[KongchangAI]
Unverified50% confidenceBenchmarkExact time

FlashAttention通过分块Tiling和重计算策略,将HBM访问次数从O(N²)降至O(N²/M),长序列场景下速度提升达2-4倍

1
Sources
50%
Confidence
Long-term
Relevance
7/15/2026
First Seen

Sources

Related Claims

Cite This Claim

Stable URI
https://kongchang.com/claim/512121
API
curl https://kongchang.com/api/v1/knowledge/claims/512121
MCP
get_claim(id=512121)