Long-Context Attention from Kernel Efficiency to Distributed Context Parallelism

1 points | by PaulHoule 10 hours ago