[GH]score: 0.50
MoonshotAI / FlashKDA
July 29, 2026
FlashKDA provides high-performance Kimi Delta Attention kernels designed to optimize attention mechanisms. The repository implements specialized CUDA kernels to accelerate delta-based attention computations, focusing on reducing latency and increasing throughput for large-scale model inference.
DAILY DIGEST
you don't check 9 sources — we do. one email every morning, read in 2 min. free. unsubscribe anytime. privacy