mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-04-20 06:49:15 +00:00
Add a scale op, related instances and examples (#1242)
* Add a scale op * Update the element op * Add instances * Add an example * Add a client example * Add a flag check * Revert flag check addition * Fix flag check * Update d strides in example * Update d strides in client example * Apply suggestions from code review Update copyright header Co-authored-by: Bartłomiej Kocot <barkocot@amd.com> * Move the example * Move the client example * Update element op * Update example with the new element op * Add scalar layout * Update example * Update kernel for scalar Ds * Revert kernel changes * Update element op * Update example to use scales' pointers * Format * Update instances * Update client example * Move element op to unary elements * Update element op to work with values instead of pointers * Update instances to take element op as an argument * Update examples to use random scale values --------- Co-authored-by: Bartłomiej Kocot <barkocot@amd.com>
This commit is contained in:
Rostyslav Geyyer
@@ -1,4 +1,5 @@
|
||||
add_subdirectory(binary)
|
||||
add_subdirectory(convscale)
|
||||
add_subdirectory(multi_AB)
|
||||
add_subdirectory(unary)
|
||||
|
||||
|
||||
10
example/62_convnd_activ/convscale/CMakeLists.txt
Normal file
10
example/62_convnd_activ/convscale/CMakeLists.txt
Normal file
@@ -0,0 +1,10 @@
|
||||
list(APPEND gpu_list gfx908 gfx90a gfx940 gfx941 gfx942)
|
||||
set(target 0)
|
||||
foreach(gpu IN LISTS GPU_TARGETS)
|
||||
if(gpu IN_LIST gpu_list AND target EQUAL 0)
|
||||
add_custom_target(example_convnd_activ_xdl_convscale)
|
||||
add_example_executable(example_convnd_fwd_xdl_convscale_fp8 convnd_fwd_xdl_convscale_fp8.cpp)
|
||||
add_example_dependencies(example_convnd_activ_xdl_convscale example_convnd_fwd_xdl_convscale_fp8)
|
||||
set(target 1)
|
||||
endif()
|
||||
endforeach()
|
||||
@@ -0,0 +1,301 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include <cstdlib>
|
||||
#include <iostream>
|
||||
#include <numeric>
|
||||
#include <type_traits>
|
||||
|
||||
#include "ck/ck.hpp"
|
||||
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
|
||||
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
|
||||
|
||||
#include "ck/library/utility/algorithm.hpp"
|
||||
#include "ck/library/utility/check_err.hpp"
|
||||
#include "ck/library/utility/device_memory.hpp"
|
||||
#include "ck/library/utility/host_tensor.hpp"
|
||||
#include "ck/library/utility/host_tensor_generator.hpp"
|
||||
#include "ck/library/utility/convolution_parameter.hpp"
|
||||
#include "ck/library/utility/convolution_host_tensor_descriptor_helper.hpp"
|
||||
#include "ck/library/reference_tensor_operation/cpu/reference_conv_fwd.hpp"
|
||||
|
||||
using PassThrough = ck::tensor_operation::element_wise::PassThrough;
|
||||
using ConvScale = ck::tensor_operation::element_wise::ConvScale;
|
||||
|
||||
void print_helper_msg()
|
||||
{
|
||||
std::cout << "arg1: verification (0=no, 1=yes)\n"
|
||||
<< "arg2: initialization (0=no init, 1=integer value, 2=decimal value)\n"
|
||||
<< "arg3: time kernel (0=no, 1=yes)\n"
|
||||
<< ck::utils::conv::get_conv_param_parser_helper_msg() << std::endl;
|
||||
}
|
||||
|
||||
template <typename DataType>
|
||||
inline __host__ __device__ constexpr double get_rtol()
|
||||
{
|
||||
if constexpr(std::is_same_v<DataType, float>)
|
||||
{
|
||||
return 1e-3;
|
||||
}
|
||||
else if constexpr(std::is_same_v<DataType, double>)
|
||||
{
|
||||
return 1e-6;
|
||||
}
|
||||
else if constexpr(std::is_same_v<DataType, ck::half_t>)
|
||||
{
|
||||
return 1e-3;
|
||||
}
|
||||
else if constexpr(std::is_same_v<DataType, ck::bhalf_t>)
|
||||
{
|
||||
return 5e-2;
|
||||
}
|
||||
else if constexpr(std::is_same_v<DataType, int32_t>)
|
||||
{
|
||||
return 1e-1;
|
||||
}
|
||||
else if constexpr(std::is_same_v<DataType, int8_t>)
|
||||
{
|
||||
return 1e-1;
|
||||
}
|
||||
else if constexpr(std::is_same_v<DataType, ck::f8_t>)
|
||||
{
|
||||
return 1e-1; // 240 and 224 are acceptable
|
||||
}
|
||||
else if constexpr(std::is_same_v<DataType, ck::bf8_t>)
|
||||
{
|
||||
return 1.5e-1; // 57344 and 49152 are acceptable
|
||||
}
|
||||
else
|
||||
{
|
||||
return 1e-3;
|
||||
}
|
||||
}
|
||||
|
||||
template <typename DataType>
|
||||
inline __host__ __device__ constexpr double get_atol()
|
||||
{
|
||||
if constexpr(std::is_same_v<DataType, float>)
|
||||
{
|
||||
return 1e-3;
|
||||
}
|
||||
else if constexpr(std::is_same_v<DataType, double>)
|
||||
{
|
||||
return 1e-6;
|
||||
}
|
||||
else if constexpr(std::is_same_v<DataType, ck::half_t>)
|
||||
{
|
||||
return 1e-3;
|
||||
}
|
||||
else if constexpr(std::is_same_v<DataType, ck::bhalf_t>)
|
||||
{
|
||||
return 5e-2;
|
||||
}
|
||||
else if constexpr(std::is_same_v<DataType, int32_t>)
|
||||
{
|
||||
return 1e-1;
|
||||
}
|
||||
else if constexpr(std::is_same_v<DataType, int8_t>)
|
||||
{
|
||||
return 1e-1;
|
||||
}
|
||||
else if constexpr(std::is_same_v<DataType, ck::f8_t>)
|
||||
{
|
||||
return 16.1; // 240 and 224 are acceptable
|
||||
}
|
||||
else if constexpr(std::is_same_v<DataType, ck::bf8_t>)
|
||||
{
|
||||
return 8192.1; // 57344 and 49152 are acceptable
|
||||
}
|
||||
else
|
||||
{
|
||||
return 1e-3;
|
||||
}
|
||||
}
|
||||
|
||||
template <ck::index_t NumDimSpatial, ck::index_t NumNonSpatialDim = 3>
|
||||
std::size_t
|
||||
GetFlops(const std::array<ck::index_t, NumDimSpatial + NumNonSpatialDim>& output_lengths,
|
||||
const std::array<ck::index_t, NumDimSpatial + NumNonSpatialDim>& weights_lengths,
|
||||
const std::size_t& ds_size)
|
||||
{
|
||||
// G * N * C * <output spatial lengths product> * (2 * K * <filter spatial lengths product> +
|
||||
// <number of scale factors>)
|
||||
ck::index_t G = weights_lengths[0];
|
||||
ck::index_t N = output_lengths[1];
|
||||
ck::index_t K = weights_lengths[1];
|
||||
ck::index_t C = weights_lengths[2];
|
||||
|
||||
return G * N * C *
|
||||
std::accumulate(std::next(std::begin(output_lengths), NumNonSpatialDim),
|
||||
std::end(output_lengths),
|
||||
static_cast<std::size_t>(1),
|
||||
std::multiplies<>()) *
|
||||
(static_cast<std::size_t>(2) * K *
|
||||
std::accumulate(std::next(std::begin(weights_lengths), NumNonSpatialDim),
|
||||
std::end(weights_lengths),
|
||||
static_cast<std::size_t>(1),
|
||||
std::multiplies<>()) +
|
||||
ds_size);
|
||||
}
|
||||
|
||||
template <ck::index_t NDimSpatial,
|
||||
typename InDataType,
|
||||
typename WeiDataType,
|
||||
typename CShuffleDataType,
|
||||
typename DsDataType,
|
||||
typename OutDataType,
|
||||
typename InElementOp,
|
||||
typename WeiElementOp,
|
||||
typename OutElementOp,
|
||||
typename DeviceConvNDFwdInstance>
|
||||
bool run_grouped_conv_fwd(bool do_verification,
|
||||
int init_method,
|
||||
bool time_kernel,
|
||||
const ck::utils::conv::ConvParam& conv_param,
|
||||
const HostTensorDescriptor& in_g_n_c_wis_desc,
|
||||
const HostTensorDescriptor& wei_g_k_c_xs_desc,
|
||||
const HostTensorDescriptor& out_g_n_k_wos_desc,
|
||||
const InElementOp& in_element_op,
|
||||
const WeiElementOp& wei_element_op)
|
||||
{
|
||||
Tensor<InDataType> in(in_g_n_c_wis_desc);
|
||||
Tensor<WeiDataType> wei(wei_g_k_c_xs_desc);
|
||||
Tensor<CShuffleDataType> c(out_g_n_k_wos_desc);
|
||||
Tensor<OutDataType> out_host(out_g_n_k_wos_desc);
|
||||
Tensor<OutDataType> out_device(out_g_n_k_wos_desc);
|
||||
|
||||
std::cout << "in: " << in.mDesc << std::endl;
|
||||
std::cout << "wei: " << wei.mDesc << std::endl;
|
||||
std::cout << "out: " << out_host.mDesc << std::endl;
|
||||
|
||||
switch(init_method)
|
||||
{
|
||||
case 0: break;
|
||||
case 1:
|
||||
in.GenerateTensorValue(GeneratorTensor_2<InDataType>{-5, 5});
|
||||
wei.GenerateTensorValue(GeneratorTensor_2<WeiDataType>{-5, 5});
|
||||
break;
|
||||
default:
|
||||
in.GenerateTensorValue(GeneratorTensor_3<InDataType>{0.0, 1.0});
|
||||
wei.GenerateTensorValue(GeneratorTensor_3<WeiDataType>{-0.5, 0.5});
|
||||
}
|
||||
|
||||
DeviceMem in_device_buf(sizeof(InDataType) * in.mDesc.GetElementSpaceSize());
|
||||
DeviceMem wei_device_buf(sizeof(WeiDataType) * wei.mDesc.GetElementSpaceSize());
|
||||
DeviceMem out_device_buf(sizeof(OutDataType) * out_device.mDesc.GetElementSpaceSize());
|
||||
|
||||
in_device_buf.ToDevice(in.mData.data());
|
||||
wei_device_buf.ToDevice(wei.mData.data());
|
||||
|
||||
std::array<ck::index_t, NDimSpatial + 3> a_g_n_c_wis_lengths{};
|
||||
std::array<ck::index_t, NDimSpatial + 3> a_g_n_c_wis_strides{};
|
||||
std::array<ck::index_t, NDimSpatial + 3> b_g_k_c_xs_lengths{};
|
||||
std::array<ck::index_t, NDimSpatial + 3> b_g_k_c_xs_strides{};
|
||||
std::array<ck::index_t, NDimSpatial + 3> e_g_n_k_wos_lengths{};
|
||||
std::array<ck::index_t, NDimSpatial + 3> e_g_n_k_wos_strides{};
|
||||
std::array<ck::index_t, NDimSpatial> conv_filter_strides{};
|
||||
std::array<ck::index_t, NDimSpatial> conv_filter_dilations{};
|
||||
std::array<ck::index_t, NDimSpatial> input_left_pads{};
|
||||
std::array<ck::index_t, NDimSpatial> input_right_pads{};
|
||||
|
||||
auto copy = [](const auto& x, auto& y) { ck::ranges::copy(x, y.begin()); };
|
||||
|
||||
copy(in_g_n_c_wis_desc.GetLengths(), a_g_n_c_wis_lengths);
|
||||
copy(in_g_n_c_wis_desc.GetStrides(), a_g_n_c_wis_strides);
|
||||
copy(wei_g_k_c_xs_desc.GetLengths(), b_g_k_c_xs_lengths);
|
||||
copy(wei_g_k_c_xs_desc.GetStrides(), b_g_k_c_xs_strides);
|
||||
copy(out_g_n_k_wos_desc.GetLengths(), e_g_n_k_wos_lengths);
|
||||
copy(out_g_n_k_wos_desc.GetStrides(), e_g_n_k_wos_strides);
|
||||
copy(conv_param.conv_filter_strides_, conv_filter_strides);
|
||||
copy(conv_param.conv_filter_dilations_, conv_filter_dilations);
|
||||
copy(conv_param.input_left_pads_, input_left_pads);
|
||||
copy(conv_param.input_right_pads_, input_right_pads);
|
||||
|
||||
// random scale values
|
||||
float scale_in = float(std::rand()) / float(RAND_MAX);
|
||||
float scale_wei = float(std::rand()) / float(RAND_MAX);
|
||||
float scale_out = float(std::rand()) / float(RAND_MAX);
|
||||
|
||||
// initialize out_element_op for each iteration
|
||||
const auto out_element_op = OutElementOp{scale_in, scale_wei, scale_out};
|
||||
|
||||
// do Conv
|
||||
auto conv = DeviceConvNDFwdInstance{};
|
||||
auto invoker = conv.MakeInvoker();
|
||||
auto argument = conv.MakeArgument(in_device_buf.GetDeviceBuffer(),
|
||||
wei_device_buf.GetDeviceBuffer(),
|
||||
std::array<const void*, 0>{},
|
||||
out_device_buf.GetDeviceBuffer(),
|
||||
a_g_n_c_wis_lengths,
|
||||
a_g_n_c_wis_strides,
|
||||
b_g_k_c_xs_lengths,
|
||||
b_g_k_c_xs_strides,
|
||||
std::array<std::array<ck::index_t, NDimSpatial + 3>, 0>{},
|
||||
std::array<std::array<ck::index_t, NDimSpatial + 3>, 0>{},
|
||||
e_g_n_k_wos_lengths,
|
||||
e_g_n_k_wos_strides,
|
||||
conv_filter_strides,
|
||||
conv_filter_dilations,
|
||||
input_left_pads,
|
||||
input_right_pads,
|
||||
in_element_op,
|
||||
wei_element_op,
|
||||
out_element_op);
|
||||
|
||||
if(!conv.IsSupportedArgument(argument))
|
||||
{
|
||||
throw std::runtime_error(
|
||||
"wrong! device_conv with the specified compilation parameters does "
|
||||
"not support this Conv problem");
|
||||
}
|
||||
|
||||
float avg_time = invoker.Run(argument, StreamConfig{nullptr, time_kernel});
|
||||
|
||||
std::size_t ds_size = 3; // 3 element-wise scale multipliers
|
||||
std::size_t flop = GetFlops<NDimSpatial>(e_g_n_k_wos_lengths, b_g_k_c_xs_lengths, ds_size);
|
||||
std::size_t num_btype = conv_param.GetInputByte<InDataType>() +
|
||||
conv_param.GetWeightByte<WeiDataType>() + sizeof(float) +
|
||||
sizeof(float) + sizeof(float) + conv_param.GetOutputByte<OutDataType>();
|
||||
|
||||
float tflops = static_cast<float>(flop) / 1.E9 / avg_time;
|
||||
float gb_per_sec = num_btype / 1.E6 / avg_time;
|
||||
std::cout << "Perf: " << avg_time << " ms, " << tflops << " TFlops, " << gb_per_sec << " GB/s, "
|
||||
<< conv.GetTypeString() << std::endl;
|
||||
|
||||
if(do_verification)
|
||||
{
|
||||
auto ref_conv = ck::tensor_operation::host::ReferenceConvFwd<NDimSpatial,
|
||||
InDataType,
|
||||
WeiDataType,
|
||||
CShuffleDataType,
|
||||
InElementOp,
|
||||
WeiElementOp,
|
||||
PassThrough>();
|
||||
|
||||
auto ref_invoker = ref_conv.MakeInvoker();
|
||||
auto ref_argument = ref_conv.MakeArgument(in,
|
||||
wei,
|
||||
c,
|
||||
conv_param.conv_filter_strides_,
|
||||
conv_param.conv_filter_dilations_,
|
||||
conv_param.input_left_pads_,
|
||||
conv_param.input_right_pads_,
|
||||
in_element_op,
|
||||
wei_element_op,
|
||||
PassThrough{});
|
||||
|
||||
ref_invoker.Run(ref_argument);
|
||||
|
||||
out_host.ForEach([&](auto&, auto idx) { out_element_op(out_host(idx), c(idx)); });
|
||||
|
||||
out_device_buf.FromDevice(out_device.mData.data());
|
||||
|
||||
return ck::utils::check_err(out_device,
|
||||
out_host,
|
||||
"Error: incorrect results!",
|
||||
get_rtol<OutDataType>(),
|
||||
get_atol<OutDataType>());
|
||||
}
|
||||
|
||||
return true;
|
||||
}
|
||||
@@ -0,0 +1,88 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "convnd_fwd_convscale_common.hpp"
|
||||
|
||||
#include "ck/tensor_operation/gpu/device/impl/device_grouped_conv_fwd_multiple_abd_xdl_cshuffle.hpp"
|
||||
|
||||
#include "ck/library/utility/convolution_host_tensor_descriptor_helper.hpp"
|
||||
|
||||
using InDataType = ck::f8_t;
|
||||
using WeiDataType = ck::f8_t;
|
||||
using AccDataType = float;
|
||||
using CShuffleDataType = float;
|
||||
using DsDataType = ck::Tuple<>;
|
||||
using OutDataType = ck::f8_t;
|
||||
using AComputeDataType = ck::f8_t;
|
||||
using BComputeDataType = ck::f8_t;
|
||||
|
||||
template <ck::index_t... Is>
|
||||
using S = ck::Sequence<Is...>;
|
||||
|
||||
using InElementOp = PassThrough;
|
||||
using WeiElementOp = PassThrough;
|
||||
using OutElementOp = ConvScale;
|
||||
|
||||
static constexpr auto ConvSpec =
|
||||
ck::tensor_operation::device::ConvolutionForwardSpecialization::Default;
|
||||
|
||||
static constexpr auto GemmSpec = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
|
||||
|
||||
template <ck::index_t NDimSpatial,
|
||||
typename InLayout,
|
||||
typename WeiLayout,
|
||||
typename DsLayout,
|
||||
typename OutLayout>
|
||||
using DeviceGroupedConvNDFwdInstance =
|
||||
ck::tensor_operation::device::DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<
|
||||
NDimSpatial,
|
||||
InLayout,
|
||||
WeiLayout,
|
||||
DsLayout,
|
||||
OutLayout,
|
||||
InDataType,
|
||||
WeiDataType,
|
||||
AccDataType,
|
||||
CShuffleDataType,
|
||||
DsDataType,
|
||||
OutDataType,
|
||||
InElementOp,
|
||||
WeiElementOp,
|
||||
OutElementOp,
|
||||
ConvSpec, // ConvForwardSpecialization
|
||||
GemmSpec, // GemmSpecialization
|
||||
1, //
|
||||
256, // BlockSize
|
||||
128, // MPerBlock
|
||||
256, // NPerBlock
|
||||
32, // KPerBlock
|
||||
8, // AK1
|
||||
8, // BK1
|
||||
32, // MPerXdl
|
||||
32, // NPerXdl
|
||||
2, // MXdlPerWave
|
||||
4, // NXdlPerWave
|
||||
S<4, 64, 1>, // ABlockTransferThreadClusterLengths_AK0_M_AK1
|
||||
S<1, 0, 2>, // ABlockTransferThreadClusterArrangeOrder
|
||||
S<1, 0, 2>, // ABlockTransferSrcAccessOrder
|
||||
2, // ABlockTransferSrcVectorDim
|
||||
8, // ABlockTransferSrcScalarPerVector
|
||||
8, // ABlockTransferDstScalarPerVector_AK1
|
||||
1, // ABlockLdsExtraM
|
||||
S<4, 64, 1>, // BBlockTransferThreadClusterLengths_BK0_N_BK1
|
||||
S<1, 0, 2>, // BBlockTransferThreadClusterArrangeOrder
|
||||
S<1, 0, 2>, // BBlockTransferSrcAccessOrder
|
||||
2, // BBlockTransferSrcVectorDim
|
||||
8, // BBlockTransferSrcScalarPerVector
|
||||
8, // BBlockTransferDstScalarPerVector_BK1
|
||||
1, // BBlockLdsExtraN
|
||||
1,
|
||||
1,
|
||||
S<1, 32, 1, 8>,
|
||||
8,
|
||||
AComputeDataType,
|
||||
BComputeDataType>;
|
||||
|
||||
#include "run_convnd_fwd_convscale_example.inc"
|
||||
|
||||
int main(int argc, char* argv[]) { return run_convnd_fwd_example(argc, argv) ? 0 : 1; }
|
||||
@@ -0,0 +1,104 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#pragma once
|
||||
|
||||
bool run_convnd_fwd_example(int argc, char* argv[])
|
||||
{
|
||||
print_helper_msg();
|
||||
|
||||
bool do_verification = true;
|
||||
int init_method = 1;
|
||||
bool time_kernel = false;
|
||||
|
||||
ck::utils::conv::ConvParam conv_param{
|
||||
2, 1, 128, 256, 192, {3, 3}, {71, 71}, {2, 2}, {1, 1}, {1, 1}, {1, 1}};
|
||||
|
||||
if(argc == 1)
|
||||
{
|
||||
// use default
|
||||
}
|
||||
else if(argc == 4)
|
||||
{
|
||||
do_verification = std::stoi(argv[1]);
|
||||
init_method = std::stoi(argv[2]);
|
||||
time_kernel = std::stoi(argv[3]);
|
||||
}
|
||||
else
|
||||
{
|
||||
do_verification = std::stoi(argv[1]);
|
||||
init_method = std::stoi(argv[2]);
|
||||
time_kernel = std::stoi(argv[3]);
|
||||
const ck::index_t num_dim_spatial = std::stoi(argv[4]);
|
||||
|
||||
conv_param = ck::utils::conv::parse_conv_param(num_dim_spatial, 5, argv);
|
||||
}
|
||||
|
||||
// instantiate in and wei element ops, will
|
||||
// instantiate out_element_op below for every iteration
|
||||
const auto in_element_op = InElementOp{};
|
||||
const auto wei_element_op = WeiElementOp{};
|
||||
|
||||
const auto run =
|
||||
[&](auto ndim_spatial, auto in_layout, auto wei_layout, auto ds_layout, auto out_layout) {
|
||||
constexpr ck::index_t ndim_spatial_value = ndim_spatial.value;
|
||||
|
||||
using InLayout = decltype(in_layout);
|
||||
using WeiLayout = decltype(wei_layout);
|
||||
using DsLayout = decltype(ds_layout);
|
||||
using OutLayout = decltype(out_layout);
|
||||
|
||||
const auto in_g_n_c_wis_desc =
|
||||
ck::utils::conv::make_input_host_tensor_descriptor_g_n_c_wis_packed<InLayout>(
|
||||
conv_param);
|
||||
|
||||
const auto wei_g_k_c_xs_desc =
|
||||
ck::utils::conv::make_weight_host_tensor_descriptor_g_k_c_xs_packed<WeiLayout>(
|
||||
conv_param);
|
||||
|
||||
const auto out_g_n_k_wos_desc =
|
||||
ck::utils::conv::make_output_host_tensor_descriptor_g_n_k_wos_packed<OutLayout>(
|
||||
conv_param);
|
||||
|
||||
return run_grouped_conv_fwd<ndim_spatial_value,
|
||||
InDataType,
|
||||
WeiDataType,
|
||||
CShuffleDataType,
|
||||
DsDataType,
|
||||
OutDataType,
|
||||
InElementOp,
|
||||
WeiElementOp,
|
||||
OutElementOp,
|
||||
DeviceGroupedConvNDFwdInstance<ndim_spatial_value,
|
||||
InLayout,
|
||||
WeiLayout,
|
||||
DsLayout,
|
||||
OutLayout>>(
|
||||
do_verification,
|
||||
init_method,
|
||||
time_kernel,
|
||||
conv_param,
|
||||
in_g_n_c_wis_desc,
|
||||
wei_g_k_c_xs_desc,
|
||||
out_g_n_k_wos_desc,
|
||||
in_element_op,
|
||||
wei_element_op);
|
||||
};
|
||||
|
||||
namespace ctc = ck::tensor_layout::convolution;
|
||||
|
||||
if(conv_param.num_dim_spatial_ == 1)
|
||||
{
|
||||
return run(ck::Number<1>{}, ctc::GNWC{}, ctc::GKXC{}, ck::Tuple<>{}, ctc::GNWK{});
|
||||
}
|
||||
else if(conv_param.num_dim_spatial_ == 2)
|
||||
{
|
||||
return run(ck::Number<2>{}, ctc::GNHWC{}, ctc::GKYXC{}, ck::Tuple<>{}, ctc::GNHWK{});
|
||||
}
|
||||
else if(conv_param.num_dim_spatial_ == 3)
|
||||
{
|
||||
return run(ck::Number<3>{}, ctc::GNDHWC{}, ctc::GKZYXC{}, ck::Tuple<>{}, ctc::GNDHWK{});
|
||||
}
|
||||
|
||||
return true;
|
||||
}
|
||||
Reference in New Issue
Block a user