Graphcore and Hugging Face have released a new lineup of transformers optimized for Graphcore's Intelligence Processing Unit (IPU). These models are designed to take advantage of the IPU's architecture, which is specifically tailored for machine learning workloads. The new models include several popular transformer variants, such as BERT and RoBERTa, with performance improvements up to 2x compared to their non-IPU-optimized counterparts. --- Why it matters: This matters because it provides researchers and developers with faster and more efficient ways to train and deploy AI models, which can lead to breakthroughs in areas like natural language processing and computer vision. Source: https://huggingface.co/blog/graphcore-update