A production-grade block-scaled FP8 GEMM + FlashAttention library for NVIDIA Hopper (H100 / H200, sm_90a) and Blackwell (sm_120a) GPUs. Beats PyTorch / cuBLAS / FlashInfer on the shapes that matter ...
When “Spamalot” rides into St. Paul this week with antics like knights who say “Ni!,” fish slapping and flying cows, it might come across as just another super silly musical. But it wasn’t that long ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results