arXiv · cs.AI · Momin Abbas, Sadia Asif, Asad Ul Haq, Kaoutar El Maghraoui · 11 小时前模型评分33
Turbo-GMM：受 FlashAttention 启发的 图形处理器（GPU，Graphics Processing Unit） 加速高斯混合模型
Turbo-GMM: FlashAttention-Inspired GPU-Accelerated Gaussian Mixture Model
导读
Turbo-GMM 提出单次遍历的融合 EM 内核，在一次遍历组件维度内完成责任度归一化与充分统计量累加，省去 Flash-GMM 两遍扫描中的第二次打分计算。
分类置信度
中 · 待人工复核
- 存在未解决的相邻类别：compute-hardware，需要人工复核。
- 只有单一媒体来源，需要人工复核。
来源级别：媒体与个人
事件类型：研究发布
证据摘录
- “We introduce Turbo-GMM, a single-sweep fused EM kernel that completes responsibility normalization and sufficient-statistic accumulation within one traversal of the component dimension.”
- “Its applicability is limited by register pressure, occupancy, and the component and feature dimensions, making it particularly suitable for large datasets with relatively small numbers of components.”
- “For larger configurations, Turbo-GMM employs a two-pass fallback following the established fused design.”
来源：arXiv · cs.AI · arxiv.org