Kimi.ai(@Kimi_Moonshot):We're open-sourcing FlashKDA — our high-performance CUTLASS-based implementation of Kimi Delta Attention kernels. Achieves 1.72×–2.22× prefill speedup over the flash-linear-attention baseline on H20, and works as a drop-in backend for flash-linear-attention. Explore on github:

Kimi.ai

@Kimi_Moonshot

Built by Moonshot AI to empower everyone to be superhuman. PR: globalpr@moonshot.ai DC:

加入 December 2024

137 正在關注 355.4K 粉絲

Kimi.ai@Kimi_Moonshot

2026.04.21 15:12

We're open-sourcing FlashKDA — our high-performance CUTLASS-based implementation of Kimi Delta Attention kernels. Achieves 1.72×–2.22× prefill speedup over the flash-linear-attention baseline on H20, and works as a drop-in backend for flash-linear-attention. Explore on github: