From 11ec07e9d13c41ea8c1512f86414fd0096a0e095 Mon Sep 17 00:00:00 2001 From: Chao Liu Date: Thu, 1 Jul 2021 16:50:57 -0500 Subject: [PATCH] fix complain about divide by zero (#40) --- ...volution_forward_implicit_gemm_v4r4_nchw_kcyx_nkhw.cpp | 8 ++++---- ...volution_forward_implicit_gemm_v4r5_nchw_kcyx_nkhw.cpp | 8 ++++---- 2 files changed, 8 insertions(+), 8 deletions(-) diff --git a/composable_kernel/src/kernel_wrapper/dynamic_convolution_forward_implicit_gemm_v4r4_nchw_kcyx_nkhw.cpp b/composable_kernel/src/kernel_wrapper/dynamic_convolution_forward_implicit_gemm_v4r4_nchw_kcyx_nkhw.cpp index 19501b3fcd..8dc473ec3f 100644 --- a/composable_kernel/src/kernel_wrapper/dynamic_convolution_forward_implicit_gemm_v4r4_nchw_kcyx_nkhw.cpp +++ b/composable_kernel/src/kernel_wrapper/dynamic_convolution_forward_implicit_gemm_v4r4_nchw_kcyx_nkhw.cpp @@ -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)); diff --git a/composable_kernel/src/kernel_wrapper/dynamic_convolution_forward_implicit_gemm_v4r5_nchw_kcyx_nkhw.cpp b/composable_kernel/src/kernel_wrapper/dynamic_convolution_forward_implicit_gemm_v4r5_nchw_kcyx_nkhw.cpp index 503da231e9..fc27016624 100644 --- a/composable_kernel/src/kernel_wrapper/dynamic_convolution_forward_implicit_gemm_v4r5_nchw_kcyx_nkhw.cpp +++ b/composable_kernel/src/kernel_wrapper/dynamic_convolution_forward_implicit_gemm_v4r5_nchw_kcyx_nkhw.cpp @@ -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( 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));