mirror of
https://github.com/NVIDIA/cutlass.git
synced 2026-05-25 07:14:30 +00:00
* Support half precision sigmoid activation * introduce a vectorized variant using fast_tanh * move the math to fast_math.h * fixed compile * .raw() -> .to_half() Co-authored-by: Haicheng Wu <haichengw@nvidia.com>