ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-04-19 22:39:03 +00:00
Code Issues Packages Projects Releases Wiki Activity

Profile resnet layout fixes (#3360)

Browse Source
This commit is contained in:
Bartłomiej Kocot
2025-12-05 17:20:46 +01:00
committed by GitHub
parent f5b0af2272
commit 82f796a1f0
2 changed files with 4 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
profiler/include/profiler/profile_conv_fwd_bias_relu_add_impl.hpp
View File

@@ -75,13 +75,13 @@ void profile_conv_fwd_bias_relu_add_impl(int do_verification,
is_same<decltype(layout), ck::tensor_layout::convolution::KCYX>::value ||
is_same<decltype(layout), ck::tensor_layout::convolution::NKHW>::value)
{
return HostTensorDescriptor({N_, C_, H, W}, {C_ * H * W, H * W, W, 1_uz});
return HostTensorDescriptor({N_, C_, H, W}, {C_ * H * W, H * W, W, 1_uz}, layout);
}
else if constexpr(is_same<decltype(layout), tensor_layout::convolution::NHWC>::value ||
is_same<decltype(layout), tensor_layout::convolution::KYXC>::value ||
is_same<decltype(layout), tensor_layout::convolution::NHWK>::value)
{
return HostTensorDescriptor({N_, C_, H, W}, {C_ * H * W, 1_uz, W * C_, C_});
return HostTensorDescriptor({N_, C_, H, W}, {C_ * H * W, 1_uz, W * C_, C_}, layout);
}
};

4
profiler/include/profiler/profile_conv_fwd_bias_relu_impl.hpp
View File

@@ -75,13 +75,13 @@ void profile_conv_fwd_bias_relu_impl(int do_verification,
is_same<decltype(layout), ck::tensor_layout::convolution::KCYX>::value ||
is_same<decltype(layout), ck::tensor_layout::convolution::NKHW>::value)
{
return HostTensorDescriptor({N_, C_, H, W}, {C_ * H * W, H * W, W, 1_uz});
return HostTensorDescriptor({N_, C_, H, W}, {C_ * H * W, H * W, W, 1_uz}, layout);
}
else if constexpr(is_same<decltype(layout), tensor_layout::convolution::NHWC>::value ||
is_same<decltype(layout), tensor_layout::convolution::KYXC>::value ||
is_same<decltype(layout), tensor_layout::convolution::NHWK>::value)
{
return HostTensorDescriptor({N_, C_, H, W}, {C_ * H * W, 1_uz, W * C_, C_});
return HostTensorDescriptor({N_, C_, H, W}, {C_ * H * W, 1_uz, W * C_, C_}, layout);
}
};