Thanks for sharing the work! Do you have any plan to support split-k grouped gemm? TRT-LLM has implemented it: https://github.com/NVIDIA/TensorRT-LLM/blob/4420547017fd31dc3568f0cebe13be08501d30db/cpp/tensorrt_llm/cutlass_extensions/include/cutlass_extensions/gemm/kernel/splitk_gemm_grouped.h