zero dim support for tensorwise float8 training#1352

Open

vkuzo wants to merge 1 commit intomainfrom 20241118_zero_dim

+41-1

Commits on Dec 2, 2024

[wip] zero dim support for float8 training
vkuzo
committed