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

1 pointsposted an hour ago
by matt_d
(danielvegamyhre.github.io)

No comments yet