ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-04-20 14:59:17 +00:00
Code Issues Packages Projects Releases Wiki Activity

Replace grouped conv bwd wei wmmaV3 bilin/scale bf16f32bf16 support with bf16bf16bf16 (#3470)

Browse Source 
* Replace grouped convolution bwd weight wmma v3 bilinear and scale bf16f32bf16 support with bf16bf16bf16 support. Update tests.

* Tentative fix for bwd weight bilinear bf16bf16bf16, seems like the bilinear elementwise overload for this case (bf16, f32 accu, bf16) was wrong.
This commit is contained in:
Kiefer van Teutem
2025-12-29 12:58:29 +01:00
committed by GitHub
parent b0ea67e377
commit 88ae445580
10 changed files with 47 additions and 46 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
include/ck/tensor_operation/gpu/element/binary_element_wise_operation.hpp
View File

@@ -348,9 +348,7 @@ struct Bilinear
__host__ __device__ constexpr void
operator()<bhalf_t, float, bhalf_t>(bhalf_t& y, const float& x0, const bhalf_t& x1) const
{
const float x1_tmp = ck::type_convert<float>(x1);
const float y_tmp = alpha_ * x0 + beta_ * x1_tmp;
y = y_tmp;
y = type_convert<bhalf_t>(alpha_ * x0 + beta_ * ck::type_convert<float>(x1));
};
template <>