This is your work, valued
LLM, CUDA
DeepGEMM. DeepGEMM: clean and efficient FP8 GEMM kernels with fine-grained scaling
