Are GPUs the best hardware choice for GNNs? In a new post authored with @emaros96@Daniels_Data we explore the new @graphcoreai IPU architecture for Temporal GNNs showing it produces 10x gains in performance
https://t.co/qwHblByY1l
You may be aware we have two @graphcoreai Pod16s we are evaluating at UCL. Well Graphcore have launched a new initiative called the Accelerated Computing Academy to help scientists learn and exchange knowledge on ways to use the IPU outside of AI/ML https://t.co/1lu9dDFKx1
The authors of the paper "Improving Neural Network Training in Low-Dimensional Random Bases" will be presenting their poster today at #NeurIPS2020.
Join @graphcoreai's Frithjof, Zach & Carlo at 9AM PST 🕘https://t.co/QEYfzKUa08
Read the paper 📄 https://t.co/vs2fQ8MrzS
IPUs are here to stay! Today’s mk2 benchmarks announcement demonstrates once again the fundamental architectural advantage of our MIMD, massively parallel processor. Go team @graphcoreai 💪🏻🎉
Brand new performance results for @graphcoreai MK2 IPU systems are now available.
The results feature orders of magnitude improvements for some models - and outperform Nvidia A100 DGX across the board.
https://t.co/TneUs4PMxz
Excited to share a paper on local updates as an alternative to global backprop, co-led with @Luke_Metz + @graphcoreai@GoogleAI & @berkeley_ai.
tl;dr - Local updates can improve the efficiency of training deep nets in the high-compute regime.
👉 https://t.co/viEiwBx6Wf
1/N
Really great history + position paper by @sarahookr:
The Hardware Lottery (https://t.co/AJMgLaFyPJ)
"This essay introduces the term HW lottery to describe when a research idea wins b/c it suits the available SW/HW, not because the idea is superior to alt. research directions."
So very excited about my first live theatre experience since February. @oldvictheatre pushing the boundaries (again!) with a live streamed session on Zoom. #FaithHealer
Join @graphcoreai's Helen Byrne at the AI Applications Virtual Summit for insights into new approaches to financial modelling and accelerating machine intelligence workloads.
Register now: https://t.co/6vnOkyfR0b
@teamrework#Finance#Quant#AI#MachineIntelligence#IPU
Introducing our Second Generation of IPU Systems | Watch as @graphcoreai CEO Nigel Toon announces three new breakthrough technologies for machine intelligence – the IPU-Machine, 7nm GC200 IPU Processor and IPU-POD64
https://t.co/nDrgDff4Xw
AI currents: AI research you should know about. 1st edition!
Sifting the paper deluge with @azeem@marijagavrilovv@ExponentialView
https://t.co/ipzrtF57B4