Skip to content

P
Paddle

wmsofts / Paddle
与 Fork 源项目一致

Fork自 PaddlePaddle / Paddle

通知 1
Star 0
Fork 0
P
Paddle

前往新版Gitcode，体验更适合开发者的 AI 搜索 >>

未验证 提交 433cef03 编写于 作者: T tianshuo78520a 提交者： GitHub
浏览文件
操作

fix typo word (#22784)

上级 ebc7ffc3
隐藏空白更改
内联 并排
Showing with 190 addition and 190 deletion
cmake/configure.cmake
浏览文件 @ 433cef03
......@@ -48,7 +48,7 @@ if(WIN32)
SET(CMAKE_C_RESPONSE_FILE_LINK_FLAG "@")
SET(CMAKE_CXX_RESPONSE_FILE_LINK_FLAG "@")
# set defination for the dll export
# set definition for the dll export
if (NOT MSVC)
message(FATAL "Windows build only support msvc. Which was binded by the nvcc compiler of NVIDIA.")
endif(NOT MSVC)
......
cmake/third_party.cmake
浏览文件 @ 433cef03
......@@ -174,7 +174,7 @@ if(${CMAKE_VERSION} VERSION_GREATER "3.5.2")
set(SHALLOW_CLONE "GIT_SHALLOW TRUE") # adds --depth=1 arg to git clone of External_Projects
endif()
########################### include third_party accoring to flags ###############################
########################### include third_party according to flags ###############################
include(external/zlib) # download, build, install zlib
include(external/gflags) # download, build, install gflags
include(external/glog) # download, build, install glog
......
paddle/fluid/framework/ir/multi_devices_graph_pass/multi_devices_graph_pass.cc
浏览文件 @ 433cef03
......@@ -857,7 +857,7 @@ int DistSSAGraphBuilder::CreateRPCOp(ir::Graph *result, ir::Node *node) const {
op_dev_id = GetVarDeviceID(node->inputs[0]->Name());
PADDLE_ENFORCE(!ir::IsControlDepVar(*node->inputs[0]),
"This hack no longer holds, please fix.");
// the variable name which contains .block means it was splited by
// the variable name which contains .block means it was split by
// split_byref op
if (strategy_.reduce_ ==
details::BuildStrategy::ReduceStrategy::kAllReduce &&
......
paddle/fluid/framework/operator.cc
浏览文件 @ 433cef03
......@@ -990,7 +990,7 @@ void OperatorWithKernel::RunImpl(const Scope& scope,
}
if (!transfered_inplace_vars.empty()) {
// there is inplace variable has been transfered.
// there is inplace variable has been transferred.
TransferInplaceVarsBack(scope, transfered_inplace_vars, *transfer_scope);
}
if (FLAGS_enable_unused_var_check) {
......
paddle/fluid/framework/operator.h
浏览文件 @ 433cef03
......@@ -517,7 +517,8 @@ class OperatorWithKernel : public OperatorBase {
RuntimeContext* runtime_ctx) const;
/**
* Transfer data from scope to a transfered scope. If there is no data need to
* Transfer data from scope to a transferred scope. If there is no data need
* to
* be tranfered, it returns nullptr.
*
* * transfered_inplace_vars is a output vector.
......
paddle/fluid/framework/parallel_executor.cc
浏览文件 @ 433cef03
......@@ -87,18 +87,18 @@ class ParallelExecutorPrivate {
inline bool HasGarbageCollectors() const { return !gcs_.empty(); }
/**
* NOTE(zengjinle): the feeded variables of users should not be reused,
* because users may feed them into another network. Changing the feeded
* NOTE(zengjinle): the fed variables of users should not be reused,
* because users may feed them into another network. Changing the fed
* variables that users can visit may cause calculation wrong, which is
* a very subtle bug when traning networks. However, these variables
* can be garbage collected.
*
* ParallelExecutor provides 2 methods to feed variables:
*
* - FeedTensorsIntoLocalScopes: this method would share memory of feeded
* - FeedTensorsIntoLocalScopes: this method would share memory of fed
* variables, so we have to skip these.
*
* - FeedAndSplitTensorIntoLocalScopes: this method would copy data of feeded
* - FeedAndSplitTensorIntoLocalScopes: this method would copy data of fed
* variables, so we do not need to skip
* them.
*/
......
paddle/fluid/framework/reader.h
浏览文件 @ 433cef03
......@@ -53,10 +53,10 @@ class ReaderBase {
// they are readers just before read op.
std::unordered_set<ReaderBase*> GetEndPoints();
// Returns the shapes of the feeded variables
// Returns the shapes of the fed variables
const std::vector<DDim>& Shapes() const { return shapes_; }
// Returns the dtypes of the feeded variables
// Returns the dtypes of the fed variables
const std::vector<proto::VarType::Type>& VarTypes() const {
return var_types_;
}
......@@ -80,13 +80,13 @@ class ReaderBase {
mutable std::mutex mu_;
// The shapes of the feeded variables.
// The shapes of the fed variables.
std::vector<DDim> shapes_;
// The dtypes of the feeded variables.
// The dtypes of the fed variables.
std::vector<proto::VarType::Type> var_types_;
// Whether to check the shape and dtype of feeded variables.
// Whether to check the shape and dtype of fed variables.
// For Backward compatibility, variables created by old API fluid.layers.data
// doesn't check shape but fluid.data checks.
std::vector<bool> need_check_feed_;
......
paddle/fluid/imperative/tests/test_prepare_op.cc
浏览文件 @ 433cef03
......@@ -210,7 +210,7 @@ TEST(test_prepare_op, test_prepare_data_same_place) {
attr_map);
framework::RuntimeContext ctx = PrepareRuntimeContext(ins, outs);
// test if it never transfered on GPU place
// test if it never transferred on GPU place
PreparedOp prepared_op = PreparedOp::Prepare(
ins, outs, dynamic_cast<framework::OperatorWithKernel&>(*op), cpu_place,
&attr_map);
......
paddle/fluid/inference/CMakeLists.txt
浏览文件 @ 433cef03
......@@ -14,7 +14,7 @@
#
if(WITH_TESTING)
include(tests/test.cmake) # some generic cmake funtion for inference
include(tests/test.cmake) # some generic cmake function for inference
endif()
# TODO(panyx0718): Should this be called paddle_fluid_inference_api_internal?
......
paddle/fluid/inference/api/demo_ci/README.md
浏览文件 @ 433cef03
......@@ -12,7 +12,7 @@ There are several demos:
- Each line contains a single record
- Each record's format is
```
<space splitted floats as data>\t<space splitted ints as shape>
<space split floats as data>\t<space split ints as shape>
```
To build and execute the demos, simply run
......
paddle/fluid/inference/api/demo_ci/trt_mobilenet_demo.cc
浏览文件 @ 433cef03
......@@ -23,10 +23,9 @@ limitations under the License. */
DECLARE_double(fraction_of_gpu_memory_to_use);
DEFINE_string(modeldir, "", "Directory of the inference model.");
DEFINE_string(refer, "", "path to reference result for comparison.");
DEFINE_string(
data, "",
"path of data; each line is a record, format is "
"'<space splitted floats as data>\t<space splitted ints as shape'");
DEFINE_string(data, "",
"path of data; each line is a record, format is "
"'<space split floats as data>\t<space split ints as shape'");
namespace paddle {
namespace demo {
......
paddle/fluid/inference/api/demo_ci/vis_demo.cc
浏览文件 @ 433cef03
......@@ -25,10 +25,9 @@ DECLARE_double(fraction_of_gpu_memory_to_use);
#endif
DEFINE_string(modeldir, "", "Directory of the inference model.");
DEFINE_string(refer, "", "path to reference result for comparison.");
DEFINE_string(
data, "",
"path of data; each line is a record, format is "
"'<space splitted floats as data>\t<space splitted ints as shape'");
DEFINE_string(data, "",
"path of data; each line is a record, format is "
"'<space split floats as data>\t<space split ints as shape'");
DEFINE_bool(use_gpu, false, "Whether use gpu.");
#ifdef PADDLE_WITH_SHARED_LIB
DECLARE_bool(profile);
......
paddle/fluid/operators/beam_search_op.cc
浏览文件 @ 433cef03
......@@ -52,7 +52,7 @@ class BeamSearchOpMaker : public framework::OpProtoAndCheckerMaker {
"A LoDTensor containing the accumulated scores corresponding to "
"Output(selected_ids).");
AddOutput("parent_idx",
"A Tensor preserving the selected_ids' parent indice in pre_ids.")
"A Tensor preserving the selected_ids' parent index in pre_ids.")
.AsDispensable();
// Attributes stored in AttributeMap
......
paddle/fluid/operators/cudnn_lstm_op.cc
浏览文件 @ 433cef03
......@@ -119,7 +119,7 @@ class CudnnLSTMOpMaker : public framework::OpProtoAndCheckerMaker {
.SetDefault(0.0);
AddAttr<bool>("is_bidirec",
"is_bidirec"
"if it is bidirection rnn"
"if it is bidirectional rnn"
"The will affect the shape of the Out, last_h, and last_c")
.SetDefault(false);
AddAttr<int>("input_size", "input size ot the Input Tensor").SetDefault(10);
......
paddle/fluid/operators/detection/density_prior_box_op.cc
浏览文件 @ 433cef03
......@@ -35,7 +35,7 @@ class DensityPriorBoxOp : public framework::OperatorWithKernel {
platform::errors::InvalidArgument(
"The input tensor Input's height"
"of DensityPriorBoxOp should be smaller than input tensor Image's"
"hight. But received Input's height = %d, Image's height = %d",
"height. But received Input's height = %d, Image's height = %d",
input_dims[2], image_dims[2]));
PADDLE_ENFORCE_LT(
......
paddle/fluid/operators/distributed/communicator.cc
浏览文件 @ 433cef03
......@@ -543,7 +543,7 @@ std::unordered_set<int64_t> GeoSgdCommunicator::SparseIdsMerge(
const std::string &splited_var_name) {
// every batch has some sparse id, merge them into one unoredered_set
VLOG(4) << "Sparse Ids merge var: " << var_name
<< " splited var: " << splited_var_name;
<< " split var: " << splited_var_name;
auto before_run_ids_merge_ = GetCurrentUS();
auto origin_var_name = DeltaVarToVar(var_name);
auto splited_var_index = GetSplitedVarIndex(var_name, splited_var_name);
......@@ -567,9 +567,8 @@ void GeoSgdCommunicator::SendUpdateDenseVars(
// var_name: param.delta
auto origin_var_name = DeltaVarToVar(var_name);
auto splited_var_index = GetSplitedVarIndex(var_name, splited_var_name);
VLOG(4) << "Dense var: " << var_name
<< " 's splited var: " << splited_var_name
<< " splited var index: " << splited_var_index;
VLOG(4) << "Dense var: " << var_name << " 's split var: " << splited_var_name
<< " split var index: " << splited_var_index;
auto before_run_send_dense = GetCurrentUS();
auto cpu_ctx = paddle::platform::CPUDeviceContext();
......@@ -592,7 +591,7 @@ void GeoSgdCommunicator::SendUpdateDenseVars(
begin_loc = absolute_section_[origin_var_name][splited_var_index];
dimension = total_element / vars_first_dimension_[origin_var_name];
total_element = section * dimension;
VLOG(4) << "Dense splited var: " << splited_var_name
VLOG(4) << "Dense split var: " << splited_var_name
<< " section: " << section << " dimension: " << dimension
<< " begin loc: " << begin_loc << " total_element "
<< total_element;
......@@ -600,12 +599,12 @@ void GeoSgdCommunicator::SendUpdateDenseVars(
auto *var_x_data = var_x_tensor.mutable_data<float>(var_x_tensor.place()) +
begin_loc * dimension;
VLOG(4) << "Dense splited var: " << splited_var_name << " var_x_data[0] "
VLOG(4) << "Dense split var: " << splited_var_name << " var_x_data[0] "
<< var_x_data[0] << " var_x_data[end] "
<< var_x_data[total_element - 1];
auto *var_y_data = var_y_tensor.mutable_data<float>(var_y_tensor.place()) +
begin_loc * dimension;
VLOG(4) << "Dense splited var: " << splited_var_name << " var_y_data[0] "
VLOG(4) << "Dense split var: " << splited_var_name << " var_y_data[0] "
<< var_y_data[0] << " var_y_data[end] "
<< var_y_data[total_element - 1];
......@@ -616,14 +615,14 @@ void GeoSgdCommunicator::SendUpdateDenseVars(
var_z_tensor->mutable_data<float>(dims, cpu_ctx.GetPlace());
auto *var_z_data = var_z_tensor->mutable_data<float>(cpu_ctx.GetPlace());
VLOG(4) << "Dense splited var: " << splited_var_name << "var_z_data[0] "
VLOG(4) << "Dense split var: " << splited_var_name << "var_z_data[0] "
<< var_z_data[0] << " var_z_data[end] "
<< var_z_data[total_element - 1];
// calc sub = var_training - var_old
auto blas = math::GetBlas<paddle::platform::CPUDeviceContext, float>(cpu_ctx);
blas.VSUB(total_element, var_x_data, var_y_data, var_z_data);
VLOG(4) << "Dense splited var: " << splited_var_name << " var_z_data[0] "
VLOG(4) << "Dense split var: " << splited_var_name << " var_z_data[0] "
<< var_z_data[0] << " var_z_data[end] "
<< var_z_data[total_element - 1];
......@@ -633,7 +632,7 @@ void GeoSgdCommunicator::SendUpdateDenseVars(
// calc var_old += var_delta
blas.VADD(total_element, var_y_data, var_z_data, var_y_data);
VLOG(4) << "Dense splited var: " << splited_var_name << " var_y_data[0] "
VLOG(4) << "Dense split var: " << splited_var_name << " var_y_data[0] "
<< var_y_data[0] << " var_y_data[end] "
<< var_y_data[total_element - 1];
......@@ -763,7 +762,7 @@ void GeoSgdCommunicator::RecvUpdateDenseVars(
section = dims[0];
begin_loc = absolute_section_[origin_var_name][splited_var_index];
dimension = total_element / section;
VLOG(4) << "Dense splited var: " << splited_var_name
VLOG(4) << "Dense split var: " << splited_var_name
<< " section: " << section << " dimension: " << dimension
<< " begin loc: " << begin_loc << " total_element "
<< total_element;
......@@ -771,18 +770,18 @@ void GeoSgdCommunicator::RecvUpdateDenseVars(
auto *var_x_data = var_x_tensor.mutable_data<float>(var_x_tensor.place()) +
begin_loc * dimension;
VLOG(4) << "Dense splited var: " << splited_var_name << " var_x_data[0] "
VLOG(4) << "Dense split var: " << splited_var_name << " var_x_data[0] "
<< var_x_data[0] << " var_x_data[end] "
<< var_x_data[total_element - 1];
auto *var_y_data = var_y_tensor.mutable_data<float>(var_y_tensor.place()) +
begin_loc * dimension;
VLOG(4) << "Dense splited var: " << splited_var_name << " var_y_data[0] "
VLOG(4) << "Dense split var: " << splited_var_name << " var_y_data[0] "
<< var_y_data[0] << " var_y_data[end] "
<< var_y_data[total_element - 1];
auto *var_z_data = var_z_tensor.mutable_data<float>(cpu_ctx.GetPlace());
VLOG(4) << "Dense splited var: " << splited_var_name << " var_z_data[0] "
VLOG(4) << "Dense split var: " << splited_var_name << " var_z_data[0] "
<< var_z_data[0] << " var_z_data[end] "
<< var_z_data[total_element - 1];
......@@ -793,7 +792,7 @@ void GeoSgdCommunicator::RecvUpdateDenseVars(
auto *var_y_sub_data =
var_y_sub_tensor->mutable_data<float>(cpu_ctx.GetPlace());
VLOG(4) << "Dense splited var: " << splited_var_name << " var_y_sub_data[0] "
VLOG(4) << "Dense split var: " << splited_var_name << " var_y_sub_data[0] "
<< var_y_sub_data[0] << " var_y_sub_data[end] "
<< var_y_sub_data[total_element - 1];
......@@ -801,19 +800,19 @@ void GeoSgdCommunicator::RecvUpdateDenseVars(
// calc sub = pserver - old
blas.VSUB(total_element, var_z_data, var_y_data, var_y_sub_data);
VLOG(4) << "Dense splited var: " << splited_var_name << " var_y_sub_data[0] "
VLOG(4) << "Dense split var: " << splited_var_name << " var_y_sub_data[0] "
<< var_y_sub_data[0] << " var_y_sub_data[end] "
<< var_y_sub_data[total_element - 1];
// calc train += sub
blas.VADD(total_element, var_x_data, var_y_sub_data, var_x_data);
VLOG(4) << "Dense splited var: " << splited_var_name << " var_x_data[0] "
VLOG(4) << "Dense split var: " << splited_var_name << " var_x_data[0] "
<< var_x_data[0] << " var_x_data[end] "
<< var_x_data[total_element - 1];
// calc old = pserver
blas.VCOPY(total_element, var_z_data, var_y_data);
VLOG(4) << "Dense splited var: " << splited_var_name << " var_y_data[0] "
VLOG(4) << "Dense split var: " << splited_var_name << " var_y_data[0] "
<< var_y_data[0] << " var_y_data[end] "
<< var_y_data[total_element - 1];
......@@ -824,7 +823,7 @@ void GeoSgdCommunicator::RecvUpdateDenseVars(
void GeoSgdCommunicator::RecvUpdateSparseVars(
const std::string &var_name, const std::string &splited_var_name) {
// step 1: recv splited var from pserver
// step 1: recv split var from pserver
auto splited_var_index = GetSplitedVarIndex(var_name, splited_var_name);
auto origin_var_name = DeltaVarToVar(var_name);
auto origin_splited_var_name = DeltaVarToVar(splited_var_name);
......
paddle/fluid/operators/distributed_ops/recv_op.cc
浏览文件 @ 433cef03
......@@ -122,7 +122,7 @@ This operator can get variables from server side.
AddAttr<std::vector<std::string>>(
"recv_varnames",
"(vector<string>) "
"the splited parameter varnames to be recved from pserver")
"the split parameter varnames to be recved from pserver")
.SetDefault(std::vector<std::string>{});
AddAttr<int>("do_not_run", "if recv need to really run").SetDefault(0);
}
......
paddle/fluid/operators/distributed_ops/send_op.cc
浏览文件 @ 433cef03
......@@ -116,7 +116,7 @@ This operator will send variables to listen_and_serve op at the parameter server
AddAttr<std::vector<std::string>>(
"send_varnames",
"(vector<string>) "
"the splited output varnames to send to pserver")
"the split output varnames to send to pserver")
.SetDefault(std::vector<std::string>{});
AddAttr<int>("num",
"(int, default 0)"
......
paddle/fluid/operators/fused/conv_fusion_op.cc
浏览文件 @ 433cef03
......@@ -28,7 +28,7 @@ namespace operators {
// x is Input,
// z is ResidualData,
// bias is Bias
// When `split_channels` is set, y will be splitted into multiple outputs,
// When `split_channels` is set, y will be split into multiple outputs,
// each output has split_channels[i] number of channels.
class Conv2DFusionOpMaker : public Conv2DOpMaker {
protected:
......
paddle/fluid/operators/hierarchical_sigmoid_op.cc
浏览文件 @ 433cef03
......@@ -162,7 +162,7 @@ class HierarchicalSigmoidOpMaker : public framework::OpProtoAndCheckerMaker {
.SetDefault({});
AddAttr<std::vector<std::string>>(
"table_names",
"(string vector, the splited table names that will be fetched from "
"(string vector, the split table names that will be fetched from "
"parameter server)"
"in the order of input variables for mapping")
.SetDefault({});
......
paddle/fluid/operators/lookup_table_op.cc
浏览文件 @ 433cef03
......@@ -113,7 +113,7 @@ class LookupTableOpMaker : public framework::OpProtoAndCheckerMaker {
.SetDefault({});
AddAttr<std::vector<std::string>>(
"table_names",
"(string vector, the splited table names that will be fetched from "
"(string vector, the split table names that will be fetched from "
"parameter server)"
"in the order of input variables for mapping")
.SetDefault({});
......
paddle/fluid/operators/lookup_table_v2_op.cc
浏览文件 @ 433cef03
......@@ -100,7 +100,7 @@ class LookupTableV2OpMaker : public framework::OpProtoAndCheckerMaker {
.SetDefault({});
AddAttr<std::vector<std::string>>(
"table_names",
"(string vector, the splited table names that will be fetched from "
"(string vector, the split table names that will be fetched from "
"parameter server)"
"in the order of input variables for mapping")
.SetDefault({});
......
paddle/fluid/operators/math/blas_impl.h
浏览文件 @ 433cef03
......@@ -777,11 +777,11 @@ void Blas<DeviceContext>::MatMul(const framework::Tensor &mat_a,
* When user calls this API, the multiplication of two big matrixes is split
* into multiplication of several (head_number_) small matrixes. e.g. if Mat A
* is [3, 24] and Mat B is [24, 4], when multiple A and B with head_number as
* 4, Mat A will be splitted as 4 matrix of [3, 6] and Mat B will be
* (horizontally) splitted as 4 matrix of [6, 4]. The result of final matrix
* 4, Mat A will be split as 4 matrix of [3, 6] and Mat B will be
* (horizontally) split as 4 matrix of [6, 4]. The result of final matrix
* will be 4 matrix of [3, 4], i.e. [3, 16].
* Another example is A is [3, 8], B is [2, 16], head_number is 4. In this
* case, A will be splitted as [3, 2], B will be (vertically) splitted as
* case, A will be split as [3, 2], B will be (vertically) split as
* [2, 4]. The final result will be 4 matrix of 4 matrix of [3,4], i.e. [3, 16]
*/
template <typename DeviceContext>
......
paddle/fluid/operators/math/matrix_bit_code.h
浏览文件 @ 433cef03
......@@ -106,7 +106,7 @@ class SimpleCode {
: c_(static_cast<size_t>(ids[code]) + num_classes) {}
/**
* Here the id of root should be 1 rather than 0, thus the encoding of class c
* is `c + num_classes` and all siblings can get the same weight indice using
* is `c + num_classes` and all siblings can get the same weight index using
* prefixes.
* Weight index is the prefixes of encoding, thus leave out the right most
* bit in calc_index.
......@@ -133,7 +133,7 @@ class CustomCode {
}
/**
* Here the id of root should be 1 rather than 0, thus the encoding of class c
* is `c + num_classes` and all siblings can get the same weight indice using
* is `c + num_classes` and all siblings can get the same weight index using
* prefixes.
* Weight index is the prefixes of encoding, thus leave out the right most
* bit in calc_index.
......
paddle/fluid/operators/mkldnn/fc_mkldnn_op.cc
浏览文件 @ 433cef03
......@@ -55,11 +55,12 @@ class FCPrimitiveFactory {
} // Otherwise, create a new one.
auto in_col_dims = ctx.Attr<int>("in_num_col_dims");
PADDLE_ENFORCE_LE(in_col_dims, 2,
platform::errors::Unimplemented(
"DNNL FC doesn't support in_num_col_dims paramter to "
"be higher than "
"2."));
PADDLE_ENFORCE_LE(
in_col_dims, 2,
platform::errors::Unimplemented(
"DNNL FC doesn't support in_num_col_dims parameter to "
"be higher than "
"2."));
if (in_col_dims == 2) {
PADDLE_ENFORCE_EQ(
input->dims().size(), 3,
......
paddle/fluid/operators/nce_op.cc
浏览文件 @ 433cef03
......@@ -192,7 +192,7 @@ class NCEOpMaker : public framework::OpProtoAndCheckerMaker {
.SetDefault({});
AddAttr<std::vector<std::string>>(
"table_names",
"(string vector, the splited table names that will be fetched from "
"(string vector, the split table names that will be fetched from "
"parameter server)"
"in the order of input variables for mapping")
.SetDefault({});
......
paddle/fluid/operators/pad2d_op.cc
浏览文件 @ 433cef03
......@@ -563,7 +563,7 @@ class Pad2dOpMaker : public framework::OpProtoAndCheckerMaker {
.SetDefault("NCHW");
AddComment(R"DOC(
Pad2d Operator.
Pad 2-d images accordding to 'paddings' and 'mode'.
Pad 2-d images according to 'paddings' and 'mode'.
If mode is 'reflect', paddings[0] and paddings[1] must be no greater
than height-1. And the width dimension has the same condition.
......
paddle/fluid/operators/reader/read_op.cc
浏览文件 @ 433cef03
......@@ -118,7 +118,7 @@ class ReadOp : public framework::OperatorBase {
reader->VarTypes();
const std::vector<bool>& need_check_feed = reader->NeedCheckFeed();
PADDLE_ENFORCE_EQ(out_arg_names.size(), need_check_feed.size(),
"output size of read_op and the number of feeded "
"output size of read_op and the number of fed "
"variables of reader do not match");
for (size_t i = 0; i < out_arg_names.size(); ++i) {
......@@ -127,13 +127,13 @@ class ReadOp : public framework::OperatorBase {
if (need_check_feed[i]) {
auto in_dims = ins[i].dims();
PADDLE_ENFORCE_EQ(DimensionIsCompatibleWith(shapes[i], in_dims), true,
"The feeded Variable %s should have dimensions = %d, "
"shape = [%s], but received feeded shape [%s]",
"The fed Variable %s should have dimensions = %d, "
"shape = [%s], but received fed shape [%s]",
out_arg_names[i], shapes[i].size(), shapes[i],
in_dims);
PADDLE_ENFORCE_EQ(
ins[i].type(), var_types[i],
"The data type of feeded Variable %s must be %s, but received %s",
"The data type of fed Variable %s must be %s, but received %s",
out_arg_names[i], var_types[i], ins[i].type());
}
out->ShareDataWith(ins[i]);
......
paddle/fluid/operators/sequence_ops/sequence_expand_as_op.cu
浏览文件 @ 433cef03
......@@ -67,8 +67,8 @@ struct SequenceExpandFunctor<platform::CUDADeviceContext, T> {
const platform::CUDADeviceContext &context, const LoDTensor &x,
const framework::Vector<size_t> &ref_lod, /*expand referenced lod*/
LoDTensor *out) {
int hight = x.dims()[0];
int width = framework::product(x.dims()) / hight;
int height = x.dims()[0];
int width = framework::product(x.dims()) / height;
const int kThreadsPerBlock = 1024;
int thread_x = kThreadsPerBlock;
......@@ -82,7 +82,7 @@ struct SequenceExpandFunctor<platform::CUDADeviceContext, T> {
dim3 block_size(thread_x);
dim3 grid_size(block_x);
sequence_expand_as_kernel<<<grid_size, block_size, 0, context.stream()>>>(
x.data<T>(), ref_lod.CUDAData(context.GetPlace()), hight, width,
x.data<T>(), ref_lod.CUDAData(context.GetPlace()), height, width,
out->mutable_data<T>(context.GetPlace()));
}
};
......@@ -93,8 +93,8 @@ struct SequenceExpandAsGradFunctor<platform::CUDADeviceContext, T> {
const LoDTensor &dout,
const framework::Vector<size_t> &ref_lod, /*expand based lod*/
LoDTensor *dx) {
int hight = dx->dims()[0];
int width = framework::product(dx->dims()) / hight;
int height = dx->dims()[0];
int width = framework::product(dx->dims()) / height;
const int kThreadsPerBlock = 1024;
int thread_x = kThreadsPerBlock;
......@@ -109,7 +109,7 @@ struct SequenceExpandAsGradFunctor<platform::CUDADeviceContext, T> {
dim3 grid_size(block_x);
sequence_expand_as_grad_kernel<<<grid_size, block_size, 0,
context.stream()>>>(
dout.data<T>(), ref_lod.CUDAData(context.GetPlace()), hight, width,
dout.data<T>(), ref_lod.CUDAData(context.GetPlace()), height, width,
dx->mutable_data<T>(context.GetPlace()));
}
};
......
paddle/fluid/operators/sequence_ops/sequence_expand_as_op.h
浏览文件 @ 433cef03
......@@ -46,13 +46,13 @@ struct SequenceExpandFunctor<platform::CPUDeviceContext, T> {
const platform::CPUDeviceContext &context, const framework::LoDTensor &x,
const framework::Vector<size_t> &ref_lod, /*expand referenced lod*/
framework::LoDTensor *out) {
int64_t hight = x.dims()[0];
int64_t width = framework::product(x.dims()) / hight;
int64_t height = x.dims()[0];
int64_t width = framework::product(x.dims()) / height;
const T *in_data = x.data<T>();
T *out_data = out->mutable_data<T>(context.GetPlace());
for (int h_id = 0; h_id < hight; ++h_id) {
for (int h_id = 0; h_id < height; ++h_id) {
size_t span = ref_lod[h_id + 1] - ref_lod[h_id];
if (span == 0) continue;
const T *src = in_data + h_id * width;
......@@ -109,13 +109,13 @@ struct SequenceExpandAsGradFunctor<platform::CPUDeviceContext, T> {
const framework::LoDTensor &dout,
const framework::Vector<size_t> &ref_lod, /*expand referenced lod*/
framework::LoDTensor *dx) {
int64_t hight = dx->dims()[0];
int64_t width = framework::product(dx->dims()) / hight;
int64_t height = dx->dims()[0];
int64_t width = framework::product(dx->dims()) / height;
const T *dout_data = dout.data<T>();
T *dx_data = dx->mutable_data<T>(context.GetPlace());
for (int64_t h_id = 0; h_id < hight; ++h_id) {
for (int64_t h_id = 0; h_id < height; ++h_id) {
T *dst = dx_data + h_id * width;
size_t span = ref_lod[h_id + 1] - ref_lod[h_id];
for (int64_t w_id = 0; w_id < width; ++w_id) {
......
paddle/fluid/operators/sequence_ops/sequence_softmax_op.cu
浏览文件 @ 433cef03
......@@ -115,7 +115,7 @@ struct SequenceSoftmaxFunctor<platform::CUDADeviceContext, T> {
const LoDTensor &x,
const framework::Vector<size_t> &ref_lod, /*referenced lod*/
LoDTensor *out) {
int hight = ref_lod.size() - 1;
int height = ref_lod.size() - 1;
const int kThreadsPerBlock = 32;
int thread_x = kThreadsPerBlock;
......@@ -126,7 +126,7 @@ struct SequenceSoftmaxFunctor<platform::CUDADeviceContext, T> {
dim3 grid_size(max_blocks);
sequence_softmax_kernel<
T, kThreadsPerBlock><<<grid_size, block_size, 0, context.stream()>>>(
x.data<T>(), ref_lod.CUDAData(context.GetPlace()), hight,
x.data<T>(), ref_lod.CUDAData(context.GetPlace()), height,
out->mutable_data<T>(context.GetPlace()));
}
};
......@@ -137,7 +137,7 @@ struct SequenceSoftmaxGradFunctor<platform::CUDADeviceContext, T> {
const LoDTensor &dout, const LoDTensor &out,
const framework::Vector<size_t> &ref_lod, /*referenced lod*/
LoDTensor *dx) {
size_t hight = ref_lod.size() - 1;
size_t height = ref_lod.size() - 1;
const int kThreadsPerBlock = 32;
int thread_x = kThreadsPerBlock;
......@@ -150,7 +150,7 @@ struct SequenceSoftmaxGradFunctor<platform::CUDADeviceContext, T> {
sequence_softmax_grad_kernel<
T, kThreadsPerBlock><<<grid_size, block_size, 0, context.stream()>>>(
dout.data<T>(), out.data<T>(), ref_lod.CUDAData(context.GetPlace()),
hight, dx->mutable_data<T>(context.GetPlace()));
height, dx->mutable_data<T>(context.GetPlace()));
}
};
......
paddle/fluid/operators/sequence_ops/sequence_softmax_op.h
浏览文件 @ 433cef03
......@@ -43,10 +43,10 @@ struct SequenceSoftmaxFunctor<platform::CPUDeviceContext, T> {
void operator()(const platform::CPUDeviceContext &ctx, const LoDTensor &x,
const framework::Vector<size_t> &ref_lod, /*referenced lod*/
LoDTensor *out) {
size_t hight = ref_lod.size() - 1;
size_t height = ref_lod.size() - 1;
const T *in_data = x.data<T>();
T *out_data = out->mutable_data<T>(ctx.GetPlace());
for (size_t i = 0; i < hight; ++i) {
for (size_t i = 0; i < height; ++i) {
size_t span = ref_lod[i + 1] - ref_lod[i];
T result = 0;
for (size_t j = 0; j < span; ++j) {
......@@ -65,13 +65,13 @@ struct SequenceSoftmaxGradFunctor<platform::CPUDeviceContext, T> {
const LoDTensor &out,
const framework::Vector<size_t> &ref_lod, /*referenced lod*/
LoDTensor *dx) {
size_t hight = ref_lod.size() - 1;
size_t height = ref_lod.size() - 1;
const T *softmax_grad_data = dout.data<T>();
const T *softmax = out.data<T>();
T *dx_data = dx->mutable_data<T>(ctx.GetPlace());
for (size_t i = 0; i < hight; ++i) {
for (size_t i = 0; i < height; ++i) {
size_t span = ref_lod[i + 1] - ref_lod[i];
T result = 0;
for (size_t j = 0; j < span; ++j) {
......
paddle/fluid/operators/split_op.cc
浏览文件 @ 433cef03
......@@ -90,7 +90,7 @@ class SplitOpMaker : public framework::OpProtoAndCheckerMaker {
void Make() override {
AddInput("X", "(Tensor) Input tensor of the split operator.");
AddInput("AxisTensor",
"(Tensor) The axis which the input will be splited on. "
"(Tensor) The axis which the input will be split on. "
"It has higher priority than Attr(axis). "
"The shape of AxisTensor must be [1]")
.AsDispensable();
......@@ -131,7 +131,7 @@ Example:
.SetDefault(0);
AddAttr<int>("axis",
"(int, default 0) "
"The axis which the input will be splited on.")
"The axis which the input will be split on.")
.SetDefault(0);
}
};
......
paddle/fluid/operators/unfold_op.cc
浏览文件 @ 433cef03
......@@ -76,7 +76,7 @@ class UnfoldOp : public framework::OperatorWithKernel {
// Only [N, C, H, W] input supported now
PADDLE_ENFORCE(
in_dims.size() == 4,
"Input shold be 4-D tensor of format [N, C, H, W], but get %u",
"Input should be 4-D tensor of format [N, C, H, W], but get %u",
in_dims.size());
PADDLE_ENFORCE(
in_dims.size() - kernel_sizes.size() == 2U,
......@@ -86,7 +86,7 @@ class UnfoldOp : public framework::OperatorWithKernel {
in_dims.size(), kernel_sizes.size());
PADDLE_ENFORCE_EQ(
strides.size(), kernel_sizes.size(),
"The dims of strides shold be the same with that of kernel_sizes. "
"The dims of strides should be the same with that of kernel_sizes. "
"But recieved dims(strides: %u) != dims(kernel_sizes: %u).",
strides.size(), kernel_sizes.size());
PADDLE_ENFORCE_EQ(
......@@ -96,7 +96,7 @@ class UnfoldOp : public framework::OperatorWithKernel {
paddings.size(), strides.size());
PADDLE_ENFORCE_EQ(
strides.size(), dilations.size(),
"The dims of strides shold be the same with that of dilations. "
"The dims of strides should be the same with that of dilations. "
"But recieved dims(strides: %u) != dims(dilations: %u).",
strides.size(), dilations.size());
......
python/paddle/compat.py
浏览文件 @ 433cef03
......@@ -202,7 +202,7 @@ def round(x, d=0):
"""
if six.PY3:
# The official walkaround of round in Python3 is incorrect
# we implement accroding this answer: https://www.techforgeek.info/round_python.html
# we implement according this answer: https://www.techforgeek.info/round_python.html
if x > 0.0:
p = 10**d
return float(math.floor((x * p) + math.copysign(0.5, x))) / p
......
python/paddle/dataset/cifar.py
浏览文件 @ 433cef03
......@@ -17,7 +17,7 @@ CIFAR dataset.
This module will download dataset from https://dataset.bj.bcebos.com/cifar/cifar-10-python.tar.gz and https://dataset.bj.bcebos.com/cifar/cifar-100-python.tar.gz, parse train/test set into
paddle reader creators.
The CIFAR-10 dataset consists of 60000 32x32 colour images in 10 classes,
The CIFAR-10 dataset consists of 60000 32x32 color images in 10 classes,
with 6000 images per class. There are 50000 training images and 10000 test
images.
......
python/paddle/dataset/image.py
浏览文件 @ 433cef03
......@@ -27,7 +27,7 @@ the image layout as follows.
OpenCV use BGR color format. PIL use RGB color format. Both
formats can be used for training. Noted that, the format should
be keep consistent between the training and inference peroid.
be keep consistent between the training and inference period.
"""
from __future__ import print_function
......
python/paddle/dataset/wmt16.py
浏览文件 @ 433cef03
......@@ -112,7 +112,7 @@ def reader_creator(tar_file, file_name, src_dict_size, trg_dict_size, src_lang):
trg_dict = __load_dict(tar_file, trg_dict_size,
("de" if src_lang == "en" else "en"))
# the indice for start mark, end mark, and unk are the same in source
# the index for start mark, end mark, and unk are the same in source
# language and target language. Here uses the source language
# dictionary to determine their indices.
start_id = src_dict[START_MARK]
......
python/paddle/fluid/clip.py
浏览文件 @ 433cef03
......@@ -192,7 +192,7 @@ class GradientClipByNorm(BaseGradientClipAttr):
"""
Convert the input multidimensional Tensor :math:`X` to a multidimensional Tensor whose L2 norm does not exceed the given two-norm maximum ( :math:`clip\_norm` ).
The tensor is not passed through this class, but passed through the parametre of ``main_program`` in ``fluid.program_guard``.
The tensor is not passed through this class, but passed through the parameter of ``main_program`` in ``fluid.program_guard``.
This class limits the L2 norm of the input :math:`X` within :math:`clip\_norm`.
......
python/paddle/fluid/contrib/layers/rnn_impl.py
浏览文件 @ 433cef03
......@@ -156,7 +156,7 @@ def basic_gru(input,
dtype='float32',
name='basic_gru'):
"""
GRU implementation using basic operator, supports multiple layers and bidirection gru.
GRU implementation using basic operator, supports multiple layers and bidirectional gru.
.. math::
u_t & = actGate(W_ux xu_{t} + W_uh h_{t-1} + b_u)
......@@ -377,7 +377,7 @@ def basic_lstm(input,
dtype='float32',
name='basic_lstm'):
"""
LSTM implementation using basic operators, supports multiple layers and bidirection LSTM.
LSTM implementation using basic operators, supports multiple layers and bidirectional LSTM.
.. math::
i_t &= \sigma(W_{ix}x_{t} + W_{ih}h_{t-1} + b_i)
......
python/paddle/fluid/contrib/tests/test_image_classification_fp16.py
浏览文件 @ 433cef03
......@@ -236,7 +236,7 @@ def infer(use_cuda, save_dirname=None):
inference_scope = fluid.core.Scope()
with fluid.scope_guard(inference_scope):
# Use fluid.io.load_inference_model to obtain the inference program desc,
# the feed_target_names (the names of variables that will be feeded
# the feed_target_names (the names of variables that will be fed
# data using feed operators), and the fetch_targets (variables that
# we want to obtain data from using fetch operators).
[inference_program, feed_target_names,
......
python/paddle/fluid/contrib/utils/lookup_table_utils.py
浏览文件 @ 433cef03
......@@ -420,7 +420,7 @@ def get_inference_model(main_program, feeded_var_names, target_vars):
build the inference model. If is set None,
the default main program will be used.
Default: None.
feeded_var_names(list[str]): Names of variables that need to be feeded data
feeded_var_names(list[str]): Names of variables that need to be fed data
during inference.
target_vars(list[Variable]): Variables from which we can get inference
results.
......
python/paddle/fluid/data.py
浏览文件 @ 433cef03
......@@ -27,7 +27,7 @@ def data(name, shape, dtype='float32', lod_level=0):
This function creates a variable on the global block. The global variable
can be accessed by all the following operators in the graph. The variable
is a placeholder that could be feeded with input, such as Executor can feed
is a placeholder that could be fed with input, such as Executor can feed
input into the variable.
Note:
......@@ -35,8 +35,8 @@ def data(name, shape, dtype='float32', lod_level=0):
future version. Please use this `paddle.fluid.data`.
The `paddle.fluid.layers.data` set shape and dtype at compile time but
does NOT check the shape or the dtype of feeded data, this
`paddle.fluid.data` checks the shape and the dtype of data feeded by
does NOT check the shape or the dtype of fed data, this
`paddle.fluid.data` checks the shape and the dtype of data fed by
Executor or ParallelExecutor during run time.
To feed variable size inputs, users can set -1 on the variable
......
python/paddle/fluid/dataset.py
浏览文件 @ 433cef03
......@@ -11,7 +11,7 @@
# 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.
"""This is defination of dataset class, which is high performance IO."""
"""This is definition of dataset class, which is high performance IO."""
from paddle.fluid.proto import data_feed_pb2
from google.protobuf import text_format
......
python/paddle/fluid/executor.py
浏览文件 @ 433cef03
......@@ -196,7 +196,7 @@ def dimension_is_compatible_with(first, second):
def check_feed_shape_type(var, feed, num_places=1):
"""
Returns True if the variable doesn't require feed check or it is compatible
with the shape and have same dtype as the feeded value.
with the shape and have same dtype as the fed value.
A dimension is compatible with the other if:
1. The length of the dimensions are same.
......@@ -206,7 +206,7 @@ def check_feed_shape_type(var, feed, num_places=1):
Args:
var (Variable): the Variable object
feed (LoDTensor): the feeded value, which must be a LoDTensor
feed (LoDTensor): the fed value, which must be a LoDTensor
num_places: an integer value indicating the number of places.
ParallelExecutor will divide data into devices (CPU/GPU) evenly.
Returns:
......@@ -225,8 +225,8 @@ def check_feed_shape_type(var, feed, num_places=1):
num_places) if len(feed.lod()) == 0 else -1
if not dimension_is_compatible_with(feed_shape, var.shape):
raise ValueError(
'The feeded Variable %r should have dimensions = %d, shape = '
'%r, but received feeded shape %r on each device' %
'The fed Variable %r should have dimensions = %d, shape = '
'%r, but received fed shape %r on each device' %
(var.name, len(var.shape), var.shape, feed_shape))
if not dtype_is_compatible_with(feed._dtype(), var.dtype):
var_dtype_format = convert_dtype(var.dtype) if isinstance(
......@@ -234,8 +234,8 @@ def check_feed_shape_type(var, feed, num_places=1):
feed_dtype_format = convert_dtype(feed._dtype()) if isinstance(
feed._dtype(), core.VarDesc.VarType) else feed._dtype()
raise ValueError(
'The data type of feeded Variable %r must be %r, but received %r'
% (var.name, var_dtype_format, feed_dtype_format))
'The data type of fed Variable %r must be %r, but received %r' %
(var.name, var_dtype_format, feed_dtype_format))
return True
......
python/paddle/fluid/incubate/fleet/parameter_server/pslib/__init__.py
浏览文件 @ 433cef03
......@@ -448,7 +448,7 @@ class PSLib(Fleet):
model_proto_file(str): path of program desc proto binary
file, can be local or hdfs/afs file
var_names(list): var name list
load_combine(bool): load from a file or splited param files
load_combine(bool): load from a file or split param files
default False.
Examples:
......@@ -502,7 +502,7 @@ class PSLib(Fleet):
model_proto_file(str): path of program desc proto binary file,
can be local or hdfs/afs file
var_names(list): load var names
load_combine(bool): load from a file or splited param files
load_combine(bool): load from a file or split param files
"""
self._role_maker._barrier_worker()
......
python/paddle/fluid/io.py
浏览文件 @ 433cef03
......@@ -1043,7 +1043,7 @@ def save_inference_model(dirname,
Args:
dirname(str): The directory path to save the inference model.
feeded_var_names(list[str]): list of string. Names of variables that need to be feeded
feeded_var_names(list[str]): list of string. Names of variables that need to be fed
data during inference.
target_vars(list[Variable]): list of Variable. Variables from which we can get
inference results.
......
python/paddle/fluid/layer_helper_base.py
浏览文件 @ 433cef03
......@@ -285,7 +285,7 @@ class LayerHelperBase(object):
Args:
attr: [ParamAttr] should be the parameter attribute for this parameter
shape: shape of the paramter
shape: shape of the parameter
dtype: data type of this parameter
is_bias: if this is a bias parameter
default_initializer: set the default initializer for this parameter
......
python/paddle/fluid/layers/io.py
浏览文件 @ 433cef03
......@@ -56,8 +56,8 @@ def data(name,
a later version. Please use :code:`paddle.fluid.data` .
This :code:`paddle.fluid.layers.data` set shape and dtype at compile
time but does NOT check the shape or the dtype of feeded data, the
:code:`paddle.fluid.data` checks the shape and the dtype of data feeded
time but does NOT check the shape or the dtype of fed data, the
:code:`paddle.fluid.data` checks the shape and the dtype of data fed
by Executor or ParallelExecutor during run time.
To feed variable size inputs, users can feed variable size inputs
......@@ -760,7 +760,7 @@ def create_py_reader_by_data(capacity,
reader.decorate_paddle_reader(
paddle.reader.shuffle(paddle.batch(mnist.train(), batch_size=5), buf_size=500))
img, label = fluid.layers.read_file(reader)
loss = network(img, label) # The definition of custom network and the loss funtion
loss = network(img, label) # The definition of custom network and the loss function
place = fluid.CUDAPlace(0) if USE_CUDA else fluid.CPUPlace()
exe = fluid.Executor(place)
......
python/paddle/fluid/layers/nn.py
浏览文件 @ 433cef03
......@@ -4914,7 +4914,7 @@ def transpose(x, perm, name=None):
Args:
x (Variable): The input Tensor. It is a N-D Tensor of data types float32, float64, int32.
perm (list): Permute the input accoring to the data of perm.
perm (list): Permute the input according to the data of perm.
name (str): The name of this layer. It is optional.
Returns:
......@@ -5488,7 +5488,7 @@ def reshape(x, shape, actual_shape=None, act=None, inplace=False, name=None):
be set -1.
2. 0 means the actual dimension value is going to be copied from the
corresponding dimension of x. The indice of 0s in shape can not exceed
corresponding dimension of x. The index of 0s in shape can not exceed
the dimension of x.
Here are some examples to explain it.
......@@ -6484,7 +6484,7 @@ def image_resize(input,
The input must be a 4-D Tensor of the shape (num_batches, channels, in_h, in_w)
or (num_batches, in_h, in_w, channels), or a 5-D Tensor of the shape
(num_batches, channels, in_d, in_h, in_w) or (num_batches, in_d, in_h, in_w, channels),
and the resizing only applies on the three dimensions(depth, hight and width).
and the resizing only applies on the three dimensions(depth, height and width).
**Warning:** the parameter :attr:`actual_shape` will be deprecated in the
future and only use :attr:`out_shape` instead.
......@@ -8396,7 +8396,7 @@ def pad2d(input,
data_format="NCHW",
name=None):
"""
Pad 2-d images accordding to 'paddings' and 'mode'.
Pad 2-d images according to 'paddings' and 'mode'.
If mode is 'reflect', paddings[0] and paddings[1] must be no greater
than height-1. And the width dimension has the same condition.
......@@ -8418,7 +8418,7 @@ def pad2d(input,
name (str, optional) : The default value is None. Normally there is no need for
user to set this property. For more information, please refer to :ref:`api_guide_Name` .
Returns: a 4-D Tensor padded accordding to paddings and mode and data type is same as input.
Returns: a 4-D Tensor padded according to paddings and mode and data type is same as input.
Return Type: Variable
......@@ -13346,7 +13346,7 @@ def unfold(x, kernel_sizes, strides=1, paddings=0, dilations=1, name=None):
[padding_h, padding_w, padding_h, padding_w]. If an integer
padding was given, [padding, padding, padding, padding] will
be used. For default, paddings will be [0, 0, 0, 0]
dilations(int|list): the dilations of convolution kernel, shold be
dilations(int|list): the dilations of convolution kernel, should be
[dilation_h, dilation_w], or an integer dilation treated as
[dilation, dilation]. For default, it will be [1, 1].
name(str, optional): The default value is None.
......
python/paddle/fluid/layers/rnn.py
浏览文件 @ 433cef03
......@@ -2469,10 +2469,10 @@ def dynamic_gru(input,
See usage for details in :ref:`api_fluid_ParamAttr` .
is_reverse(bool, optional): Whether to compute in the reversed order of
input sequences. Default False.
gate_activation(str, optional): The activation fuction corresponding to
gate_activation(str, optional): The activation function corresponding to
:math:`act_g` in the formula. "sigmoid", "tanh", "relu" and "identity"
are supported. Default "sigmoid".
candidate_activation(str, optional): The activation fuction corresponding to
candidate_activation(str, optional): The activation function corresponding to
:math:`act_c` in the formula. "sigmoid", "tanh", "relu" and "identity"
are supported. Default "tanh".
h_0 (Variable, optional): A Tensor representing the initial hidden state.
......@@ -2618,10 +2618,10 @@ def gru_unit(input,
bias_attr (ParamAttr, optional): To specify the bias parameter property.
Default: None, which means the default bias parameter property is used.
See usage for details in :ref:`api_fluid_ParamAttr` .
activation(str, optional): The activation fuction corresponding to
activation(str, optional): The activation function corresponding to
:math:`act_c` in the formula. "sigmoid", "tanh", "relu" and "identity"
are supported. Default "tanh".
gate_activation(str, optional): The activation fuction corresponding to
gate_activation(str, optional): The activation function corresponding to
:math:`act_g` in the formula. "sigmoid", "tanh", "relu" and "identity"
are supported. Default "sigmoid".
......@@ -2746,7 +2746,7 @@ def beam_search(pre_ids,
`[batch_size * beam_size, K]`, where `K` supposed to be greater than
``beam_size`` and the first dimension size (decrease as samples reach
to the end) should be same as that of ``pre_ids`` . The data type
should be int64. It can be None, which use indice in ``scores`` as
should be int64. It can be None, which use index in ``scores`` as
ids.
scores(Variable): A LodTensor variable containing the accumulated
scores corresponding to ``ids`` . Both its shape and lod are same as
......@@ -2765,7 +2765,7 @@ def beam_search(pre_ids,
to :ref:`api_guide_Name`. Usually name is no need to set and
None by default.
return_parent_idx(bool, optional): Whether to return an extra Tensor variable
in output, which stores the selected ids' parent indice in
in output, which stores the selected ids' parent index in
``pre_ids`` and can be used to update RNN's states by gather operator.
Default False.
......@@ -2774,7 +2774,7 @@ def beam_search(pre_ids,
representing the selected ids and the corresponding accumulated scores of \
current step, have the same shape `[batch_size, beam_size]` and lod with 2 levels, \
and have data types int64 and float32. If ``return_parent_idx`` is True, \
an extra Tensor variable preserving the selected ids' parent indice \
an extra Tensor variable preserving the selected ids' parent index \
is included, whose shape is `[batch_size * beam_size]` and data type \
is int64.
......
python/paddle/fluid/layers/tensor.py
浏览文件 @ 433cef03
......@@ -668,7 +668,7 @@ def fill_constant_batch_size_like(input,
output_dim_idx=0,
force_cpu=False):
"""
This OP creates a Tesnor accroding the shape and dtype, and initializes the
This OP creates a Tesnor according the shape and dtype, and initializes the
Tensor with the constants provided in ``value``. When the input is LoDTensor
and the input_dim_idx is 0, the output_dim_idx dimension is set to the value
of the batch_size input by the input, the Stop_gradient attribute of the created
......
python/paddle/fluid/metrics.py
浏览文件 @ 433cef03
......@@ -261,7 +261,7 @@ class CompositeMetric(MetricBase):
Returns:
list: results of all added metrics.
The shape and dtype of each result depend on the defination of its metric.
The shape and dtype of each result depend on the definition of its metric.
"""
ans = []
for m in self._metrics:
......
python/paddle/fluid/optimizer.py
浏览文件 @ 433cef03
......@@ -3378,10 +3378,10 @@ class PipelineOptimizer(object):
"""
Pipeline Optimizer
Train with pipeline mode. The program will be splited by cut_list.
Train with pipeline mode. The program will be split by cut_list.
If the len of cut_list is k, then the whole program (including \
backward part) will be splited to 2*k-1 sections.
backward part) will be split to 2*k-1 sections.
So the length of place_list and concurrency_list must be also 2*k-1.
......
python/paddle/fluid/parallel_executor.py
浏览文件 @ 433cef03
......@@ -287,7 +287,7 @@ class ParallelExecutor(object):
loss_name=loss.name)
# If the feed is a dict:
# the image will be splitted into devices. If there is two devices
# the image will be split into devices. If there is two devices
# each device will process an image with shape (5, 1)
x = numpy.random.random(size=(10, 1)).astype('float32')
loss_data, = train_exe.run(feed={"X": x},
......
python/paddle/fluid/reader.py
浏览文件 @ 433cef03
......@@ -125,7 +125,7 @@ class DataLoader(object):
presented as a list. It is only valid when iterable=True.
If return_list=False, the return value on each device would
be a dict of str -> LoDTensor, where the key of the dict is
the name of each feeded variables. If return_list=True, the
the name of each fed variables. If return_list=True, the
return value on each device would be a list(LoDTensor). It is
recommended to use return_list=False in static graph mode and
use return_list=True in dygraph mode.
......@@ -891,7 +891,7 @@ class PyReader(DataLoaderBase):
presented as a list. It is only valid when iterable=True.
If return_list=False, the return value on each device would
be a dict of str -> LoDTensor, where the key of the dict is
the name of each feeded variables. If return_list=True, the
the name of each fed variables. If return_list=True, the
return value on each device would be a list(LoDTensor). It is
recommended to use return_list=False in static graph mode and
use return_list=True in dygraph mode.
......
python/paddle/fluid/tests/book/high-level-api/cifar10_small_test_set.py
浏览文件 @ 433cef03
......@@ -18,7 +18,7 @@ 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,
The CIFAR-10 dataset consists of 60000 32x32 color images in 10 classes,
with 6000 images per class. There are 50000 training images and 10000 test
images.
......
python/paddle/fluid/tests/book/high-level-api/test_fit_a_line_new_api.py
浏览文件 @ 433cef03
......@@ -121,7 +121,7 @@ def infer_by_saved_model(use_cuda, save_dirname=None):
inference_scope = fluid.core.Scope()
with fluid.scope_guard(inference_scope):
# Use fluid.io.load_inference_model to obtain the inference program desc,
# the feed_target_names (the names of variables that will be feeded
# the feed_target_names (the names of variables that will be fed
# data using feed operators), and the fetch_targets (variables that
# we want to obtain data from using fetch operators).
[inference_program, feed_target_names,
......
python/paddle/fluid/tests/book/notest_understand_sentiment.py
浏览文件 @ 433cef03
......@@ -213,7 +213,7 @@ def infer(word_dict, use_cuda, save_dirname=None):
inference_scope = fluid.core.Scope()
with fluid.scope_guard(inference_scope):
# Use fluid.io.load_inference_model to obtain the inference program desc,
# the feed_target_names (the names of variables that will be feeded
# the feed_target_names (the names of variables that will be fed
# data using feed operators), and the fetch_targets (variables that
# we want to obtain data from using fetch operators).
[inference_program, feed_target_names,
......
python/paddle/fluid/tests/book/test_fit_a_line.py
浏览文件 @ 433cef03
......@@ -103,7 +103,7 @@ def infer(use_cuda, save_dirname=None):
inference_scope = fluid.core.Scope()
with fluid.scope_guard(inference_scope):
# Use fluid.io.load_inference_model to obtain the inference program desc,
# the feed_target_names (the names of variables that will be feeded
# the feed_target_names (the names of variables that will be fed
# data using feed operators), and the fetch_targets (variables that
# we want to obtain data from using fetch operators).
[inference_program, feed_target_names,
......
python/paddle/fluid/tests/book/test_image_classification.py
浏览文件 @ 433cef03
......@@ -210,7 +210,7 @@ def infer(use_cuda, save_dirname=None):
inference_scope = fluid.core.Scope()
with fluid.scope_guard(inference_scope):
# Use fluid.io.load_inference_model to obtain the inference program desc,
# the feed_target_names (the names of variables that will be feeded
# the feed_target_names (the names of variables that will be fed
# data using feed operators), and the fetch_targets (variables that
# we want to obtain data from using fetch operators).
[inference_program, feed_target_names,
......
python/paddle/fluid/tests/book/test_label_semantic_roles.py
浏览文件 @ 433cef03
......@@ -247,7 +247,7 @@ def infer(use_cuda, save_dirname=None):
inference_scope = fluid.core.Scope()
with fluid.scope_guard(inference_scope):
# Use fluid.io.load_inference_model to obtain the inference program desc,
# the feed_target_names (the names of variables that will be feeded
# the feed_target_names (the names of variables that will be fed
# data using feed operators), and the fetch_targets (variables that
# we want to obtain data from using fetch operators).
[inference_program, feed_target_names,
......
python/paddle/fluid/tests/book/test_recommender_system.py
浏览文件 @ 433cef03
......@@ -254,7 +254,7 @@ def infer(use_cuda, save_dirname=None):
inference_scope = fluid.core.Scope()
with fluid.scope_guard(inference_scope):
# Use fluid.io.load_inference_model to obtain the inference program desc,
# the feed_target_names (the names of variables that will be feeded
# the feed_target_names (the names of variables that will be fed
# data using feed operators), and the fetch_targets (variables that
# we want to obtain data from using fetch operators).
[inference_program, feed_target_names,
......
python/paddle/fluid/tests/book/test_rnn_encoder_decoder.py
浏览文件 @ 433cef03
......@@ -208,7 +208,7 @@ def infer(use_cuda, save_dirname=None):
inference_scope = fluid.core.Scope()
with fluid.scope_guard(inference_scope):
# Use fluid.io.load_inference_model to obtain the inference program desc,
# the feed_target_names (the names of variables that will be feeded
# the feed_target_names (the names of variables that will be fed
# data using feed operators), and the fetch_targets (variables that
# we want to obtain data from using fetch operators).
[inference_program, feed_target_names,
......
python/paddle/fluid/tests/book/test_word2vec.py
浏览文件 @ 433cef03
......@@ -151,7 +151,7 @@ def infer(use_cuda, save_dirname=None):
inference_scope = fluid.core.Scope()
with fluid.scope_guard(inference_scope):
# Use fluid.io.load_inference_model to obtain the inference program desc,
# the feed_target_names (the names of variables that will be feeded
# the feed_target_names (the names of variables that will be fed
# data using feed operators), and the fetch_targets (variables that
# we want to obtain data from using fetch operators).
[inference_program, feed_target_names,
......
python/paddle/fluid/tests/unittests/dist_transformer.py
浏览文件 @ 433cef03
......@@ -272,7 +272,7 @@ class LearningRateScheduler(object):
"""
Wrapper for learning rate scheduling as described in the Transformer paper.
LearningRateScheduler adapts the learning rate externally and the adapted
learning rate will be feeded into the main_program as input data.
learning rate will be fed into the main_program as input data.
"""
def __init__(self,
......
python/paddle/fluid/tests/unittests/ir/pass_test.py
浏览文件 @ 433cef03
......@@ -130,7 +130,7 @@ class PassTest(unittest.TestCase):
if startup_on_cpu and not isinstance(place, fluid.CPUPlace):
warnings.warn(
"Parameters are on CPU, and will be transfered to GPU "
"Parameters are on CPU, and will be transferred to GPU "
"automatically by data transform.")
outs_opt, lods_opt = self._run_program(executor, opt_program)
......
python/paddle/fluid/tests/unittests/test_dist_transpiler.py
浏览文件 @ 433cef03
......@@ -118,7 +118,7 @@ class TestBasicModel(TranspilerTest):
trainer, trainer_startup = self.get_trainer()
# splited var blocks should be in startup program
# split var blocks should be in startup program
self.assertTrue("fc_w.block0" in trainer_startup.global_block().vars)
self.assertTrue("fc_w.block1" in trainer_startup.global_block().vars)
self.assertTrue("fc_w" in trainer_startup.global_block().vars)
......
python/paddle/fluid/tests/unittests/test_feed_data_check_shape_type.py
浏览文件 @ 433cef03
......@@ -46,7 +46,7 @@ class TestFeedData(unittest.TestCase):
def _get_feed_batch_size(self, use_cuda, use_parallel_executor):
"""
Returns actual feeded data size. We should multiple the number of
Returns actual fed data size. We should multiple the number of
devices when it is using ParallelExecutor
"""
return self.data_batch_size * self._get_device_count(
......@@ -100,8 +100,8 @@ class TestFeedData(unittest.TestCase):
use_parallel_executor)
self.assertEqual(
str(shape_mismatch_err.exception),
"The feeded Variable %r should have dimensions = %r, "
"shape = %r, but received feeded shape %r on each device" %
"The fed Variable %r should have dimensions = %r, "
"shape = %r, but received fed shape %r on each device" %
(u'data', len(in_shape_tuple), in_shape_tuple,
error_shape_list))
......@@ -110,7 +110,7 @@ class TestFeedData(unittest.TestCase):
use_parallel_executor)
self.assertEqual(
str(dtype_mismatch_err.exception),
"The data type of feeded Variable %r must be 'int64', but "
"The data type of fed Variable %r must be 'int64', but "
"received 'float64'" % (u'label'))
def _test_feed_data_dtype_mismatch(self, use_cuda, use_parallel_executor):
......
python/paddle/fluid/tests/unittests/test_static_save_load.py
浏览文件 @ 433cef03
......@@ -905,7 +905,7 @@ class TestLoadFromOldInterface(unittest.TestCase):
with self.assertRaises(RuntimeError):
fluid.load(main_program, "test_path", exe)
# check unused paramter
# check unused parameter
fluid.load(test_clone_program, "test_path", exe)
......
python/paddle/fluid/transpiler/distribute_transpiler.py
浏览文件 @ 433cef03
......@@ -16,11 +16,11 @@ from __future__ import print_function
"""
Steps to transpile trainer:
1. split variable to multiple blocks, aligned by product(dim[1:]) (width).
2. rename splited grad variables to add trainer_id suffix ".trainer_%d".
2. rename split grad variables to add trainer_id suffix ".trainer_%d".
3. modify trainer program add split_op to each grad variable.
4. append send_op to send splited variables to server and
5. add recv_op to fetch params(splited blocks or origin param) from server.
6. append concat_op to merge splited blocks to update local weights.
4. append send_op to send split variables to server and
5. add recv_op to fetch params(split blocks or origin param) from server.
6. append concat_op to merge split blocks to update local weights.
Steps to transpile pserver:
1. create new program for parameter server.
......@@ -106,7 +106,7 @@ def slice_variable(var_list, slice_count, min_block_size):
var_list (list): List of variables.
slice_count (int): Numel of count that variables will be sliced, which
could be the pserver services' count.
min_block_size (int): Minimum splitted block size.
min_block_size (int): Minimum split block size.
Returns:
blocks (list[(varname, block_id, current_block_size)]): A list
of VarBlocks. Each VarBlock specifies a shard of the var.
......@@ -157,7 +157,7 @@ class DistributeTranspilerConfig(object):
.. py:attribute:: min_block_size (int)
Minimum number of splitted elements in block, default is 8192.
Minimum number of split elements in block, default is 8192.
According to : https://github.com/PaddlePaddle/Paddle/issues/8638#issuecomment-369912156
We can use bandwidth efficiently when data size is larger than 2MB.If you
......@@ -667,8 +667,8 @@ WIKI: https://github.com/PaddlePaddle/Fleet/blob/develop/markdown_doc/transpiler
self.origin_program._is_chief = self.trainer_id == 0
self.origin_program._distributed_lookup_table = self.table_name if self.table_name else None
# split and create vars, then put splited vars in dicts for later use.
# step 1: split and create vars, then put splited vars in dicts for later use.
# split and create vars, then put split vars in dicts for later use.
# step 1: split and create vars, then put split vars in dicts for later use.
self._init_splited_vars()
# step 2: insert send op to send gradient vars to parameter servers
......@@ -742,8 +742,8 @@ WIKI: https://github.com/PaddlePaddle/Fleet/blob/develop/markdown_doc/transpiler
sections = []
send_varnames = []
# get send op_role_var, if not splited, the grad should have .trainer suffix
# if splited, grad should be the original grad var name (split_by_ref and send
# get send op_role_var, if not split, the grad should have .trainer suffix
# if split, grad should be the original grad var name (split_by_ref and send
# will be on the same place). ParallelExecutor
# will use op_role_var to get expected device place to run this op.
program.global_block()._insert_op(
......@@ -860,8 +860,8 @@ WIKI: https://github.com/PaddlePaddle/Fleet/blob/develop/markdown_doc/transpiler
recv_dep_in = self.grad_name_to_send_dummy_out[
self.param_name_to_grad_name[param_varname]]
# get recv op_role_var, if not splited, the grad should have .trainer suffix
# if splited, grad should be the original grad var name. ParallelExecutor
# get recv op_role_var, if not split, the grad should have .trainer suffix
# if split, grad should be the original grad var name. ParallelExecutor
# will use op_role_var to get expected device place to run this op.
orig_grad_name = self.param_name_to_grad_name[param_varname]
recv_op_role_var_name = orig_grad_name
......@@ -1120,7 +1120,7 @@ WIKI: https://github.com/PaddlePaddle/Fleet/blob/develop/markdown_doc/transpiler
for varname, splited_var in six.iteritems(self.param_var_mapping):
if varname in sparse_table_names:
continue
# add concat ops to merge splited parameters received from parameter servers.
# add concat ops to merge split parameters received from parameter servers.
if len(splited_var) <= 1:
continue
# NOTE: if enable memory optimization, origin vars maybe removed.
......@@ -1670,8 +1670,8 @@ WIKI: https://github.com/PaddlePaddle/Fleet/blob/develop/markdown_doc/transpiler
def _init_splited_vars(self):
# update these mappings for further transpile:
# 1. param_var_mapping: param var name -> [splited params vars]
# 2. grad_var_mapping: grad var name -> [splited grads vars]
# 1. param_var_mapping: param var name -> [split params vars]
# 2. grad_var_mapping: grad var name -> [split grads vars]
# 3. grad_param_mapping: grad.blockx -> param.blockx
# 4. param_grad_ep_mapping: ep -> {"params": [], "grads": []}
......@@ -1966,7 +1966,7 @@ WIKI: https://github.com/PaddlePaddle/Fleet/blob/develop/markdown_doc/transpiler
outputs={"Out": [grad_var]},
attrs={"use_mkldnn": False})
else:
# in async_mode, for table gradient, it also need to be splited to each parameter server
# in async_mode, for table gradient, it also need to be split to each parameter server
origin_grad_name = grad_var.name
splited_grad_name = self.trainer_side_table_grad_list[
pserver_index].name
......@@ -2040,9 +2040,9 @@ WIKI: https://github.com/PaddlePaddle/Fleet/blob/develop/markdown_doc/transpiler
block_map[varname] = []
block_map[varname].append((int(offset), int(size)))
for varname, splited in six.iteritems(block_map):
for varname, split in six.iteritems(block_map):
orig_var = program.global_block().var(varname)
if len(splited) == 1:
if len(split) == 1:
if self.sync_mode and add_trainer_suffix:
new_var_name = "%s.trainer_%d" % \
(orig_var.name, self.trainer_id)
......@@ -2059,7 +2059,7 @@ WIKI: https://github.com/PaddlePaddle/Fleet/blob/develop/markdown_doc/transpiler
if len(orig_shape) >= 2:
orig_dim1_flatten = reduce(lambda x, y: x * y, orig_shape[1:])
for i, block in enumerate(splited):
for i, block in enumerate(split):
size = block[1]
rows = size // orig_dim1_flatten
splited_shape = [rows]
......@@ -2077,7 +2077,7 @@ WIKI: https://github.com/PaddlePaddle/Fleet/blob/develop/markdown_doc/transpiler
persistable=False,
dtype=orig_var.dtype,
type=orig_var.type,
shape=splited_shape) # flattend splited var
shape=splited_shape) # flattend split var
var_mapping[varname].append(var)
program.global_block()._sync_with_cpp()
return var_mapping
......@@ -2393,9 +2393,9 @@ WIKI: https://github.com/PaddlePaddle/Fleet/blob/develop/markdown_doc/transpiler
if the variable is not grad/param, e.g.
a@GRAD -> a@GRAD.block0
a@GRAD -> a@GRAD (a is not splited)
a@GRAD -> a@GRAD (a is not split)
fc_0.w_0 -> fc_0.w_0.block_0
fc_0.w_0 -> fc_0.w_0 (weight is not splited)
fc_0.w_0 -> fc_0.w_0 (weight is not split)
_generated_var_123 -> None
"""
grad_block = None
......@@ -2403,7 +2403,7 @@ WIKI: https://github.com/PaddlePaddle/Fleet/blob/develop/markdown_doc/transpiler
if self._orig_varname(g.name) == self._orig_varname(var.name):
# skip per trainer vars
if g.name.find(".trainer_") == -1:
# only param or grads have splited blocks
# only param or grads have split blocks
if self._orig_varname(g.name) in self.grad_name_to_param_name or \
self._orig_varname(g.name) in self.param_name_to_grad_name:
grad_block = g
......@@ -2442,7 +2442,7 @@ WIKI: https://github.com/PaddlePaddle/Fleet/blob/develop/markdown_doc/transpiler
varlist = [varlist]
for i in range(len(varlist)):
var = varlist[i]
# for ops like clipping and weight decay, get the splited var (xxx.block0)
# for ops like clipping and weight decay, get the split var (xxx.block0)
# for inputs/outputs
grad_block = self._get_pserver_grad_param_var(
var, program.global_block().vars)
......
python/paddle/fluid/transpiler/geo_sgd_transpiler.py
浏览文件 @ 433cef03
......@@ -108,15 +108,15 @@ class GeoSgdTranspiler(DistributeTranspiler):
self.sparse_var_list = []
self.sparse_var_splited_list = []
# split and create vars, then put splited vars in dicts for later use.
# step 1. split and create vars, then put splited vars in dicts for later use.
# split and create vars, then put split vars in dicts for later use.
# step 1. split and create vars, then put split vars in dicts for later use.
self._init_splited_vars()
# step 3. create send recv var (param after optimize)
send_vars = []
ps_dispatcher.reset()
param_var_mapping_items = list(six.iteritems(self.param_var_mapping))
# send_vars is the parameter which splited by communicator and send to pserver,not the origin parameter
# send_vars is the parameter which split by communicator and send to pserver,not the origin parameter
for _, splited_vars in param_var_mapping_items:
for _, var in enumerate(splited_vars):
send_vars.append(var)
......@@ -292,7 +292,7 @@ class GeoSgdTranspiler(DistributeTranspiler):
len(self.pserver_endpoints),
self.config.min_block_size)
# step 3. Create splited param from split blocks
# step 3. Create split param from split blocks
# origin_param_name -> [splited_param_vars]
# Todo: update _create_vars_from_blocklist
self.param_var_mapping = self._create_vars_from_blocklist(
......
python/paddle/utils/preprocess_util.py
浏览文件 @ 433cef03
......@@ -301,9 +301,9 @@ class DatasetCreater(object):
Create a data set object from a path.
It will use directory structure or a file list to determine dataset if
self.from_list is True. Otherwise, it will uses a file list to
determine the datset.
determine the dataset.
path: the path of the dataset.
return a tuple of Dataset object, and a mapping from lable set
return a tuple of Dataset object, and a mapping from label set
to label id.
"""
if self.from_list:
......@@ -314,9 +314,9 @@ class DatasetCreater(object):
def create_dataset_from_list(self, path):
"""
Create a data set object from a path.
It will uses a file list to determine the datset.
It will uses a file list to determine the dataset.
path: the path of the dataset.
return a tuple of Dataset object, and a mapping from lable set
return a tuple of Dataset object, and a mapping from label set
to label id
"""
raise NotImplementedError
......@@ -327,7 +327,7 @@ class DatasetCreater(object):
It will use directory structure or a file list to determine dataset if
self.from_list is True.
path: the path of the dataset.
return a tuple of Dataset object, and a mapping from lable set
return a tuple of Dataset object, and a mapping from label set
to label id
"""
raise NotImplementedError
......
tools/check_api_approvals.sh
浏览文件 @ 433cef03
#!/bin/bash
if [ -z ${BRANCH} ]; then
BRANCH="develop"
fi
......
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