Graphcore produced a series of striking images of computational graphs mapped to its "Intelligent Processing Unit."
The graph compiler builds up an intermediate representation of the computational graph to be scheduled and deployed across one or many IPU devices. The compiler can display this computational graph, so an application written at the level of a machine learning framework reveals an image of the computational graph which runs on the IPU.
The image below shows the graph for the full forward and backward training loop of AlexNet, generated from a TensorFlow description.
Our Poplar graph compiler has converted a description of the network into a computational graph of 18.7 million vertices and 115.8 million edges. This graph represents AlexNet as a highly-parallel execution plan for the IPU. The vertices of the graph represent computation processes and the edges represent communication between processes. The layers in the graph are labelled with the corresponding layers from the high level description of the network. The clearly visible clustering is the result of intensive communication between processes in each layer of the network, with lighter communication between layers.
I'm heading back to Austin for the SXSW Interactive festival and you can catch me three times this weekend: first on the Untold AI panel with Malka Older, Rashida Richardson and Christopher Noessel (5-6PM, Fairmont Manchester AB); then at the EFF Austin Party with Cindy Cohn and Bruce Sterling (7PM, 1309 Bonham Terrace); and on […]
There aren’t many details in Trump’s “American A.I. Initiative,” but the point appears to be: send a message of technological dominance to China.
Stanford folklorist and science historian Adrienne Mayor has a fascinating-sounding new book out, titled “Gods and Robots: Myths, Machines, and Ancient Dreams of Technology.” It’s a survey of how ancient Greeks, Romans, Indian, and Chinese myths imagined and grappled with visions of synthetic life, artificial intelligence, and autonomous robots. From Mayor’s interview at Princeton University […]
Big companies want automation on a big scale. Doing that means diving into the tricky world of machine learning and data science. And no matter what platform you’ll be implementing it on, you can learn how with the Machine Learning & Data Science Certification Training Bundle. In 48 hours and through eight courses, this bundle […]
Big systems need tight security – and the experts who can implement it. Cisco Networking Systems are the go-to providers for network infrastructure, but maintaining it takes a lot of up-to-date knowledge. If you want that knowledge right from the source, there’s an online course that can get you certified painlessly: The Foundational Cisco CCNA […]
Computer slowing down? There are a ton of reasons why that might be, especially if your unit has a few years on it. Junk files and programs can accumulate over time, some even left over from otherwise uninstalled software. This virtual debris can slow your PC down dramatically, but there’s a surprisingly quick fix. Lauded […]