Blogs

Blogs

• MagiAttention
• Long-Context Attention Benchmark
• Support Native Group Collective Based on DeepEP
• Support Blackwell with FFA_FA4 Backend
• Support Learnable Attention Sink
• Support Muon QK-Clip
• Optimize Sparse Attention in FFA
• Dynamic Attention Solver
• How to Ensure Kernels Actually Overlapped
• Distributed-Native FFA
• Attention Engine for Inference
• Flash Attention 2 Math Derivation
• Support JIT Compilation in FFA