Skip to content

P
Paddle

Crayon鑫 / Paddle
与 Fork 源项目一致

Fork自 PaddlePaddle / Paddle

通知 1
Star 1
Fork 0
P
Paddle
提交 06aa23b0 编写于 作者: T tangwei12
浏览文件
操作

Merge branch 'develop' of github.com:PaddlePaddle/Paddle into checkpoint

上级 be050565 35e55636
隐藏空白更改
内联 并排
Showing with 635 addition and 275 deletion
Dockerfile
浏览文件 @ 06aa23b0
......@@ -70,7 +70,7 @@ RUN localedef -i en_US -f UTF-8 en_US.UTF-8
# specify sphinx version as 1.5.6 and remove -U option for [pip install -U
# sphinx-rtd-theme] since -U option will cause sphinx being updated to newest
# version(1.7.1 for now), which causes building documentation failed.
RUN pip install --upgrade pip==9.0.3 && \
RUN easy_install -U pip && \
pip install -U wheel && \
pip install -U docopt PyYAML sphinx==1.5.6 && \
pip install sphinx-rtd-theme==0.1.9 recommonmark
......
doc/fluid/howto/optimization/cpu_profiling_cn.md
浏览文件 @ 06aa23b0
# CPU性能调优
此教程会介绍如何使用Python的cProfile包、Python库yep、Google perftools来进行性能分析 (profiling) 与调优（performance tuning）。
Profling 指发现性能瓶颈。系统中的瓶颈可能和程序员开发过程中想象的瓶颈相去甚远。Tuning 指消除瓶颈。性能优化的过程通常是不断重复地 profiling 和 tuning。
......@@ -8,7 +10,7 @@ PaddlePaddle 用户一般通过调用 Python API 编写深度学习程序。大
* Python 与 C++ 混合代码的性能分析
# Python代码的性能分析
## Python代码的性能分析
### 生成性能分析文件
......
doc/fluid/howto/optimization/cpu_profiling_en.md
浏览文件 @ 06aa23b0
# Tune CPU performance
This tutorial introduces techniques we use to profile and tune the
CPU performance of PaddlePaddle. We will use Python packages
`cProfile` and `yep`, and Google's `perftools`.
......@@ -14,7 +16,7 @@ the profiling and tuning of
1. the Python code and
1. the mixture of Python and C++ code.
# Profiling the Python Code
## Profiling the Python Code
### Generate the Performance Profiling File
......
doc/v2/build_and_install/pip_install_cn.rst
浏览文件 @ 06aa23b0
......@@ -37,12 +37,11 @@ PaddlePaddle可以使用常用的Python包管理工具
:header: "版本说明", "cp27-cp27mu", "cp27-cp27m"
:widths: 1, 3, 3
"cpu_avx_mkl", "`paddlepaddle-0.11.0-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuAvxCp27cp27mu/.lastSuccessful/paddlepaddle-0.11.0-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle-0.11.0-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuAvxCp27cp27mu/.lastSuccessful/paddlepaddle-0.11.0-cp27-cp27m-linux_x86_64.whl>`_"
"cpu_avx_openblas", "`paddlepaddle-0.11.0-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuAvxOpenblas/.lastSuccessful/paddlepaddle-0.11.0-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle-0.11.0-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuAvxOpenblas/.lastSuccessful/paddlepaddle-0.11.0-cp27-cp27m-linux_x86_64.whl>`_"
"cpu_noavx_openblas", "`paddlepaddle-0.11.0-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuNoavxOpenblas/.lastSuccessful/paddlepaddle-0.11.0-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle-0.11.0-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuNoavxOpenblas/.lastSuccessful/paddlepaddle-0.11.0-cp27-cp27m-linux_x86_64.whl>`_"
"cuda7.5_cudnn5_avx_mkl", "`paddlepaddle_gpu-0.11.0-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda75cudnn5cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-0.11.0-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle_gpu-0.11.0-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda75cudnn5cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-0.11.0-cp27-cp27m-linux_x86_64.whl>`_"
"cuda8.0_cudnn5_avx_mkl", "`paddlepaddle_gpu-0.11.0-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda80cudnn5cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-0.11.0-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle_gpu-0.11.0-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda80cudnn5cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-0.11.0-cp27-cp27m-linux_x86_64.whl>`_"
"cuda8.0_cudnn7_avx_mkl", "`paddlepaddle_gpu-0.11.0-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda8cudnn7cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-0.11.0-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle_gpu-0.11.0-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda8cudnn7cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-0.11.0-cp27-cp27m-linux_x86_64.whl>`_"
"cpu_avx_mkl", "`paddlepaddle-latest-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuAvxCp27cp27mu/.lastSuccessful/paddlepaddle-latest-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle-latest-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuAvxCp27cp27mu/.lastSuccessful/paddlepaddle-latest-cp27-cp27m-linux_x86_64.whl>`_"
"cpu_avx_openblas", "`paddlepaddle-latest-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuAvxOpenblas/.lastSuccessful/paddlepaddle-latest-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle-latest-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuAvxOpenblas/.lastSuccessful/paddlepaddle-latest-cp27-cp27m-linux_x86_64.whl>`_"
"cpu_noavx_openblas", "`paddlepaddle-latest-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuNoavxOpenblas/.lastSuccessful/paddlepaddle-latest-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle-latest-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuNoavxOpenblas/.lastSuccessful/paddlepaddle-latest-cp27-cp27m-linux_x86_64.whl>`_"
"cuda8.0_cudnn5_avx_mkl", "`paddlepaddle_gpu-latest-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda80cudnn5cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-latest-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle_gpu-latest-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda80cudnn5cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-latest-cp27-cp27m-linux_x86_64.whl>`_"
"cuda8.0_cudnn7_avx_mkl", "`paddlepaddle_gpu-latest-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda8cudnn7cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-latest-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle_gpu-latest-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda8cudnn7cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-latest-cp27-cp27m-linux_x86_64.whl>`_"
.. _pip_dependency:
......
doc/v2/build_and_install/pip_install_en.rst
浏览文件 @ 06aa23b0
......@@ -40,12 +40,11 @@ If the links below shows up the login form, just click "Log in as guest" to star
:header: "version", "cp27-cp27mu", "cp27-cp27m"
:widths: 1, 3, 3
"cpu_avx_mkl", "`paddlepaddle-0.11.0-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuAvxCp27cp27mu/.lastSuccessful/paddlepaddle-0.11.0-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle-0.11.0-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuAvxCp27cp27mu/.lastSuccessful/paddlepaddle-0.11.0-cp27-cp27m-linux_x86_64.whl>`_"
"cpu_avx_openblas", "`paddlepaddle-0.11.0-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuAvxOpenblas/.lastSuccessful/paddlepaddle-0.11.0-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle-0.11.0-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuAvxOpenblas/.lastSuccessful/paddlepaddle-0.11.0-cp27-cp27m-linux_x86_64.whl>`_"
"cpu_noavx_openblas", "`paddlepaddle-0.11.0-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuNoavxOpenblas/.lastSuccessful/paddlepaddle-0.11.0-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle-0.11.0-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuNoavxOpenblas/.lastSuccessful/paddlepaddle-0.11.0-cp27-cp27m-linux_x86_64.whl>`_"
"cuda7.5_cudnn5_avx_mkl", "`paddlepaddle_gpu-0.11.0-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda75cudnn5cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-0.11.0-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle_gpu-0.11.0-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda75cudnn5cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-0.11.0-cp27-cp27m-linux_x86_64.whl>`_"
"cuda8.0_cudnn5_avx_mkl", "`paddlepaddle_gpu-0.11.0-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda80cudnn5cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-0.11.0-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle_gpu-0.11.0-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda80cudnn5cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-0.11.0-cp27-cp27m-linux_x86_64.whl>`_"
"cuda8.0_cudnn7_avx_mkl", "`paddlepaddle_gpu-0.11.0-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda8cudnn7cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-0.11.0-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle_gpu-0.11.0-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda8cudnn7cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-0.11.0-cp27-cp27m-linux_x86_64.whl>`_"
"cpu_avx_mkl", "`paddlepaddle-latest-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuAvxCp27cp27mu/.lastSuccessful/paddlepaddle-latest-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle-latest-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuAvxCp27cp27mu/.lastSuccessful/paddlepaddle-latest-cp27-cp27m-linux_x86_64.whl>`_"
"cpu_avx_openblas", "`paddlepaddle-latest-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuAvxOpenblas/.lastSuccessful/paddlepaddle-latest-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle-latest-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuAvxOpenblas/.lastSuccessful/paddlepaddle-latest-cp27-cp27m-linux_x86_64.whl>`_"
"cpu_noavx_openblas", "`paddlepaddle-latest-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuNoavxOpenblas/.lastSuccessful/paddlepaddle-latest-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle-latest-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuNoavxOpenblas/.lastSuccessful/paddlepaddle-latest-cp27-cp27m-linux_x86_64.whl>`_"
"cuda8.0_cudnn5_avx_mkl", "`paddlepaddle_gpu-latest-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda80cudnn5cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-latest-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle_gpu-latest-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda80cudnn5cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-latest-cp27-cp27m-linux_x86_64.whl>`_"
"cuda8.0_cudnn7_avx_mkl", "`paddlepaddle_gpu-latest-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda8cudnn7cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-latest-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle_gpu-latest-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda8cudnn7cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-latest-cp27-cp27m-linux_x86_64.whl>`_"
.. _pip_dependency:
......
paddle/fluid/framework/data_type_transform.cc
浏览文件 @ 06aa23b0
......@@ -91,6 +91,12 @@ void TransDataType(const OpKernelType& kernel_type_for_var,
case proto::VarType::BOOL:
framework::VisitDataType(dst_type, CastDataType<bool>(in, out, ctx));
break;
case proto::VarType::INT16:
framework::VisitDataType(dst_type, CastDataType<bool>(in, out, ctx));
break;
case proto::VarType::UINT8:
framework::VisitDataType(dst_type, CastDataType<bool>(in, out, ctx));
break;
default:
PADDLE_THROW("Not support type %d", src_type);
}
......
paddle/fluid/framework/details/multi_devices_graph_builder.cc
浏览文件 @ 06aa23b0
......@@ -98,7 +98,7 @@ bool MultiDevSSAGraphBuilder::IsDistTrainOp(const OpDesc &op,
return false;
};
if (op.Type() == "split") {
if (op.Type() == "split" || op.Type() == "split_byref") {
return checker(op.OutputArgumentNames(), send_op->InputArgumentNames());
} else if (op.Type() == "concat") {
return checker(op.InputArgumentNames(), send_op->OutputArgumentNames());
......
paddle/fluid/inference/tests/test_helper.h
浏览文件 @ 06aa23b0
......@@ -149,7 +149,7 @@ void TestInference(const std::string& dirname,
state = paddle::platform::ProfilerState::kCPU;
} else {
#ifdef PADDLE_WITH_CUDA
state = paddle::platform::ProfilerState::kCUDA;
state = paddle::platform::ProfilerState::kAll;
// The default device_id of paddle::platform::CUDAPlace is 0.
// Users can get the device_id using:
// int device_id = place.GetDeviceId();
......@@ -172,7 +172,7 @@ void TestInference(const std::string& dirname,
}
// Disable the profiler and print the timing information
paddle::platform::DisableProfiler(paddle::platform::EventSortingKey::kDefault,
"load_program_profiler.txt");
"load_program_profiler");
paddle::platform::ResetProfiler();
// 3. Get the feed_target_names and fetch_target_names
......@@ -236,8 +236,7 @@ void TestInference(const std::string& dirname,
// Disable the profiler and print the timing information
paddle::platform::DisableProfiler(
paddle::platform::EventSortingKey::kDefault,
"run_inference_profiler.txt");
paddle::platform::EventSortingKey::kDefault, "run_inference_profiler");
paddle::platform::ResetProfiler();
}
......
paddle/fluid/operators/CMakeLists.txt
浏览文件 @ 06aa23b0
......@@ -204,6 +204,7 @@ if(WITH_DISTRIBUTE)
set_source_files_properties(send_recv_op_test.cc PROPERTIES COMPILE_FLAGS ${DISTRIBUTE_COMPILE_FLAGS})
cc_test(test_send_recv SRCS send_recv_op_test.cc DEPS prefetch_op send_op listen_and_serv_op sum_op executor)
if(WITH_GPU)
set_source_files_properties(test_send_nccl_id.cc PROPERTIES COMPILE_FLAGS ${DISTRIBUTE_COMPILE_FLAGS})
cc_test(test_send_nccl_id SRCS test_send_nccl_id.cc DEPS send_op listen_and_serv_op executor)
op_library(gen_nccl_id_op DEPS nccl_common sendrecvop_grpc)
set_source_files_properties(gen_nccl_id_op.cc PROPERTIES COMPILE_FLAGS ${DISTRIBUTE_COMPILE_FLAGS})
......
paddle/fluid/operators/beam_search_op.h
浏览文件 @ 06aa23b0
......@@ -14,10 +14,6 @@ limitations under the License. */
#pragma once
#ifdef PADDLE_WITH_TESTING
#include "gtest/gtest.h"
#endif
#include <string>
#include <vector>
#include "paddle/fluid/framework/lod_tensor.h"
......
paddle/fluid/operators/is_empty_op.cc
浏览文件 @ 06aa23b0
......@@ -12,45 +12,41 @@ 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/is_empty_op.h"
#include "paddle/fluid/framework/op_registry.h"
#include "paddle/fluid/framework/operator.h"
namespace paddle {
namespace operators {
constexpr char kInput[] = "X";
constexpr char kOutput[] = "Out";
class IsEmptyOp : public framework::OperatorBase {
class IsEmptyOp : public framework::OperatorWithKernel {
public:
IsEmptyOp(const std::string &type, const framework::VariableNameMap &inputs,
const framework::VariableNameMap &outputs,
const framework::AttributeMap &attrs)
: OperatorBase(type, inputs, outputs, attrs) {}
using framework::OperatorWithKernel::OperatorWithKernel;
private:
void RunImpl(const framework::Scope &scope,
const platform::Place &place) const override {
// get input
auto *var = scope.FindVar(Input(kInput));
PADDLE_ENFORCE_NOT_NULL(var);
auto &tensor = var->Get<framework::LoDTensor>();
// get output
auto *out = scope.FindVar(Output(kOutput));
PADDLE_ENFORCE_NOT_NULL(out);
auto *out_tensor = out->GetMutable<framework::LoDTensor>();
protected:
void InferShape(framework::InferShapeContext *ctx) const override {
PADDLE_ENFORCE(ctx->HasInput("X"),
"Input(X) of IsEmptyOp should not be null.");
PADDLE_ENFORCE(ctx->HasOutput("Out"),
"Output(Out) of IsEmptyOp should not be null.");
ctx->SetOutputDim("Out", {1});
}
out_tensor->Resize({1});
out_tensor->mutable_data<bool>(platform::CPUPlace())[0] =
framework::product(tensor.dims()) == 0;
framework::OpKernelType GetExpectedKernelType(
const framework::ExecutionContext &ctx) const override {
framework::OpKernelType kt = framework::OpKernelType(
framework::ToDataType(ctx.Input<framework::LoDTensor>("X")->type()),
platform::CPUPlace());
return kt;
}
};
class IsEmptyOpProtoMaker : public framework::OpProtoAndCheckerMaker {
class IsEmptyOpMaker : public framework::OpProtoAndCheckerMaker {
public:
void Make() override {
AddInput(kInput, "(Tensor) Tensor which is to be checked.");
AddOutput(kOutput, "(Tensor) a boolean Tensor that indicate empty or not.");
AddInput("X", "(LoDTensor) Tensor which is to be checked.");
AddOutput("Out",
"(LoDTensor) a boolean Tensor that indicate empty or not.");
AddComment(R"DOC(
IsEmpty Operator which checks whether a tensor is empty.
......@@ -62,5 +58,12 @@ It will just return product(tensor.ddims()) > 0;
} // namespace operators
} // namespace paddle
REGISTER_OP_WITHOUT_GRADIENT(is_empty, paddle::operators::IsEmptyOp,
paddle::operators::IsEmptyOpProtoMaker);
namespace ops = paddle::operators;
REGISTER_OPERATOR(is_empty, ops::IsEmptyOp, ops::IsEmptyOpMaker,
paddle::framework::EmptyGradOpMaker);
REGISTER_OP_CPU_KERNEL(
is_empty, ops::IsEmptyOpKernel<paddle::platform::CPUDeviceContext, float>,
ops::IsEmptyOpKernel<paddle::platform::CPUDeviceContext, double>,
ops::IsEmptyOpKernel<paddle::platform::CPUDeviceContext, int>,
ops::IsEmptyOpKernel<paddle::platform::CPUDeviceContext, int64_t>);
paddle/fluid/operators/is_empty_op.h 0 → 100644
浏览文件 @ 06aa23b0
/* 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 "paddle/fluid/framework/op_registry.h"
#include "paddle/fluid/framework/operator.h"
namespace paddle {
namespace operators {
template <typename DeviceContext, typename T>
class IsEmptyOpKernel : public framework::OpKernel<T> {
public:
void Compute(const framework::ExecutionContext& context) const override {
// get input
auto* input_tensor = context.Input<framework::LoDTensor>("X");
// get output
auto* output_tensor = context.Output<framework::LoDTensor>("Out");
output_tensor->mutable_data<bool>(platform::CPUPlace())[0] =
framework::product(input_tensor->dims()) == 0;
}
};
} // namespace operators
} // namespace paddle
paddle/fluid/operators/pool_mkldnn_op.cc
浏览文件 @ 06aa23b0
......@@ -18,6 +18,26 @@ limitations under the License. */
namespace paddle {
namespace operators {
using mkldnn::memory; // Note: paddle has also "memory" namespace
using mkldnn::pooling_forward;
using mkldnn::pooling_backward;
// Generate keys for storing/retriving primitives for this operator
// TODO(jczaja): Make hashing function more optimial
static std::string gethash(memory::dims& input_dims, std::string& pooling_type,
std::vector<int>& ksize, std::vector<int>& strides,
std::vector<int>& paddings, std::string suffix) {
auto dims2str = [](memory::dims& operand_dims) {
std::string dstr = "";
for (size_t i = 0; i < operand_dims.size(); ++i) {
dstr += std::to_string(operand_dims[i]) + "-";
}
return dstr;
};
return dims2str(input_dims) + dims2str(ksize) + dims2str(strides) +
dims2str(paddings) + pooling_type + suffix;
}
template <typename T>
class PoolMKLDNNOpKernel : public paddle::framework::OpKernel<T> {
public:
......@@ -34,10 +54,6 @@ class PoolMKLDNNOpKernel : public paddle::framework::OpKernel<T> {
// Get an unique name from "argument" name of "Out" variable
// This name will be used as key when saving info into device context
const std::string key = ctx.op().Output("Out");
const std::string key_pool_pd = key + "@pool_pd";
const std::string key_pool_workspace_memory =
key + "@pool_workspace_memory";
std::string pooling_type = ctx.Attr<std::string>("pooling_type");
std::vector<int> ksize = ctx.Attr<std::vector<int>>("ksize");
......@@ -63,37 +79,71 @@ class PoolMKLDNNOpKernel : public paddle::framework::OpKernel<T> {
std::vector<int> src_tz = paddle::framework::vectorize2int(input->dims());
std::vector<int> dst_tz = paddle::framework::vectorize2int(output->dims());
// TODO(pzelazko-intel): support more formats
auto src_md = platform::MKLDNNMemDesc(src_tz, mkldnn::memory::f32,
mkldnn::memory::format::nchw);
auto dst_md = platform::MKLDNNMemDesc(dst_tz, mkldnn::memory::f32,
mkldnn::memory::format::nchw);
std::shared_ptr<mkldnn::pooling_forward::primitive_desc> pool_pd =
CreatePrimitiveDesc(src_md, dst_md, strides, paddings, ksize,
pooling_type, mkldnn_engine);
// save pool_pd into global device context to be referred in backward path
dev_ctx.SetBlob(key_pool_pd, pool_pd);
std::shared_ptr<mkldnn::memory> workspace_memory =
CreateWorkspaceMemory(pool_pd, pooling_type, mkldnn_engine);
// save pool_workspace_memory to be referred in backward path
dev_ctx.SetBlob(key_pool_workspace_memory, workspace_memory);
auto src_memory =
mkldnn::memory({src_md, mkldnn_engine},
static_cast<void*>(const_cast<T*>(input_data)));
auto dst_memory =
mkldnn::memory({dst_md, mkldnn_engine},
static_cast<void*>(const_cast<T*>(output_data)));
const std::string key = gethash(src_tz, pooling_type, ksize, strides,
paddings, ctx.op().Output("Out"));
const std::string key_pool_p = key + "@pool_p";
const std::string key_pool_pd = key + "@pool_pd";
const std::string key_pool_src_mem_p = key + "@pool_src_mem_p";
const std::string key_pool_dst_mem_p = key + "@pool_dst_mem_p";
const std::string key_pool_workspace_memory =
key + "@pool_workspace_memory";
auto pool_prim = mkldnn::pooling_forward(*pool_pd, src_memory, dst_memory,
*workspace_memory);
auto pool_p =
std::static_pointer_cast<pooling_forward>(dev_ctx.GetBlob(key_pool_p));
if (pool_p == nullptr) {
// TODO(pzelazko-intel): support more formats
auto src_md =
platform::MKLDNNMemDesc(src_tz, platform::MKLDNNGetDataType<T>(),
mkldnn::memory::format::nchw);
auto dst_md =
platform::MKLDNNMemDesc(dst_tz, platform::MKLDNNGetDataType<T>(),
mkldnn::memory::format::nchw);
std::shared_ptr<pooling_forward::primitive_desc> pool_pd =
CreatePrimitiveDesc(src_md, dst_md, strides, paddings, ksize,
pooling_type, mkldnn_engine);
// save pool_pd into global device context to be referred in backward path
dev_ctx.SetBlob(key_pool_pd, pool_pd);
std::shared_ptr<mkldnn::memory> workspace_memory =
CreateWorkspaceMemory(pool_pd, pooling_type, mkldnn_engine);
// save pool_workspace_memory to be referred in backward path
dev_ctx.SetBlob(key_pool_workspace_memory, workspace_memory);
auto pool_src_memory_p = std::make_shared<memory>(
memory::primitive_desc{src_md, mkldnn_engine},
static_cast<void*>(const_cast<T*>(input_data)));
dev_ctx.SetBlob(key_pool_src_mem_p, pool_src_memory_p);
auto pool_dst_memory_p = std::make_shared<memory>(
memory::primitive_desc{dst_md, mkldnn_engine},
static_cast<void*>(output_data));
dev_ctx.SetBlob(key_pool_dst_mem_p, pool_dst_memory_p);
pool_p = std::make_shared<pooling_forward>(
*pool_pd, *(pool_src_memory_p.get()), *(pool_dst_memory_p.get()),
*workspace_memory);
dev_ctx.SetBlob(key_pool_p, pool_p);
} else {
// Primitives already exist
auto pool_src_memory_p =
std::static_pointer_cast<memory>(dev_ctx.GetBlob(key_pool_src_mem_p));
PADDLE_ENFORCE(pool_src_memory_p != nullptr,
"Fail to find pooling src mem_p in device context");
auto pool_dst_memory_p =
std::static_pointer_cast<memory>(dev_ctx.GetBlob(key_pool_dst_mem_p));
PADDLE_ENFORCE(pool_dst_memory_p != nullptr,
"Fail to find pooling dst mem_p in device context");
pool_src_memory_p->set_data_handle(
reinterpret_cast<void*>(const_cast<T*>(input_data)));
pool_dst_memory_p->set_data_handle(output_data);
}
// push primitive to stream and wait until it's executed
std::vector<mkldnn::primitive> pipeline{pool_prim};
std::vector<mkldnn::primitive> pipeline{*(pool_p.get())};
mkldnn::stream(mkldnn::stream::kind::eager).submit(pipeline).wait();
}
......@@ -120,9 +170,10 @@ class PoolMKLDNNOpKernel : public paddle::framework::OpKernel<T> {
mkldnn::memory::primitive_desc workspace_md =
pooling_type == "max"
? pool_pd->workspace_primitive_desc()
: mkldnn::memory::primitive_desc(
{{}, mkldnn::memory::f32, mkldnn::memory::format::nchw},
engine);
: mkldnn::memory::primitive_desc({{},
platform::MKLDNNGetDataType<T>(),
mkldnn::memory::format::nchw},
engine);
auto p_workspace_memory = new mkldnn::memory(workspace_md);
return std::unique_ptr<mkldnn::memory>(p_workspace_memory);
......@@ -140,13 +191,6 @@ class PoolMKLDNNGradOpKernel : public paddle::framework::OpKernel<T> {
const Tensor* out_grad = ctx.Input<Tensor>(framework::GradVarName("Out"));
Tensor* in_x_grad = ctx.Output<Tensor>(framework::GradVarName("X"));
// Get an unique name from "argument" name of "Out" variable
// This name will be used as key when referring info from device context
const std::string key = ctx.op().Input("Out");
const std::string key_pool_pd = key + "@pool_pd";
const std::string key_pool_workspace_memory =
key + "@pool_workspace_memory";
std::string pooling_type = ctx.Attr<std::string>("pooling_type");
std::vector<int> ksize = ctx.Attr<std::vector<int>>("ksize");
std::vector<int> strides = ctx.Attr<std::vector<int>>("strides");
......@@ -171,43 +215,76 @@ class PoolMKLDNNGradOpKernel : public paddle::framework::OpKernel<T> {
std::vector<int> diff_dst_tz =
paddle::framework::vectorize2int(out_grad->dims());
auto diff_src_md = platform::MKLDNNMemDesc(diff_src_tz, mkldnn::memory::f32,
mkldnn::memory::format::nchw);
auto diff_dst_md = platform::MKLDNNMemDesc(diff_dst_tz, mkldnn::memory::f32,
mkldnn::memory::format::nchw);
// Retrieve pool_pd/pool_workspace_memory from device context
auto pool_pd =
std::static_pointer_cast<mkldnn::pooling_forward::primitive_desc>(
dev_ctx.GetBlob(key_pool_pd));
PADDLE_ENFORCE(pool_pd != nullptr,
"Fail to find pool_pd in device context");
auto workspace_memory = std::static_pointer_cast<mkldnn::memory>(
dev_ctx.GetBlob(key_pool_workspace_memory));
PADDLE_ENFORCE(workspace_memory != nullptr,
"Fail to find workspace_memory in device context");
auto pool_bwd_desc = mkldnn::pooling_backward::desc(
pooling_type == "max" ? mkldnn::algorithm::pooling_max
: mkldnn::algorithm::pooling_avg,
diff_src_md, diff_dst_md, strides, ksize, paddings, paddings,
mkldnn::padding_kind::zero);
auto pool_bwd_pd = mkldnn::pooling_backward::primitive_desc(
pool_bwd_desc, mkldnn_engine, *pool_pd);
auto diff_src_memory =
mkldnn::memory({diff_src_md, mkldnn_engine},
static_cast<void*>(const_cast<T*>(in_x_grad_data)));
auto diff_dst_memory =
mkldnn::memory({diff_dst_md, mkldnn_engine},
static_cast<void*>(const_cast<T*>(out_grad_data)));
// Get an unique name from "argument" name of "Out" variable
// This name will be used as key when referring info from device context
const std::string key = gethash(diff_src_tz, pooling_type, ksize, strides,
paddings, ctx.op().Input("Out"));
const std::string key_pool_bwd_p = key + "@pool_bwd_p";
const std::string key_pool_diff_src_mem_p = key + "@pool_diff_src_mem_p";
const std::string key_pool_diff_dst_mem_p = key + "@pool_diff_dst_mem_p";
const std::string key_pool_pd = key + "@pool_pd";
const std::string key_pool_workspace_memory =
key + "@pool_workspace_memory";
auto bwd_prim = mkldnn::pooling_backward(
pool_bwd_pd, diff_dst_memory, *workspace_memory, diff_src_memory);
auto pool_bwd_p = std::static_pointer_cast<pooling_backward>(
dev_ctx.GetBlob(key_pool_bwd_p));
if (pool_bwd_p == nullptr) {
auto diff_src_md =
platform::MKLDNNMemDesc(diff_src_tz, platform::MKLDNNGetDataType<T>(),
mkldnn::memory::format::nchw);
auto diff_dst_md =
platform::MKLDNNMemDesc(diff_dst_tz, platform::MKLDNNGetDataType<T>(),
mkldnn::memory::format::nchw);
// Retrieve pool_pd/pool_workspace_memory from device context
auto pool_pd =
std::static_pointer_cast<mkldnn::pooling_forward::primitive_desc>(
dev_ctx.GetBlob(key_pool_pd));
PADDLE_ENFORCE(pool_pd != nullptr,
"Fail to find pool_pd in device context");
auto workspace_memory = std::static_pointer_cast<mkldnn::memory>(
dev_ctx.GetBlob(key_pool_workspace_memory));
PADDLE_ENFORCE(workspace_memory != nullptr,
"Fail to find workspace_memory in device context");
auto pool_diff_src_memory_p = std::make_shared<memory>(memory(
{diff_src_md, mkldnn_engine}, static_cast<void*>(in_x_grad_data)));
dev_ctx.SetBlob(key_pool_diff_src_mem_p, pool_diff_src_memory_p);
auto pool_diff_dst_memory_p = std::make_shared<memory>(
memory({diff_dst_md, mkldnn_engine},
static_cast<void*>(const_cast<T*>(out_grad_data))));
dev_ctx.SetBlob(key_pool_diff_dst_mem_p, pool_diff_dst_memory_p);
auto pool_bwd_desc = mkldnn::pooling_backward::desc(
pooling_type == "max" ? mkldnn::algorithm::pooling_max
: mkldnn::algorithm::pooling_avg,
diff_src_md, diff_dst_md, strides, ksize, paddings, paddings,
mkldnn::padding_kind::zero);
auto pool_bwd_pd = mkldnn::pooling_backward::primitive_desc(
pool_bwd_desc, mkldnn_engine, *pool_pd);
pool_bwd_p = std::make_shared<pooling_backward>(
pool_bwd_pd, *(pool_diff_dst_memory_p.get()), *workspace_memory,
*(pool_diff_src_memory_p));
dev_ctx.SetBlob(key_pool_bwd_p, pool_bwd_p);
} else {
// Primitives already exist
auto pool_diff_src_memory_p = std::static_pointer_cast<memory>(
dev_ctx.GetBlob(key_pool_diff_src_mem_p));
PADDLE_ENFORCE(pool_diff_src_memory_p != nullptr,
"Fail to find pooling src mem_p in device context");
auto pool_diff_dst_memory_p = std::static_pointer_cast<memory>(
dev_ctx.GetBlob(key_pool_diff_dst_mem_p));
PADDLE_ENFORCE(pool_diff_dst_memory_p != nullptr,
"Fail to find pooling dst mem_p in device context");
pool_diff_src_memory_p->set_data_handle(
reinterpret_cast<void*>(in_x_grad_data));
pool_diff_dst_memory_p->set_data_handle(const_cast<T*>(out_grad_data));
}
// push primitive to stream and wait until it's executed
std::vector<mkldnn::primitive> pipeline{bwd_prim};
std::vector<mkldnn::primitive> pipeline{*(pool_bwd_p.get())};
mkldnn::stream(mkldnn::stream::kind::eager).submit(pipeline).wait();
} // Compute()
};
......
paddle/fluid/platform/CMakeLists.txt
浏览文件 @ 06aa23b0
......@@ -49,7 +49,7 @@ nv_test(device_context_test SRCS device_context_test.cu DEPS device_context gpu_
nv_test(cudnn_helper_test SRCS cudnn_helper_test.cc DEPS dynload_cuda)
nv_test(transform_test SRCS transform_test.cu DEPS memory place device_context)
cc_library(device_tracer SRCS device_tracer.cc DEPS boost profiler_proto ${GPU_CTX_DEPS})
cc_library(device_tracer SRCS device_tracer.cc DEPS boost profiler_proto framework_proto ${GPU_CTX_DEPS})
cc_library(profiler SRCS profiler.cc DEPS device_context device_tracer)
cc_test(profiler_test SRCS profiler_test.cc DEPS profiler)
......
paddle/fluid/platform/mkldnn_helper.h
浏览文件 @ 06aa23b0
......@@ -71,5 +71,15 @@ inline bool CanMKLDNNBeUsed(const framework::ExecutionContext& ctx) {
return use_mkldnn && platform::is_cpu_place(ctx.GetPlace());
}
template <typename Type>
mkldnn::memory::data_type MKLDNNGetDataType() {
return mkldnn::memory::data_undef;
}
template <>
inline mkldnn::memory::data_type MKLDNNGetDataType<float>() {
return mkldnn::memory::f32;
}
} // namespace platform
} // namespace paddle
paddle/fluid/platform/profiler.cc
浏览文件 @ 06aa23b0
......@@ -173,8 +173,9 @@ void PopEvent(const std::string& name, const DeviceContext* dev_ctx) {
}
RecordEvent::RecordEvent(const std::string& name, const DeviceContext* dev_ctx)
: start_ns_(PosixInNsec()) {
: is_enabled_(false), start_ns_(PosixInNsec()) {
if (g_state == ProfilerState::kDisabled) return;
is_enabled_ = true;
dev_ctx_ = dev_ctx;
name_ = name;
PushEvent(name_, dev_ctx_);
......@@ -183,7 +184,7 @@ RecordEvent::RecordEvent(const std::string& name, const DeviceContext* dev_ctx)
}
RecordEvent::~RecordEvent() {
if (g_state == ProfilerState::kDisabled) return;
if (g_state == ProfilerState::kDisabled || !is_enabled_) return;
DeviceTracer* tracer = GetDeviceTracer();
if (tracer) {
tracer->AddCPURecords(CurAnnotation(), start_ns_, PosixInNsec(),
......@@ -193,14 +194,16 @@ RecordEvent::~RecordEvent() {
PopEvent(name_, dev_ctx_);
}
RecordBlock::RecordBlock(int block_id) : start_ns_(PosixInNsec()) {
RecordBlock::RecordBlock(int block_id)
: is_enabled_(false), start_ns_(PosixInNsec()) {
if (g_state == ProfilerState::kDisabled) return;
is_enabled_ = true;
SetCurBlock(block_id);
name_ = string::Sprintf("block_%d", block_id);
}
RecordBlock::~RecordBlock() {
if (g_state == ProfilerState::kDisabled) return;
if (g_state == ProfilerState::kDisabled || !is_enabled_) return;
DeviceTracer* tracer = GetDeviceTracer();
if (tracer) {
// We try to put all blocks at the same nested depth in the
......
paddle/fluid/platform/profiler.h
浏览文件 @ 06aa23b0
......@@ -74,6 +74,7 @@ struct RecordEvent {
~RecordEvent();
bool is_enabled_;
uint64_t start_ns_;
// The device context is used by Event to get the current cuda stream.
const DeviceContext* dev_ctx_;
......@@ -89,6 +90,7 @@ struct RecordBlock {
~RecordBlock();
private:
bool is_enabled_;
std::string name_;
uint64_t start_ns_;
};
......
paddle/fluid/pybind/protobuf.cc
浏览文件 @ 06aa23b0
......@@ -238,6 +238,7 @@ void BindVarDsec(pybind11::module *m) {
pybind11::enum_<pd::proto::VarType::Type>(var_desc, "VarType", "")
.value("BOOL", pd::proto::VarType::BOOL)
.value("UINT8", pd::proto::VarType::UINT8)
.value("INT16", pd::proto::VarType::INT16)
.value("INT32", pd::proto::VarType::INT32)
.value("INT64", pd::proto::VarType::INT64)
......
paddle/scripts/docker/build.sh
浏览文件 @ 06aa23b0
......@@ -198,7 +198,7 @@ EOF
# run paddle version to install python packages first
RUN apt-get update &&\
${NCCL_DEPS}\
apt-get install -y wget python-pip dmidecode python-tk && pip install -U pip==9.0.3 && \
apt-get install -y wget python-pip dmidecode python-tk && easy_install -U pip && \
pip install /*.whl; apt-get install -f -y && \
apt-get clean -y && \
rm -f /*.whl && \
......
paddle/scripts/paddle_build.sh
浏览文件 @ 06aa23b0
......@@ -405,17 +405,19 @@ EOF
function gen_dockerfile() {
# Set BASE_IMAGE according to env variables
CUDA_MAJOR="$(echo $CUDA_VERSION | cut -d '.' -f 1).$(echo $CUDA_VERSION | cut -d '.' -f 2)"
CUDNN_MAJOR=$(echo $CUDNN_VERSION | cut -d '.' -f 1)
if [[ ${WITH_GPU} == "ON" ]]; then
BASE_IMAGE="nvidia/cuda:8.0-cudnn5-runtime-ubuntu16.04"
BASE_IMAGE="nvidia/cuda:${CUDA_MAJOR}-cudnn${CUDNN_MAJOR}-runtime-ubuntu16.04"
else
BASE_IMAGE="ubuntu:16.04"
BASE_IMAGE="ubuntu:16.04"
fi
DOCKERFILE_GPU_ENV=""
DOCKERFILE_CUDNN_DSO=""
if [[ ${WITH_GPU:-OFF} == 'ON' ]]; then
DOCKERFILE_GPU_ENV="ENV LD_LIBRARY_PATH /usr/lib/x86_64-linux-gnu:\${LD_LIBRARY_PATH}"
DOCKERFILE_CUDNN_DSO="RUN ln -s /usr/lib/x86_64-linux-gnu/libcudnn.so.5 /usr/lib/x86_64-linux-gnu/libcudnn.so"
DOCKERFILE_CUDNN_DSO="RUN ln -s /usr/lib/x86_64-linux-gnu/libcudnn.so.${CUDNN_MAJOR} /usr/lib/x86_64-linux-gnu/libcudnn.so"
fi
cat <<EOF
......@@ -449,7 +451,7 @@ EOF
# run paddle version to install python packages first
RUN apt-get update &&\
${NCCL_DEPS}\
apt-get install -y wget python-pip dmidecode python-tk && pip install -U pip==9.0.3 && \
apt-get install -y wget python-pip dmidecode python-tk && easy_install -U pip && \
pip install /*.whl; apt-get install -f -y && \
apt-get clean -y && \
rm -f /*.whl && \
......@@ -490,7 +492,7 @@ function gen_fluid_inference_lib() {
Deploying fluid inference library ...
========================================
EOF
make inference_lib_dist
make -j `nproc` inference_lib_dist
fi
}
......
python/paddle/fluid/framework.py
浏览文件 @ 06aa23b0
......@@ -72,6 +72,8 @@ def convert_np_dtype_to_dtype_(np_dtype):
return core.VarDesc.VarType.INT64
elif dtype == np.bool:
return core.VarDesc.VarType.BOOL
elif dtype == np.uint8:
return core.VarDesc.VarType.UINT8
else:
raise ValueError("Not supported numpy dtype " + str(dtype))
......
python/paddle/fluid/layers/control_flow.py
浏览文件 @ 06aa23b0
......@@ -49,6 +49,7 @@ __all__ = [
'reorder_lod_tensor_by_rank',
'ParallelDo',
'Print',
'is_empty',
]
......@@ -1562,3 +1563,40 @@ def reorder_lod_tensor_by_rank(x, rank_table):
'RankTable': [rank_table]},
outputs={'Out': [out]})
return out
def is_empty(x, cond=None, **ignored):
"""
**Is Empty**
This layer returns the truth value of whether the variable is empty.
Args:
x(Variable): Operand of *is_empty*
cond(Variable|None): Optional output variable to store the result
of *is_empty*
Returns:
Variable: The tensor variable storing the output of *is_empty*.
Raises:
TypeError: If input cond is not a variable, or cond's dtype is
not bool
Examples:
.. code-block:: python
less = fluid.layers.is_empty(x=input)
"""
helper = LayerHelper("is_empty", **locals())
if cond is None:
cond = helper.create_tmp_variable(dtype='bool')
cond.stop_gradient = True
elif not isinstance(cond, Variable):
raise TypeError("cond takes a variable")
elif cond.dtype != 'bool':
raise TypeError("The data type of cond must be bool")
helper.append_op(
type='is_empty', inputs={'X': [x]}, outputs={'Out': [cond]})
return cond
python/paddle/fluid/layers/detection.py
浏览文件 @ 06aa23b0
......@@ -23,6 +23,7 @@ import nn
import math
__all__ = [
'prior_box',
'multi_box_head',
'bipartite_match',
'target_assign',
......@@ -564,6 +565,98 @@ def ssd_loss(location,
return loss
def prior_box(input,
image,
min_sizes,
max_sizes=None,
aspect_ratios=None,
variance=[0.1, 0.1, 0.2, 0.2],
flip=False,
clip=False,
steps=[0.0, 0.0],
offset=0.5,
name=None):
"""
**Prior box operator**
Generate prior boxes for SSD(Single Shot MultiBox Detector) algorithm.
Each position of the input produce N prior boxes, N is determined by
the count of min_sizes, max_sizes and aspect_ratios, The size of the
box is in range(min_size, max_size) interval, which is generated in
sequence according to the aspect_ratios.
Args:
input(Variable): The Input Variables, the format is NCHW.
image(Variable): The input image data of PriorBoxOp,
the layout is NCHW.
min_sizes(list|tuple): min sizes of generated prior boxes.
max_sizes(list|tuple|None): max sizes of generated prior boxes.
Default: None.
aspect_ratios(list|tuple): the aspect ratios of generated prior
boxes. Default: None.
variance(list|tuple): the variances to be encoded in prior boxes.
Default:[0.1, 0.1, 0.2, 0.2].
flip(bool): Whether to flip aspect ratios. Default:False.
clip(bool): Whether to clip out-of-boundary boxes. Default: False.
step(list|turple): Prior boxes step across weight and height, If
step[0] == 0.0/step[1] == 0.0, the prior boxes step across
height/weight of the input will be automatically calculated.
Default: [0.0]
offset(float): Prior boxes center offset. Default: 0.5
name(str): Name of the prior box op. Default: None.
Returns:
boxes(Variable): the output prior boxes of PriorBox.
The layout is [H, W, num_priors, 4].
H is the height of input, W is the width of input,
num_priors is the total
box count of each position of input.
Variances(Variable): the expanded variances of PriorBox.
The layout is [H, W, num_priors, 4].
H is the height of input, W is the width of input
num_priors is the total
box count of each position of input
Examples:
.. code-block:: python
box, var = prior_box(
input=conv1,
image=images,
min_sizes=[100.],
flip=True,
clip=True)
"""
helper = LayerHelper("prior_box", **locals())
dtype = helper.input_dtype()
attrs = {
'min_sizes': min_sizes,
'aspect_ratios': aspect_ratios,
'variances': variance,
'flip': flip,
'clip': clip,
'step_w': steps[0],
'step_h': steps[1],
'offset': offset
}
if max_sizes is not None and len(max_sizes) > 0 and max_sizes[0] > 0:
attrs['max_sizes'] = max_sizes
box = helper.create_tmp_variable(dtype)
var = helper.create_tmp_variable(dtype)
helper.append_op(
type="prior_box",
inputs={"Input": input,
"Image": image},
outputs={"Boxes": box,
"Variances": var},
attrs=attrs, )
box.stop_gradient = True
var.stop_gradient = True
return box, var
def multi_box_head(inputs,
image,
base_size,
......@@ -660,47 +753,6 @@ def multi_box_head(inputs,
clip=True)
"""
def _prior_box_(input,
image,
min_sizes,
max_sizes,
aspect_ratios,
variance,
flip=False,
clip=False,
step_w=0.0,
step_h=0.0,
offset=0.5,
name=None):
helper = LayerHelper("prior_box", **locals())
dtype = helper.input_dtype()
attrs = {
'min_sizes': min_sizes,
'aspect_ratios': aspect_ratios,
'variances': variance,
'flip': flip,
'clip': clip,
'step_w': step_w,
'step_h': step_h,
'offset': offset
}
if len(max_sizes) > 0 and max_sizes[0] > 0:
attrs['max_sizes'] = max_sizes
box = helper.create_tmp_variable(dtype)
var = helper.create_tmp_variable(dtype)
helper.append_op(
type="prior_box",
inputs={"Input": input,
"Image": image},
outputs={"Boxes": box,
"Variances": var},
attrs=attrs, )
box.stop_gradient = True
var.stop_gradient = True
return box, var
def _reshape_with_axis_(input, axis=1):
if not (axis > 0 and axis < len(input.shape)):
raise ValueError("The axis should be smaller than "
......@@ -777,11 +829,10 @@ def multi_box_head(inputs,
aspect_ratio = aspect_ratios[i]
if not _is_list_or_tuple_(aspect_ratio):
aspect_ratio = [aspect_ratio]
step = [step_w[i] if step_w else 0.0, step_h[i] if step_w else 0.0]
box, var = _prior_box_(input, image, min_size, max_size, aspect_ratio,
variance, flip, clip, step_w[i]
if step_w else 0.0, step_h[i]
if step_w else 0.0, offset)
box, var = prior_box(input, image, min_size, max_size, aspect_ratio,
variance, flip, clip, step, offset)
box_results.append(box)
var_results.append(var)
......
python/paddle/fluid/tests/book/high-level-api/CMakeLists.txt
浏览文件 @ 06aa23b0
......@@ -8,3 +8,4 @@ endforeach()
add_subdirectory(fit_a_line)
add_subdirectory(recognize_digits)
add_subdirectory(image_classification)
python/paddle/fluid/tests/book/high-level-api/image_classification/CMakeLists.txt 0 → 100644
浏览文件 @ 06aa23b0
file(GLOB TEST_OPS RELATIVE "${CMAKE_CURRENT_SOURCE_DIR}" "test_*.py")
string(REPLACE ".py" "" TEST_OPS "${TEST_OPS}")
# default test
foreach(src ${TEST_OPS})
py_test(${src} SRCS ${src}.py)
endforeach()
python/paddle/fluid/tests/book/high-level-api/image_classification/cifar10_small_test_set.py 0 → 100644
浏览文件 @ 06aa23b0
# 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.
"""
CIFAR dataset.
This module will download dataset from
https://www.cs.toronto.edu/~kriz/cifar.html and parse train/test set into
paddle reader creators.
The CIFAR-10 dataset consists of 60000 32x32 colour images in 10 classes,
with 6000 images per class. There are 50000 training images and 10000 test
images.
The CIFAR-100 dataset is just like the CIFAR-10, except it has 100 classes
containing 600 images each. There are 500 training images and 100 testing
images per class.
"""
import cPickle
import itertools
import numpy
import paddle.v2.dataset.common
import tarfile
__all__ = ['train10']
URL_PREFIX = 'https://www.cs.toronto.edu/~kriz/'
CIFAR10_URL = URL_PREFIX + 'cifar-10-python.tar.gz'
CIFAR10_MD5 = 'c58f30108f718f92721af3b95e74349a'
def reader_creator(filename, sub_name, batch_size=None):
def read_batch(batch):
data = batch['data']
labels = batch.get('labels', batch.get('fine_labels', None))
assert labels is not None
for sample, label in itertools.izip(data, labels):
yield (sample / 255.0).astype(numpy.float32), int(label)
def reader():
with tarfile.open(filename, mode='r') as f:
names = (each_item.name for each_item in f
if sub_name in each_item.name)
batch_count = 0
for name in names:
batch = cPickle.load(f.extractfile(name))
for item in read_batch(batch):
if isinstance(batch_size, int) and batch_count > batch_size:
break
batch_count += 1
yield item
return reader
def train10(batch_size=None):
"""
CIFAR-10 training set creator.
It returns a reader creator, each sample in the reader is image pixels in
[0, 1] and label in [0, 9].
:return: Training reader creator
:rtype: callable
"""
return reader_creator(
paddle.v2.dataset.common.download(CIFAR10_URL, 'cifar', CIFAR10_MD5),
'data_batch',
batch_size=batch_size)
python/paddle/fluid/tests/book/high-level-api/image_classification/notest_image_classification_resnet.py → python/paddle/fluid/tests/book/high-level-api/image_classification/test_image_classification_resnet.py
浏览文件 @ 06aa23b0
......@@ -17,6 +17,7 @@ from __future__ import print_function
import paddle
import paddle.fluid as fluid
import numpy
import cifar10_small_test_set
def resnet_cifar10(input, depth=32):
......@@ -81,46 +82,50 @@ def train_network():
cost = fluid.layers.cross_entropy(input=predict, label=label)
avg_cost = fluid.layers.mean(cost)
accuracy = fluid.layers.accuracy(input=predict, label=label)
return avg_cost, accuracy
return [avg_cost, accuracy]
def train(use_cuda, save_path):
def train(use_cuda, train_program, save_dirname):
BATCH_SIZE = 128
EPOCH_NUM = 1
train_reader = paddle.batch(
paddle.reader.shuffle(
paddle.dataset.cifar.train10(), buf_size=128 * 10),
cifar10_small_test_set.train10(batch_size=10), buf_size=128 * 10),
batch_size=BATCH_SIZE)
test_reader = paddle.batch(
paddle.dataset.cifar.test10(), batch_size=BATCH_SIZE)
def event_handler(event):
if isinstance(event, fluid.EndIteration):
if (event.batch_id % 10) == 0:
avg_cost, accuracy = trainer.test(reader=test_reader)
if isinstance(event, fluid.EndStepEvent):
avg_cost, accuracy = trainer.test(
reader=test_reader, feed_order=['pixel', 'label'])
print('BatchID {1:04}, Loss {2:2.2}, Acc {3:2.2}'.format(
event.batch_id + 1, avg_cost, accuracy))
print('Loss {0:2.2}, Acc {1:2.2}'.format(avg_cost, accuracy))
if accuracy > 0.01: # Low threshold for speeding up CI
trainer.params.save(save_path)
return
if accuracy > 0.01: # Low threshold for speeding up CI
if save_dirname is not None:
trainer.save_params(save_dirname)
return
place = fluid.CUDAPlace(0) if use_cuda else fluid.CPUPlace()
trainer = fluid.Trainer(
train_network,
train_func=train_program,
optimizer=fluid.optimizer.Adam(learning_rate=0.001),
place=place,
event_handler=event_handler)
trainer.train(train_reader, EPOCH_NUM, event_handler=event_handler)
place=place)
trainer.train(
reader=train_reader,
num_epochs=EPOCH_NUM,
event_handler=event_handler,
feed_order=['pixel', 'label'])
def infer(use_cuda, save_path):
params = fluid.Params(save_path)
def infer(use_cuda, inference_program, save_dirname=None):
place = fluid.CUDAPlace(0) if use_cuda else fluid.CPUPlace()
inferencer = fluid.Inferencer(inference_network, params, place=place)
inferencer = fluid.Inferencer(
infer_func=inference_program, param_path=save_dirname, place=place)
# The input's dimension of conv should be 4-D or 5-D.
# Use normilized image pixels as input data, which should be in the range
......@@ -135,8 +140,14 @@ def main(use_cuda):
if use_cuda and not fluid.core.is_compiled_with_cuda():
return
save_path = "image_classification_resnet.inference.model"
train(use_cuda, save_path)
infer(use_cuda, save_path)
train(
use_cuda=use_cuda, train_program=train_network, save_dirname=save_path)
infer(
use_cuda=use_cuda,
inference_program=inference_network,
save_dirname=save_path)
if __name__ == '__main__':
......
python/paddle/fluid/tests/book/high-level-api/image_classification/notest_image_classification_vgg.py → python/paddle/fluid/tests/book/high-level-api/image_classification/test_image_classification_vgg.py
浏览文件 @ 06aa23b0
......@@ -17,6 +17,7 @@ from __future__ import print_function
import paddle
import paddle.fluid as fluid
import numpy
import cifar10_small_test_set
def vgg16_bn_drop(input):
......@@ -60,46 +61,48 @@ def train_network():
cost = fluid.layers.cross_entropy(input=predict, label=label)
avg_cost = fluid.layers.mean(cost)
accuracy = fluid.layers.accuracy(input=predict, label=label)
return avg_cost, accuracy
return [avg_cost, accuracy]
def train(use_cuda, save_path):
def train(use_cuda, train_program, save_dirname):
BATCH_SIZE = 128
EPOCH_NUM = 1
train_reader = paddle.batch(
paddle.reader.shuffle(
paddle.dataset.cifar.train10(), buf_size=128 * 10),
cifar10_small_test_set.train10(batch_size=10), buf_size=128 * 10),
batch_size=BATCH_SIZE)
test_reader = paddle.batch(
paddle.dataset.cifar.test10(), batch_size=BATCH_SIZE)
def event_handler(event):
if isinstance(event, fluid.EndIteration):
if (event.batch_id % 10) == 0:
avg_cost, accuracy = trainer.test(reader=test_reader)
if isinstance(event, fluid.EndStepEvent):
avg_cost, accuracy = trainer.test(
reader=test_reader, feed_order=['pixel', 'label'])
print('BatchID {1:04}, Loss {2:2.2}, Acc {3:2.2}'.format(
event.batch_id + 1, avg_cost, accuracy))
print('Loss {0:2.2}, Acc {1:2.2}'.format(avg_cost, accuracy))
if accuracy > 0.01: # Low threshold for speeding up CI
trainer.params.save(save_path)
return
if accuracy > 0.01: # Low threshold for speeding up CI
if save_dirname is not None:
trainer.save_params(save_dirname)
return
place = fluid.CUDAPlace(0) if use_cuda else fluid.CPUPlace()
trainer = fluid.Trainer(
train_network,
optimizer=fluid.optimizer.Adam(learning_rate=0.001),
train_func=train_program,
place=place,
event_handler=event_handler)
trainer.train(train_reader, EPOCH_NUM, event_handler=event_handler)
optimizer=fluid.optimizer.Adam(learning_rate=0.001))
trainer.train(
reader=train_reader,
num_epochs=1,
event_handler=event_handler,
feed_order=['pixel', 'label'])
def infer(use_cuda, save_path):
params = fluid.Params(save_path)
def infer(use_cuda, inference_program, save_dirname=None):
place = fluid.CUDAPlace(0) if use_cuda else fluid.CPUPlace()
inferencer = fluid.Inferencer(inference_network, params, place=place)
inferencer = fluid.Inferencer(
infer_func=inference_program, param_path=save_dirname, place=place)
# The input's dimension of conv should be 4-D or 5-D.
# Use normilized image pixels as input data, which should be in the range
......@@ -114,8 +117,14 @@ def main(use_cuda):
if use_cuda and not fluid.core.is_compiled_with_cuda():
return
save_path = "image_classification_vgg.inference.model"
train(use_cuda, save_path)
infer(use_cuda, save_path)
train(
use_cuda=use_cuda, train_program=train_network, save_dirname=save_path)
infer(
use_cuda=use_cuda,
inference_program=inference_network,
save_dirname=save_path)
if __name__ == '__main__':
......
python/paddle/fluid/tests/book/high-level-api/word2vec/no_test_word2vec_new_api.py → python/paddle/fluid/tests/book/high-level-api/word2vec/test_word2vec_new_api.py
浏览文件 @ 06aa23b0
......@@ -90,7 +90,7 @@ def train_program(is_sparse):
return avg_cost
def train(use_cuda, train_program, save_path):
def train(use_cuda, train_program, save_dirname):
train_reader = paddle.batch(
paddle.dataset.imikolov.train(word_dict, N), BATCH_SIZE)
test_reader = paddle.batch(
......@@ -99,27 +99,36 @@ def train(use_cuda, train_program, save_path):
place = fluid.CUDAPlace(0) if use_cuda else fluid.CPUPlace()
def event_handler(event):
if isinstance(event, fluid.EndEpochEvent):
outs = trainer.test(reader=test_reader)
if isinstance(event, fluid.EndStepEvent):
outs = trainer.test(
reader=test_reader,
feed_order=['firstw', 'secondw', 'thirdw', 'forthw', 'nextw'])
avg_cost = outs[0]
print("loss= ", avg_cost)
if avg_cost < 5.0:
trainer.save_params(save_path)
return
if avg_cost < 10.0:
trainer.save_params(save_dirname)
trainer.stop()
if math.isnan(avg_cost):
sys.exit("got NaN loss, training failed.")
trainer = fluid.Trainer(
train_program, fluid.optimizer.SGD(learning_rate=0.001), place=place)
train_func=train_program,
optimizer=fluid.optimizer.SGD(learning_rate=0.001),
place=place)
trainer.train(
reader=train_reader, num_epochs=1, event_handler=event_handler)
reader=train_reader,
num_epochs=1,
event_handler=event_handler,
feed_order=['firstw', 'secondw', 'thirdw', 'forthw', 'nextw'])
def infer(use_cuda, inference_program, save_path):
def infer(use_cuda, inference_program, save_dirname=None):
place = fluid.CUDAPlace(0) if use_cuda else fluid.CPUPlace()
inferencer = fluid.Inferencer(
infer_func=inference_program, param_path=save_path, place=place)
infer_func=inference_program, param_path=save_dirname, place=place)
lod = [0, 1]
first_word = create_random_lodtensor(lod, place, low=0, high=dict_size - 1)
......@@ -127,12 +136,14 @@ def infer(use_cuda, inference_program, save_path):
third_word = create_random_lodtensor(lod, place, low=0, high=dict_size - 1)
fourth_word = create_random_lodtensor(lod, place, low=0, high=dict_size - 1)
result = inferencer.infer({
'firstw': first_word,
'secondw': second_word,
'thirdw': third_word,
'forthw': fourth_word
})
result = inferencer.infer(
{
'firstw': first_word,
'secondw': second_word,
'thirdw': third_word,
'forthw': fourth_word
},
return_numpy=False)
print(np.array(result[0]))
......@@ -140,9 +151,17 @@ def main(use_cuda, is_sparse):
if use_cuda and not fluid.core.is_compiled_with_cuda():
return
save_path = "word2vec.params"
train(use_cuda, partial(train_program, is_sparse), save_path)
infer(use_cuda, partial(inference_program, is_sparse), save_path)
save_path = "word2vec.inference.model"
train(
use_cuda=use_cuda,
train_program=partial(train_program, is_sparse),
save_dirname=save_path)
infer(
use_cuda=use_cuda,
inference_program=partial(inference_program, is_sparse),
save_dirname=save_path)
if __name__ == '__main__':
......
python/paddle/fluid/tests/test_detection.py
浏览文件 @ 06aa23b0
......@@ -109,6 +109,24 @@ class TestDetection(unittest.TestCase):
print(str(program))
class TestPriorBox(unittest.TestCase):
def test_prior_box(self):
data_shape = [3, 224, 224]
images = fluid.layers.data(
name='pixel', shape=data_shape, dtype='float32')
conv1 = fluid.layers.conv2d(images, 3, 3, 2)
box, var = layers.prior_box(
input=conv1,
image=images,
min_sizes=[100.0],
aspect_ratios=[1.],
flip=True,
clip=True)
assert len(box.shape) == 4
assert box.shape == var.shape
assert box.shape[3] == 4
class TestMultiBoxHead(unittest.TestCase):
def test_multi_box_head(self):
data_shape = [3, 224, 224]
......
python/paddle/fluid/tests/unittests/test_is_empty_op.py
浏览文件 @ 06aa23b0
......@@ -14,42 +14,24 @@
import unittest
import numpy as np
from paddle.fluid.op import Operator
import paddle.fluid.core as core
from op_test import OpTest
def create_tensor(scope, name, np_data):
tensor = scope.var(name).get_tensor()
tensor.set_dims(np_data.shape)
tensor.set(np_data, core.CPUPlace())
return tensor
class TestIsEmptyOp(unittest.TestCase):
class TestEmpty(OpTest):
def setUp(self):
self.scope = core.Scope()
# create input variables
np_data0 = np.array([0, 1, 2])
create_tensor(self.scope, "X0", np_data0)
np_data1 = np.array([1])
t = create_tensor(self.scope, "X1", np_data1)
t.set_dims([0])
self.op_type = "is_empty"
self.inputs = {'X': np.array([1, 2, 3])}
self.outputs = {'Out': np.array([False])}
# create output variables
self.scope.var("out")
def test_check_output(self):
self.check_output()
def test_no_empty(self):
self.one_case("X0", False)
def test_empty(self):
self.one_case("X1", True)
def one_case(self, input, target):
op = Operator(type="is_empty", X=input, Out="out")
op.run(self.scope, core.CPUPlace())
out = self.scope.var("out").get_tensor()
self.assertEqual(np.array(out)[0], target)
class TestNotEmpty(TestEmpty):
def setUp(self):
self.op_type = "is_empty"
self.inputs = {'X': np.array([])}
self.outputs = {'Out': np.array([True])}
if __name__ == "__main__":
......
python/paddle/fluid/transpiler/memory_optimization_transpiler.py
浏览文件 @ 06aa23b0
......@@ -24,7 +24,8 @@ dtype_to_size = {
core.VarDesc.VarType.INT16: 2,
core.VarDesc.VarType.INT32: 4,
core.VarDesc.VarType.INT64: 8,
core.VarDesc.VarType.BOOL: 1
core.VarDesc.VarType.BOOL: 1,
core.VarDesc.VarType.UINT8: 1,
}
SUB_BLOCK_OPS = [
......
Markdown is supported
0% 或 .
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 或