Generative AI is moving from a cool plaything to a serious commercial tool. We outline the three tests that users should apply when considering how they might deploy it professionally. https://www.graphcore.ai/posts/using-generative-ai-commercially-check-it-passes-these-three-tests

Whisper it... Open AI's groundbreaking speech recognition model Whisper is now available to run as a Paperspace Gradient Notebook, powered by Graphcore IPUs .https://www.graphcore.ai/posts/how-to-use-openais-whisper-for-speech-recognition?utm_content=246290773&utm_medium=social&utm_source=facebook&hss_channel=fbp-890447934394683

It's a GNN match made in heaven: #PyG - the leading GNN development framework now works with
Graphcore IPUs - the best performing GNN compute platform.
https://www.graphcore.ai/posts/graphcore-users-can-now-build-and-run-gnns-with-pytorch-geometric

Join the Graphcore and Kumo community meetup to discuss the power of Graph Neural Networks (GNN), and the future of enterprise deep learning. Sign up to attend in person - London, Graphcore office, 13 April. https://www.graphcore.ai/meetup-amplify-your-graph-neural-network-workflows

Do you need #LLMs like #GPT-4 and #GPT-3 for every language task? No... fine-tuning open source #GPT-J may be a more cost-effective alternative.
https://www.graphcore.ai/posts/fine-tuned-gpt-j-a-cost-effective-alternative-to-gpt-4-for-nlp-tasks

Graphcore's partnership with PNNL (part of the US Department of Energy) unlocks dramatically faster 3D molecular modelling using SchNet Graph Neural Network #GNN https://t.co/Lv2dpaKUMK

GNNs are shaking-up AI, and it's no coincidence that Graph Neural Networks perform way better on Graphcore IPUs. https://www.graphcore.ai/posts/what-gnns-are-great-at-and-why-graphcore-ipus-are-great-at-gnns

Paperspace and Graphcore launch pay-as-you-grow Gradient Notebooks – powered by the Intelligence Processing Unit. Let’s start building! https://www.graphcore.ai/posts/paperspace-and-graphcore-launch-pay-as-you-grow-gradient-notebooks

Following Graphcore’s double win in this year’s Open Graph Benchmark Large Scale Challenge, team member Daniel Justus breaks down our approach to fast and accurate knowledge graph embedding models. https://towardsdatascience.com/large-scale-knowledge-graph-completion-on-ipu-4cf386dfa826

If anyone wants to try out Graphcore's OGB-LSC winning GPS++ model on IPUs, they can run it for free now on Paperspace.

Inference: https://console.paperspace.com/github/graphcore/ogb-lsc-pcqm4mv2?machine=Free-IPU-POD16&container=graphcore%2Ftensorflow-jupyter%3A2-amd-3.0.0-ubuntu-18.04-20221025&file=%2Fnotebook_inference.ipynb

Training: https://console.paperspace.com/github/graphcore/ogb-lsc-pcqm4mv2?machine=Free-IPU-POD16&container=graphcore%2Ftensorflow-jupyter%3A2-amd-3.0.0-ubuntu-18.04-20221025&file=%2Fnotebook_training.ipynb

#introduction

Howdy folks! I'm Luke, an #AI Research Engineer in the UK at @Graphcore. Currently I'm specialising in #multimodal training and figuring out how to build computers that can digest any data thrown their way.

Outside of work, I'm interested in all things #twistypuzzles and #speedcubing with an official 3x3 single/average of 7.16/9.42 seconds and numerous custom puzzles that I've designed and #3DPrinted (see pics below).

We're really thrilled with this result. Graphcore and our partners at Valence Discovery, Mila and Université de Montréal secured a double win on the Open Graph Benchmark Large Scale Challenge. More proof of the power of IPUs for Graph Neural Networks. https://www.graphcore.ai/posts/graphcore-claims-double-win-in-open-graph-benchmark-challenge

Hi Mastodon and the Sigmoid community. This felt like the natural server for us. Pop by and say hi. We'll be Tooting about interesting things going on in AI compute hardware, software and research at Graphcore.