Michael Diggin·Jul 20, 2025Understand PyTorch’s DDP by Implementing itIteratively coding up DistributedDataParallel to see what makes it work
Michael Diggin·May 18, 2025Reduce GPU idle time with greedy data loadingSacrifice exact reproducibility for performance improvements with PyTorch’s DataLoader
Michael Diggin·Apr 13, 2025Side Stepping Head of Line Blocking in Kafka ConsumersEnabling faster streaming applications without sacrificing consistencyA response icon1A response icon1
Michael Diggin·Mar 22, 2025OpenAI’s Triton: An end to end exampleAn introduction to Triton, writing custom GPU kernels and using them to accelerate machine learning
Michael Diggin·Mar 1, 2025Implementing a Split-K Matrix Multiplication Kernel in TritonIteratively writing a cuBLAS level kernel in TritonA response icon3A response icon3
Michael Diggin·Dec 27, 2024The Power of 8: Getting the most out of Tensor CoresMaximising matrix multiplication throughput