MXFP8 GEMM: Up to 99% of cuBLAS Performance Using CUDA and PTX

 (danielvegamyhre.github.io)

1 points | by matt_d 11 hours ago ago

No comments yet.