diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml index 83fe9af768964003130d02b7d913ad1c2102dd1d..59661c9c1da53a2ddac0127ed1827fedde811a1d 100644 --- a/.pre-commit-config.yaml +++ b/.pre-commit-config.yaml @@ -31,6 +31,3 @@ - id: go-fmt types: - go - - id: gometalinter - types: - - go diff --git a/CMakeLists.txt b/CMakeLists.txt index 4783095194dc9c6409dc31c95588f46c9bee7c61..1252e7539816016dfdf1b90b8941fa42e6bb85e0 100644 --- a/CMakeLists.txt +++ b/CMakeLists.txt @@ -105,6 +105,12 @@ if (WITH_C_API AND WITH_PYTHON) "different Python interpreter from compiling.") endif() +if(MOBILE_INFERENCE) + set(THIRD_PARTY_BUILD_TYPE MinSizeRel) +else() + set(THIRD_PARTY_BUILD_TYPE Release) +endif() + ######################################################################################## include(external/mklml) # download mklml package diff --git a/cmake/external/eigen.cmake b/cmake/external/eigen.cmake index f7483f6be9169eb58f0148cd3a956a8c881e1fe3..bd853d921b4362ac7ac5e17e629552b2a200f08a 100644 --- a/cmake/external/eigen.cmake +++ b/cmake/external/eigen.cmake @@ -8,7 +8,7 @@ ExternalProject_Add( extern_eigen3 ${EXTERNAL_PROJECT_LOG_ARGS} GIT_REPOSITORY "https://github.com/RLovelett/eigen.git" - GIT_TAG "master" + GIT_TAG 4e79cb69b9425f5f8c3a84be4350d4ab75b5fd9d PREFIX ${EIGEN_SOURCE_DIR} UPDATE_COMMAND "" CONFIGURE_COMMAND "" diff --git a/cmake/external/gflags.cmake b/cmake/external/gflags.cmake index 957f8271e4841836956b0c3f2cf3d8c88a31192a..c819eb4d70898e48eab499c666168d78262d4240 100644 --- a/cmake/external/gflags.cmake +++ b/cmake/external/gflags.cmake @@ -36,6 +36,7 @@ ExternalProject_Add( # change this back to the official Github repo once my PR is # merged. GIT_REPOSITORY "https://github.com/wangkuiyi/gflags.git" + GIT_TAG 986964c07427ecb9cdb5bd73f73ebbd40e54dadb PREFIX ${GFLAGS_SOURCES_DIR} UPDATE_COMMAND "" CMAKE_ARGS -DCMAKE_CXX_COMPILER=${CMAKE_CXX_COMPILER} @@ -45,11 +46,11 @@ ExternalProject_Add( -DCMAKE_INSTALL_PREFIX=${GFLAGS_INSTALL_DIR} -DCMAKE_POSITION_INDEPENDENT_CODE=ON -DBUILD_TESTING=OFF - -DCMAKE_BUILD_TYPE=Release + -DCMAKE_BUILD_TYPE=${THIRD_PARTY_BUILD_TYPE} ${EXTERNAL_OPTIONAL_ARGS} CMAKE_CACHE_ARGS -DCMAKE_INSTALL_PREFIX:PATH=${GFLAGS_INSTALL_DIR} -DCMAKE_POSITION_INDEPENDENT_CODE:BOOL=ON - -DCMAKE_BUILD_TYPE:STRING=Release + -DCMAKE_BUILD_TYPE:STRING=${THIRD_PARTY_BUILD_TYPE} ) ADD_LIBRARY(gflags STATIC IMPORTED GLOBAL) diff --git a/cmake/external/glog.cmake b/cmake/external/glog.cmake index b3fef738ccc0b5886bb0a32501bb7b7adade0ff1..08bdc1e1623b0d917061c7368e9b2a8f7e9517fd 100644 --- a/cmake/external/glog.cmake +++ b/cmake/external/glog.cmake @@ -31,6 +31,7 @@ ExternalProject_Add( ${EXTERNAL_PROJECT_LOG_ARGS} DEPENDS gflags GIT_REPOSITORY "https://github.com/google/glog.git" + GIT_TAG v0.3.5 PREFIX ${GLOG_SOURCES_DIR} UPDATE_COMMAND "" CMAKE_ARGS -DCMAKE_CXX_COMPILER=${CMAKE_CXX_COMPILER} @@ -43,12 +44,12 @@ ExternalProject_Add( -DWITH_GFLAGS=ON -Dgflags_DIR=${GFLAGS_INSTALL_DIR}/lib/cmake/gflags -DBUILD_TESTING=OFF - -DCMAKE_BUILD_TYPE=Release + -DCMAKE_BUILD_TYPE=${THIRD_PARTY_BUILD_TYPE} ${EXTERNAL_OPTIONAL_ARGS} CMAKE_CACHE_ARGS -DCMAKE_INSTALL_PREFIX:PATH=${GLOG_INSTALL_DIR} -DCMAKE_INSTALL_LIBDIR:PATH=${GLOG_INSTALL_DIR}/lib -DCMAKE_POSITION_INDEPENDENT_CODE:BOOL=ON - -DCMAKE_BUILD_TYPE:STRING=Release + -DCMAKE_BUILD_TYPE:STRING=${THIRD_PARTY_BUILD_TYPE} ) ADD_LIBRARY(glog STATIC IMPORTED GLOBAL) diff --git a/cmake/external/gtest.cmake b/cmake/external/gtest.cmake index 6a2a79b7631b32e8a099797de509af64533bbb95..5a4aa7a5b71a4fdfd556a46037e6d1846d668fc4 100644 --- a/cmake/external/gtest.cmake +++ b/cmake/external/gtest.cmake @@ -56,11 +56,11 @@ IF(WITH_TESTING) -DBUILD_GMOCK=ON -Dgtest_disable_pthreads=ON -Dgtest_force_shared_crt=ON - -DCMAKE_BUILD_TYPE=Release + -DCMAKE_BUILD_TYPE=${THIRD_PARTY_BUILD_TYPE} ${EXTERNAL_OPTIONAL_ARGS} CMAKE_CACHE_ARGS -DCMAKE_INSTALL_PREFIX:PATH=${GTEST_INSTALL_DIR} -DCMAKE_POSITION_INDEPENDENT_CODE:BOOL=ON - -DCMAKE_BUILD_TYPE:STRING=Release + -DCMAKE_BUILD_TYPE:STRING=${THIRD_PARTY_BUILD_TYPE} ) ADD_LIBRARY(gtest STATIC IMPORTED GLOBAL) diff --git a/cmake/external/protobuf.cmake b/cmake/external/protobuf.cmake index 7cf7ba85cca4c248dcc74e078124c0b3815ee380..be7f6a9465970711170bd15dcecaadeaa8a55f86 100644 --- a/cmake/external/protobuf.cmake +++ b/cmake/external/protobuf.cmake @@ -191,12 +191,12 @@ FUNCTION(build_protobuf TARGET_NAME BUILD_FOR_HOST) ${OPTIONAL_ARGS} -Dprotobuf_BUILD_TESTS=OFF -DCMAKE_POSITION_INDEPENDENT_CODE=ON - -DCMAKE_BUILD_TYPE=Release + -DCMAKE_BUILD_TYPE=${THIRD_PARTY_BUILD_TYPE} -DCMAKE_INSTALL_PREFIX=${PROTOBUF_INSTALL_DIR} -DCMAKE_INSTALL_LIBDIR=lib CMAKE_CACHE_ARGS -DCMAKE_INSTALL_PREFIX:PATH=${PROTOBUF_INSTALL_DIR} - -DCMAKE_BUILD_TYPE:STRING=Release + -DCMAKE_BUILD_TYPE:STRING=${THIRD_PARTY_BUILD_TYPE} -DCMAKE_VERBOSE_MAKEFILE:BOOL=OFF -DCMAKE_POSITION_INDEPENDENT_CODE:BOOL=ON ${OPTIONAL_CACHE_ARGS} diff --git a/cmake/external/warpctc.cmake b/cmake/external/warpctc.cmake index bb258c7b5581fc22b44f4fe15c119f8081f4767e..8bd058222880b4df3b08da09c02f9fe7f1d0ee66 100644 --- a/cmake/external/warpctc.cmake +++ b/cmake/external/warpctc.cmake @@ -35,6 +35,7 @@ ExternalProject_Add( extern_warpctc ${EXTERNAL_PROJECT_LOG_ARGS} GIT_REPOSITORY "https://github.com/gangliao/warp-ctc.git" + GIT_TAG b63a0644654a3e0ed624c85a1767bc8193aead09 PREFIX ${WARPCTC_SOURCES_DIR} UPDATE_COMMAND "" CMAKE_ARGS -DCMAKE_CXX_COMPILER=${CMAKE_CXX_COMPILER} @@ -48,9 +49,9 @@ ExternalProject_Add( -DCMAKE_DISABLE_FIND_PACKAGE_Torch=ON -DBUILD_SHARED=ON -DCMAKE_POSITION_INDEPENDENT_CODE=ON - -DCMAKE_BUILD_TYPE=Release + -DCMAKE_BUILD_TYPE=${THIRD_PARTY_BUILD_TYPE} ${EXTERNAL_OPTIONAL_ARGS} - CMAKE_CACHE_ARGS -DCMAKE_BUILD_TYPE:STRING=Release + CMAKE_CACHE_ARGS -DCMAKE_BUILD_TYPE:STRING=${THIRD_PARTY_BUILD_TYPE} -DCMAKE_POSITION_INDEPENDENT_CODE:BOOL=ON -DCMAKE_INSTALL_PREFIX:PATH=${WARPCTC_INSTALL_DIR} ) diff --git a/cmake/external/zlib.cmake b/cmake/external/zlib.cmake index c496a52b780364f3014f8fa3dfbc944a7aa7430e..e2c9fe56f335ae5b627b4d8d4bb17e4a2a466677 100644 --- a/cmake/external/zlib.cmake +++ b/cmake/external/zlib.cmake @@ -42,11 +42,11 @@ ExternalProject_Add( -DBUILD_SHARED_LIBS=OFF -DCMAKE_POSITION_INDEPENDENT_CODE=ON -DCMAKE_MACOSX_RPATH=ON - -DCMAKE_BUILD_TYPE=Release + -DCMAKE_BUILD_TYPE=${THIRD_PARTY_BUILD_TYPE} ${EXTERNAL_OPTIONAL_ARGS} CMAKE_CACHE_ARGS -DCMAKE_INSTALL_PREFIX:PATH=${ZLIB_INSTALL_DIR} -DCMAKE_POSITION_INDEPENDENT_CODE:BOOL=ON - -DCMAKE_BUILD_TYPE:STRING=Release + -DCMAKE_BUILD_TYPE:STRING=${THIRD_PARTY_BUILD_TYPE} ) LIST(APPEND external_project_dependencies zlib) diff --git a/cmake/generic.cmake b/cmake/generic.cmake index ff9868fc4e0d970b11e4763d2e0c8581f4f85907..c311783aa3187678c31c27ddbbd074790ca444f3 100644 --- a/cmake/generic.cmake +++ b/cmake/generic.cmake @@ -389,13 +389,60 @@ function(go_test TARGET_NAME) WORKING_DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR}) endfunction(go_test) +# Modification of standard 'protobuf_generate_cpp()' with protobuf-lite support +# Usage: +# paddle_protobuf_generate_cpp( ) + +function(paddle_protobuf_generate_cpp SRCS HDRS) + if(NOT ARGN) + message(SEND_ERROR "Error: paddle_protobuf_generate_cpp() called without any proto files") + return() + endif() + + set(${SRCS}) + set(${HDRS}) + + if (MOBILE_INFERENCE) + set(EXTRA_FLAG "lite:") + else() + set(EXTRA_FLAG "") + endif() + + foreach(FIL ${ARGN}) + get_filename_component(ABS_FIL ${FIL} ABSOLUTE) + get_filename_component(FIL_WE ${FIL} NAME_WE) + + set(_protobuf_protoc_src "${CMAKE_CURRENT_BINARY_DIR}/${FIL_WE}.pb.cc") + set(_protobuf_protoc_hdr "${CMAKE_CURRENT_BINARY_DIR}/${FIL_WE}.pb.h") + list(APPEND ${SRCS} "${_protobuf_protoc_src}") + list(APPEND ${HDRS} "${_protobuf_protoc_hdr}") + + add_custom_command( + OUTPUT "${_protobuf_protoc_src}" + "${_protobuf_protoc_hdr}" + + COMMAND ${CMAKE_COMMAND} -E make_directory "${CMAKE_CURRENT_BINARY_DIR}" + COMMAND ${PROTOBUF_PROTOC_EXECUTABLE} + -I${CMAKE_CURRENT_SOURCE_DIR} + --cpp_out "${EXTRA_FLAG}${CMAKE_CURRENT_BINARY_DIR}" ${ABS_FIL} + DEPENDS ${ABS_FIL} protoc + COMMENT "Running C++ protocol buffer compiler on ${FIL}" + VERBATIM ) + endforeach() + + set_source_files_properties(${${SRCS}} ${${HDRS}} PROPERTIES GENERATED TRUE) + set(${SRCS} ${${SRCS}} PARENT_SCOPE) + set(${HDRS} ${${HDRS}} PARENT_SCOPE) +endfunction() + + function(proto_library TARGET_NAME) set(oneValueArgs "") set(multiValueArgs SRCS DEPS) cmake_parse_arguments(proto_library "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN}) set(proto_srcs) set(proto_hdrs) - protobuf_generate_cpp(proto_srcs proto_hdrs ${proto_library_SRCS}) + paddle_protobuf_generate_cpp(proto_srcs proto_hdrs ${proto_library_SRCS}) cc_library(${TARGET_NAME} SRCS ${proto_srcs} DEPS ${proto_library_DEPS} protobuf) endfunction() diff --git a/doc/design/block.md b/doc/design/block.md index 9c812732d6ead76eb3aa2d1b617449c96807f21a..7cbf0d55b1faeb2093ee7cf234d1c2ad1905885b 100644 --- a/doc/design/block.md +++ b/doc/design/block.md @@ -243,7 +243,7 @@ class SymbolTable { // TODO determine whether name is generated by python or C++. // Currently assume that a unique name will be generated by C++ if the // argument name is left default. - VarDesc* NewVar(const string& name=""); + VarDesc* Var(const string& name=""); // find a VarDesc by name, if recursive is true, find parent's SymbolTable // recursively. diff --git a/doc/design/executor.md b/doc/design/executor.md new file mode 100644 index 0000000000000000000000000000000000000000..b5fb6c5c3c1da3c112ce63878322083dd5c42b70 --- /dev/null +++ b/doc/design/executor.md @@ -0,0 +1,23 @@ +# Executor Design Doc + +## Motivation + +We use executor to do the runtime evaluation of a `ProgramDesc`. + +## Overview + +An executor takes a `ProgramDesc`, a `block_id` and a `Scope`. The `ProgramDesc` is a list of blocks and each block contains the protobuf definition of all the parameters and operators. The `block_id` specifies the entrance block. And the `Scope` is the container of all the variable instance, which is persistent throughout different runs. + +### What does executor do? + +It evaluates all the operators in the `block_id`th block of a `ProgramDesc`. + +### What does executor NOT do? + +It does not do runtime optimization, meaning intelligently parse the dependency of each op a choose which one to be run and in which order they should be run. + +It does not do graph partitioning, meaning dividing the `ProgramDesc` into several small pieces and executing them on different devices. + +## Implementation + +`Executor` evaluates a `ProgramDesc`. Essentially, it instantiates Variables and Operators, then run all the operators in sequence. [[code]](https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/framework/executor.cc) diff --git a/doc/design/images/graph_construction_example.dot b/doc/design/images/graph_construction_example.dot index 8d1b673abf6b78c851676fa379dc850c4818f0e5..e115f9844bae6ad24f638c8ed4749cea8aff06a9 100644 --- a/doc/design/images/graph_construction_example.dot +++ b/doc/design/images/graph_construction_example.dot @@ -33,7 +33,6 @@ digraph ImageClassificationGraph { cost -> MSE_Grad [color=red]; d_cost -> MSE_Grad [color=red]; - x -> MSE_Grad [color=red]; l -> MSE_Grad [color=red]; y -> MSE_Grad -> d_y [color=red]; diff --git a/doc/design/images/graph_construction_example_all.png b/doc/design/images/graph_construction_example_all.png index 181187503472d15779b87284105841168b3945c4..261611a5721f9aa97874f7e6d897fe48cf667db2 100644 Binary files a/doc/design/images/graph_construction_example_all.png and b/doc/design/images/graph_construction_example_all.png differ diff --git a/doc/design/images/graph_construction_example_forward_backward.png b/doc/design/images/graph_construction_example_forward_backward.png index 3049a9315fd616464dec54e33064cb75598ca536..4c69687f4a6a181138f3df72ce5e8aa48487b5be 100644 Binary files a/doc/design/images/graph_construction_example_forward_backward.png and b/doc/design/images/graph_construction_example_forward_backward.png differ diff --git a/doc/design/images/graph_construction_example_forward_only.png b/doc/design/images/graph_construction_example_forward_only.png index 25d19088cbf0b5f68cf734f2ff21eba8af4a2860..e668c16e0cac73acb4e5dc2b1827557ae77126b4 100644 Binary files a/doc/design/images/graph_construction_example_forward_only.png and b/doc/design/images/graph_construction_example_forward_only.png differ diff --git a/doc/design/infer_var_type.md b/doc/design/infer_var_type.md new file mode 100644 index 0000000000000000000000000000000000000000..d9d5397becba2ef1806d9341cd49cd9aabbf4a6a --- /dev/null +++ b/doc/design/infer_var_type.md @@ -0,0 +1,78 @@ +# Design Doc: InferVarType + +## The Problem Posed + +The variable in our design can hold variant types. Such as `LoDTensor` and `SelectedRows`. An operator should be able to inference the variable types of its output. + +For example, a `lookup table` operator takes two `LoDTensor`; one is a float tensor as the embedding table, the other is an int tensor as word ID. The gradient operator of `lookup table` will generate a `SelectedRows` as its output. A `sum` operator can take both `LoDTensor` and `SelectedRows` as its inputs and will generate a `LoDTensor` if any of its inputs is `LoDTensor`, otherwise, the `sum` operator will generate `SelectedRows` as its output. + +The variable type will be constant at runtime. Every variable's type can either be set by the user (input data and parameter) or be inferred by the operator in compile time. + +## Proposed Solution + +The `InferVarType` is a compile-time function which is registered to each operator. The inferface of that function is: + + +```c++ +using InferVarTypeFN = std::function< + void (const OpDescBind& /*op_desc*/, BlockDescBind* /*block*/)>; +``` + +It takes an operator description as its input and will write the output variable type and store them in block description. + +The `InferVarTypeFN` will be registered in `OpInfo`, to replace `infer_var_type_` field. The `OpInfo` should be + +```cpp +struct OpInfo { + InferVarTypeFN infer_var_type_; + ... +}; +``` + +The default `InferVarType` will set output type as `LoDTensor`. It can be done by `GetInferVarType()`. + +```cpp +void DefaultInferVarType(const OpDescBind& op_desc, BlockDescBind* block) { + // set the output type of variable as `LoDTensor`. + // ... +} + +struct OpInfo { + InferVarTypeFN infer_var_type_; + InferVarTypeFN GetInferVarType() const { + if (infer_var_type_) { + return infer_var_type_; + } else { + return DefaultInferVarType; + } + } +}; +``` + +## Register InferVarType + +We provide a thin base class for registering an `InferVarTypeFN`. To use a base class will ease the implementation of registry since we can detect the registry entry is an `InferVarTypeFN` or not. + +```cpp +class VarTypeInferer { +public: + virtual void operator()(const OpDescBind& op_desc, BlockDescBind* block) const = 0; +} +``` + +Operator developers can write the specialize `VarTypeInferer` as follow. + +```cpp +class SpecialVarTypeInferer : public VarTypeInferer { +public: + virtual void operator()(const OpDescBind& op_desc, BlockDescBind* block) const { + // .. own logic + } +} +``` + +Then user can register the `InferVarType` just like `GradOpDescMaker` and `OpInfoMaker`. + +``` +REGISTER_OPERATOR(some_op, OpType, SpecialVarTypeInferer, ...); +``` diff --git a/doc/design/python_api.md b/doc/design/python_api.md index 56ae1d925a96622b5576013f38e33e5f92cbbb90..cb5fdc765b7126fc66a1c8978d4b96c0dc5a9f2c 100644 --- a/doc/design/python_api.md +++ b/doc/design/python_api.md @@ -179,40 +179,104 @@ init_attr={ `optimize_op_attrs` is not in the `VarDesc` message, but kept in the Python instance, as it will be used in the Python space when creating the optimize operator's `OpDesc`, and will be in the `OpDesc` message. -## Layer Functions +## Layer Function -A layer is a Python function that creates some operators and variables. Layers simplify the work of application programmers. +A layer is a Python function that creates some operators and variables. Layers simplify the work of application programmers. -### Data Layer +Layer functions take `Variable` and configuration parameters as its input and return the output variable(s). + +For example, `FullyConnected` take one or more variable as its input. The input could be input data or another layer's output. There are many configuration options for a `FullyConnected` layer, such as layer size, activation, parameter names, initialization strategies of parameters, and so on. The `FullyConnected` layer will return an output variable. + + +### Necessity for reusing code between layer functions + +There are a lot of code that can be reused. Such as + +* Give the default value of configuration. e.g., default initialize strategy for parameters is uniform random with `min = -1.0`, `max = 1.0`. and default initialize strategy for bias is to fill zero. +* Append the activation operator. +* Create a temporary variable. +* Create parameter. +* Generate a unique name. +* Add a bias. +* ... + +A mechanism to reuse code between layer functions is necessary. It will be around [150 lines of code](https://github.com/PaddlePaddle/Paddle/pull/4724/files#diff-823b27e07e93914ada859232ae23f846R12) if we write a `FullyConnected` layer without any helper functions. + + + +### Comparision between global functions and helper class + +The `FullyConnected` layer will be as follow when we provide global functions: ```python -def data_layer(name, type, column_name): - block = the_current_program.glolal_block() - var = block.create_global_var( - name=name, - shape=[None] + type.dims(), - dtype=type.dtype) - block.prepend_operator(block, - type="Feed", - inputs = None, - outputs = [var], - {column_name: column_name}) - return var +def fc_layer(input, size, param_attr=None, bias_attr=None, act=None, name=None): + if name is None: + name = unique_name("fc") + input = multiple_input(input) + param_attr = default_param_attr(param_attr) + param_attr = multiple_param_attr(param_attr, len(input)) + + # mul + mul_results = [] + for ipt, attr in zip(input, param_attr): + shape = ipt.shape[1:] + [size] + w = g_program.global_block().create_parameter(shape, ipt.dtype, name, attr) + tmp = create_tmp_var(name) + g_program.current_block().append_op("mul", {ipt, w}, {tmp}) + mul_results.append(tmp) + + # add sum + ... + # add bias + ... + # add activation + ... + return out ``` -The input to the feed operator is a special variable in the global scope, which is the output of [Python readers](https://github.com/PaddlePaddle/Paddle/blob/develop/doc/design/reader/README.md). +We can provide many helpers functions for layer developers. However, there are several disadvantages for global helper functions: + +1. We need a namespace for these methods, then layer developers can quickly figure out what method they can use. +2. Global functions will force layer developers to pass its parameter time by time. + +So we provide a helper class, `LayerHelper`, to share code between layer functions. The `FullyConnected` Layer will be as follow. + +```python +def fc_layer(input, size, param_attr=None, bias_attr=None, act=None, name=None): + helper = LayerHelper(locals()) # pass all parameter to LayerHelper + + mul_results = [] + for ipt, param in helper.iter_multiple_input_and_param(): + w = helper.create_parameter(shape=ipt.shape[1:] + [size], dtype = ipt.dtype) + tmp = helper.create_tmp_variable() + helper.append_op('mul', {ipt, w}, {tmp}) + mul_results.append(tmp) + + pre_bias = helper.add_sum(mul_results) + pre_activation = helper.add_bias(pre_bias) + return helper.add_activation(pre_activation) +``` + +We not only use the fewer lines of code to write `fc_layer` but also make the code clearer to understand. At the same time, layer developers can figure out what function they can invoke by typing `helper.` in a python editor. + + +### Implementation of layer helper -### FC Layer +We just keep all parameters of a layer function as a dictionary in layer helper as a private data member. Every method of layer helper will look up the dictionary after it is invoked. In that way, we can implement a layer helper for all layer functions even some layer does not contain some operator. For example, The `activation` is used by the FullyConnected layer or convolution layers, but a cross-entropy layer does not use it. The example code of `add_activation` are: ```python -def fc_layer(input, size, ...): - block = program.current_block() - w = block.create_parameter(...) - b = block.create_parameter(...) - out = block.create_var() - op = block.append_operator("FC", X=input, W=w, b=b, out=out) - out.writer = op - return out +class LayerHelper(object): + def __init__(self, **kwargs): # kwargs is short for `keyword arguments` + self.kwargs = kwargs + + def add_activation(self, input_var): + act = self.kwargs.get("act", None) # default value is None + if act is None: # do nothing if no act + return input_var + + tmp = self.create_tmp_var(self) + self.append_op(type=act, input=input_var, output=tmp) + return tmp ``` ## Optimizer diff --git a/doc/design/register_grad_op.md b/doc/design/register_grad_op.md index 3cf8a59446d244bb3a388b87b14273d9096c839a..9f1ce4bae7b393cb9f04909e5e4917b8d660771c 100644 --- a/doc/design/register_grad_op.md +++ b/doc/design/register_grad_op.md @@ -3,15 +3,17 @@ ## The Problem Posed -In our current operator registration mechanism, for each operator, the programmer should register a *gradient operator creator* function, which takes a C++ operator instance, and returns the corresponding gradient instance. +Currently, for each C++ operator class definition, there registers a *gradient operator creator* function, which takes a C++ operator instance and returns the corresponding gradient operator instance. -However, as we decided to separate the *compilation* and *execution* of DL models, we need to reshape the creator to take a protobuf `OpDesc` message, and returns a corresponding message. +However, we noticed two problems with the current deisgn: -More than that, the new registration mechanism need to support the fact that an operators' gradient computation might be a composition of operators. +1. As we decided to separate the *compilation* and *execution* phases, we need to change the creator to take an `OpDesc` protobuf message in a `ProgramDesc` and inserts corresponding `OpDesc` messages into the `ProgramDesc` message. -## Current Implementation +1. Some operator's gradient computation requires more than one gradient operators. For example, the gradient of *minus* consists of two operators -- an identity operaotr and a scale operator. So we need to make the registration mechanism to support the mapping from an operator to a set of operators for gradient computation. -OpInfos store in a association map which key is the operator type. The `grad_op_type` indicate associated gradient operator type. Operator can create gradient operator by `OpInfo::creator_` of gradient. The pseudo code is +## The Current Implementation + +The C++ class `OpInfos` store in a association map which key is the operator type. The `grad_op_type` indicate associated gradient operator type. Operator can create gradient operator by `OpInfo::creator_` of gradient. The pseudo code is ```cpp struct OpInfo { diff --git a/doc/design/scope.md b/doc/design/scope.md index b1f9bb4378eb5ec6926f1e53f7c1f4fd5674064c..4da76eebb74abcd26ec2b8671399e6bc4fb58574 100644 --- a/doc/design/scope.md +++ b/doc/design/scope.md @@ -37,7 +37,7 @@ Scope is an association of a name to variable. All variables belong to `Scope`. ```cpp class Scope { public: - Variable* NewVar(const std::string& name); + Variable* Var(const std::string& name); const Variable* FindVar(const std::string& name) const; private: @@ -98,7 +98,7 @@ class Scope { Variable* FindVar(const std::string& name) const; // return if already contains same name variable. - Variable* NewVar(const std::string& name); + Variable* Var(const std::string& name); private: std::shared_ptr parent_; @@ -107,7 +107,7 @@ class Scope { ``` ## Only scope can create a variable -To ensure `only scope can create a variable`, we should mark `Variable`'s constructor as a private member function, and Scope is a friend class of Variable. And then only `NewVar` can construct `Variable`. +To ensure `only scope can create a variable`, we should mark `Variable`'s constructor as a private member function, and Scope is a friend class of Variable. And then only `Var` can construct `Variable`. ## When scope destroyed, all variables inside this scope should be destroyed together @@ -121,4 +121,4 @@ Also, as the parent scope is a `shared_ptr`, we can only `Create()` a scope shar ## Orthogonal interface -`FindVar` will return `nullptr` when `name` is not found. It can be used as `Contains` method. `NewVar` will return an `Error` when there is a name conflict locally. Combine `FindVar` and `NewVar`, we can implement `NewVar` easily. +`FindVar` will return `nullptr` when `name` is not found. It can be used as `Contains` method. `Var` will return an `Error` when there is a name conflict locally. Combine `FindVar` and `Var`, we can implement `Var` easily. diff --git a/doc/design/tensor_array.md b/doc/design/tensor_array.md index 8378e97bf7cfaae54c36b1b92e202b16e4fe1e28..37e4f7b90f94fa3eb015e733999cd84c96b2239c 100644 --- a/doc/design/tensor_array.md +++ b/doc/design/tensor_array.md @@ -161,7 +161,7 @@ class TensorArray: @name: str the name of the variable to output. ''' - tensor = NewVar(name) + tensor = Var(name) tensor_array_stack(self.name, tensor) return tensor diff --git a/doc/design/var_desc.md b/doc/design/var_desc.md index bfbbdd0578ebc69ea4b49ade9b041573a9e9ad55..0b2958c1b10ef6a6ce51aa75f61e15a7f2d94b3f 100644 --- a/doc/design/var_desc.md +++ b/doc/design/var_desc.md @@ -16,16 +16,23 @@ The computation graph is constructed by Data Node and Operation Node. The concep ## Definition of VarDesc -A VarDesc should have a name and value, in PaddlePaddle, the value will always be a tensor. Since we use LoDTensor most of the time. We add a LoDTesnorDesc to represent it. +A VarDesc should have a name, and value. The are two kinds of variable type in compile time, they are `LoDTensor` and `SelectedRows`. ```proto message VarDesc { required string name = 1; - optional LoDTensorDesc lod_tensor = 2; + enum VarType { + LOD_TENSOR = 0; + SELECTED_ROWS = 1; + } + required VarType type = 2; + optional LoDTensorDesc lod_desc = 3; + optional TensorDesc selected_rows_desc = 4; + optional bool persistable = 5 [ default = false ]; } ``` -## Definition of LodTensorDesc +## Definition of TensorDesc ```proto enum DataType { @@ -38,87 +45,25 @@ enum DataType { FP64 = 6; } -message LoDTensorDesc { +message TensorDesc { required DataType data_type = 1; - repeated int32 dims = 2; // [UNK, 640, 480] is saved as [-1, 640, 480] - optional int32 lod_level = 3 [default=0]; + repeated int64 dims = 2; // [UNK, 640, 480] is saved as [-1, 640, 480] } ``` -## Definition of Variable in Python - -In Python API, layer will take Variable as Input, and return Variable as Output. There should be a class `Variable` in python to help create and manage Variable. - -```python -image = Variable(dims=[-1, 640, 480]) -# fc1 and fc2 are both Variable -fc1 = layer.fc(input=image, output_size=10) -fc2 = layer.fc(input=fc1, output_size=20) -``` -### what should class `Variable` Have -1. `name`.a name of string type is used to mark the value of the Variable. -1. `initializer`. Since our Tensor does not have value. we will always use some Operator to fullfill it when run. So we should have a initialize method to help add the init operator. -1. `operator`. Variable should record which operator produce itself. The reaon is: - - we use pd.eval(targets=[var1, var2]) to run the related ops to get the value of var1 and var2. var.op is used to trace the dependency of the current variable. - -In PaddlePaddle, we use Block to describe Computation Graph, so in the code we will use Block but not Graph. - -```python -import VarDesc -import LoDTensorDesc -import framework - -def AddInitialOperator(variable, initializer): - # add an initialize Operator to block to init this Variable - -class Variable(object): - def __init__(self, name, dims, type, initializer): - self._block = get_default_block() - self._name = name - self.op = None - - tensor_desc = LoDTensorDesc(data_type=type, dims=dims) - _var_desc = VarDesc(name=name, lod_tensor=tensor_desc) - self._var = framework.CreateVar(_var_desc) - self._block.add_var(self) +A TensorDesc describes `SelectedRows` and `LoDTensor`. For details of `SelectedRows`, please reference [`SelectedRows`](./selected_rows.md). - # add initial op according to initializer - if initializer is not None: - AddInitialOperator(self, initializer) - - def dims(self): - return self._var.dims() - - def data_type(self): - return self._var.data_type() +## Definition of LodTensorDesc - def to_proto(self): - pass +```proto +message LoDTensorDesc { + required TensorDesc tensor = 1; + optional int lod_level = 2; +} ``` -Then we can use this Variable to create a fc layer in Python. +A LoDTensorDesc contains a tensor and a lod_level. -```python -import paddle as pd - -def flatten_size(X, num_flatten_dims): - prod = 1 # of last num_flatten_dims - for i in xrange(num_flatten_dims): - prod = prod * X.dims[-i-1] - return prod - -def layer.fc(X, output_size, num_flatten_dims): - W = Variable(pd.random_uniform(), type=FP32, dims=[flatten_size(X, num_flatten_dims), output_size]) - b = Variable(pd.random_uniform(), type=FP32, dims=[output_size]) - out = Variable(type=FP32) - y = operator.fc(X, W, b, output=out) # fc will put fc op input into out - pd.InferShape(y) - return out - -x = Variable(dims=[-1, 640, 480]) -y = layer.fc(x, output_size=100) -z = layer.fc(y, output_size=200) +## Definition of Variable in Python -paddle.eval(targets=[z], ...) -print(z) -``` +For Variable in Python, please reference [`Python API`](./python_api.md). diff --git a/doc/howto/deep_model/rnn/rnn_config_cn.rst b/doc/howto/deep_model/rnn/rnn_config_cn.rst index 4d684cf8ad5a8082cf31fb27027119b3d3e700b6..63fa161fafed0f3a8ec8799af21304cbec62d813 100644 --- a/doc/howto/deep_model/rnn/rnn_config_cn.rst +++ b/doc/howto/deep_model/rnn/rnn_config_cn.rst @@ -21,7 +21,7 @@ wmt14数据的提供文件在 `python/paddle/v2/dataset/wmt14.py