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

7 years ago · 4fef60298f
parent fe93cce387 5c58eee552
commit 4fef60298f
13 changed files with 44 additions and 22 deletions
--- a/cmake/external/snappy.cmake
+++ b/cmake/external/snappy.cmake
@ -47,8 +47,6 @@ ExternalProject_Add(
                     -DCMAKE_INSTALL_LIBDIR:PATH=${SNAPPY_INSTALL_DIR}/lib
                     -DCMAKE_POSITION_INDEPENDENT_CODE:BOOL=ON
                     -DCMAKE_BUILD_TYPE:STRING=${THIRD_PARTY_BUILD_TYPE}
-    BUILD_COMMAND   make -j8
-    INSTALL_COMMAND make install
 )

 add_library(snappy STATIC IMPORTED GLOBAL)
--- a/cmake/external/snappystream.cmake
+++ b/cmake/external/snappystream.cmake
@ -46,8 +46,6 @@ ExternalProject_Add(
                        -DCMAKE_INSTALL_PREFIX:PATH=${SNAPPYSTREAM_INSTALL_DIR}
                        -DCMAKE_INSTALL_LIBDIR:PATH=${SNAPPYSTREAM_INSTALL_DIR}/lib
                        -DCMAKE_BUILD_TYPE:STRING=${THIRD_PARTY_BUILD_TYPE}
-        BUILD_COMMAND   make -j8
-        INSTALL_COMMAND make install
        DEPENDS snappy
 )

--- a/cmake/inference_lib.cmake
+++ b/cmake/inference_lib.cmake
@ -148,4 +148,10 @@ copy(string_lib
  DSTS ${dst_dir}/${module} ${dst_dir}/${module}/tinyformat
 )

+set(module "pybind")
+copy(pybind_lib
+  SRCS ${CMAKE_CURRENT_BINARY_DIR}/paddle/fluid/${module}/pybind.h
+  DSTS ${dst_dir}/${module}
+)
+
 add_custom_target(inference_lib_dist DEPENDS ${inference_lib_dist_dep}) 
--- a/paddle/CMakeLists.txt
+++ b/paddle/CMakeLists.txt
@ -24,6 +24,6 @@ if(NOT WITH_FLUID_ONLY)
 endif()

 add_subdirectory(testing)
-if(NOT MOBILE_INFERENCE AND NOT RPI)
+if(NOT MOBILE_INFERENCE AND NOT RPI AND NOT WITH_C_API)
  add_subdirectory(fluid)
 endif()
--- a/paddle/fluid/framework/data_type.cc
+++ b/paddle/fluid/framework/data_type.cc
@ -58,6 +58,7 @@ static DataTypeMap* InitDataTypeMap() {
  RegType(bool, proto::VarType::BOOL);
  RegType(size_t, proto::VarType::SIZE_T);
  RegType(int16_t, proto::VarType::INT16);
+  RegType(uint8_t, proto::VarType::UINT8);

 #undef RegType
  return retv;
--- a/paddle/fluid/framework/data_type.h
+++ b/paddle/fluid/framework/data_type.h
@ -47,8 +47,14 @@ inline void VisitDataType(proto::VarType::Type type, Visitor visitor) {
    case proto::VarType::BOOL:
      visitor.template operator()<bool>();
      break;
+    case proto::VarType::UINT8:
+      visitor.template operator()<uint8_t>();
+      break;
+    case proto::VarType::INT16:
+      visitor.template operator()<int16_t>();
+      break;
    default:
-      PADDLE_THROW("Not supported");
+      PADDLE_THROW("Not supported %d", type);
  }
 }

--- a/paddle/fluid/framework/details/fetch_op_handle.cc
+++ b/paddle/fluid/framework/details/fetch_op_handle.cc
@ -48,17 +48,18 @@ void FetchOpHandle::RunImpl() {
  WaitInputVarGenerated(platform::CPUPlace());

  tensors_.resize(inputs_.size());
-  auto *var_handle = static_cast<VarHandle *>(inputs_[0]);
-  auto &var_name = var_handle->name_;
  platform::CPUPlace cpu;
  auto &scopes = *local_scopes_;

-  for (size_t i = 0; i < scopes.size(); ++i) {
-    auto &scope = scopes[i];
-    auto *var =
-        scope->FindVar(kLocalExecScopeName)->Get<Scope *>()->FindVar(var_name);
+  for (size_t i = 0; i < inputs_.size(); ++i) {
+    auto *var_handle = static_cast<VarHandle *>(inputs_[i]);
+    auto &scope = scopes.at(var_handle->scope_idx_);
+    auto *var = scope->FindVar(kLocalExecScopeName)
+                    ->Get<Scope *>()
+                    ->FindVar(var_handle->name_);
    PADDLE_ENFORCE_NOT_NULL(var, "Cannot find variable %s in execution scope",
-                            var_name);
+                            var_handle->name_);
+
    auto &t = var->Get<framework::LoDTensor>();
    if (platform::is_gpu_place(t.place())) {
 #ifdef PADDLE_WITH_CUDA
--- a/paddle/fluid/framework/framework.proto
+++ b/paddle/fluid/framework/framework.proto
@ -103,6 +103,7 @@ message VarType {
    FP64 = 6;
    // Tensor<size_t> is used in C++.
    SIZE_T = 19;
+    UINT8 = 20;

    // Other types that may need additional descriptions
    LOD_TENSOR = 7;
--- a/paddle/fluid/framework/lod_tensor_test.cc
+++ b/paddle/fluid/framework/lod_tensor_test.cc
@ -228,11 +228,12 @@ TEST(LoD, CheckAbsLoD) {
  ASSERT_FALSE(CheckAbsLoD(abs_lod0));
 }

-TEST(LoDTensor, RecordIO) {
+template <typename T>
+static void TestRecordIO() {
  LoDTensor tensor;
-  int* tmp = tensor.mutable_data<int>(make_ddim({4, 5}), platform::CPUPlace());
+  T* tmp = tensor.mutable_data<T>(make_ddim({4, 5}), platform::CPUPlace());
  for (int i = 0; i < 20; ++i) {
-    tmp[i] = i;
+    tmp[i] = static_cast<T>(i);
  }

  std::stringstream* stream = new std::stringstream();
@ -247,7 +248,7 @@ TEST(LoDTensor, RecordIO) {

  auto assert_tensor_ok = [](const LoDTensor& tensor) {
    for (int i = 0; i < 20; ++i) {
-      ASSERT_EQ(tensor.data<int>()[i], i);
+      ASSERT_EQ(tensor.data<T>()[i], static_cast<T>(i));
    }
  };

@ -265,5 +266,13 @@ TEST(LoDTensor, RecordIO) {
  }
 }

+TEST(LoDTensor, RecordIO) {
+  TestRecordIO<int>();
+  TestRecordIO<int16_t>();
+  TestRecordIO<uint8_t>();
+  TestRecordIO<float>();
+  TestRecordIO<double>();
+}
+
 }  // namespace framework
 }  // namespace paddle
--- a/paddle/fluid/operators/math/math_function.cc
+++ b/paddle/fluid/operators/math/math_function.cc
@ -38,7 +38,9 @@ template struct SetConstant<platform::CPUDeviceContext, bool>;
  template struct Transpose<platform::CPUDeviceContext, double, RANK>;     \
  template struct Transpose<platform::CPUDeviceContext, int, RANK>;        \
  template struct Transpose<platform::CPUDeviceContext, int64_t, RANK>;    \
-  template struct Transpose<platform::CPUDeviceContext, bool, RANK>;
+  template struct Transpose<platform::CPUDeviceContext, bool, RANK>;       \
+  template struct Transpose<platform::CPUDeviceContext, int16_t, RANK>;    \
+  template struct Transpose<platform::CPUDeviceContext, uint8_t, RANK>;

 DEFINE_CPU_TRANS(1);
 DEFINE_CPU_TRANS(2);
--- a/paddle/fluid/platform/CMakeLists.txt
+++ b/paddle/fluid/platform/CMakeLists.txt
@ -1,4 +1,4 @@
-proto_library(profiler_proto SRCS profiler.proto)
+proto_library(profiler_proto SRCS profiler.proto DEPS framework_proto)
 py_proto_compile(profiler_py_proto SRCS profiler.proto)

 add_custom_target(profiler_py_proto_init ALL COMMAND ${CMAKE_COMMAND} -E touch __init__.py)
--- a/paddle/scripts/paddle_build.sh
+++ b/paddle/scripts/paddle_build.sh
@ -504,6 +504,7 @@ function main() {
        ;;
      capi)
        cmake_gen ${PYTHON_ABI:-""}
+        build
        gen_capi_package
        ;;
      fluid_inference_lib)
--- a/python/paddle/fluid/tests/unittests/test_parallel_executor.py
+++ b/python/paddle/fluid/tests/unittests/test_parallel_executor.py
@ -775,7 +775,7 @@ class TestCRFModel(unittest.TestCase):
        build_strategy = fluid.BuildStrategy()
        build_strategy.reduce_strategy = fluid.BuildStrategy.ReduceStrategy.Reduce
        self.check_network_convergence(
-            is_sparse=False, build_strategy=build_strategy)
+            is_sparse=True, build_strategy=build_strategy)

    def test_update_dense_parameter_reduce(self):
        build_strategy = fluid.BuildStrategy()
@ -849,8 +849,7 @@ class TestFetchOp(unittest.TestCase):
                    assert not math.isnan(np.sum(ret[i])) and \
                           not math.isinf(np.sum(ret[i]))

-    @unittest.skip("this test is buggy")
-    def test_feed(self):
+    def test_fetch_op(self):
        tst_reader = paddle.batch(flowers.test(use_xmap=False), batch_size=16)
        tst_reader_iter = tst_reader()