BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:AI Training Series - Scaling Deep Learning - From Single GPU to Cl
 usters
DTSTART:20260518T080000Z
DTEND:20260520T150000Z
DTSTAMP:20260608T234000Z
UID:indico-event-362@events.hpc-portal.eu
DESCRIPTION:Contents\nThis 3-days block course combines lectures about Fu
 ndamentals of Deep Learning\, Data Parallelism and Model Parallelism. 
 \nThe lectures are interleaved with many demos and hands-on sessions using
  Jupyter Notebooks.\nThe course is co-organised by LRZ and NVIDIA Deep Le
 arning Institute (DLI).  All instructors are NVIDIA certified University
  Ambassadors.\nDay 1: Fundamentals of Deep Learning\nBusinesses worldwide 
 are using artificial intelligence to solve their greatest challenges. Heal
 thcare professionals use AI to enable more accurate\, faster diagnoses in 
 patients. Retail businesses use it to offer personalised customer shopping
  experiences. Automakers use it to make personal vehicles\, shared mobilit
 y\, and delivery services safer and more efficient. Deep learning is a pow
 erful AI approach that uses multi-layered artificial neural networks to de
 liver state-of-the-art accuracy in tasks such as object detection\, speech
  recognition\, and language translation. Using deep learning\, computers c
 an learn and recognise patterns from data that are considered too complex 
 or subtle for expert-written software.\nIn this lecture\, you’ll learn h
 ow deep learning works through hands-on exercises in computer vision and n
 atural language processing. You’ll train deep learning models from scrat
 ch\, learning tools and tricks to achieve highly accurate results. You’l
 l also learn to leverage freely available\, state-of-the-art pre-trained m
 odels to save time and get your deep learning application up and running q
 uickly.\nBy participating in this lecture\, you will:\n\nLearn the fundame
 ntal techniques and tools required to train a deep learning model\nGain ex
 perience with common deep learning data types and model architectures\nEnh
 ance datasets through data augmentation to improve model accuracy\nLeverag
 e transfer learning between models to achieve efficient results with less 
 data and computation\nBuild confidence to take on your own project with a 
 modern deep learning framework\n\nDay 2: Data Parallelism\nModern deep lea
 rning challenges leverage increasingly larger datasets and more complex mo
 dels. As a result\, significant computational power is required to train m
 odels effectively and efficiently. Learning to distribute data across mult
 iple GPUs during deep learning model training makes possible an incredible
  wealth of new applications utilizing deep learning.\nAdditionally\, the e
 ffective use of systems with multiple GPUs reduces training time\, allowin
 g for faster application development and much faster iteration cycles. Tea
 ms who are able to perform training using multiple GPUs will have an edge\
 , building models trained on more data in shorter periods of time and with
  greater engineer productivity.\nThis lecture teaches you techniques for d
 ata-parallel deep learning training on multiple GPUs to shorten the traini
 ng time required for data-intensive applications. Working with deep learni
 ng tools\, frameworks\, and workflows to perform neural network training\,
  you’ll learn how to decrease model training time by distributing data t
 o multiple GPUs\, while retaining the accuracy of training on a single GPU
 .\nBy participating in this lecture\, you’ll:\n\nUnderstand how data par
 allel deep learning training is performed using multiple GPUs\nAchieve max
 imum throughput when training\, for the best use of multiple GPUs\nDistrib
 ute training to multiple GPUs using Pytorch Distributed Data Parallel\nUnd
 erstand and utilize algorithmic considerations specific to multi-GPU train
 ing performance and accuracy\n\nDay 3: Model Parallelism\nLarge language m
 odels (LLMs) and deep neural networks (DNNs)\, whether applied to natural 
 language processing (e.g.\, GPT-3)\, computer vision (e.g.\, huge Vision T
 ransformers)\, or speech AI (e.g.\, Wave2Vec 2)\, have certain properties 
 that set them apart from their smaller counterparts. As LLMs and DNNs beco
 me larger and are trained on progressively larger datasets\, they can adap
 t to new tasks with just a handful of training examples\, accelerating the
  route toward general artificial intelligence. Training models that contai
 n tens to hundreds of billions of parameters on vast datasets isn’t triv
 ial and requires a unique combination of AI\, high-performance computing (
 HPC)\, and systems knowledge. The goal of this course is to demonstrate ho
 w to train the largest of neural networks and deploy them to production.\n
 By participating in this lecture\, you’ll learn how to:\n\nScale trainin
 g and deployment of LLMs and neural networks across multiple nodes.\nUse t
 echniques such as activation checkpointing\, gradient accumulation\, and v
 arious forms of model parallelism to overcome the challenges associated wi
 th large-model memory footprint.\nCapture and understand training performa
 nce characteristics to optimize model architecture.\nDeploy very large mul
 ti-GPU\, multi-node models to production using NVIDIA Triton™ Inference 
 Server.\n\nImportant information\nAfter you are accepted\, please create a
 n account under https://learn.nvidia.com/join.\nEnsure your laptop / PC w
 ill run smoothly by going to http://websocketstest.com/ Make sure that W
 ebSockets work for you by seeing under Environment\, WebSockets is support
 ed and Data Receive\, Send and Echo Test all check Yes under WebSockets (P
 ort 80).If there are issues with WebSockets\, try updating your browser.\n
 NVIDIA Deep Learning Institute\nThe NVIDIA Deep Learning Institute deliv
 ers hands-on training for developers\, data scientists\, and engineers. Th
 e program is designed to help you get started with training\, optimising\,
  and deploying neural networks to solve real-world problems across diverse
  industries such as self-driving cars\, healthcare\, online services\, and
  robotics.\nPrerequisites\n\nDay 1: An understanding of fundamental progra
 mming concepts in Python 3\, such as functions\, loops\, dictionaries\, a
 nd arrays\; familiarity with Pandas data structures\; and an understandin
 g of how to compute a regression line.\nDay 2: Experience with deep lear
 ning training using Python.\nDay 3: Good understanding of PyTorch\, good 
 understanding of deep learning and data parallel training concepts\, pract
 ice with natural language processing are useful\, but optional\n\nHands-On
 \nThe lectures are interleaved with many hands-on sessions using Jupyter N
 otebooks. The exercises will be done on a fully configured GPU-accelerated
  workstation in the cloud.\nLanguage\nEnglish\nLecturers\nAjay Navilarekal
  Rajgopal (LRZ\, NVIDIA certified University Ambassador)\nPrices and Elig
 ibility\nThe course is open and free of charge for people from academia fr
 om the Member States of the European Union (EU) and Associated Countries 
 to the Horizon 2020 programme.\nRegistration\nPlease register with your of
 ficial e-mail address to prove your affiliation.\nWithdrawal Policy\nSee 
 Withdrawal\nLegal Notices\nFor registration for LRZ courses and workshops 
 we use the service edoobox from Etzensperger Informatik AG (www.edoobox.co
 m). Etzensperger Informatik AG acts as processor and we have concluded a D
 ata Processing Agreement with them.\nSee Legal Notices\n\nhttps://events.
 hpc-portal.eu/event/362/
LOCATION:Seminarraum 2 ( 	Leibniz Rechenzentrum Boltzmannstr. 1 85748 Garc
 hing b. München)
URL:https://events.hpc-portal.eu/event/362/
END:VEVENT
END:VCALENDAR