Skip to content

GitLab

Unverified Commit 984889fb authored by Illia Silin's avatar Illia Silin Committed by GitHub
Browse files

Merge pull request #13 from ROCmSoftwarePlatform/merge_from-_public_repo 

Merge from  public repo
parents a73ab0d8 4a106f7d
Hide whitespace changes
Inline Side-by-side
Showing with 185 additions and 58 deletions
example/41_grouped_conv_conv_fwd/CMakeLists.txt
View file @ 984889fb
add_example_executable(example_grouped_conv_conv_fwd_xdl_fp32 grouped_conv_conv_fwd_xdl_fp32.cpp) list(APPEND gpu_list1 gfx908 gfx90a gfx940 gfx941 gfx942)
add_example_executable(example_grouped_conv_conv_fwd_xdl_fp16 grouped_conv_conv_fwd_xdl_fp16.cpp) list(APPEND gpu_list2 gfx908 gfx90a)
add_example_executable(example_grouped_conv_conv_fwd_xdl_bf16 grouped_conv_conv_fwd_xdl_bf16.cpp) set(target 0)
if(NOT GPU_TARGETS MATCHES "gfx940") foreach(gpu IN LISTS GPU_TARGETS)
add_example_executable(example_grouped_conv_conv_fwd_xdl_int8 grouped_conv_conv_fwd_xdl_int8.cpp) if(gpu IN_LIST gpu_list1 AND target EQUAL 0)
add_example_executable(example_grouped_conv_conv_fwd_xdl_fp32 grouped_conv_conv_fwd_xdl_fp32.cpp)
add_example_executable(example_grouped_conv_conv_fwd_xdl_fp16 grouped_conv_conv_fwd_xdl_fp16.cpp)
add_example_executable(example_grouped_conv_conv_fwd_xdl_bf16 grouped_conv_conv_fwd_xdl_bf16.cpp)
if(USE_BITINT_EXTENSION_INT4)
add_example_executable(example_grouped_conv_conv_fwd_xdl_int4 grouped_conv_conv_fwd_xdl_int4.cpp)
endif(USE_BITINT_EXTENSION_INT4)
set(target 1)
endif()
endforeach()
if(NOT GPU_TARGETS MATCHES "gfx94" AND NOT GPU_TARGETS MATCHES "gfx1")
add_example_executable(example_grouped_conv_conv_fwd_xdl_int8 grouped_conv_conv_fwd_xdl_int8.cpp)
endif() endif()
if(USE_BITINT_EXTENSION_INT4)
add_example_executable(example_grouped_conv_conv_fwd_xdl_int4 grouped_conv_conv_fwd_xdl_int4.cpp)
endif(USE_BITINT_EXTENSION_INT4)
example/41_grouped_conv_conv_fwd/grouped_conv_conv_fwd_xdl_bf16.cpp
View file @ 984889fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#include <cstdlib> #include <cstdlib>
#include <iostream> #include <iostream>
......
example/41_grouped_conv_conv_fwd/grouped_conv_conv_fwd_xdl_fp16.cpp
View file @ 984889fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#include <cstdlib> #include <cstdlib>
#include <iostream> #include <iostream>
......
example/41_grouped_conv_conv_fwd/grouped_conv_conv_fwd_xdl_fp32.cpp
View file @ 984889fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#include <cstdlib> #include <cstdlib>
#include <iostream> #include <iostream>
......
example/41_grouped_conv_conv_fwd/grouped_conv_conv_fwd_xdl_int4.cpp
View file @ 984889fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#ifndef CK_EXPERIMENTAL_BIT_INT_EXTENSION_INT4 #ifndef CK_EXPERIMENTAL_BIT_INT_EXTENSION_INT4
#error Should compile this file with ck::int4_t support #error Should compile this file with ck::int4_t support
......
example/41_grouped_conv_conv_fwd/grouped_conv_conv_fwd_xdl_int8.cpp
View file @ 984889fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#include <cstdlib> #include <cstdlib>
#include <iostream> #include <iostream>
......
example/41_grouped_conv_conv_fwd/run_grouped_conv_conv_fwd_example.inc
View file @ 984889fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once #pragma once
......
example/42_groupnorm/CMakeLists.txt
View file @ 984889fb
add_example_executable(example_groupnorm_sigmoid_mul_fp16 groupnorm_sigmoid_mul_fp16.cpp) add_example_executable(example_groupnorm_sigmoid_mul_fp16 groupnorm_sigmoid_mul_fp16.cpp)
add_example_executable(example_groupnorm_splitk_fp16 groupnorm_splitk_fp16.cpp)
add_example_executable(example_groupnorm_swish_fp16 groupnorm_swish_fp16.cpp) add_example_executable(example_groupnorm_swish_fp16 groupnorm_swish_fp16.cpp)
example/42_groupnorm/common.hpp
View file @ 984889fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once #pragma once
...@@ -12,6 +12,7 @@ ...@@ -12,6 +12,7 @@
#include "ck/ck.hpp" #include "ck/ck.hpp"
#include "ck/utility/reduction_enums.hpp" #include "ck/utility/reduction_enums.hpp"
#include "ck/tensor_operation/gpu/device/impl/device_normalization_impl.hpp" #include "ck/tensor_operation/gpu/device/impl/device_normalization_impl.hpp"
#include "ck/tensor_operation/gpu/device/impl/device_normalization_splitk_impl.hpp"
#include "ck/tensor_operation/gpu/device/reduction_operator_mapping.hpp" #include "ck/tensor_operation/gpu/device/reduction_operator_mapping.hpp"
#include "ck/library/utility/fill.hpp" #include "ck/library/utility/fill.hpp"
......
example/42_groupnorm/groupnorm_sigmoid_mul_fp16.cpp
View file @ 984889fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#include "common.hpp" #include "common.hpp"
constexpr int Rank = 5; constexpr int Rank = 5;
constexpr int NumReduceDim = 3; constexpr int NumReduceDim = 3;
using XDataType = ck::half_t; using XDataType = ck::half_t;
using GammaDataType = ck::half_t; using GammaDataType = ck::half_t;
using BetaDataType = ck::half_t; using BetaDataType = ck::half_t;
using YDataType = ck::half_t; using YDataType = ck::half_t;
using ComputeDataType = float; using SaveMeanInvStdDataType = float;
using ComputeDataType = float;
#define SAVE_MEAN_INV_STD
struct YElementOp struct YElementOp
{ {
template <typename T> template <typename Y, typename X>
__host__ __device__ void operator()(T& y, const T& x) const __host__ __device__ void operator()(Y& y, const X& x) const
{ {
static_assert(ck::is_same<T, float>::value || ck::is_same<T, double>::value || static_assert(ck::is_same<X, float>::value || ck::is_same<X, double>::value ||
ck::is_same<T, ck::half_t>::value, ck::is_same<X, ck::half_t>::value,
"Data type is not supported by this operation!");
static_assert(ck::is_same<Y, float>::value || ck::is_same<Y, double>::value ||
ck::is_same<Y, ck::half_t>::value,
"Data type is not supported by this operation!"); "Data type is not supported by this operation!");
T a; X a;
ck::tensor_operation::element_wise::Sigmoid{}(a, x); ck::tensor_operation::element_wise::Sigmoid{}(a, x);
y = x * a; y = ck::type_convert<Y>(x * a);
}; };
}; };
...@@ -35,6 +42,7 @@ using DeviceInstance = ...@@ -35,6 +42,7 @@ using DeviceInstance =
BetaDataType, BetaDataType,
ComputeDataType, ComputeDataType,
YDataType, YDataType,
SaveMeanInvStdDataType,
YElementOp, YElementOp,
Rank, Rank,
NumReduceDim, NumReduceDim,
...@@ -49,7 +57,8 @@ using DeviceInstance = ...@@ -49,7 +57,8 @@ using DeviceInstance =
2, // GammaScalarPerVector 2, // GammaScalarPerVector
1, // BetaVecDim (0=M, 1=K) 1, // BetaVecDim (0=M, 1=K)
2, // BetaScalarPerVector 2, // BetaScalarPerVector
2>; // OutScalarPerVector 2, // YScalarPerVector
1>; // SaveMeanInvStdScalarPerVector
#include "run_groupnorm_example.inc" #include "run_groupnorm_example.inc"
......
example/42_groupnorm/groupnorm_splitk_fp16.cpp 0 → 100644
View file @ 984889fb
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#include "common.hpp"
constexpr int Rank = 5;
constexpr int NumReduceDim = 3;
using XDataType = ck::half_t;
using GammaDataType = ck::half_t;
using BetaDataType = ck::half_t;
using YDataType = ck::half_t;
using SaveMeanInvStdDataType = float;
using ComputeDataType = float;
using YElementOp = ck::tensor_operation::element_wise::Swish;
#define SAVE_MEAN_INV_STD
using DeviceInstance =
ck::tensor_operation::device::DeviceNormalizationSplitKImpl<XDataType,
GammaDataType,
BetaDataType,
ComputeDataType,
YDataType,
SaveMeanInvStdDataType,
YElementOp,
Rank,
NumReduceDim,
256, // BlockSize
1, // ClusterM
256, // ClusterK
1, // SliceM
16, // SliceK
1, // SrcVecDim (0=M, 1=K)
2, // SrcScalarPerVector
1, // GammaVecDim (0=M, 1=K)
2, // GammaScalarPerVector
1, // BetaVecDim (0=M, 1=K)
2, // BetaScalarPerVector
2, // YScalarPerVector
1>; // SaveMeanInvStdScalarPerVector
#include "run_groupnorm_example.inc"
int main(int argc, char* argv[]) { run_groupnorm_example(argc, argv); }
example/42_groupnorm/groupnorm_swish_fp16.cpp
View file @ 984889fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#include "common.hpp" #include "common.hpp"
constexpr int Rank = 5; constexpr int Rank = 5;
constexpr int NumReduceDim = 3; constexpr int NumReduceDim = 3;
using XDataType = ck::half_t; using XDataType = ck::half_t;
using GammaDataType = ck::half_t; using GammaDataType = ck::half_t;
using BetaDataType = ck::half_t; using BetaDataType = ck::half_t;
using YDataType = ck::half_t; using YDataType = ck::half_t;
using ComputeDataType = float; using SaveMeanInvStdDataType = float;
using YElementOp = ck::tensor_operation::element_wise::Swish; using ComputeDataType = float;
using YElementOp = ck::tensor_operation::element_wise::Swish;
#define SAVE_MEAN_INV_STD
using DeviceInstance = using DeviceInstance =
ck::tensor_operation::device::DeviceNormalizationImpl<XDataType, ck::tensor_operation::device::DeviceNormalizationImpl<XDataType,
...@@ -19,6 +22,7 @@ using DeviceInstance = ...@@ -19,6 +22,7 @@ using DeviceInstance =
BetaDataType, BetaDataType,
ComputeDataType, ComputeDataType,
YDataType, YDataType,
SaveMeanInvStdDataType,
YElementOp, YElementOp,
Rank, Rank,
NumReduceDim, NumReduceDim,
...@@ -33,7 +37,8 @@ using DeviceInstance = ...@@ -33,7 +37,8 @@ using DeviceInstance =
2, // GammaScalarPerVector 2, // GammaScalarPerVector
1, // BetaVecDim (0=M, 1=K) 1, // BetaVecDim (0=M, 1=K)
2, // BetaScalarPerVector 2, // BetaScalarPerVector
2>; // OutScalarPerVector 2, // YScalarPerVector
1>; // SaveMeanInvStdScalarPerVector
#include "run_groupnorm_example.inc" #include "run_groupnorm_example.inc"
......
example/42_groupnorm/run_groupnorm_example.inc
View file @ 984889fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once #pragma once
...@@ -34,6 +34,8 @@ int run_groupnorm_example(int argc, char* argv[]) ...@@ -34,6 +34,8 @@ int run_groupnorm_example(int argc, char* argv[])
Tensor<YDataType> y({N, H, W, G, C}); Tensor<YDataType> y({N, H, W, G, C});
Tensor<GammaDataType> gamma({G, C}); Tensor<GammaDataType> gamma({G, C});
Tensor<BetaDataType> beta({G, C}); Tensor<BetaDataType> beta({G, C});
Tensor<SaveMeanInvStdDataType> save_mean({N, G});
Tensor<SaveMeanInvStdDataType> save_inv_std({N, G});
ck::utils::FillUniformDistribution<XDataType>{0.f, 1.f}(x); ck::utils::FillUniformDistribution<XDataType>{0.f, 1.f}(x);
ck::utils::FillUniformDistribution<GammaDataType>{0.f, 1.f}(gamma); ck::utils::FillUniformDistribution<GammaDataType>{0.f, 1.f}(gamma);
...@@ -43,6 +45,11 @@ int run_groupnorm_example(int argc, char* argv[]) ...@@ -43,6 +45,11 @@ int run_groupnorm_example(int argc, char* argv[])
DeviceMem gamma_dev(sizeof(GammaDataType) * gamma.mDesc.GetElementSpaceSize()); DeviceMem gamma_dev(sizeof(GammaDataType) * gamma.mDesc.GetElementSpaceSize());
DeviceMem beta_dev(sizeof(BetaDataType) * beta.mDesc.GetElementSpaceSize()); DeviceMem beta_dev(sizeof(BetaDataType) * beta.mDesc.GetElementSpaceSize());
DeviceMem y_dev(sizeof(YDataType) * y.mDesc.GetElementSpaceSize()); DeviceMem y_dev(sizeof(YDataType) * y.mDesc.GetElementSpaceSize());
#ifdef SAVE_MEAN_INV_STD
DeviceMem save_mean_dev(sizeof(SaveMeanInvStdDataType) * save_mean.mDesc.GetElementSpaceSize());
DeviceMem save_inv_std_dev(sizeof(SaveMeanInvStdDataType) *
save_inv_std.mDesc.GetElementSpaceSize());
#endif
x_dev.ToDevice(x.mData.data()); x_dev.ToDevice(x.mData.data());
gamma_dev.ToDevice(gamma.mData.data()); gamma_dev.ToDevice(gamma.mData.data());
...@@ -57,14 +64,23 @@ int run_groupnorm_example(int argc, char* argv[]) ...@@ -57,14 +64,23 @@ int run_groupnorm_example(int argc, char* argv[])
{0, 0, 0, C, 1}, {0, 0, 0, C, 1},
{0, 0, 0, C, 1}, {0, 0, 0, C, 1},
std::vector<ck::index_t>{y.mDesc.GetStrides().begin(), y.mDesc.GetStrides().end()}, std::vector<ck::index_t>{y.mDesc.GetStrides().begin(), y.mDesc.GetStrides().end()},
std::vector<ck::index_t>{save_mean.mDesc.GetStrides().begin(),
save_mean.mDesc.GetStrides().end()},
std::vector<ck::index_t>{save_mean.mDesc.GetStrides().begin(),
save_mean.mDesc.GetStrides().end()},
{1, 2, 4}, // reduction dimension: [H, W, C] {1, 2, 4}, // reduction dimension: [H, W, C]
1e-6, 1e-6,
x_dev.GetDeviceBuffer(), x_dev.GetDeviceBuffer(),
gamma_dev.GetDeviceBuffer(), gamma_dev.GetDeviceBuffer(),
beta_dev.GetDeviceBuffer(), beta_dev.GetDeviceBuffer(),
y_dev.GetDeviceBuffer(), y_dev.GetDeviceBuffer(),
#ifdef SAVE_MEAN_INV_STD
save_mean_dev.GetDeviceBuffer(),
save_inv_std_dev.GetDeviceBuffer(),
#else
nullptr, nullptr,
nullptr, nullptr,
#endif
y_element_op); y_element_op);
if(!device_instance.IsSupportedArgument(argument_ptr.get())) if(!device_instance.IsSupportedArgument(argument_ptr.get()))
...@@ -73,6 +89,10 @@ int run_groupnorm_example(int argc, char* argv[]) ...@@ -73,6 +89,10 @@ int run_groupnorm_example(int argc, char* argv[])
return 1; return 1;
}; };
size_t workspace_sz = device_instance.GetWorkSpaceSize(argument_ptr.get());
DeviceMem workspace_dev(workspace_sz);
device_instance.SetWorkSpacePointer(argument_ptr.get(), workspace_dev.GetDeviceBuffer());
auto invoker_ptr = device_instance.MakeInvokerPointer(); auto invoker_ptr = device_instance.MakeInvokerPointer();
float ave_time = invoker_ptr->Run(argument_ptr.get(), StreamConfig{nullptr, true, true}); float ave_time = invoker_ptr->Run(argument_ptr.get(), StreamConfig{nullptr, true, true});
...@@ -88,21 +108,40 @@ int run_groupnorm_example(int argc, char* argv[]) ...@@ -88,21 +108,40 @@ int run_groupnorm_example(int argc, char* argv[])
bool pass = true; bool pass = true;
{ {
Tensor<YDataType> host_y({N, H, W, G, C}); Tensor<YDataType> host_y({N, H, W, G, C});
using ReferenceInstance = ck::tensor_operation::host::ReferenceGroupnorm<XDataType, Tensor<SaveMeanInvStdDataType> host_save_mean(HostTensorDescriptor{N, G});
GammaDataType, Tensor<SaveMeanInvStdDataType> host_save_inv_std(HostTensorDescriptor{N, G});
BetaDataType, using ReferenceInstance =
YDataType, ck::tensor_operation::host::ReferenceGroupnorm<XDataType,
ComputeDataType, GammaDataType,
YElementOp>; BetaDataType,
YDataType,
SaveMeanInvStdDataType,
ComputeDataType,
YElementOp>;
ReferenceInstance ref; ReferenceInstance ref;
auto ref_argument = auto ref_argument = ref.MakeArgument(x,
ref.MakeArgument(x, gamma, beta, host_y, y_element_op, {N, H, W, G, C}, 1e-6); gamma,
auto ref_invoker = ref.MakeInvoker(); beta,
host_y,
host_save_mean,
host_save_inv_std,
y_element_op,
{N, H, W, G, C},
1e-6);
auto ref_invoker = ref.MakeInvoker();
ref_invoker.Run(ref_argument); ref_invoker.Run(ref_argument);
y_dev.FromDevice(y.mData.data()); y_dev.FromDevice(y.mData.data());
pass &= ck::utils::check_err(y, host_y, "Error: Incorrect results", 1e-3, 1e-3); pass &= ck::utils::check_err(y, host_y, "Error: Incorrect results", 1e-3, 1e-3);
#ifdef SAVE_MEAN_INV_STD
save_mean_dev.FromDevice(save_mean.mData.data());
save_inv_std_dev.FromDevice(save_inv_std.mData.data());
pass &= ck::utils::check_err(
save_mean, host_save_mean, "Error: Incorrect results (mean)", 1e-3, 1e-3);
pass &= ck::utils::check_err(
save_inv_std, host_save_inv_std, "Error: Incorrect results (inv_std)", 1e-3, 1e-3);
#endif
} }
return (pass ? 0 : 1); return (pass ? 0 : 1);
......
example/43_splitk_gemm_bias_e_permute/splitk_gemm_bias_e_permute_xdl_fp16.cpp
View file @ 984889fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#include <iostream> #include <iostream>
#include <numeric> #include <numeric>
...@@ -8,7 +8,7 @@ ...@@ -8,7 +8,7 @@
#include "ck/ck.hpp" #include "ck/ck.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp" #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/device/device_splitk_contraction_multiple_d_xdl_cshuffle.hpp" #include "ck/tensor_operation/gpu/device/impl/device_splitk_contraction_multiple_d_xdl_cshuffle.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp" #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
#include "ck/library/utility/check_err.hpp" #include "ck/library/utility/check_err.hpp"
......
example/43_splitk_gemm_bias_e_permute/splitk_gemm_bias_e_permute_xdl_fp32.cpp
View file @ 984889fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#include <iostream> #include <iostream>
#include <numeric> #include <numeric>
...@@ -8,7 +8,7 @@ ...@@ -8,7 +8,7 @@
#include "ck/ck.hpp" #include "ck/ck.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp" #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/device/device_splitk_contraction_multiple_d_xdl_cshuffle.hpp" #include "ck/tensor_operation/gpu/device/impl/device_splitk_contraction_multiple_d_xdl_cshuffle.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp" #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
#include "ck/library/utility/check_err.hpp" #include "ck/library/utility/check_err.hpp"
......
example/45_elementwise_normalization/elementwise_layernorm_blockwise.cpp
View file @ 984889fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#include <iostream> #include <iostream>
#include <numeric> #include <numeric>
...@@ -167,20 +167,31 @@ int main() ...@@ -167,20 +167,31 @@ int main()
XElementwiseOperation>(x, a, b, mn, XElementwiseOperation{}); XElementwiseOperation>(x, a, b, mn, XElementwiseOperation{});
Tensor<YDataType> host_y(f_host_tensor_descriptor2d(M, N, Stride)); Tensor<YDataType> host_y(f_host_tensor_descriptor2d(M, N, Stride));
Tensor<AccDataType> host_save_mean({M});
Tensor<AccDataType> host_save_inv_std({M});
using ReferenceInstance = using ReferenceInstance =
ck::tensor_operation::host::ReferenceLayernorm<XDataType, ck::tensor_operation::host::ReferenceLayernorm<XDataType,
GammaDataType, GammaDataType,
BetaDataType, BetaDataType,
YDataType, YDataType,
AccDataType, AccDataType,
AccDataType,
YElementwiseOperation, YElementwiseOperation,
Rank, Rank,
NumReduceDim>; NumReduceDim>;
ReferenceInstance ref; ReferenceInstance ref;
auto ref_argument = auto ref_argument = ref.MakeArgument(x,
ref.MakeArgument(x, gamma, beta, host_y, YElementwiseOperation{}, {M, N}, {1}, 1e-4); gamma,
auto ref_invoker = ref.MakeInvoker(); beta,
host_y,
host_save_mean,
host_save_inv_std,
YElementwiseOperation{},
{M, N},
{1},
1e-4);
auto ref_invoker = ref.MakeInvoker();
ref_invoker.Run(ref_argument); ref_invoker.Run(ref_argument);
y_dev.FromDevice(y.mData.data()); y_dev.FromDevice(y.mData.data());
......
example/46_gemm_add_multiply/common.hpp
View file @ 984889fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once #pragma once
......
example/46_gemm_add_multiply/gemm_add_multiply_dl_fp16.cpp
View file @ 984889fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#include "common.hpp" #include "common.hpp"
#include "ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_dl.hpp" #include "ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_dl.hpp"
......
example/46_gemm_add_multiply/gemm_add_multiply_xdl_fp16.cpp
View file @ 984889fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#include "common.hpp" #include "common.hpp"
#include "ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_xdl_cshuffle.hpp" #include "ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_xdl_cshuffle.hpp"
......
example/47_gemm_bias_softmax_gemm_permute/CMakeLists.txt
View file @ 984889fb
add_example_executable(example_gemm_bias_softmax_gemm_permute gemm_bias_softmax_gemm_permute.cpp) 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_example_executable(example_gemm_bias_softmax_gemm_permute gemm_bias_softmax_gemm_permute.cpp)
set(target 1)
endif()
endforeach()
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment