry Tools TensorFlow OpenSeq2seq: https://nvidia.github.io/OpenSeq2Seq/html/mixed-precision.html & arVix paper PyTorch Apex: https://nvidia.github.io/apex/fp16_utils.html & NVIDIA developer news article Examples New mixed-precision model examples: https://developer.nvidia.com/deep-learning-examples GitHub: https://github.com/NVIDIA/DeepLearningExamples
89 NVSWITCH WORLD’S HIGHEST BANDWIDTH ON-NODE SWITCH 7.2 Terabits/sec or 900 GB/sec 18 NVLINK ports | 50GB/s per port bi-directional Fully-connected crossbar 2 billion transistors | 47.5mm x 47.5mm package
ANNOUNCING NVIDIA DGX-2 THE LARGEST GPU EVER CREATED 2 PFLOPS | 512GB HBM2 | 10 kW | 350 lbs
91 GROUNDBREAKING AI AT YOUR DESK Designed for your office DGX Station is the world’s first personal supercomputer for leading- edge AI development. Built on the same Deep Learning Stack powering all NVIDIA DGX™ Systems, you can now experiment at your desk and extend your work across DGX Systems and the cloud. Now with 32GB per GPU, same cost
NVIDIA GPU CLOUD Optimized Stacks for Every Cloud 20,000+ Registered Organizations | 30 Containers NOW on AWS, GCP, AliCloud, Oracle Cloud, DGX
93 Containerized Applications TF Tuned SW NVIDIA Docker CNTK Tuned SW NVIDIA Docker Caffe2 Tuned SW NVIDIA Docker PyTorch Tuned SW NVIDIA Docker CUDA RTCUDA RTCUDA RTCUDA RT Linux Kernel + CUDA Driver Tuned SW NVIDIA Docker CUDA RT Other Frameworks and Apps. . . ALWAYS UP-TO-DATE Monthly Updates from NVIDIA to Frameworks and Containers
94 NVIDIA-DOCKER 1.0 github.com/NVIDIA/nvidia-docker Open-source Project on Github since 2016 >1.5MM downloads >5K stars >4.5 MM pulls of CUDA images Enabled Various Use-cases DGX / NGC containers Major deep learning frameworks Customers Adopted
95 ANNOUNCING: JETSON XAVIER Computer for Autonomous Machines AI Server Performance in 30W  15W  10W 512 Volta CUDA Cores  2x NVDLA 8 core CPU 30 DL TOPS
96
97 Sample Code Deep Learning CUDA, Linux4Tegra, ROS Multimedia API MediaComputer Vision Graphics Nsight Developer Tools Jetson Embedded Supercomputer: Advanced GPU, 64-bit CPU, Video CODEC, VIC, ISP JETPACK SDK FOR AI @ THE EDGE DEVELOPER.NVIDIA.COM/EMBEDDED-COMPUTING TensorRT cuDNN VisionWorks OpenCV Vulkan OpenGL libargus Video API
98 Available to Instructors Now! developer.nvidia.com/teaching-kits Robotics Teaching Kit with ‘Jet’ - ServoCity D E E P L E A R N I N G
Fundamentals Accelerated Computing Game Development & Digital Content Finance NVIDIA DEEP LEARNING INSTITUTE Online self-paced labs and instructor-led workshops on deep learning and accelerated computing Take self-paced labs at www.nvidia.co.uk/dlilabs View upcoming workshops and request a workshop onsite at www.nvidia.co.uk/dli Educators can join the University Ambassador Program to teach DLI courses on campus and access resources. Learn more at www.nvidia.com/dli Intelligent Video Analytics Healthcare Robotics Autonomous Vehicles Virtual Reality
100 NVIDIA INCEPTION PROGRAM Accelerates AI startups with a boost of GPU tools, tech and deep learning expertise Startup Qualifications Driving advances in the field of AI Business plan Incorporated Web presence Technology DL startup kit* Pascal Titan X Deep Learning Institute (DLI) credit Connect with a DL tech expert DGX-1 ISV discount* Software release notification Live webinar and office hours *By application Marketing Inclusion in NVIDIA marketing efforts GPU Technology Conference (GTC) discount Emerging Company Summit (ECS) participation+ Marketing kit One-page story template eBook template Inception web badge and banners Social promotion request form Event opportunities list Promotion at industry events GPU ventures+ +By invitation www.nvidia.com/inception
101

More Related Content

PDF
The What, Who & Why of NVIDIA
byAlison B. Lowndes
 
PDF
Talk on using AI to address some of humanities problems
byAlison B. Lowndes
 
PDF
Tales of AI agents saving the human race!
byAlison B. Lowndes
 
PDF
Enabling Artificial Intelligence - Alison B. Lowndes
byWithTheBest
 
PDF
NVIDIA Corporation Brochure: Who We Are
byNVIDIA
 
PDF
Fuelling the AI Revolution with Gaming
byAlison B. Lowndes
 
PDF
NVIDIA DataArt IT
byAlison B. Lowndes
 
PDF
Innovation Roundtable
byAlison B. Lowndes
 
The What, Who & Why of NVIDIA
byAlison B. Lowndes
 
Talk on using AI to address some of humanities problems
byAlison B. Lowndes
 
Tales of AI agents saving the human race!
byAlison B. Lowndes
 
Enabling Artificial Intelligence - Alison B. Lowndes
byWithTheBest
 
NVIDIA Corporation Brochure: Who We Are
byNVIDIA
 
Fuelling the AI Revolution with Gaming
byAlison B. Lowndes
 
NVIDIA DataArt IT
byAlison B. Lowndes
 
Innovation Roundtable
byAlison B. Lowndes
 

What's hot

PDF
NVIDIA
byKevin Berce
 
PDF
Talk on commercialising space data
byAlison B. Lowndes
 
PDF
GTC World Tour 2017 highlights
byShanker Trivedi
 
PDF
Harnessing the virtual realm
byAlison B. Lowndes
 
PDF
BAT40 NVIDIA Stampfli Künstliche Intelligenz, Roboter und autonome Fahrzeuge ...
byBATbern
 
PPTX
PGI Compilers & Tools Update- March 2018
byNVIDIA
 
PDF
EPSRC CDT Conference
byAlison B. Lowndes
 
PPTX
NVIDIA 2017 Overview
byNVIDIA
 
PDF
GTC 2018: A New AI Era Dawns
byNVIDIA
 
PDF
Building New Realities in AEC with NVIDIA Quadro VR Webinar
byNVIDIA
 
PPTX
OpenACC Monthly Highlights - March 2018
byNVIDIA
 
PDF
NVIDIA Keynote #GTC21
byAlison B. Lowndes
 
PDF
NVIDIA – Inventor of the GPU
byNVIDIA
 
PPTX
OpenACC Monthly Highlights - February 2018
byNVIDIA
 
PPTX
Simplifying AI Infrastructure: Lessons in Scaling on DGX Systems
byRenee Yao
 
PDF
Hire a Machine to Code - Michael Arthur Bucko & Aurélien Nicolas
byWithTheBest
 
PDF
NVIDIA Overview 2015
byNVIDIA
 
PDF
Fuelling the AI Revolution with Gaming
byC4Media
 
PDF
GTC 2012 Jen-Hsun Huang Keynote
byNVIDIA
 
PDF
NVIDIA CES 2016 Highlights
byNVIDIA
 
NVIDIA
byKevin Berce
 
Talk on commercialising space data
byAlison B. Lowndes
 
GTC World Tour 2017 highlights
byShanker Trivedi
 
Harnessing the virtual realm
byAlison B. Lowndes
 
BAT40 NVIDIA Stampfli Künstliche Intelligenz, Roboter und autonome Fahrzeuge ...
byBATbern
 
PGI Compilers & Tools Update- March 2018
byNVIDIA
 
EPSRC CDT Conference
byAlison B. Lowndes
 
NVIDIA 2017 Overview
byNVIDIA
 
GTC 2018: A New AI Era Dawns
byNVIDIA
 
Building New Realities in AEC with NVIDIA Quadro VR Webinar
byNVIDIA
 
OpenACC Monthly Highlights - March 2018
byNVIDIA
 
NVIDIA Keynote #GTC21
byAlison B. Lowndes
 
NVIDIA – Inventor of the GPU
byNVIDIA
 
OpenACC Monthly Highlights - February 2018
byNVIDIA
 
Simplifying AI Infrastructure: Lessons in Scaling on DGX Systems
byRenee Yao
 
Hire a Machine to Code - Michael Arthur Bucko & Aurélien Nicolas
byWithTheBest
 
NVIDIA Overview 2015
byNVIDIA
 
Fuelling the AI Revolution with Gaming
byC4Media
 
GTC 2012 Jen-Hsun Huang Keynote
byNVIDIA
 
NVIDIA CES 2016 Highlights
byNVIDIA
 

Similar to NVIDIA @ Infinite Conference, London

PDF
Alison Lowndes, Artificial Intelligence DevRel, Nvidia – Fueling the Artifici...
byTechsylvania
 
PDF
Alison B Lowndes - Fueling the Artificial Intelligence Revolution with Gaming...
byCodemotion
 
PDF
Fueling the AI Revolution with Gaming
byAlison B. Lowndes
 
PDF
GTC Europe 2017 Keynote
byNVIDIA
 
PDF
GTC Taiwan 2017 主題演說
byNVIDIA Taiwan
 
PDF
Nvidia why every industry should be thinking about AI today
byJustin Hayward
 
PDF
Introduction to Deep Learning (NVIDIA)
byRakuten Group, Inc.
 
PDF
AI talk at CogX 2018
byAlison B. Lowndes
 
PPTX
Deep Learning Workflows: Training and Inference
byNVIDIA
 
PPTX
abelbrownnvidiarakuten2016-170208065814 (1).pptx
bygopikahari7
 
PDF
AI in the Financial Services Industry
byAlison B. Lowndes
 
PDF
Exploring solutions for humanity's greatest challenges
byAlison B. Lowndes
 
PDF
GTC 2016 Opening Keynote
byNVIDIA
 
PPTX
The-Backbone-of-the-AI-Revolution .pptx
bymru846747
 
PDF
Harnessing AI for the Benefit of All.
byAlison B. Lowndes
 
PDF
GTC China 2016
byNVIDIA
 
PDF
Introduction to the Artificial Intelligence and Computer Vision revolution
byDarian Frajberg
 
PDF
GTC 2017: Powering the AI Revolution
byNVIDIA
 
PDF
AI + E-commerce
byAlison B. Lowndes
 
PPTX
Deep Learning on Qubole Data Platform
byShivaji Dutta
 
Alison Lowndes, Artificial Intelligence DevRel, Nvidia – Fueling the Artifici...
byTechsylvania
 
Alison B Lowndes - Fueling the Artificial Intelligence Revolution with Gaming...
byCodemotion
 
Fueling the AI Revolution with Gaming
byAlison B. Lowndes
 
GTC Europe 2017 Keynote
byNVIDIA
 
GTC Taiwan 2017 主題演說
byNVIDIA Taiwan
 
Nvidia why every industry should be thinking about AI today
byJustin Hayward
 
Introduction to Deep Learning (NVIDIA)
byRakuten Group, Inc.
 
AI talk at CogX 2018
byAlison B. Lowndes
 
Deep Learning Workflows: Training and Inference
byNVIDIA
 
abelbrownnvidiarakuten2016-170208065814 (1).pptx
bygopikahari7
 
AI in the Financial Services Industry
byAlison B. Lowndes
 
Exploring solutions for humanity's greatest challenges
byAlison B. Lowndes
 
GTC 2016 Opening Keynote
byNVIDIA
 
The-Backbone-of-the-AI-Revolution .pptx
bymru846747
 
Harnessing AI for the Benefit of All.
byAlison B. Lowndes
 
GTC China 2016
byNVIDIA
 
Introduction to the Artificial Intelligence and Computer Vision revolution
byDarian Frajberg
 
GTC 2017: Powering the AI Revolution
byNVIDIA
 
AI + E-commerce
byAlison B. Lowndes
 
Deep Learning on Qubole Data Platform
byShivaji Dutta
 

More from Alison B. Lowndes

PDF
AI Star Keynote on Artificial Intelligence. OnWorld. OffWorld. InWorld
byAlison B. Lowndes
 
PDF
Bits and pixels. Efficiency for a multi-planet society.
byAlison B. Lowndes
 
PDF
NVIDIA at Breakthrough Discuss for Space Exploration
byAlison B. Lowndes
 
PDF
Bits & Pixels using AI for Good.........
byAlison B. Lowndes
 
PDF
Future of Skills
byAlison B. Lowndes
 
PDF
NVIDIA @ AI FEST
byAlison B. Lowndes
 
PDF
MAXSS & NVIDIA
byAlison B. Lowndes
 
PDF
Omniverse for the Metaverse
byAlison B. Lowndes
 
PDF
DataArt
byAlison B. Lowndes
 
PDF
From gaming to the metaverse
byAlison B. Lowndes
 
PDF
Breaking RSA & the internet
byAlison B. Lowndes
 
PDF
Decision-ready climate data
byAlison B. Lowndes
 
PDF
GTC 2022 Keynote
byAlison B. Lowndes
 
PDF
Harnessing the virtual realm for successful real world artificial intelligence
byAlison B. Lowndes
 
PDF
Hardware in Space
byAlison B. Lowndes
 
PDF
GTC Fall 2020 Keynote
byAlison B. Lowndes
 
PDF
Possibilities of generative models
byAlison B. Lowndes
 
PDF
Phi Week 2019
byAlison B. Lowndes
 
AI Star Keynote on Artificial Intelligence. OnWorld. OffWorld. InWorld
byAlison B. Lowndes
 
Bits and pixels. Efficiency for a multi-planet society.
byAlison B. Lowndes
 
NVIDIA at Breakthrough Discuss for Space Exploration
byAlison B. Lowndes
 
Bits & Pixels using AI for Good.........
byAlison B. Lowndes
 
Future of Skills
byAlison B. Lowndes
 
NVIDIA @ AI FEST
byAlison B. Lowndes
 
MAXSS & NVIDIA
byAlison B. Lowndes
 
Omniverse for the Metaverse
byAlison B. Lowndes
 
DataArt
byAlison B. Lowndes
 
From gaming to the metaverse
byAlison B. Lowndes
 
Breaking RSA & the internet
byAlison B. Lowndes
 
Decision-ready climate data
byAlison B. Lowndes
 
GTC 2022 Keynote
byAlison B. Lowndes
 
Harnessing the virtual realm for successful real world artificial intelligence
byAlison B. Lowndes
 
Hardware in Space
byAlison B. Lowndes
 
GTC Fall 2020 Keynote
byAlison B. Lowndes
 
Possibilities of generative models
byAlison B. Lowndes
 
Phi Week 2019
byAlison B. Lowndes
 

Recently uploaded

PDF
final.pdf
byomarbishtawi04
 
PDF
shayk.online - Anonymous chat with Sinatra and WebSockets
byEleanor McHugh
 
PDF
Chapter 5 Security Mechanisms iin computer security.pdf
byGetnet Tigabie Askale -(GM)
 
PPTX
apidays Paris 2025 | Integration is Feminist: Building Peace in Distributed S...
byapidays
 
PPTX
Tech Trends 2026: AI Agents, Quantum Computing, Robotics & Cybersecurity
byridwansassman
 
PDF
Automated Governance for FME Flow: Smarter Admin at Scale
bySafe Software
 
PPTX
Introducing VisualSim 2610 The Next Leap in System Level Modeling
byDeepak Shankar
 
PDF
Security-Fails-in-the-Gaps-Between-Systems.pptx.pdf
byOhhproJunction
 
PDF
AI Vector Search Best Practices Multicloud Feb 2026
bySandesh Rao
 
PDF
Odoo Implementation Checklist: A Strategic ERP Blueprint for Business-Ready D...
byBANIBRO IT SOLUTION
 
PDF
The new book “Marketing in the Metaverse” spotlights Scott M. Graffius’ research
byScott M. Graffius
 
PPTX
PPTX game guess the logo with a twistppt
byAdrianAnig
 
PDF
Antminer S23Hyd-580_Manual-V1.1-oneminers.pdf
byFranzhLawrenceBataan1
 
PDF
NTG -Company Profile_Software_Vendor_Company_in_MENA
byMustafa Kuğu
 
PPTX
Do You Control the AI, or Does the AI Control You?
bymedhateltoukhy
 
PDF
Session 1/5: Enhancing Automation with Screenplay & Business Rules
bysuhanisingh58689
 
PDF
How does MES(Manufacturing Execution System) work?
byakipii ogaoga
 
PDF
Founder & Tech Lead | Web Development & Digital Growth Consultant | Helping B...
byShyamal Das
 
PDF
20260212 Security-JAWS activity results for 2025 and activity goals for 2026
byTyphon 666
 
PDF
How to Install XRDP on CentOS 10 .pdf
byGreen Webpage
 
final.pdf
byomarbishtawi04
 
shayk.online - Anonymous chat with Sinatra and WebSockets
byEleanor McHugh
 
Chapter 5 Security Mechanisms iin computer security.pdf
byGetnet Tigabie Askale -(GM)
 
apidays Paris 2025 | Integration is Feminist: Building Peace in Distributed S...
byapidays
 
Tech Trends 2026: AI Agents, Quantum Computing, Robotics & Cybersecurity
byridwansassman
 
Automated Governance for FME Flow: Smarter Admin at Scale
bySafe Software
 
Introducing VisualSim 2610 The Next Leap in System Level Modeling
byDeepak Shankar
 
Security-Fails-in-the-Gaps-Between-Systems.pptx.pdf
byOhhproJunction
 
AI Vector Search Best Practices Multicloud Feb 2026
bySandesh Rao
 
Odoo Implementation Checklist: A Strategic ERP Blueprint for Business-Ready D...
byBANIBRO IT SOLUTION
 
The new book “Marketing in the Metaverse” spotlights Scott M. Graffius’ research
byScott M. Graffius
 
PPTX game guess the logo with a twistppt
byAdrianAnig
 
Antminer S23Hyd-580_Manual-V1.1-oneminers.pdf
byFranzhLawrenceBataan1
 
NTG -Company Profile_Software_Vendor_Company_in_MENA
byMustafa Kuğu
 
Do You Control the AI, or Does the AI Control You?
bymedhateltoukhy
 
Session 1/5: Enhancing Automation with Screenplay & Business Rules
bysuhanisingh58689
 
How does MES(Manufacturing Execution System) work?
byakipii ogaoga
 
Founder & Tech Lead | Web Development & Digital Growth Consultant | Helping B...
byShyamal Das
 
20260212 Security-JAWS activity results for 2025 and activity goals for 2026
byTyphon 666
 
How to Install XRDP on CentOS 10 .pdf
byGreen Webpage
 

NVIDIA @ Infinite Conference, London

  • 1.
    ALISON B LOWNDES AIDevRel | EMEA Fuelling the AI Revolution with Gaming @alisonblowndes
  • 2.
    2 The day job AUTOMOTIVE Autosensors reporting location, problems COMMUNICATIONS Location-based advertising CONSUMER PACKAGED GOODS Sentiment analysis of what’s hot, problems $ FINANCIAL SERVICES Risk & portfolio analysis New products EDUCATION & RESEARCH Experiment sensor analysis HIGH TECHNOLOGY / INDUSTRIAL MFG. Mfg. quality Warranty analysis LIFE SCIENCES MEDIA/ENTERTAINMENT Viewers / advertising effectiveness ON-LINE SERVICES / SOCIAL MEDIA People & career matching HEALTH CARE Patient sensors, monitoring, EHRs OIL & GAS Drilling exploration sensor analysis RETAIL Consumer sentiment TRAVEL & TRANSPORTATION Sensor analysis for optimal traffic flows UTILITIES Smart Meter analysis for network capacity, LAW ENFORCEMENT & DEFENSE Threat analysis - social media monitoring, photo analysis
  • 3.
  • 4.
    A MISSION CONTROL FORSPACESHIP EARTH KICK OFF WORKSHOP ESA ESRIN, ROME 25th - 29th June ‘18 RESEARCH SPRINT 30th June - 17th August
  • 5.
  • 6.
  • 7.
  • 8.
  • 9.
    9 TITAN V THE MOSTPOWERFUL PC GPU EVER 5,120 CUDA cores 640 NEW Tensor cores 12GB HBM2 Memory 1.7Gbps 110 Tensor TFLOPS 13.8 TFLOPS fp16 | 6.9 fp32
  • 10.
    10 THE RISE OFGPU COMPUTING Big Data Needs Algorithms and Compute That Scales 1980 1990 2000 2010 2020 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 103 105 107 1.5X per year END OF MOORE’S LAW Single-threaded perf GPU-Computing perf 1.5X per year 1.1X per year CPU vs. GPU
  • 11.
  • 12.
  • 14.
  • 15.
    15 CONVOLUTION 0 0 0 0 0 0 0 0 1 1 1 0 0 0 0 1 2 2 1 1 1 0 1 2 2 2 1 1 0 1 2 2 2 1 1 0 0 1 1 1 1 1 0 0 0 0 0 0 0 4 0 0 0 0 0 0 0 -4 1 0 -8 Source Pixel Convolution kernel New pixel value (destination pixel) Centreelement of the kernel is placed over the source pixel. The source pixel is then replaced with a weighted sum of itself and nearby pixels. http://www.biomachina.org/
  • 16.
  • 17.
    DEEP LEARNING NEW PROGRAMMINGMODEL DEVICE EDGE CLOUD
  • 19.
    19 Describe the differencesbetween these 2 bikes……..
  • 20.
    20 Image “Volvo XC90” Imagesource: “Unsupervised Learning of Hierarchical Representations with Convolutional Deep Belief Networks” ICML 2009 & Comm. ACM 2011. Honglak Lee, Roger Grosse, Rajesh Ranganath, and Andrew Ng. CONVOLUTIONAL NEURAL NETWORKS
  • 21.
  • 22.
    22 Li et al,University of Maryland & US Naval Academy https://arxiv.org/pdf/1712.09913.pdf
  • 23.
  • 24.
    24 ResNets vs HighwayNets (IDSIA) https://arxiv.org/pdf/1612.07771.pdf Klaus Greff, Rupesh K. Srivastava Really great explanation of “representation” Compares the two.. shows for language modelling, translation HN >> RN. Not quite as simple as each layer building a new level of representation from the previous - since removing any layer doesn’t critically disrupt.
  • 25.
    25 Capsules & routingwith EM, Hinton et al https://arxiv.org/pdf/1710.09829.pdf [NIPS2017] SegCaps: capsules for object segmentation https://arxiv.org/pdf/1804.04241.pdf University of Central Florida
  • 26.
  • 27.
  • 28.
    28 Long short-term memory(LSTM) Hochreiter (1991) analysed vanishing gradient “LSTM falls out of this almost naturally” Gates control importance of the corresponding activations Training via backprop unfolded in time LSTM: input gate output gate Long time dependencies are preserved until input gate is closed (-) and forget gate is open (O) forget gate Fig from Vinyals et al, Google April 2015 NIC Generator Fig from Graves, Schmidhuber et al, Supervised Sequence Labelling with RNNs
  • 29.
  • 30.
    CAMBRIAN EXPLOSION Convolutional Networks ReLuEncoder/Decoder Dropout PoolingConcat BatchNorm Recurrent Networks GRULSTM CTC BeamSearch WaveNet Attention Generative Adversarial Networks 3D-GAN Speech Enhancement GANCoupled GAN Conditional GANMedGAN Reinforcement Learning DQN Simulation DDPG New Species Neural Collaborative FilteringMixture of Experts Block Sparse LSTM Capsule Nets
  • 31.
    31 Shakir Mohamed andDanilo Rezende, DeepMind, UAI 2017
  • 32.
    32 IMAGINE NO LIMITATIONS TO DRUGDISCOVERY It takes 12 years and $2.6B to bring a new drug to market. Insilico Medicine takes an innovative approach to speed drug discovery and uses GPU-accelerated Generative Adversarial Networks (GANs) to "invent" new molecular structures on demand. With GANs, in only a few months’ time, Insilico did what would have taken years with conventional research methods – identified 69 new molecules. To date, Insilico has generated 5,000+ new molecules, integrated GANs into a comprehensive drug discovery framework, and established working collaborations with large pharmaceutical companies. This unique approach could dramatically reduce the time and cost of developing novel substances with medicinal properties.
  • 33.
  • 34.
  • 35.
    35 “After finishing thisbook, you will have a deep understanding of how to set technical direction for a machine learning project.” www.mlyearning.org
  • 36.
  • 37.
  • 38.
    38 Brain Computer Interfaces Focusedon treatment for disease and dysfunction eg epilepsy, depression, Parkinsons but ultimately to advance human intelligence by restoring and extending cognitive vibrancy. “We’re either going to have to merge with AI or be left behind”; Elon Musk
  • 39.
  • 40.
    40 RL TIMELINE •1984: RichSuttons thesis •1989: Q-learning •1991: REINFORCE •2006: MCTS (Szepesvari) •2009: David Silver’s thesis •2013: DQN •2015: DQN in Nature •2016: A3C & DRL •2017: Soft Q-learning/ AlphaGo Zero •2018: GA3C
  • 41.
    41 Tabish Rashid etal, Oxford/DeepMind QMIX: https://arxiv.org/pdf/1803.11485.pdf
  • 42.
    42 DEEPMIND ALPHA* 1.0=> 2.0 => 0.0 * ..a deeply structured hybrid (Gary Marcus, Jan 2018)
  • 43.
  • 44.
    44 ATLAS, BOSTON DYNAMICS Insertingvideo: Insert/Video/Video from File. Insert video by browsing your directory and selecting OK. File types that works best in PowerPoint are mp4 or wmv
  • 45.
  • 46.
    46 Holodeck: photorealistic, collaborativeVR environment Real-world experience through sight, sound, and haptics Explore high-fidelity, full-resolution models THE DESIGN LAB OF THE FUTURE IMMERSIVE AI - HOLODECK
  • 47.
  • 48.
  • 49.
    49 NVIDIA ISAAC ROBOTICSPLATFORM SIMULATION TRAINING DEPLOYMENT SDK https://developer.nvidia.com/isaac-sdk
  • 50.
  • 51.
    51 INSERT TELEPORTATION VIDEOOF TIM DRIVING A VR VERSION OF A REAL SDC
  • 52.
    A massive DeepLearning challenge
  • 53.
  • 54.
    54 BRIDGING THE REALITYGAP https://arxiv.org/pdf/1804.06516.pdf NVR+University of Toronto “leveraging the power of synthetic data improves upon results obtained using real data alone”
  • 55.
    55 END TO ENDSYSTEM FOR AV Sources: NVIDIA, RAND Corporation SIMULATETRAIN MODELSCOLLECT DATA RE-SIMULATE Lanes Lights Path Signs PedestriansCars Up to 2 petabytes per car per year 20 DNNs 1+ million images per DNN 10B+ miles to ensure safe driving Validate and verify for self-driving cars by 2020 MAPPING 5 + DNNs Create and update HD maps
  • 56.
    56 LEVEL 5 DEMANDS EXTREME COMPUTING +10X camera resolution + Surround LIDAR point-cloud processing + Camera & LIDAR localization to HD map + Tracking all surrounding objects + New map generation + Sophisticated path planning & control + Algorithm diversity + Sensor & computing fail-operate + ASIL-D Functional Safety + Excess computing capacity Level 2 Level 5
  • 57.
    57 COMPUTATIONAL SCALE REQUIRED 3million labeled images 1 DGX-1 trains 300k labeled images on 1 DNN in 1 day 10 DNNs required for self-driving 10 parallel experiments at all times 100 DGX-1 per car
  • 58.
    58 Xavier (ASIL-D rated) “Ourin-depth technical assessment confirms the Xavier SoC architecture is suitable for use in autonomous driving applications and highlights NVIDIA’s commitment to enable safe autonomous driving”; TÜV SÜD, Germany
  • 59.
    59 XAVIER DLA NOW OPENSOURCE Command Interface Tensor Execution Micro-controller Memory Interface Input DMA (Activations and Weights) Unified 512KB Input Buffer Activations and Weights Sparse Weight Decompression Native Winograd Input Transform MAC Array 2048 Int8 or 1024 Int16 or 1024 FP16 Output Accumulators Output Postprocess or (Activation Function, Pooling etc.) Output DMA WWW.NVDLA.ORG
  • 60.
  • 61.
  • 62.
    62 Dynamic Memory Networks MetaMindnow Salesforce (2yrs ago) https://arxiv.org/pdf/1603.01417v1.pdf
  • 63.
    63 “The Building Blocksof Interpretability” https://distill.pub/2018/building-blocks/ Chris Olah, Google et al, March 2018
  • 64.
  • 66.
    66 A PLETHORA OFHEALTHCARE STORIES Molecular Energetics For Drug Discovery AI for Drug Discovery Medical Decision Making Treatment Outcomes Reducing Cancer Diagnosis Errors by 85% Predicting Toxicology Predicting Growth Problems Image Processing Gene Mutations Detect Colon Polyps Predicting Disease from Medical Records Enabling Detection of Fatty Acid Liver Disease
  • 67.
  • 68.
    68 Radiation Therapy withAI ViewRay MRIdian (imaging & radiation dose)
  • 69.
    V-NET CINEMATIC RENDERINGAUTOMAP GPUDEEP LEARNING REVOLUTIONIZING MEDICAL IMAGING CUDA Source: https://arxiv.org/pdf/1704.08841.pdf Source: https://arxiv.org/pdf/1606.04797.pdf Source: Elliot Fishman, M.D. & Siemens
  • 70.
  • 71.
    71 COMPUTATIONAL & DATASCIENCE AI SUPERCOMPUTING WILL TRANSFORM HPC Extending Reach of HPC By Combining Computational & Data Science DATA SCIENCECOMPUTATIONAL SCIENCE Turbulent Flow Molecular Dynamics Structural Analysis N-body Simulation “Next move?” “Is there cancer?”“What’s happening?” “What does she mean?” Understanding Universe Clean EnergyDrug Discovery Monitoring Climate Change
  • 72.
    72 GPU READY APPS AWARENESS: ADOPTION : UTILIZATION ✓ Easy setup ✓ Best app practices ✓ Optimal results ✓ Higher productivity ✓ Faster discoveries App Users Life Science Developers Deep Learning IT Managers HPC/Data Centers Simple, ISV-approved, step-by-step instructions 1. 2. 3. 4. 5. Target Customers Quick Start Guide User Benefits www.nvidia.com/gpu-ready-apps
  • 73.
  • 74.
  • 75.
    75 POWERING THE DEEPLEARNING ECOSYSTEM NVIDIA SDK accelerates every major framework COMPUTER VISION OBJECT DETECTION IMAGE CLASSIFICATION SPEECH & AUDIO VOICE RECOGNITION LANGUAGE TRANSLATION NATURAL LANGUAGE PROCESSING RECOMMENDATION ENGINES SENTIMENT ANALYSIS DEEP LEARNING FRAMEWORKS NVIDIA DEEP LEARNING SDK developer.nvidia.com/deep-learning-software
  • 76.
  • 77.
    May 8-11, 2017| Silicon Valley CUDA 9 https://devblogs.nvidia.com/parallelforall/cuda-9-features-revealed/
  • 78.
    78NVIDIA CONFIDENTIAL. DONOT DISTRIBUTE. NVIDIA DEEP LEARNING SDK UPDATE GPU-accelerated DL Primitives Faster training Optimizations for RNNs Leading frameworks support cuDNN 7 Multi-node distributed training (multiple machines) Leading frameworks support Multi-GPU & Multi-node NCCL 2 TensorFlow model reader Object detection INT8 RNNs support High-performance Inference Engine TensorRT 4
  • 79.
    79 Circa 2000 -Torch7 - 4th (using odd numbers only 1,3,5,7) Web-scale learning in speech, image and video applications Maintained by top researchers including Soumith Chintala - Research Engineer @ Facebook All the goodness of Torch7 with an intuitive Python frontend that focuses on rapid prototyping, readable code & support for a wide variety of deep learning models. https://pytorch.org/2018/05/02/road-to-1.0.html
  • 80.
  • 81.
    81 Apex - APyTorch Extension ● Goal: Raise PyTorch customer awareness and increase adoption of NVIDIA Tensor Cores ● Content: Provide an easy to use set of utility functions in PyTorch for mixed-precision optimizations ● Benefit: Few lines of code to achieve improved training speed while maintaining accuracy and stability of single precision (Tensor Cores) ● Target audience: Deep learning researchers and developers of PyTorch with NVIDIA Volta ● Key Features: AMP (Auditor for mixed-precision) and Optimizer Wrapper (Dynamic loss scaling and master parameters) ● Teams Involved: Leading NVIDIA PyTorch team and collaboration with external FB PyTorch team Overview
  • 82.
    82 DALI Full input pipelineacceleration including data loading and augmentation Drop-in integration with direct plugins to frameworks – MxNet, TensorFlow, PyTorch Portable workflows through multiple input formats and configurable graphs Unblock CPU with GPU-accelerated DL pre-processing library Version 0.1 supports: • Resnet-50 image classification & segmentation training • Input formats – JPEG, LMDB, RecordIO, TFRecord • Python APIs to define, build and run an input pipelineAvailable as open source - https://github.com/NVIDIA/DALI • DALI Samples & Tutorial: https://github.com/NVIDIA/DALI/blob/master/examples/Getting%20started.ipynb • nvJPEG (Webpage, Documentation) - https://developer.nvidia.com/nvjpeg
  • 83.
    NEW NVIDIA TENSORRT4 DRIVE PX 2 JETSON TX2 NVIDIA DLA TESLA P4 TensorRT TESLA V100 Programmable Inference Accelerator Compile and Optimize Neural Networks | Support for Every Framework Optimize for Each Target Platform
  • 84.
    84 TensorRT INTEGRATED WITHTensorFlow 8x faster Inference Than TensorFlow Only * 14 86 705 0 100 200 300 400 500 600 700 800 CPU Only FP32 P4 FP32 P4 INT8 TensorRT Images/sec Throughput at < 7ms latency (TensorFlow ResNet-50) * Available in TensorFlow 1.7 https://github.com/tensorflow/tensorflow * Min CPU latency measured was 70 ms. It is not < 7 ms. CPU: Skylake Gold 6140, 2.5GHz, Ubuntu 16.04; 18 CPU threads. Pascal P4; CUDA (384.111; v9.0.176); Batch size: CPU=1, TF_GPU=1 (latency 12 ms) , TF-TRT=4 w/ latency=6ms
  • 85.
    85 NVIDIA TensorRT 4 Recommender, Speech& Machine Translation RNN and MLP Layers ⚫ ONNX Import ⚫ NVIDIA DRIVE Support Free download to members of NVIDIA Developer Program soon at developer.nvidia.com/tensorrt 50x Faster Inference for ONNX Models Easily import and accelerate inference for ONNX frameworks with native ONNX parser in TensorRT Accelerate inference of recommenders, speech and machine translation apps with new layers and optimizations Deploy optimized deep learning inference models NVIDIA DRIVE Xavier Support for NVIDIA DRIVE Xavier 1 45x 0X 10X 20X 30X 40X 50X CPU TensorRT
  • 86.
    86 TESLA V100 32GB THEMOST ADVANCED DATA CENTER GPU EVER BUILT 5,120 CUDA cores 640 NEW Tensor cores 7.5 FP64 TFLOPS | 15 FP32 TFLOPS 120 Tensor TFLOP 20MB SM RF | 16MB Cache | 32GB HBM2 @ 900 GB/s 300 GB/s NVLink https://devblogs.nvidia.com/parallelforall/inside-volta/
  • 87.
    87 TENSORCORES D = A* B + C (4x4)
  • 88.
    88 Tensor Cores: Moreresources Further information Mixed-precision blog: https://devblogs.nvidia.com/mixed-precision-training-deep-neural-networks/ Mixed-precision best practices: https://docs.nvidia.com/deeplearning/sdk/mixed-precision-training/index.html Mixed-precision arVix paper: https://arxiv.org/abs/1710.03740 GTC 2018 Sessions: Training with Mixed Precision: Theory and Practice and Training with Mixed Precision: Real Examples Call to action: 1) Share these resources with your customers and partners 2) Provide feedback to help us prioritize new examples and improve examples DGX/NGC deep learning framework containers: Latest versions of software stack and Tensor Core optimized examples - DGX/NGC Registry Tools TensorFlow OpenSeq2seq: https://nvidia.github.io/OpenSeq2Seq/html/mixed-precision.html & arVix paper PyTorch Apex: https://nvidia.github.io/apex/fp16_utils.html & NVIDIA developer news article Examples New mixed-precision model examples: https://developer.nvidia.com/deep-learning-examples GitHub: https://github.com/NVIDIA/DeepLearningExamples
  • 89.
    89 NVSWITCH WORLD’S HIGHEST BANDWIDTHON-NODE SWITCH 7.2 Terabits/sec or 900 GB/sec 18 NVLINK ports | 50GB/s per port bi-directional Fully-connected crossbar 2 billion transistors | 47.5mm x 47.5mm package
  • 90.
    ANNOUNCING NVIDIA DGX-2 THELARGEST GPU EVER CREATED 2 PFLOPS | 512GB HBM2 | 10 kW | 350 lbs
  • 91.
    91 GROUNDBREAKING AI AT YOURDESK Designed for your office DGX Station is the world’s first personal supercomputer for leading- edge AI development. Built on the same Deep Learning Stack powering all NVIDIA DGX™ Systems, you can now experiment at your desk and extend your work across DGX Systems and the cloud. Now with 32GB per GPU, same cost
  • 92.
    NVIDIA GPU CLOUD OptimizedStacks for Every Cloud 20,000+ Registered Organizations | 30 Containers NOW on AWS, GCP, AliCloud, Oracle Cloud, DGX
  • 93.
    93 Containerized Applications TF TunedSW NVIDIA Docker CNTK Tuned SW NVIDIA Docker Caffe2 Tuned SW NVIDIA Docker PyTorch Tuned SW NVIDIA Docker CUDA RTCUDA RTCUDA RTCUDA RT Linux Kernel + CUDA Driver Tuned SW NVIDIA Docker CUDA RT Other Frameworks and Apps. . . ALWAYS UP-TO-DATE Monthly Updates from NVIDIA to Frameworks and Containers
  • 94.
    94 NVIDIA-DOCKER 1.0 github.com/NVIDIA/nvidia-docker Open-source Projecton Github since 2016 >1.5MM downloads >5K stars >4.5 MM pulls of CUDA images Enabled Various Use-cases DGX / NGC containers Major deep learning frameworks Customers Adopted
  • 95.
    95 ANNOUNCING: JETSON XAVIER Computer forAutonomous Machines AI Server Performance in 30W  15W  10W 512 Volta CUDA Cores  2x NVDLA 8 core CPU 30 DL TOPS
  • 96.
  • 97.
    97 Sample Code Deep Learning CUDA,Linux4Tegra, ROS Multimedia API MediaComputer Vision Graphics Nsight Developer Tools Jetson Embedded Supercomputer: Advanced GPU, 64-bit CPU, Video CODEC, VIC, ISP JETPACK SDK FOR AI @ THE EDGE DEVELOPER.NVIDIA.COM/EMBEDDED-COMPUTING TensorRT cuDNN VisionWorks OpenCV Vulkan OpenGL libargus Video API
  • 98.
    98 Available to InstructorsNow! developer.nvidia.com/teaching-kits Robotics Teaching Kit with ‘Jet’ - ServoCity D E E P L E A R N I N G
  • 99.
    Fundamentals Accelerated Computing Game Development& Digital Content Finance NVIDIA DEEP LEARNING INSTITUTE Online self-paced labs and instructor-led workshops on deep learning and accelerated computing Take self-paced labs at www.nvidia.co.uk/dlilabs View upcoming workshops and request a workshop onsite at www.nvidia.co.uk/dli Educators can join the University Ambassador Program to teach DLI courses on campus and access resources. Learn more at www.nvidia.com/dli Intelligent Video Analytics Healthcare Robotics Autonomous Vehicles Virtual Reality
  • 100.
    100 NVIDIA INCEPTION PROGRAM Accelerates AI startupswith a boost of GPU tools, tech and deep learning expertise Startup Qualifications Driving advances in the field of AI Business plan Incorporated Web presence Technology DL startup kit* Pascal Titan X Deep Learning Institute (DLI) credit Connect with a DL tech expert DGX-1 ISV discount* Software release notification Live webinar and office hours *By application Marketing Inclusion in NVIDIA marketing efforts GPU Technology Conference (GTC) discount Emerging Company Summit (ECS) participation+ Marketing kit One-page story template eBook template Inception web badge and banners Social promotion request form Event opportunities list Promotion at industry events GPU ventures+ +By invitation www.nvidia.com/inception
  • 101.