提交 f75815b7 编写于 作者: N nhzlx

add prelu gpu inference

上级 1b9753d1
develop 2.0.1-rocm-post Ligoml-patch-1 OliverLPH-patch-1 OliverLPH-patch-2 PaddlePM-patch-1 PaddlePM-patch-2 ZHUI-patch-1 add_default_att add_model_benchmark_ci add_some_yaml_config addfile all_new_design_exec ascendrc ascendrelease cherry_undefined_var compile_windows delete_2.0.1-rocm-post delete_add_default_att delete_all_new_design_exec delete_ascendrc delete_compile_windows delete_delete_addfile delete_disable_iterable_dataset_unittest delete_fix_dataloader_memory_leak delete_fix_imperative_dygraph_error delete_fix_retry_ci delete_fix_undefined_var delete_improve_sccache delete_incubate/lite delete_paddle_tiny_install delete_paralleltest delete_prv-disable-more-cache delete_revert-31068-fix_conv3d_windows delete_revert-31562-mean delete_revert-33630-bug-fix delete_revert-34159-add_npu_bce_logical_dev delete_revert-34910-spinlocks_for_allocator delete_revert-35069-revert-34910-spinlocks_for_allocator delete_revert-36057-dev/read_flags_in_ut dingjiaweiww-patch-1 disable_iterable_dataset_unittest dy2static enable_eager_model_test final_state_gen_python_c final_state_intermediate fix-numpy-issue fix_concat_slice fix_dataloader_memory_leak fix_imperative_dygraph_error fix_npu_ci fix_op_flops fix_retry_ci fix_rnn_docs fix_tensor_type fix_undefined_var fixiscan fixiscan1 fixiscan2 fixiscan3 github/fork/123malin/netifaces github/fork/123malin/tdm_abacus github/fork/AshburnLee/dev_unique github/fork/ForFishes/fix_memory_matmul github/fork/ForFishes/rm_fluid github/fork/LielinJiang/move-2.0-api github/fork/LielinJiang/visual-dl-cb github/fork/LiuChiachi/add-transformer-generate-square-subsequent-mask-api github/fork/LiuChiachi/fix-example-code-for-hapi-Model github/fork/LiuChiachi/remove-input-requirment-in-dygraph-Model github/fork/MrChengmo/fix_ps_profiler github/fork/MrChengmo/update_ps_heter github/fork/PWhiddy/patch-1 github/fork/Shixiaowei02/dev/save_load_upgrade github/fork/TCChenlong/fix_hapi github/fork/TCChenlong/fix_inden github/fork/Thunderbrook/xpu_slice github/fork/XieYunshen/disable_ut_test_parallel_executor_fetch_isolated_var github/fork/XieYunshen/disable_ut_test_parallel_executor_fetch_isolated_var_2 github/fork/XieYunshen/disable_ut_test_parallel_executor_fetch_isolated_var_3 github/fork/XieYunshen/timeout_20S_ut github/fork/ZeyuChen/remove-nltk github/fork/arlesniak/arlesniak/selective__mkldnn_flags github/fork/baiyfbupt/code_doc_mig github/fork/chalsliu/set_timeout github/fork/chen-zhiyu/develop github/fork/chenwhql/ci/try_to_find_test_buffer_shared_memory_reuse_pass_error github/fork/chenwhql/dygraph/remove_scale_loss_and_apply_collective_grads github/fork/chenwhql/saveload/add_get_inference_program github/fork/chenwhql/saveload/remove_save_load_config github/fork/cryoco/pass-compatibility-trt github/fork/danleifeng/isempty_api2.0 github/fork/frankwhzhang/api_transfer github/fork/hbwx24/error_msg/cuda_kernel_error_msg github/fork/heavengate/cherry_yolo_box github/fork/heavengate/update_yolo_box github/fork/iclementine/rnn_fix github/fork/iducn/testestse github/fork/jczaja/prv-25537-fix github/fork/jeff41404/release/1.8 github/fork/jiweibo/api_2.0 github/fork/jiweibo/fix_lite_resnet50_test github/fork/juncaipeng/fix_doc_1 github/fork/lfchener/sample_code github/fork/littletomatodonkey/fix_reg_doc github/fork/liym27/dy2stat_update_assign_to_rc20 github/fork/luotao1/profiler_ut github/fork/mapingshuo/add_wait github/fork/mapingshuo/doc_2.0 github/fork/mapingshuo/zero-0.5 github/fork/miraiwk/dev github/fork/pangyoki/add-Categorical-class-branch github/fork/pangyoki/add-multinomial-op-branch github/fork/pangyoki/fix-test_distritbution-CI github/fork/qjing666/doublegrad github/fork/qjing666/fix_hdfs_download github/fork/sandyhouse/add_gather_etc github/fork/sandyhouse/add_send_recv_alltoall_etc github/fork/sandyhouse/pipeline_exe_run github/fork/seiriosPlus/feature/large_scale_kv_save_delta github/fork/seiriosPlus/fix/paddle_errors_fix github/fork/seiriosPlus/fix/paddle_op_errors github/fork/shangzhizhou/fix_test_activation_op_random_bug github/fork/smallv0221/yxp0924 github/fork/smallv0221/yxp0925 github/fork/swtkiwi/del-matplotlib github/fork/tianshuo78520a/kunlun_test github/fork/tianshuo78520a/update_dockerfile github/fork/wanghaoshuang/bert_fuse github/fork/wanghaoshuang/label_smooth github/fork/wanghuancoder/develop_CUDASynchronize github/fork/wanghuancoder/develop_Layer_doc github/fork/wanghuancoder/develop_ParameterList_doc github/fork/wanghuancoder/develop_Sequential_doc github/fork/wanghuancoder/develop_bilinear_tensor_product github/fork/wanghuancoder/develop_coverage_build_sh github/fork/wanghuancoder/develop_in_dynamic_mode_doc github/fork/wanghuancoder/develop_unique_name_doc github/fork/wangxicoding/fleet_meta_combine github/fork/wawltor/error_message_fix_5 github/fork/willthefrog/remove_l2_norm github/fork/windstamp/momentum_op github/fork/windstamp/mv_op_5 github/fork/windstamp/normal_api github/fork/wojtuss/wojtuss/fusion_gru_quantization github/fork/wojtuss/wojtuss/quantization-with-shift github/fork/wzzju/fix_err_info github/fork/wzzju/pure_fp16 github/fork/xiemoyuan/op_error_message github/fork/xiemoyuan/optimize_error_message github/fork/yaoxuefeng6/fix_doc github/fork/yaoxuefeng6/mod_dataset_v2 github/fork/yongqiangma/lod github/fork/ysh329/fix-clip-by-norm-error github/fork/ysh329/fix-error-clip-by-value github/fork/yukavio/error_info github/fork/zhangting2020/conv_filter_grad github/fork/zhangting2020/is_compile_with_cuda github/fork/zhangting2020/place_doc github/fork/zhangting2020/program github/fork/zhhsplendid/fix_any github/fork/zhhsplendid/refine_api2 github/fork/zhhsplendid/refine_api2_test github/fork/zhhsplendid/refine_api_test_ptb_lm github/fork/zhhsplendid/refine_api_test_resnet github/fork/zhhsplendid/refine_api_test_simnet github/fork/zhiqiu/dev/refine_initializer github/fork/zhiqiu/dev/remove_inplace_argument github/fork/zlsh80826/nvinfer_plugin_var_len_cuda11 improve_sccache incubate/infrt incubate/lite inplace_addto make_flag_adding_easier move_embedding_to_phi move_histogram_to_pten move_sgd_to_phi move_slice_to_pten move_temporal_shift_to_phi move_yolo_box_to_phi npu_fix_alloc numel paddle_tiny_install paralleltest preln_ernie prv-disable-more-cache prv-md-even-more prv-onednn-2.5 pten_tensor_refactor release/1.3 release/1.4 release/1.5 release/1.6 release/1.7 release/1.8 release/2.0 release/2.0-alpha release/2.0-beta release/2.0-rc release/2.0-rc1 release/2.1 release/2.2 release/2.3 release/2.3-fc-ernie-fix release/2.4 release/lite-0.1 revert-24981-add_device_attr_for_regulization revert-26856-strategy_example2 revert-27520-disable_pr revert-31068-fix_conv3d_windows revert-31562-mean revert-32290-develop-hardlabel revert-33037-forci revert-33475-fix_cifar_label_dimension revert-33630-bug-fix revert-34159-add_npu_bce_logical_dev revert-34406-add_copy_from_tensor revert-34910-spinlocks_for_allocator revert-35069-revert-34910-spinlocks_for_allocator revert-36057-dev/read_flags_in_ut revert-36201-refine_fast_threaded_ssa_graph_executor revert-36985-add_license revert-37318-refactor_dygraph_to_eager revert-37926-eager_coreops_500 revert-37956-revert-37727-pylayer_support_tuple revert-38100-mingdong revert-38301-allocation_rearrange_pr revert-38703-numpy_bf16_package_reupload revert-38732-remove_useless_header_in_elementwise_mul_grad revert-38959-Reduce_Grad revert-39143-adjust_empty revert-39227-move_trace_op_to_pten revert-39268-dev/remove_concat_fluid_kernel revert-40170-support_partial_grad revert-41056-revert-40727-move_some_activaion_to_phi revert-41065-revert-40993-mv_ele_floordiv_pow revert-41068-revert-40790-phi_new revert-41944-smaller_inference_api_test revert-42149-do-not-reset-default-stream-for-stream-safe-cuda-allocator revert-43155-fix_ut_tempfile revert-43882-revert-41944-smaller_inference_api_test revert-45808-phi/simplify_size_op revert-46827-deform_comment rocm_dev_0217 support_weight_transpose test_benchmark_ci test_feature_precision_test_c test_model_benchmark test_model_benchmark_ci zhiqiu-patch-1 v2.4.0-rc0 v2.3.2 v2.3.1 v2.3.0 v2.3.0-rc0 v2.2.2 v2.2.1 v2.2.0 v2.2.0-rc0 v2.2.0-bak0 v2.1.3 v2.1.2 v2.1.1 v2.1.0 v2.1.0-rc0 v2.0.2 v2.0.1 v2.0.0 v2.0.0-rc1 v2.0.0-rc0 v2.0.0-beta0 v2.0.0-alpha0 v1.8.5 v1.8.4 v1.8.3 v1.8.2 v1.8.1 v1.8.0 v1.7.2 v1.7.1 v1.7.0 v1.6.3 v1.6.2 v1.6.1 v1.6.0 v1.6.0-rc0 v1.5.2 v1.5.1 v1.5.0 v1.4.1 v1.4.0 v1.3.2 v1.3.1 v1.3.0 lite-v0.1
无相关合并请求
......@@ -90,5 +90,4 @@ TEST(prelu_op, test_scalar) {
} // namespace inference
} // namespace paddle
// USE_OP(prelu);
USE_CPU_ONLY_OP(prelu);
USE_OP(prelu);
nv_library(tensorrt_plugin
SRCS trt_plugin.cc split_op_plugin.cu elementwise_op_plugin.cu prelu_op_plugin.cu
avg_pool_op_plugin.cu
DEPS enforce tensorrt_engine)
DEPS enforce tensorrt_engine prelu)
......@@ -14,92 +14,16 @@
#include <stdio.h>
#include <cassert>
#include <vector>
#include "glog/logging.h"
#include "paddle/fluid/inference/tensorrt/plugin/prelu_op_plugin.h"
#include "paddle/fluid/operators/math/prelu.h"
namespace paddle {
namespace inference {
namespace tensorrt {
namespace plugin {
static const int CUDA_NUM_THREADS = 1024;
static const int CUDA_MAX_NUM_BLOCKS = 65535;
inline static int GET_NUM_BLOCKS(const int N) {
return (N + CUDA_NUM_THREADS - 1) / CUDA_NUM_THREADS;
}
__global__ void PReluChannelWiseKernel(const float *input, const float *alpha,
float *output, int channel,
size_t spatial_size) {
size_t offset = blockIdx.x * spatial_size;
const float *in = input + offset;
float *out = output + offset;
float scale = alpha[blockIdx.x % channel];
for (size_t i = threadIdx.x; i < spatial_size; i += blockDim.x) {
float x = in[i];
out[i] = (x > 0) ? x : scale * x;
}
}
__global__ void PReluElementWiseKernel(const float *input, const float *alpha,
float *output, size_t spatial_size) {
size_t offset = blockIdx.x * spatial_size;
const float *in = input + offset;
const float *scale = alpha + offset;
float *out = output + offset;
for (size_t i = threadIdx.x; i < spatial_size; i += blockDim.x) {
float x = in[i];
out[i] = (x > 0) ? x : scale[i] * x;
}
}
__global__ void PReluScalarKernel(const float *input, const float *alpha,
float *output, size_t spatial_size) {
size_t offset = blockIdx.x * spatial_size;
const float *in = input + offset;
float scale = *alpha;
float *out = output + offset;
for (size_t i = threadIdx.x; i < spatial_size; i += blockDim.x) {
float x = in[i];
out[i] = (x > 0) ? x : scale * x;
}
}
static inline void PReluChannelWise(cudaStream_t stream, const float *input,
const float *alpha, float *output,
int batch_size,
const nvinfer1::Dims &dims) {
size_t unroll = batch_size * dims.d[0];
size_t spatial_size = dims.d[1] * dims.d[2];
CHECK_LT(unroll, CUDA_MAX_NUM_BLOCKS);
PReluChannelWiseKernel<<<unroll, CUDA_NUM_THREADS, 0, stream>>>(
input, alpha, output, dims.d[0], spatial_size);
}
static inline void PReluElementWise(cudaStream_t stream, const float *input,
const float *alpha, float *output,
int batch_size,
const nvinfer1::Dims &dims) {
size_t unroll = batch_size * dims.d[0];
size_t spatial_size = dims.d[1] * dims.d[2];
CHECK_LT(unroll, CUDA_MAX_NUM_BLOCKS);
PReluElementWiseKernel<<<unroll, CUDA_NUM_THREADS, 0, stream>>>(
input, alpha, output, spatial_size);
}
static inline void PReluScalar(cudaStream_t stream, const float *input,
const float *alpha, float *output,
int batch_size, const nvinfer1::Dims &dims) {
size_t unroll = batch_size * dims.d[0];
size_t spatial_size = dims.d[1] * dims.d[2];
CHECK_LT(unroll, CUDA_MAX_NUM_BLOCKS);
PReluScalarKernel<<<unroll, CUDA_NUM_THREADS, 0, stream>>>(
input, alpha, output, spatial_size);
}
nvinfer1::Dims PReluPlugin::getOutputDimensions(int index,
const nvinfer1::Dims *inputDims,
int nbInputs) {
......@@ -110,19 +34,31 @@ nvinfer1::Dims PReluPlugin::getOutputDimensions(int index,
return output_dims;
}
int PReluPlugin::enqueue(int batchSize, const void *const *inputs,
int PReluPlugin::enqueue(int batch_size, const void *const *inputs,
void **outputs, void *workspace, cudaStream_t stream) {
// input dims is CHW.
const auto &input_dims = this->getInputDims(0);
const float *input = reinterpret_cast<const float *>(inputs[0]);
const float *alpha = reinterpret_cast<const float *>(alpha_.get().values);
float *output = reinterpret_cast<float **>(outputs)[0];
std::vector<int> input_shape;
input_shape.push_back(batch_size);
for (int i = 0; i < input_dims.nbDims; i++) {
input_shape.push_back(input_dims.d[i]);
}
if (mode_ == "channel") {
PReluChannelWise(stream, input, alpha, output, batchSize, input_dims);
operators::math::PreluChannelWiseDirectCUDAFunctor<float>
prelu_channel_wise;
prelu_channel_wise(stream, input, alpha, output, input_shape);
} else if (mode_ == "element") {
PReluElementWise(stream, input, alpha, output, batchSize, input_dims);
operators::math::PreluElementWiseDirectCUDAFunctor<float>
prelu_element_wise;
prelu_element_wise(stream, input, alpha, output, input_shape);
} else {
PReluScalar(stream, input, alpha, output, batchSize, input_dims);
operators::math::PreluScalarDirectCUDAFunctor<float> prelu_scalar;
prelu_scalar(stream, input, alpha, output, input_shape);
}
return cudaGetLastError() != cudaSuccess;
}
......
......@@ -70,7 +70,7 @@ endif()
set(COMMON_OP_DEPS ${COMMON_OP_DEPS} sequence_padding sequence_scale cos_sim_functor memory jit_kernel concat_and_split cross_entropy softmax vol2col im2col sampler)
set(COMMON_OP_DEPS ${COMMON_OP_DEPS} sequence2batch lstm_compute matrix_bit_code gru_compute activation_functions)
if (WITH_GPU)
set(COMMON_OP_DEPS ${COMMON_OP_DEPS} depthwise_conv)
set(COMMON_OP_DEPS ${COMMON_OP_DEPS} depthwise_conv prelu)
endif()
# FIXME(typhoonzero): operator deps may not needed.
......
......@@ -59,6 +59,7 @@ math_library(matrix_bit_code)
math_library(unpooling)
math_library(vol2col)
math_library(prelu)
cc_test(math_function_test SRCS math_function_test.cc DEPS math_function)
cc_test(selected_rows_functor_test SRCS selected_rows_functor_test.cc DEPS selected_rows_functor)
......
/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */
#include "paddle/fluid/operators/math/prelu.h"
namespace paddle {
namespace operators {
namespace math {
static const int CUDA_NUM_THREADS = 1024;
static const int CUDA_MAX_NUM_BLOCKS = 65535;
inline static int GET_NUM_BLOCKS(const int N) {
return (N + CUDA_NUM_THREADS - 1) / CUDA_NUM_THREADS;
}
template <typename T>
__global__ void PReluChannelWiseKernel(const T *input, const T *alpha,
T *output, int channel,
size_t spatial_size) {
size_t offset = blockIdx.x * spatial_size;
const T *in = input + offset;
T *out = output + offset;
T scale = alpha[blockIdx.x % channel];
for (size_t i = threadIdx.x; i < spatial_size; i += blockDim.x) {
T x = in[i];
out[i] = (x > 0) ? x : scale * x;
}
}
template <typename T>
__global__ void PReluElementWiseKernel(const T *input, const T *alpha,
T *output, size_t spatial_size) {
size_t offset = blockIdx.x * spatial_size;
const T *in = input + offset;
const T *scale = alpha + offset;
T *out = output + offset;
for (size_t i = threadIdx.x; i < spatial_size; i += blockDim.x) {
T x = in[i];
out[i] = (x > 0) ? x : scale[i] * x;
}
}
template <typename T>
__global__ void PReluScalarKernel(const T *input, const T *alpha, T *output,
size_t spatial_size) {
size_t offset = blockIdx.x * spatial_size;
const T *in = input + offset;
T scale = *alpha;
T *out = output + offset;
for (size_t i = threadIdx.x; i < spatial_size; i += blockDim.x) {
T x = in[i];
out[i] = (x > 0) ? x : scale * x;
}
}
template <typename T>
static inline void PReluChannelWise(cudaStream_t stream, const T *input,
const T *alpha, T *output,
std::vector<int> input_shape) {
size_t unroll = input_shape[0] * input_shape[1];
size_t spatial_size = input_shape[2] * input_shape[3];
CHECK_LT(unroll, CUDA_MAX_NUM_BLOCKS);
PReluChannelWiseKernel<<<unroll, CUDA_NUM_THREADS, 0, stream>>>(
input, alpha, output, input_shape[1], spatial_size);
}
template <typename T>
static inline void PReluElementWise(cudaStream_t stream, const T *input,
const T *alpha, T *output,
std::vector<int> input_shape) {
size_t unroll = input_shape[0] * input_shape[1];
size_t spatial_size = input_shape[2] * input_shape[3];
CHECK_LT(unroll, CUDA_MAX_NUM_BLOCKS);
PReluElementWiseKernel<<<unroll, CUDA_NUM_THREADS, 0, stream>>>(
input, alpha, output, spatial_size);
}
template <typename T>
static inline void PReluScalar(cudaStream_t stream, const T *input,
const T *alpha, T *output,
std::vector<int> input_shape) {
size_t unroll = input_shape[0] * input_shape[1];
size_t spatial_size = input_shape[2] * input_shape[3];
CHECK_LT(unroll, CUDA_MAX_NUM_BLOCKS);
PReluScalarKernel<<<unroll, CUDA_NUM_THREADS, 0, stream>>>(
input, alpha, output, spatial_size);
}
template <typename T>
void PreluChannelWiseDirectCUDAFunctor<T>::operator()(
cudaStream_t stream, const T *input, const T *alpha, T *output,
std::vector<int> input_shape) {
size_t unroll = input_shape[0] * input_shape[1];
size_t spatial_size = input_shape[2] * input_shape[3];
CHECK_LT(unroll, CUDA_MAX_NUM_BLOCKS);
PReluChannelWiseKernel<<<unroll, CUDA_NUM_THREADS, 0, stream>>>(
input, alpha, output, input_shape[1], spatial_size);
}
template <typename T>
void PreluElementWiseDirectCUDAFunctor<T>::operator()(
cudaStream_t stream, const T *input, const T *alpha, T *output,
std::vector<int> input_shape) {
size_t unroll = input_shape[0] * input_shape[1];
size_t spatial_size = input_shape[2] * input_shape[3];
CHECK_LT(unroll, CUDA_MAX_NUM_BLOCKS);
PReluElementWiseKernel<<<unroll, CUDA_NUM_THREADS, 0, stream>>>(
input, alpha, output, spatial_size);
}
template <typename T>
void PreluScalarDirectCUDAFunctor<T>::operator()(cudaStream_t stream,
const T *input, const T *alpha,
T *output,
std::vector<int> input_shape) {
size_t unroll = input_shape[0] * input_shape[1];
size_t spatial_size = input_shape[2] * input_shape[3];
CHECK_LT(unroll, CUDA_MAX_NUM_BLOCKS);
PReluScalarKernel<<<unroll, CUDA_NUM_THREADS, 0, stream>>>(
input, alpha, output, spatial_size);
}
template class PreluChannelWiseDirectCUDAFunctor<float>;
template class PreluChannelWiseDirectCUDAFunctor<double>;
template class PreluElementWiseDirectCUDAFunctor<float>;
template class PreluElementWiseDirectCUDAFunctor<double>;
template class PreluScalarDirectCUDAFunctor<float>;
template class PreluScalarDirectCUDAFunctor<double>;
} // namespace math
} // namespace operators
} // namespace paddle
/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */
#pragma once
#include <vector>
#include "paddle/fluid/operators/math/math_function.h"
#include "paddle/fluid/platform/cudnn_helper.h"
namespace paddle {
namespace operators {
namespace math {
#ifdef PADDLE_WITH_CUDA
template <typename T>
class PreluChannelWiseDirectCUDAFunctor {
public:
void operator()(cudaStream_t stream, const T *input, const T *alpha,
T *output, std::vector<int> input_shape);
};
template <typename T>
class PreluElementWiseDirectCUDAFunctor {
public:
void operator()(cudaStream_t stream, const T *input, const T *alpha,
T *output, std::vector<int> input_shape);
};
template <typename T>
class PreluScalarDirectCUDAFunctor {
public:
void operator()(cudaStream_t stream, const T *input, const T *alpha,
T *output, std::vector<int> input_shape);
};
#endif
} // namespace math
} // namespace operators
} // namespace paddle
......@@ -58,7 +58,7 @@ class PReluOp : public framework::OperatorWithKernel {
const framework::ExecutionContext &ctx) const override {
return framework::OpKernelType(
framework::ToDataType(ctx.Input<Tensor>("X")->type()),
platform::CPUPlace());
ctx.device_context());
}
};
......
/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */
#include <string>
#include <vector>
#include "paddle/fluid/framework/op_registry.h"
#include "paddle/fluid/operators/math/prelu.h"
#include "paddle/fluid/operators/prelu_op.h"
#include "paddle/fluid/platform/cuda_primitives.h"
namespace paddle {
namespace operators {
using Tensor = framework::Tensor;
template <typename DeviceContext, typename T>
class CUDAPReluKernel : public framework::OpKernel<T> {
public:
void Compute(const framework::ExecutionContext& context) const override {
auto* x = context.Input<Tensor>("X");
auto* alpha = context.Input<Tensor>("Alpha");
auto* out = context.Output<Tensor>("Out");
const T* x_ptr = x->data<T>();
T* o_ptr = out->mutable_data<T>(context.GetPlace());
const T* alpha_ptr = alpha->data<T>();
auto& mode = context.Attr<std::string>("mode");
int numel = x->numel();
auto dim = x->dims();
std::vector<int> input_shape = framework::vectorize2int(dim);
if (mode == "channel") {
math::PreluChannelWiseDirectCUDAFunctor<T> prelu_channel_wise;
prelu_channel_wise(context.cuda_device_context().stream(), x_ptr,
alpha_ptr, o_ptr, input_shape);
} else if (mode == "element") {
math::PreluElementWiseDirectCUDAFunctor<T> prelu_element_wise;
prelu_element_wise(context.cuda_device_context().stream(), x_ptr,
alpha_ptr, o_ptr, input_shape);
} else {
math::PreluScalarDirectCUDAFunctor<T> prelu_scalar;
prelu_scalar(context.cuda_device_context().stream(), x_ptr, alpha_ptr,
o_ptr, input_shape);
}
}
};
} // namespace operators
} // namespace paddle
namespace ops = paddle::operators;
REGISTER_OP_CUDA_KERNEL(
prelu, ops::CUDAPReluKernel<paddle::platform::CUDADeviceContext, float>,
ops::CUDAPReluKernel<paddle::platform::CUDADeviceContext, double>);
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册
反馈
建议
客服 返回
顶部