Hugo Rosenkranz·Jan 21GPU kernel optimization: Online SoftmaxLet’s look at the last optimization that is used by modern LLM frameworks to perform Softmax efficiently over large inputs
Hugo Rosenkranz·Sep 26, 2025GPU kernel optimization: Softmax — Part 2Softmax for Large Tensors: Multi-Block CUDA vs Triton Implementation
Hugo Rosenkranz·Jun 3, 2025GPU kernel optimization: Softmax — Part 1Unpacking GPU Reductions: The Story Behind Fast SoftmaxA response icon1A response icon1
Hugo Rosenkranz·Jul 17, 2023Deploy confidential AI applications using secure enclavesLeverage Cosmian open-sourced technologies to run your own HuggingFace models confidentially on the Cloud.
Hugo Rosenkranz·Apr 7, 2022Text classification: Create somewhat interpretable embeddings using Sentence BERT with Triplet LossHow to identify the most useful words for differentiating between classesA response icon1A response icon1