mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-04-20 06:49:15 +00:00
Convolution FWD profiler refactor. (#183)
* Convolution ND * Code unification across dimensions for generating tensor descriptors. * Example * Instances * Move convnd f32 instance file to comply with repo structure. * Conv 1D tensor layouts. * Formatting and use ReferenceConv * Reference ConvFwd supporting 1D and 2D convolution. * Debug printing TensorLayout name. * Conv fwd 1D instance f32 * Refactor conv ND example. Needed to support various conv dimensio. Needed to support various conv dimensions * Rename conv nd example director to prevent conflicts. * Refactor some common utility to single file. Plus some tests. * Refactor GetHostTensorDescriptor + UT. * Add 1D test case. * Test reference convolution 1d/2d * Remove some leftovers. * Fix convolution example error for 1D * Refactor test check errors utility function. * Test Conv2D Fwd XDL * More UT for 1D case. * Parameterize input & weight initializers. * Rename example to prevent conflicts. * Split convnd instance into separate files for 1d/2d * Address review comments. * Fix data type for flops/gbytes calculations. * Assign example number 11. * 3D cases for convolution utility functions. * 3D reference convolution. * Add support for 3D convolution. * Check for inputs bigger than 2GB. * Formatting * Support for bf16/f16/f32/i8 - conv instances + UT. * Use check_err from test_util.hpp. * Split convnd test into separate files for each dim. * Fix data generation and use proper instances. * Formatting * Skip tensor initialization if not necessary. * Fix CMakefiles. * Remove redundant conv2d_fwd test. * Lower problem size for conv3D UT. * 3D case for convnd example. * Remove leftovers after merge. * Add Conv Specialization string to GetTypeString * Skip instance causing numerical errors. * Small fixes. * Remove redundant includes. * Fix namespace name error. * Script for automatic testing and logging convolution fwd UTs * Comment out numactl cmd. * Refine weights initalization and relax rtol for fp16 * Move test_util.hpp to check_err.hpp * Refine weights initalization and relax rtol for fp16 * Refactor common part of test conv utils. * Move utility function to single common place. * Add additional common functions to utility. * Refactor convnd_fwd_xdl examples. * Remove redundant files. * Unify structure. * Add constructor to ConvParams. * And add input parameters validation. * Modify conv examples to use single utility file. * Remove check_error from host_tensor.hpp * Get rid of check_indices function. * Remove bf16_to_f32 function overload for scalars. * Fix namespace. * Add half_float::half for check_err. * Fix conv params size in UT. * Fix weights initialization for int8. * Fix weights initialization for int8. * Add type_convert when store output in ref conv 1D. * Get back old conv2d_fwd_xdl operation. * Silence conv debug print. * format * clean * clean * Fix merge. * Fix namespace for check_err * Formatting. * Fix merge artifacts. * Remove deleted header. * Fix some includes and use ck::utils::check_err. * Remove unused check_indices restored by previous merge. * Fix namespaces after merge. * Fix compilation error. * Small fixes. * Use common functions. * Fix filename * Fix namespaces. * Fix merge artifact - retrieve removed by accident fun. * Fix ConvForwardSpecialization. * Working example of OpInstanceRunEngine for conv2dfwd UT. * Adhere to coding style rules. * Formatting and adhere to coding style rules. * Fix merge artifacts. * Utility for collecting conv fwd instances. + Plus commmon part for parsing cmdline params. * Refactor FillUniform because of segfault for int8_t. * Naming convention. * Elegant version of device mem allocation. * Use OpInstanceRunEngine in conv fwd nd tests. * Multiple refinements. * conditional init * don't run reference op if not provided. * Use OpInstanceRunEngine for ckProfiler conv_fwd * Refactor common tensor fill function to separate file. * Clean up unused functions. * Support different init methods. * Create CMake target for conv_fwd_util. * Add header for profile_convnd_fwd.cpp * Fix CMakefiles to link with conv_fwd_util where needed. * Fix some clutter. Co-authored-by: Adam Osewski <aosewski@amd.com> Co-authored-by: Chao Liu <chao.liu2@amd.com>
This commit is contained in:
Adam Osewski
@@ -1,4 +1,3 @@
|
||||
#include <algorithm>
|
||||
#include <cmath>
|
||||
#include <cstdlib>
|
||||
#include <half.hpp>
|
||||
@@ -10,6 +9,7 @@
|
||||
#include "config.hpp"
|
||||
#include "conv_fwd_util.hpp"
|
||||
#include "element_wise_operation.hpp"
|
||||
#include "fill.hpp"
|
||||
#include "host_tensor.hpp"
|
||||
#include "reference_conv_fwd.hpp"
|
||||
#include "tensor_layout.hpp"
|
||||
@@ -19,35 +19,6 @@ using InElementOp = ck::tensor_operation::element_wise::PassThrough;
|
||||
using WeiElementOp = ck::tensor_operation::element_wise::PassThrough;
|
||||
using OutElementOp = ck::tensor_operation::element_wise::PassThrough;
|
||||
|
||||
template <typename T>
|
||||
struct FillMonotonicSeq
|
||||
{
|
||||
T m_init_value{0};
|
||||
T m_step{1};
|
||||
|
||||
template <typename ForwardIter>
|
||||
void operator()(ForwardIter first, ForwardIter last) const
|
||||
{
|
||||
std::generate(first, last, [=, n = m_init_value]() mutable {
|
||||
auto tmp = n;
|
||||
n += m_step;
|
||||
return tmp;
|
||||
});
|
||||
}
|
||||
};
|
||||
|
||||
template <typename T>
|
||||
struct FillConstant
|
||||
{
|
||||
T m_value{0};
|
||||
|
||||
template <typename ForwardIter>
|
||||
void operator()(ForwardIter first, ForwardIter last) const
|
||||
{
|
||||
std::fill(first, last, m_value);
|
||||
}
|
||||
};
|
||||
|
||||
template <ck::index_t NDim,
|
||||
typename InDataType = float,
|
||||
typename WeiDataType = float,
|
||||
@@ -55,8 +26,8 @@ template <ck::index_t NDim,
|
||||
typename InLayout = ck::tensor_layout::convolution::NHWC,
|
||||
typename WeiLayout = ck::tensor_layout::convolution::KYXC,
|
||||
typename OutLayout = ck::tensor_layout::convolution::NHWK,
|
||||
typename FillInputOp = FillMonotonicSeq<InDataType>,
|
||||
typename FillWeightsOp = FillConstant<WeiDataType>>
|
||||
typename FillInputOp = ck::utils::FillMonotonicSeq<InDataType>,
|
||||
typename FillWeightsOp = ck::utils::FillConstant<WeiDataType>>
|
||||
Tensor<OutDataType>
|
||||
run_reference_convolution_forward(const ck::utils::conv::ConvParams& params,
|
||||
const FillInputOp& fill_input_op = FillInputOp{},
|
||||
@@ -251,7 +222,7 @@ bool test_conv1d_nwc()
|
||||
ck::tensor_layout::convolution::NWC,
|
||||
ck::tensor_layout::convolution::KXC,
|
||||
ck::tensor_layout::convolution::NWK>(
|
||||
params, FillMonotonicSeq<float>{0.f, 0.1f});
|
||||
params, ck::utils::FillMonotonicSeq<float>{0.f, 0.1f});
|
||||
|
||||
ref_dims = std::vector<std::size_t>{2, 16, 16};
|
||||
ref_data = std::vector<float>{
|
||||
@@ -349,7 +320,7 @@ bool test_conv3d_ncdhw()
|
||||
ck::tensor_layout::convolution::NCDHW,
|
||||
ck::tensor_layout::convolution::KCZYX,
|
||||
ck::tensor_layout::convolution::NKDHW>(
|
||||
params, FillMonotonicSeq<float>{0.f, 0.1f});
|
||||
params, ck::utils::FillMonotonicSeq<float>{0.f, 0.1f});
|
||||
std::vector<std::size_t> ref_dims{1, 1, 4, 4, 4};
|
||||
std::vector<float> ref_data{
|
||||
407.7, 410.40002, 413.09998, 415.80002, 423.90002, 426.6, 429.30002, 432.,
|
||||
@@ -383,7 +354,7 @@ bool test_conv3d_ncdhw()
|
||||
ck::tensor_layout::convolution::NCDHW,
|
||||
ck::tensor_layout::convolution::KCZYX,
|
||||
ck::tensor_layout::convolution::NKDHW>(
|
||||
params, FillMonotonicSeq<float>{0.f, 0.1f});
|
||||
params, ck::utils::FillMonotonicSeq<float>{0.f, 0.1f});
|
||||
ref_dims = std::vector<std::size_t>{1, 2, 4, 4, 4};
|
||||
ref_data = std::vector<float>{
|
||||
2756.7002, 2764.7998, 2772.9001, 2781., 2853.9001, 2862., 2870.1, 2878.2002,
|
||||
|
||||
Reference in New Issue
Block a user