Revert "Revert " Fp8 block scale quantization for fmha fwd (#3330)" (#3633)" (#3635)

This reverts commit de5a1d730d.

Co-authored-by: Illia Silin <98187287+illsilin@users.noreply.github.com>
This commit is contained in:
ltqin
2026-01-24 01:03:22 +08:00
committed by GitHub
parent 2e08a7e5ab
commit 67f0b74ec6
14 changed files with 667 additions and 84 deletions

View File

@@ -37,6 +37,13 @@ struct scales
return lhs_ * rhs;
}
template <typename OtherScale>
CK_TILE_HOST_DEVICE constexpr auto operator*(OtherScale other) const
{
auto new_scale = lhs_ * other;
return scales<std::decay_t<decltype(new_scale)>>(new_scale);
}
private:
Scale lhs_;
};