2 results found Sort:
📚200+ Tensor/CUDA Cores Kernels, ⚡️flash-attn-mma, ⚡️hgemm with WMMA, MMA and CuTe (98%~100% TFLOPS of cuBLAS/FA2 🎉🎉).
Created
2022-12-17
496 commits to main branch, last one a day ago
Several optimization methods of half-precision general matrix vector multiplication (HGEMV) using CUDA core.
Created
2023-10-09
1 commits to master branch, last one 4 months ago