bringing gpu-accelerated computing, data science and …
TRANSCRIPT
Joe Bungo, Deep Learning Institute (DLI) Program Manager, NVIDIAShubbhi Teneja, Assistant Professor, Sonoma State UniversityRobert Settlage, Computational Scientist, Virginia Tech University
SIGHPC Education Webinar, April 8th, 2021
BRINGING GPU-ACCELERATED COMPUTING, DATA SCIENCE AND DEEP LEARNING TO THE CLASSROOM
2
NVIDIA DEEP LEARNING INSTITUTE (DLI)
• Build and deploy end-to-end projects across a range of technologies and domains
• Access GPU-accelerated servers in the cloud
• Take self-paced, online courses anytime, anywhere
• Join remote workshops led by DLI-certified instructors— experts with industry-leading technical knowledge
• Earn NVIDIA DLI certificates to demonstrate subject matter competency
Hands-on training in accelerated computing, AI, and accelerated data science to help you solve your most challenging problems
Learn more at: www.nvidia.com/dli
3
Rich Content portfolioFundamentals and advanced hands-on training in key technologies and application domains
AI for Healthcare
Deep Learning Fundamentals
AI for RoboticsAI for Autonomous Vehicles
AI for Intelligent Video Analytics
Accelerated Computing Fundamentals
NetworkingAI for
Predictive Maintenance
Accelerated Data Science Fundamentals
Intro to AI in the Data Center
AI for Anomaly Detection
AI for Industrial Inspection
4
DLI ECOSYSTEMLarge and Growing Worldwide
● Over 280,000 developers trained
● 33 service delivery and educational service partners
● Global distributors and resellers selling DLI
● 550+ certified instructors
● 400+ University Ambassadors across 350+ institutions
5
INSTRUCTOR-LED WORKSHOPSIn-person or Remote
AGENDAIntroduction (45 mins)
Break (15 mins)
Training task #1 (1:20 mins)
Lunch (60 mins)
Training task #2 (1:20 mins)
Break (15 mins)
Training task #3 (1:20 mins)
Summary/Q&A (15 mins)
BENEFITS OF WORKSHOPGet guidance from DLI Certified Instructors while working through material
Hear from experts during the introductory lecture
Collaborate with and learn from peers
Access fully-configured, GPU-accelerated workstations in the cloud
Earn a certificate of competency in course subject matter
6
ONLINE TRAINING
BENEFITS OF SELF-PACED TRAINING
Take training any time, anywhere. All you need is a laptop and Internet connection.
Access fully configured, GPU-accelerated workstations in the cloud for hands-on training.
Learn at your own pace with hands-on exercises and videos.
Earn a certificate of competency in 8-hr course subject matter (digital badge of completion for IT training.)
Customers can purchase bulk codes for online courses
7
DLI PROGRAMS FOR EDUCATORS
UNIVERSITY AMBASSADOR PROGRAM
• Qualified faculty and researchers can get certified to teach DLI workshops to their students at no cost
• Hundreds of universities with certified DLI Ambassadors around the world
TEACHING KITS
• Qualified university educators can download courseware across accelerated computing, deep learning, and robotics
• Kits include lecture materials, GPU cloud resources, access to self-paced DLI courses, and more
Learn more at: www.nvidia.com/dli
8
Solution: Ready-made teaching material in a variety of content types
TIMESolution: Free software tools, computing resources, hardware discounts
FUNDING
Solution: Content co-developed by NVIDIA and leading academic educators
THEORY VS APPLIEDSolution: Support from NVIDIA and educator community
CONTENT FAMILIARITY
BARRIERS TO TEACHING NEW TECHNOLOGIES
9
DLI TEACHING KITSAdvancing Global Education with GPU Acceleration
Downloadable, semester-long university curriculum course material for qualified educators and teaching assistants:
Lecture slidesLecture videosHands-on labs/solutionsLarger coding projects-solutionsQuiz/exam problem sets/solutionsText and e-booksFree DLI online courses/certificate opportunitiesFree AWS cloud creditsSyllabus with integrated online DLI courses
Different kits available for different course types:
1. Accelerated Computing (CUDA): UIUC/Prof. Wen-mei Hwu, U Delaware/Prof. Sunita Chandrasekaran2. Robotics: CalPoly/Prof John Seng3. Deep Learning: NYU/Prof. Yann LeCun4. Accelerated Data Science: Georgia Tech/Prof. PoloChau, Prarie View A&M/Prof. Xishuang Dong
developer.nvidia.com/teaching-kits
GPU TEACHING KIT FOR ACCELERATED COMPUTINGAvailable Now Free for Qualified Educators!
Co-developed with Prof. Wen-Mei Hwu (UIUC)
Comprehensive teaching materials:
3rd Ed. PMPP E-book by Hwu/Kirk Lecture slides and notesLecture videosHands-on labs/solutionsLarger coding projects/solutions Quiz/exam questions/solutions
GPU compute resources:
Free DLI online courses/certifications Free AWS cloud credits
developer.nvidia.com/teaching-kits
CUDA Libraries
CUDA CProgramming
CUDA Parallelism
Model
Data Management
CUDA Memory Models
Dynamic Parallelism
Multi-GPU Systems
Parallel Computation
Patterns
Histogram
Stencil
Reduction
Scan
Sparse Matrix
Merge Sort
Graph Search
Case Studies
Advanced MRI Reconstruction
Electrostatic Potential
Calculations
Deep Learning
Related Programming
Models
MPI
CUDA Python using Numba
OpenCL
OpenACC
OpenGL
CUDAProgramming
Model
ROBOTICS TEACHING KIT WITH‘JET’
Co-developed by Prof. John Seng (CalPoly) and NVIDIA
Comprehensive teaching materials:Lecture slidesHands-on labs/solutionsQuiz/exam problem sets/solutions Open-ended coding projects
Robot hardware kit via Build of Materials (BOM):NVIDIA Jetson TX1 and TX2 supportAll mechanical, chassis, electronics parts
Jetson Nano content also available from NVIDIA
developer.nvidia.com/teaching-kits
Learn interdisciplinary, GPU-accelerated, autonomous Robotics
Technical subjects
Sensors Computer Vision Machine Learning Dead Reckoning Path Planning Localization ControlObstacle Avoidance
ROBOTICS TEACHING KIT WITH‘JET’Module Goals
DEEP LEARNING TEACHING KITAvailable Free Now for Qualified Educators!
Co-developed with Prof. Yann LeCun (NYU)
Comprehensive teaching materials:
Lecture slidesHands-on labs/solutions with optional assessment rubric using Kaggle Quiz/exam problem sets/solutions
GPU compute resources:
Free DLI online courses/certifications Free AWS cloud credits
developer.nvidia.com/teaching-kits
DEEP LEARNING TEACHING KITModule Goals
Teach academic theory and application of DL harnessing the PyTorch and Torch frameworks
Technical subjects:
Intro to ML/DLApplied Image Classification
Applied Object DetectionConvolutional NNs
Applied Image SegmentationEnergy-based Learning
Unsupervised LearningGenerative Adversarial Networks
Optimization Techniques Recurrent NNs
Natural Language Processing And more!
16
DLI ACCELERATED DATA SCIENCE TEACHING KIT
Co-developed with Prof. Polo Chau (Georgia Tech) and Prof. Xishuang Dong (Prairie View A&M)
Comprehensive teaching materials covering fundamental and advanced topics:
Data collection and preprocessingAccelerated data science with RAPIDSDistributed computingMachine learning/GPU-accelerated MLData visualizationGraph analytics.
Content also covers culturally-responsive topics such as fairness and bias, as well as challenges and important figures from underrepresented groups
29 planned modules, 1st subset released Feb. 25th 2021
2nd release planned for early May 2021
Available Free for Qualified Educators
17
DLI PROGRAMS FOR EDUCATORS
UNIVERSITY AMBASSADOR PROGRAM
• Qualified faculty and researchers can get certified to teach DLI workshops to their students at no cost
• Hundreds of universities with certified DLI Ambassadors around the world
TEACHING KITS
• Qualified university educators can download courseware across accelerated computing, deep learning, and robotics
• Kits include lecture materials, GPU cloud resources, access to self-paced DLI courses, and more
Learn more at: www.nvidia.com/dli
5
Preparing today’s students and researchers for tomorrow’s AI computing challenges
Want to bring DLI to your campus?
DLI awards qualified educators as certified DLI Ambassadors, enabling them to teach free DLI content exclusively to university students and staff
DLI University Ambassadorship is an additional status on top of DLI Instructor Certification with additional benefits
Candidates should have relevant teaching and research/work experience. Apply here today:https://developer.nvidia.com/dli/cip/application-amb
DEEP LEARNING INSTITUTEUniversity Ambassador Program
INSTRUCTOR CERTIFICATION PROCESSUniversity Ambassador Program
https://developer.nvidia.com/dli/cip/application-amb
WHY BECOME A DLI AMBASSADOR?
DLI WORKSHOP AND CONTENT ENABLEMENT• Bring free, world-class DL training to academic communities and conferences (USD$500 value per student)• Proven, ready-made content and online training platform• Off-set for event expenses, catering, and/or travel expenses (up to USD$500 per event)• Access to workshop best practices and promotional assets via Ambassador Event Kit
TRAINING AND INSTRUCTOR CERTIFICATION AWARD
• Free DLI Instructor Certification (USD$1000 value)• Recognized and certified as an applied DL expert by NVIDIA• Formal inclusion in DLI University Ambassador and Certified Instructor programs
OTHER TEACHING RESOURCES• Early access to DLI content, DLI Teaching Kits, and other cloud-based platforms complement your
curriculum courses
OPPORTUNITY TO RUN PAID DLI TRAINING FOR INDUSTRY• Ambassadors can purchase DLI workshops from NVIDIA at a discount and resell to industry and
professional continuing education customers
INSTITUTIONS WITH AMBASSADORSDLI University Ambassadors come from hundreds of institutions worldwide (some shown below)
22
Dr. Shubbhi Taneja• Assistant Professor, Sonoma State
University (SSU)• Her research interests:
• Energy-efficient computing• High-Performance Computing• Big Data Systems
About SSU• SSU is a Hispanic-serving institute• SSU campus is located 50 miles north of
San Francisco
23
USE OF NVIDIA DLI RESOURCES
AT SSU
• Overview• NVIDIA Course Used: NVIDIA Fundamentals of
Accelerated Computing with CUDA C/C++• Course Type: Undergraduate; upper-division elective• # students: 28• Time allocated: 4-weeks (16 week long semester)• Course Mode: Mostly Asynchronous
• Strategy• Students were introduced to the fundamentals of GPU
computing and working of CUDA • Students also learned about GPU hardware• Several CUDA apps like Matrix Mul, Vector Addition
were discussed• Lastly, students were assigned the NVIDIA Accelerated
Computing course with CUDA C/C++
24
WHY USE DLI CONTENT?
BENEFITS TO INSTRUCTORS• The DLI courses provided reliable, industry-relevant, and high-quality content.• Free access to a lot of useful materials for in-class usage
BENEFITS TO STUDENTS• Students enjoyed the lessons as they were auto-graded and brought diversity to the learning material (esp. in
online courses where students mostly learn from videos)• In addition to CUDA, students could learn other skills like data visualization in these tutorials.• Free access to HPC resources• Learn more than one skills (e.g., with CUDA course, students learn data visualization)• Opportunity to earn certifications enhancing students’ portfolios
25
(SSU COLLECTS END-OF-SEMESTER EVALUATIONS FROM STUDENTS)
• “the micro lectures the professor provided were great because they were short videos on certain topics so you could always go back and watch them. The professor also did a very well job of explaining the material and giving us a lot of extra resources for help like the NVIDIA guide.”
• “Professor took the virtual teaching assignment on with enthusiasm. The lectures were formatted to resemble the youtube teaching videos we are all accustomed to and lean on for additional information. This familiar format made it easier to tackle lectures. She also got us access to NIVIDIA lessons and access to bridges supercomputing for GPU practice. I am very happy with how this class went. Wish more were formatted this way.”
STUDENTS REACTION TO DLI CONTENT
26
PLANS FOR BECOMING A
DLI UNIVERSITY
AMBASSADOR
Summer 2021• Become a DLI Ambassador
Fall 2021• Accelerated Computing Teaching Kit to be implemented
in the Fall’21 offering of CS 425 Parallel Computing at SSU
27
WE ARE VIRGINIA TECHThe only school to: Sequence and assemble the
genome/transcriptome of our mascot
Serve our mascot on Saturdays
And save said mascot in November
Robert Settlage, PhD MS: Computational/Data Scientist in ARC
Bioinformatician
HPC/Computing Advocate
UVA grad working at VT
28
Solution: Browser based access
PLATFORM ISSUESSolution: Use webforms and automation
ACCESS MODE
BARRIERS TO TEACHING AND USE OF HPC
Solution: Reservations, Public cloud
HARDWARE AVAILABILITYSolution: Teach where you compute, but don’t rely on installation on student laptop!
PLATFORM INCONSISTENCY
Solution: DLI Ambassador led courses
TIME TO CREATE MODULESolution: DLI Teaching kits
TIME TO CREATE CONTENT
29
Three use cases: two solutions
Student training for future job• Focus is on skill development and resume building
• Experience with cloud computing and less reliance on often scarce local resources are plusses
• DLI Ambassador program gives students new skills and a certificate of successful completion
Student (and faculty) training for research• Focus is on immediate needs and available resources
• Teaching modality needs to match the available research compute infrastructure
• DLI Teaching kits provide useful learning material (and possibly certificates)
• Open OnDemand allows focus on application specific topics with gradual expansion into compute issues
Student engagement and networking• Focus is on networking, cross-discipline skill development and resume building
• Multifaceted and real-world experiences a must
• DLI Teaching kits and materials gives students quick intro and reference materials
30
DLI AMBASSADOR(BOB’S PERSPECTIVE)
BENEFITS OF DLI AMBASSADOR PROGRAM TO VT
• Engaging, relevant and vetted content.
• Available to learners outside of training session such that the session time commitment can be adjusted.
• Elevate offerings at VT to include industry recognized partner.
• Addition of certificate to student resumes.
• Enhance the VT experience through variety.
• No local infrastructure required.
31
TEACHING KIT(BOB’S PERSPECTIVE)
BENEFITS OF TEACHING KITS TO VT INSTRUCTORS
• Can be used as single lesson, module or basis for course.
• Course content on demand.
• Interesting industry relevant content.
• CAN teach this on local infrastructure.
• EASY to add local relevance.
32
OPEN ONDEMAND(BOB’S PERSPECTIVE)
OPEN ONDEMAND BENEFITS TO ANY HPC INSTALLATION
• Zero install, site-configurable web-based access to HPC
• New users are instantly enabled and relevant
• Telescoping functionality to support new-intermediate users
• Reduced ticket load due to early access issues
• Integration with XDMoD to support resource utilization monitoring
DLI Ambassador Led Courses at VT
• Generally offered as ½ day courses
• Offered in conjunction with VT FDI
• Currently offering :
• Deep Learning in Computer Vision• CUDA Programming
• GPU compute resources:
• NO local GPUs needed
Future offerings: Fundamentals of Deep Learning and Fundamentals of Accelerated Data Science (Rapids)
DLI Teaching Kit use at VT• Hackathon topic introduction and supplementary material
• Class module/lecture material
• Deck of homework problems(!!)
Credits to teams VT_AOE_DL and CompuOmics
Open OnDemand at VT
• Jump start new users• Facilitate progression of user advancement• GUI support• Scripting support through browser• Job monitoring• Job resource utilization• HPC, Kubernetes, Cloud
WHY I BECAME A DLI AMBASSADOR…
DLI WORKSHOPS ARE FUN!• I personally enjoy the DLI workshops and attend them at every opportunity• GTC is my current top conference – great combination of current tech and Star Trek like future
Differentiation of my FDI workshops from other VT instructors
• Certificate by NVIDIA at completion!• High quality content
Differentiation of me• I am an AMBSSSADOR!
Networking• Adding contacts at NVIDIA as an HPC professional is a ++
37
WHAT’S NEXT
Watch “Deep Learning and Beyond”
Listen to the NVIDIA AI Podcast
Review examples of AI in action
GET THE BASICS LEARN WITH DLI
For more information and to apply visit developer.nvidia.com/dli/cip
APPLY FOR DLI INSTRUCTOR CERTIFICATION/AMBASSADORSHIP Download DLI Teaching Kits for complete course
solutions across Deep Learning, Robotics, and Accelerated Computing.
Visit developer.nvidia.com/teaching-kits for more info.
UNIVERSITY RESOURCES
Take a self-paced online training at www.nvidia.com/dli
Request an onsite or remote workshop through your account manager.
Contact us at [email protected]
THE CONFERENCE FOR AI INNOVATORS, TECHNOLOGISTS, AND CREATIVESJoin us at GTC 2021 on April 12 - 16 for the latest in AI, HPC, healthcare, game development, networking, and more.NVIDIA’s GTC brings together a global community of developers, researchers, engineers, and innovators to experience global innovation and collaboration.
Don’t miss out on the exclusive GTC keynote by Jensen Huang on April 12, available to everyone.
Register for free now at www.nvidia.com/gtc.