Meet Flash-KMeans: An IO-Aware, Exact K-Means That Runs Over 200× Faster Than FAISS on GPUs
Researchers from UC Berkeley and UT Austin have developed Flash-KMeans, an open-source library that significantly accelerates the k-means clustering algorithm for modern AI pipelines. By optimizing data movement on GPUs and restructuring the algorithm's stages, Flash-KMeans achieves substantial speedups, reportedly over 200x faster than FAISS and 33x faster than NVIDIA cuML on an NVIDIA H200 GPU. The library maintains mathematical exactness with standard k-means, focusing on IO efficiency rather than approximation, and can also handle out-of-core computations for extremely large datasets. AI
IMPACT Accelerates a core data processing step in AI pipelines, potentially reducing training and inference latency.