Long-Context Attention from Kernel Efficiency to Distributed Context Parallelism

1 points | by PaulHoule 9 hours ago