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.
Research scientist Janelle Shane has been training a neural network to generate food recipes by giving it tens of thousands of cookbook recipes. The neural net’s recipes are excellent: Beef Soup With Swamp Peef And Cheese Chocolate Chops & Chocolate Chips Crimm Grunk Garlic Cleas Beasy Mist Export Bean Spoons In Pie-Shell, Top If Spoon […]
Zuck That says, “Have you ever been on the Internet when you came across a checkbox that says “I’m not a robot?” In this video, I explain how those checkboxes (No CAPTCHA reCAPTCHAs) work as well as why they exist in the first place.” I mention CAPTCHA farms briefly, but the idea behind them is […]
Thread count isn’t like one of those deceiving metrics like camera megapixels or Facebook friends—more threads are always better if you can afford them. If price was no object, we would all be snoozing soundly bundled up in 1.8 kilo-thread sheets every single night. Guess what? Price doesn’t have to be an object with this […]
Maybe it’s entirely because of podcast ads, but drag-and-drop tools like Squarespace have gotten immensely popular in recent years. While it’s definitely a great tool for any non-coders who want to get a small website up and running quickly, managing content with a primarily visual interface can become a pain once you have more than […]
When you can’t wait for the world’s longest meeting to end, the mindless leg bouncing makes your boredom obvious and just annoys everybody else. Everyone knows the TPS reports need the damn cover sheet, but some sadistic colleague keeps forgetting, probably on purpose just to eat into your lunch hour. Enough is enough!While serving a […]