ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-11 17:00:18 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix complain about divide by zero (#40)

Browse Source
This commit is contained in:
Chao Liu
2021-07-01 16:50:57 -05:00
committed by GitHub
parent 3835318cc3
commit 11ec07e9d1
2 changed files with 8 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
composable_kernel/src/kernel_wrapper/dynamic_convolution_forward_implicit_gemm_v4r4_nchw_kcyx_nkhw.cpp
View File

@@ -226,17 +226,17 @@ extern "C" __global__ void
constexpr auto I2 = Number<2>{};
constexpr auto in_n_c_hi_wi_desc =
make_dynamic_naive_tensor_descriptor_packed_v2(make_tuple(64, 4, 35, 35));
make_dynamic_naive_tensor_descriptor_packed_v2(make_tuple(256, 256, 28, 28));
constexpr auto wei_k_c_y_x_desc =
make_dynamic_naive_tensor_descriptor_packed_v2(make_tuple(8, 4, 3, 3));
make_dynamic_naive_tensor_descriptor_packed_v2(make_tuple(256, 256, 3, 3));
constexpr auto out_n_k_ho_wo_desc =
make_dynamic_naive_tensor_descriptor_packed_v2(make_tuple(64, 8, 18, 18));
make_dynamic_naive_tensor_descriptor_packed_v2(make_tuple(256, 256, 28, 28));
constexpr auto descs =
transform_forward_convolution_into_gemm_v4r4_nchw_kcyx_nkhw_pad(wei_k_c_y_x_desc,
in_n_c_hi_wi_desc,
out_n_k_ho_wo_desc,
make_tuple(2, 2),
make_tuple(1, 1),
make_tuple(1, 1),
make_tuple(1, 1),
make_tuple(1, 1));

8
composable_kernel/src/kernel_wrapper/dynamic_convolution_forward_implicit_gemm_v4r5_nchw_kcyx_nkhw.cpp
View File

@@ -236,18 +236,18 @@ extern "C" __global__ void
constexpr auto I2 = Number<2>{};
constexpr auto in_n_c_hi_wi_desc =
make_dynamic_naive_tensor_descriptor_packed_v2(make_tuple(64, 4, 35, 35));
make_dynamic_naive_tensor_descriptor_packed_v2(make_tuple(256, 256, 28, 28));
constexpr auto wei_k_c_y_x_desc =
make_dynamic_naive_tensor_descriptor_packed_v2(make_tuple(8, 4, 3, 3));
make_dynamic_naive_tensor_descriptor_packed_v2(make_tuple(256, 256, 3, 3));
constexpr auto out_n_k_ho_wo_desc =
make_dynamic_naive_tensor_descriptor_packed_v2(make_tuple(64, 8, 18, 18));
make_dynamic_naive_tensor_descriptor_packed_v2(make_tuple(256, 256, 28, 28));
constexpr auto descs =
transform_forward_convolution_into_contraction_v4r5_nchw_kcyx_nkhw_pad<N0>(
wei_k_c_y_x_desc,
in_n_c_hi_wi_desc,
out_n_k_ho_wo_desc,
make_tuple(2, 2),
make_tuple(1, 1),
make_tuple(1, 1),
make_tuple(1, 1),
make_tuple(1, 1));