Across the A100 attention benchmarks, the paper summarizes FlashAttention-2 as 1.7–3.0× faster than FlashAttention, 1.3–2.5× faster than the Triton implementation, and 3–10× faster than standard PyTorch attention. A more operation-specific summary says 1.3–1.5× over Triton in forward and around 2× in backward, and around 2× over FlashAttention and xFormers overall.

Across the A100 attention benchmarks, the paper summarizes FlashAttention-2 as 1.7–3.0× faster than FlashAttention, 1.3–2.5× faster than the Triton implementation, and 3–10× faster than standard PyTorch attention. A more operation-specific summary says 1.3–1.5× over Triton in forward and around 2× in backward, and around 2× over FlashAttention and xFormers overall. · CiteArk

测量结果

已评估 0/9 项

获得支持 0有争议或冲突 0尚未评估 9

保留同一结论下的完整结果组。已有评估排在前面；单项判断不替代整组结论。展开行查看条件、来源与历次证据。

每行是一项测量，分别显示论文报告、最近可用实测和累计证据状态。
指标与条件	论文报告	最近可用实测	证据判断
method: FlashAttention-2 · baseline: FlashAttention	1.7 ×	—	尚未评估
method: FlashAttention-2 · baseline: FlashAttention	3 ×	—	尚未评估
method: FlashAttention-2 · baseline: FlashAttention Triton	1.3 ×	—	尚未评估
method: FlashAttention-2 · baseline: FlashAttention Triton	2.5 ×	—	尚未评估
method: FlashAttention-2 · baseline: PyTorch	3 ×	—	尚未评估
method: FlashAttention-2 · baseline: PyTorch	10 ×	—	尚未评估
method: FlashAttention-2 · baseline: FlashAttention Triton	1.3 ×	—	尚未评估
method: FlashAttention-2 · baseline: FlashAttention Triton	1.5 ×	—	尚未评估
method: FlashAttention-2 · baseline: FlashAttention Triton	2 ×	—	尚未评估

报告指标与观测值

Assessment（0）

关联运行（0）

测量结果