做大规模聚类或近似最近邻搜索的团队，Flash-GMM 让你在单卡上处理百倍数据量，速度还快 20 倍，直接替换 k-means 就能提升召回率，值得一试。
Flash-GMM 是一个融合的 Triton 内核，用于在单次 GPU 传递中高效计算高斯混合模型（GMM）。它通过避免在 GPU 内存中实例化完整的责任矩阵，实现了比现有实现快 20 倍的加速，并支持在单个设备上训练比以往大 100 倍以上的数据集。该内核被集成到 IVF 粗量化器中用于近似最近邻搜索，表明软 GMM 聚类可以替代 k-means，并利用 GMM 责任将边界向量分配到多个簇。在达到固定召回率目标时，Flash-GMM 减少了最多 1.7 倍的距离计算，或在相同计算成本下召回率提升 2-12 个点。该项目已开源。
Flash-GMM: A Memory-Efficient Kernel for Scalable Soft Clustering
We present \textbf{Flash-GMM}, a fused Triton kernel for efficient computation of Gaussian Mixture Models (GMMs) over large-scale data in a single GPU pass. By eliminating the need to materialize the full responsibility matrix in GPU memory, Flash-GMM achieves a \textbf{20$\times$} speedup over existing implementations and enables training on datasets more than \textbf{100$\times$} larger than previously feasible on one device. To demonstrate its impact, we integrate Flash-GMM into the IVF coarse quantizer for approximate nearest-neighbor (ANN) search. We show that soft GMM clustering is now a viable drop-in replacement for $k$-means, and that GMM responsibilities can be leveraged to assign border vectors to multiple clusters. Our approach reaches fixed recall targets with up to $1.7\times$ fewer distance computations, or equivalently, yields $+2$--$12$ recall@10 at matched computational cost. We release the kernel as an open-source project.