WEBINARS

In this Executive Tutorial from Pienso and Graphcore, you will find out how non-technical teams can apply AI and Machine Learning to their business strategies with no-code/low-code high-performance machine intelligence.

What you will learn in this Executive Tutorial:

• What’s a Citizen Data Scientist?
• Where to start a DIY AI/ML project?
• What’s Structured vs Unstructured data?

What you'll learn:

• Baidu PaddlePaddle deep learning framework overview, technical designs and developer community
• PaddlePaddle and Graphcore IPU's integration design
• Training & inference demonstration with MNIST model

Key takeaways:

• Findings in the price prediction research paper from the Oxford-Man Institute of Quantitative Finance
• Accelerating multi-horizon forecasting on IPUs
• Introduction to IPU hardware and its architectural advantages
• Use cases of IPU technology in finance and future applications 

What you'll learn:

• Introduction to PopTorch and the Graphcore software stack
• From PyTorch to PopTorch: Overview of PopTorch functionality and a quick guide to porting your application
• Leveraging multiple IPU devices: how to easily accelerate training using advanced execution strategies
• Implementing efficient training pipelines in distributed environments

Key takeaways:

• Convergence of HPC and AI technologies: common patterns in scientific computing and machine learning
• Insights into the IPU’s unique architecture and its flexibility to support different workloads
• Identifying which scientific applications would fit well with IPU acceleration and potential benefits of enabling sparsity 
• The latest research in harnessing IPUs for data-dependent graph computations

 

What you'll learn:

• Introduction to the IPU architecture and an understanding of the programming model structure
• Overview of the Poplar Software Stack including its latest features, tools and libraries
• How developers can port and run models built with popular ML frameworks such as PyTorch and TensorFlow to the IPU
• Different considerations compared to GPU and using IPU techniques to leverage the advantages of the IPU such as parallel execution strategies

What you’ll learn:

• Introduction to the Graphcore Academic Programme: benefits of the programme and how you can participate
• Discover research papers and case studies from some of the world's leading academics and institutions using IPU hardware
• Latest performance benchmarks on Graphcore’s Mk2 IPU processor
• How to test IPU hardware and access the latest Graphcore software tools, including Poplar and PopART

What you'll learn:

• Introduction to IPU-Fabric: enabling direct connection between large clusters of IPUs for better resiliency and flexibility
• Compatibility with existing infrastructure & industry standard tools and why it matters
• How IPU-PODs are built for virtualisation and how to manage flexible and dynamic resource allocation with Virtual-IPU

What you'll learn:

• Introduction to the Poplar Software Stack including its latest features and capabilities
• Why Poplar was co-designed with the IPU and how it enables even the most demanding AI workloads to scale seamlessly across IPU systems
• Technical deep dive into Graphcore’s second generation IPU platforms designed for AI infrastructure at scale
• How Poplar makes it possible to execute current and next generation machine intelligence applications efficiently on IPU hardware

Key Takeaways:

• Why compute, data efficiency and communications are integral to enabling innovation in machine intelligence

• Insights into Graphcore's new generation of scale-out products: IPU-Machine: M2000 & IPU-POD64

• How Graphcore's MK2 IPU architecture facilitates the research and deployment of new models

• How IPU systems enhance model deployability and efficiency

Key Takeaways:

• How machine intelligence is evolving and what this means for AI processors

• Insights into IPU technology & Poplar software delivered via IPU-Server

• Poplar software supports multi-IPU constructs to enable a world of growing model sizes and complexity

• Overview of next generation image classification models using ResNeXt as an example, showing IPU benchmarks & use-case implementation

Key Takeaways:

• How the IPU is able to achieve faster financial model accelerations than other hardware available on the market

• How to use IPUs for financial modelling training and inference

• Insights into advanced models, use cases and IPU benchmarks

通过该研讨会，您将了解到：

• 计算、数据和通信为什么是实现机器智能创新的关键；

• 深入了解Graphcore的新一代横向大规模扩展产品：IPU-M2000和IPU-POD64；

• Graphcore的MK2 IPU架构如何促进新模型的研究和部署；

• IPU系统如何增强模型的可部署性和效率

通过本次研讨会，您将学到：

• 机器智能如何演变以及这对AI处理器意味着什么

• 首款IPU服务器——戴尔EMC DSS8440

• Poplar软件支持多种IPU构造，以实现模型尺寸和复杂性不断增长的世界

• 微软借助IPU加速ResNeXt-50医学成像推理的案例

通过在线研讨会，您将学习到：

• 如何利用金融算法模型预知黑天鹅事件，规避风险

• IPU如何能够实现比其他现有硬件更快的金融模型加速

• 如何使用IPU进行模型的训练和推理

• 先进算法模型洞察，案例以及IPU benchmark

주요 내용:

• Poplar 소프트웨어 스택 및 최신 기능과 성능 소개
• Poplar를 IPU와 공동 설계한 이유 및 가장 까다로운 AI 워크로드를 충족하여 여러 IPU 시스템에서 원활하게 확장 가능한 Poplar의 특성
• 대규모 AI 인프라용으로 설계된 그래프코어 2세대 IPU 플랫폼의 기술적 특성에 대한 심층 탐구
• Poplar가 IPU 하드웨어에서 기존 및 차세대 머신 인텔리전스 애플리케이션의 효율적인 실행을 지원하는 방법

해당 웨비나에서는 다음과 같은 내용을 다룹니다:

• 머신 인텔리전스 분야 혁신을 구현하기 위해 필수적인 컴퓨팅, 데이터 효율성 및 통신

• 그래프코어의 2세대 스케일아웃 신제품 소개 - IPU-머신: M2000 & IPI-POD64

• 그래프코어의 MK2 IPU 아키텍처가 연구 및 신규 모델 구축을 활성화하는 방법

• IPU 시스템이 모델 구축 및 효율성을 강화하는 방법

해당 웨비나는 다음과 같은 내용을 다룰 예정입니다:

• 기존 하드웨어 대비 빠르게 금융 모델 가속화를 달성하는 IPU

• 금융 모델 구축 학습 및 추론을 위한 IPU 활용법

• 어드밴스드 모델, 활용 사례, IPU 벤치마크에 대한 심층 정보

重要ポイント： 

• Vision Transformer 概要とその利点
• 機械学習フレームワークPyTorchを用いたVision Transformer 移植方法
• Vision Transformer の性能改善方法
• IPU-PODアーキテクチャの利点とプロファイリングツールPopVision

重要ポイント： 

• IPUプロセッサのアーキテクチャとIPU-M2000システムおよびIPU-PODシステムの紹介 
• 最新の機能やツール、ライブラリなど、Poplarソフトウェアスタックの概要 
• PyTorch、TensorFlow、またはKerasで構築したモデルをIPUに移植して実行する方法 
• GPUとは異なる検討事項と、IPU-PODシステムから最大限のパフォーマンスを引き出す方法 

• 世界で最もAIの為に最適設計された複雑なプロセッサIPU、およびIPUシステムの概要。
• HPC Systemsの会社紹介。HPC/AIシステムの製造および開発の経験、日本でのAI技術の導入事例等。
• スケールアウト効率をサポートする新しいGraphcore技術への洞察。
• 次世代のAI大躍進に備えるために革新者ができること。