diff --git a/CMakeLists.txt b/CMakeLists.txt index c62cc9bfd70d72d926eeee5eb52a69428855eb9b..bc2ac2cd939690456930d78ed0bda39dd0953173 100644 --- a/CMakeLists.txt +++ b/CMakeLists.txt @@ -130,6 +130,21 @@ if (APPLE OR WIN32) "Disable MKL for building on mac and windows" FORCE) endif() +if (WIN32) + set(WITH_AVX OFF CACHE STRING + "Disable AVX when compiling for Windows" FORCE) + set(WITH_DSO OFF CACHE STRING + "Disable DSO when compiling for Windows" FORCE) + set(WITH_MKL OFF CACHE STRING + "Disable MKL when compiling for Windows" FORCE) + set(WITH_DISTRIBUTE OFF CACHE STRING + "Disable DISTRIBUTE when compiling for Windows" FORCE) + set(WITH_C_API OFF CACHE STRING + "Disable C_API when compiling for Windows" FORCE) + set(WITH_FLUID_ONLY ON CACHE STRING + "Enable FLUID_ONLY when compiling for Windows" FORCE) +endif() + set(THIRD_PARTY_PATH "${CMAKE_BINARY_DIR}/third_party" CACHE STRING "A path setting third party libraries download & build directories.") @@ -190,11 +205,11 @@ include(external/pybind11) # download pybind11 include(external/cares) include(external/cub) include(external/xxhash) # download xxhash - -if (NOT WIN32) -# there is no official support of snappystream, warpctc, nccl, cupti in windows include(external/snappy) # download snappy include(external/snappystream) # download snappystream + +if (NOT WIN32) +# there is no official support of warpctc, nccl, cupti in windows include(external/warpctc) # download, build, install warpctc include(cupti) endif (NOT WIN32) diff --git a/cmake/external/gtest.cmake b/cmake/external/gtest.cmake index d335298742c73bf1fe44e6a778ab3c142711063d..4fe9c13fb7f2c04ae04e985252996dfa308ac304 100644 --- a/cmake/external/gtest.cmake +++ b/cmake/external/gtest.cmake @@ -50,7 +50,11 @@ IF(WITH_TESTING) CMAKE_ARGS -DCMAKE_CXX_COMPILER=${CMAKE_CXX_COMPILER} -DCMAKE_C_COMPILER=${CMAKE_C_COMPILER} -DCMAKE_CXX_FLAGS=${CMAKE_CXX_FLAGS} + -DCMAKE_CXX_FLAGS_RELEASE=${CMAKE_CXX_FLAGS_RELEASE} + -DCMAKE_CXX_FLAGS_DEBUG=${CMAKE_CXX_FLAGS_DEBUG} -DCMAKE_C_FLAGS=${CMAKE_C_FLAGS} + -DCMAKE_C_FLAGS_DEBUG=${CMAKE_C_FLAGS_DEBUG} + -DCMAKE_C_FLAGS_RELEASE=${CMAKE_C_FLAGS_RELEASE} -DCMAKE_INSTALL_PREFIX=${GTEST_INSTALL_DIR} -DCMAKE_POSITION_INDEPENDENT_CODE=ON -DBUILD_GMOCK=ON diff --git a/cmake/external/snappy.cmake b/cmake/external/snappy.cmake index af09ed4d5d6e21cc50aba5198a7e9ea56f49451a..b30403d2d81ce471f39b4d92e24a500fe41eeebb 100644 --- a/cmake/external/snappy.cmake +++ b/cmake/external/snappy.cmake @@ -24,7 +24,11 @@ set(SNAPPY_SOURCES_DIR ${THIRD_PARTY_PATH}/snappy) set(SNAPPY_INSTALL_DIR ${THIRD_PARTY_PATH}/install/snappy) set(SNAPPY_INCLUDE_DIR "${SNAPPY_INSTALL_DIR}/include" CACHE PATH "snappy include directory." FORCE) -set(SNAPPY_LIBRARIES "${SNAPPY_INSTALL_DIR}/lib/libsnappy.a") +if (WIN32) + set(SNAPPY_LIBRARIES "${SNAPPY_INSTALL_DIR}/lib/snappy.lib") +else(WIN32) + set(SNAPPY_LIBRARIES "${SNAPPY_INSTALL_DIR}/lib/libsnappy.a") +endif (WIN32) ExternalProject_Add( extern_snappy @@ -34,8 +38,12 @@ ExternalProject_Add( UPDATE_COMMAND "" CMAKE_ARGS -DCMAKE_CXX_COMPILER=${CMAKE_CXX_COMPILER} -DCMAKE_C_COMPILER=${CMAKE_C_COMPILER} - -DCMAKE_CXX_FLAGS=${CMAKE_CXX_FLAGS} -DCMAKE_C_FLAGS=${CMAKE_C_FLAGS} + -DCMAKE_C_FLAGS_DEBUG=${CMAKE_C_FLAGS_DEBUG} + -DCMAKE_C_FLAGS_RELEASE=${CMAKE_C_FLAGS_RELEASE} + -DCMAKE_CXX_FLAGS=${CMAKE_CXX_FLAGS} + -DCMAKE_CXX_FLAGS_RELEASE=${CMAKE_CXX_FLAGS_RELEASE} + -DCMAKE_CXX_FLAGS_DEBUG=${CMAKE_CXX_FLAGS_DEBUG} -DCMAKE_INSTALL_PREFIX=${SNAPPY_INSTALL_DIR} -DCMAKE_INSTALL_LIBDIR=${SNAPPY_INSTALL_DIR}/lib -DCMAKE_POSITION_INDEPENDENT_CODE=ON diff --git a/cmake/external/snappystream.cmake b/cmake/external/snappystream.cmake index 6df636d7fa8757ade73892bda03a80ba9767472b..1ec79462c14e44f2d0abe6904383ebd91d94d35e 100644 --- a/cmake/external/snappystream.cmake +++ b/cmake/external/snappystream.cmake @@ -18,36 +18,45 @@ ENDIF() include (ExternalProject) -# NOTE: snappy is needed when linking with recordio - set(SNAPPYSTREAM_SOURCES_DIR ${THIRD_PARTY_PATH}/snappy_stream) set(SNAPPYSTREAM_INSTALL_DIR ${THIRD_PARTY_PATH}/install/snappy_stream) set(SNAPPYSTREAM_INCLUDE_DIR "${SNAPPYSTREAM_INSTALL_DIR}/include" CACHE PATH "snappy stream include directory." FORCE) -set(SNAPPYSTREAM_LIBRARIES "${SNAPPYSTREAM_INSTALL_DIR}/lib/libsnappystream.a") - -ExternalProject_Add( - extern_snappystream - GIT_REPOSITORY "https://github.com/hoxnox/snappystream.git" - GIT_TAG "0.2.8" - PREFIX ${SNAPPYSTREAM_SOURCES_DIR} - UPDATE_COMMAND "" - CMAKE_ARGS -DCMAKE_CXX_COMPILER=${CMAKE_CXX_COMPILER} - -DCMAKE_C_COMPILER=${CMAKE_C_COMPILER} - -DCMAKE_CXX_FLAGS=${CMAKE_CXX_FLAGS} - -DCMAKE_C_FLAGS=${CMAKE_C_FLAGS} - -DCMAKE_INSTALL_PREFIX=${SNAPPY_INSTALL_DIR} - -DCMAKE_INSTALL_LIBDIR=${SNAPPY_INSTALL_DIR}/lib - -DCMAKE_POSITION_INDEPENDENT_CODE=ON - -DCMAKE_BUILD_TYPE=${THIRD_PARTY_BUILD_TYPE} - -DSNAPPY_ROOT=${SNAPPY_INSTALL_DIR} - ${EXTERNAL_OPTIONAL_ARGS} - CMAKE_CACHE_ARGS - -DCMAKE_INSTALL_PREFIX:PATH=${SNAPPYSTREAM_INSTALL_DIR} - -DCMAKE_INSTALL_LIBDIR:PATH=${SNAPPYSTREAM_INSTALL_DIR}/lib - -DCMAKE_BUILD_TYPE:STRING=${THIRD_PARTY_BUILD_TYPE} - DEPENDS snappy -) +if(WIN32) + # Fix me, VS2015 come without VLA support + set(SNAPPYSTREAM_LIBRARIES "${SNAPPYSTREAM_INSTALL_DIR}/lib/snappystream.lib") + MESSAGE(WARNING, "In windows, snappystream has no compile support for windows, + please build it manually and put it at " ${SNAPPYSTREAM_INSTALL_DIR}) +else(WIN32) + set(SNAPPYSTREAM_LIBRARIES "${SNAPPYSTREAM_INSTALL_DIR}/lib/libsnappystream.a") + + ExternalProject_Add( + extern_snappystream + GIT_REPOSITORY "https://github.com/hoxnox/snappystream.git" + GIT_TAG "0.2.8" + PREFIX ${SNAPPYSTREAM_SOURCES_DIR} + UPDATE_COMMAND "" + CMAKE_ARGS -DCMAKE_CXX_COMPILER=${CMAKE_CXX_COMPILER} + -DCMAKE_C_COMPILER=${CMAKE_C_COMPILER} + -DCMAKE_C_FLAGS=${CMAKE_C_FLAGS} + -DCMAKE_C_FLAGS_DEBUG=${CMAKE_C_FLAGS_DEBUG} + -DCMAKE_C_FLAGS_RELEASE=${CMAKE_C_FLAGS_RELEASE} + -DCMAKE_CXX_FLAGS=${CMAKE_CXX_FLAGS} + -DCMAKE_CXX_FLAGS_RELEASE=${CMAKE_CXX_FLAGS_RELEASE} + -DCMAKE_CXX_FLAGS_DEBUG=${CMAKE_CXX_FLAGS_DEBUG} + -DCMAKE_INSTALL_PREFIX=${SNAPPY_INSTALL_DIR} + -DCMAKE_INSTALL_LIBDIR=${SNAPPY_INSTALL_DIR}/lib + -DCMAKE_POSITION_INDEPENDENT_CODE=ON + -DCMAKE_BUILD_TYPE=${THIRD_PARTY_BUILD_TYPE} + -DSNAPPY_ROOT=${SNAPPY_INSTALL_DIR} + ${EXTERNAL_OPTIONAL_ARGS} + CMAKE_CACHE_ARGS + -DCMAKE_INSTALL_PREFIX:PATH=${SNAPPYSTREAM_INSTALL_DIR} + -DCMAKE_INSTALL_LIBDIR:PATH=${SNAPPYSTREAM_INSTALL_DIR}/lib + -DCMAKE_BUILD_TYPE:STRING=${THIRD_PARTY_BUILD_TYPE} + DEPENDS snappy + ) +endif(WIN32) add_library(snappystream STATIC IMPORTED GLOBAL) set_property(TARGET snappystream PROPERTY IMPORTED_LOCATION ${SNAPPYSTREAM_LIBRARIES}) diff --git a/cmake/generic.cmake b/cmake/generic.cmake index e21f89c7c585053631391852522d47cd7ffa7638..111627a932afe2642312befb5d9c65e36e4bf1d0 100644 --- a/cmake/generic.cmake +++ b/cmake/generic.cmake @@ -351,6 +351,9 @@ function(cc_test TARGET_NAME) cmake_parse_arguments(cc_test "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN}) add_executable(${TARGET_NAME} ${cc_test_SRCS}) target_link_libraries(${TARGET_NAME} ${cc_test_DEPS} paddle_gtest_main lod_tensor memory gtest gflags glog) + if(WIN32) + target_link_libraries(${TARGET_NAME} shlwapi) + endif(WIN32) add_dependencies(${TARGET_NAME} ${cc_test_DEPS} paddle_gtest_main lod_tensor memory gtest gflags glog) add_test(NAME ${TARGET_NAME} COMMAND ${TARGET_NAME} ${cc_test_ARGS} diff --git a/cmake/operators.cmake b/cmake/operators.cmake index ba9c266d133b637fd99f128bbfe42253a2400aaf..17107e0698757997854e4627d30de60d9a9df11b 100644 --- a/cmake/operators.cmake +++ b/cmake/operators.cmake @@ -84,9 +84,7 @@ function(op_library TARGET) endif() if (WIN32) # remove windows unsupported op, because windows has no nccl, no warpctc such ops. - foreach(windows_unsupport_op "nccl_op" "gen_nccl_id_op" "warpctc_op" "hierarchical_sigmoid_op" - "crf_decoding_op" "select_op" "lstmp_op" "gru_op" "fusion_gru_op" "lstm_op" "fusion_lstm_op" "cumsum_op" - "fusion_seqconv_eltadd_relu_op" "channel_send_op" "channel_create_op" "channel_close_op" "channel_recv_op") + foreach(windows_unsupport_op "nccl_op" "gen_nccl_id_op" "warpctc_op") if ("${TARGET}" STREQUAL "${windows_unsupport_op}") return() endif() diff --git a/cmake/simd.cmake b/cmake/simd.cmake index 566dc75fda019eb66759eb403f60e16f18cffef1..86096d4feaace040da416a01872882456c4098fc 100644 --- a/cmake/simd.cmake +++ b/cmake/simd.cmake @@ -57,43 +57,46 @@ int main() return 0; }" SSE3_FOUND) -# Check AVX -set(CMAKE_REQUIRED_FLAGS ${AVX_FLAG}) -set(AVX_FOUND_EXITCODE 1 CACHE STRING "Result from TRY_RUN" FORCE) -CHECK_CXX_SOURCE_RUNS(" -#include -int main() -{ - __m256 a = _mm256_set_ps (-1.0f, 2.0f, -3.0f, 4.0f, -1.0f, 2.0f, -3.0f, 4.0f); - __m256 b = _mm256_set_ps (1.0f, 2.0f, 3.0f, 4.0f, 1.0f, 2.0f, 3.0f, 4.0f); - __m256 result = _mm256_add_ps (a, b); - return 0; -}" AVX_FOUND) +# disable AVX by default on windows +if(NOT WIN32) + # Check AVX + set(CMAKE_REQUIRED_FLAGS ${AVX_FLAG}) + set(AVX_FOUND_EXITCODE 1 CACHE STRING "Result from TRY_RUN" FORCE) + CHECK_CXX_SOURCE_RUNS(" + #include + int main() + { + __m256 a = _mm256_set_ps (-1.0f, 2.0f, -3.0f, 4.0f, -1.0f, 2.0f, -3.0f, 4.0f); + __m256 b = _mm256_set_ps (1.0f, 2.0f, 3.0f, 4.0f, 1.0f, 2.0f, 3.0f, 4.0f); + __m256 result = _mm256_add_ps (a, b); + return 0; + }" AVX_FOUND) -# Check AVX 2 -set(CMAKE_REQUIRED_FLAGS ${AVX2_FLAG}) -set(AVX2_FOUND_EXITCODE 1 CACHE STRING "Result from TRY_RUN" FORCE) -CHECK_CXX_SOURCE_RUNS(" -#include -int main() -{ - __m256i a = _mm256_set_epi32 (-1, 2, -3, 4, -1, 2, -3, 4); - __m256i result = _mm256_abs_epi32 (a); - return 0; -}" AVX2_FOUND) + # Check AVX 2 + set(CMAKE_REQUIRED_FLAGS ${AVX2_FLAG}) + set(AVX2_FOUND_EXITCODE 1 CACHE STRING "Result from TRY_RUN" FORCE) + CHECK_CXX_SOURCE_RUNS(" + #include + int main() + { + __m256i a = _mm256_set_epi32 (-1, 2, -3, 4, -1, 2, -3, 4); + __m256i result = _mm256_abs_epi32 (a); + return 0; + }" AVX2_FOUND) -# Check AVX512F -set(CMAKE_REQUIRED_FLAGS ${AVX512F_FLAG}) -set(AVX512F_FOUND_EXITCODE 1 CACHE STRING "Result from TRY_RUN" FORCE) -CHECK_CXX_SOURCE_RUNS(" -#include -int main() -{ - __m512i a = _mm512_set_epi32 (-1, 2, -3, 4, -1, 2, -3, 4, - 13, -5, 6, -7, 9, 2, -6, 3); - __m512i result = _mm512_abs_epi32 (a); - return 0; -}" AVX512F_FOUND) + # Check AVX512F + set(CMAKE_REQUIRED_FLAGS ${AVX512F_FLAG}) + set(AVX512F_FOUND_EXITCODE 1 CACHE STRING "Result from TRY_RUN" FORCE) + CHECK_CXX_SOURCE_RUNS(" + #include + int main() + { + __m512i a = _mm512_set_epi32 (-1, 2, -3, 4, -1, 2, -3, 4, + 13, -5, 6, -7, 9, 2, -6, 3); + __m512i result = _mm512_abs_epi32 (a); + return 0; + }" AVX512F_FOUND) +endif(NOT WIN32) set(CMAKE_REQUIRED_FLAGS ${CMAKE_REQUIRED_FLAGS_RETAINED}) mark_as_advanced(MMX_FOUND SSE2_FOUND SSE3_FOUND AVX_FOUND AVX2_FOUND AVX512F_FOUND) diff --git a/paddle/fluid/CMakeLists.txt b/paddle/fluid/CMakeLists.txt index abadda3adb00e1f41e90e07aa5e961134e69ae3d..6b526f0103ad3c530c06a68757cf89293f4fb84b 100644 --- a/paddle/fluid/CMakeLists.txt +++ b/paddle/fluid/CMakeLists.txt @@ -3,13 +3,9 @@ add_subdirectory(platform) add_subdirectory(framework) add_subdirectory(operators) add_subdirectory(string) - -add_subdirectory(pybind) -if (NOT WIN32) add_subdirectory(recordio) -endif(NOT WIN32) +add_subdirectory(pybind) # NOTE: please add subdirectory inference at last. add_subdirectory(inference) - add_subdirectory(train) diff --git a/paddle/fluid/framework/CMakeLists.txt b/paddle/fluid/framework/CMakeLists.txt index cb9057672cc2c29af21b662edc189004bb0a4866..43e1bc6b2efecd24d2d5bade5b6a7727bfb0a607 100644 --- a/paddle/fluid/framework/CMakeLists.txt +++ b/paddle/fluid/framework/CMakeLists.txt @@ -31,9 +31,7 @@ function(windows_symbolic TARGET) endfunction() add_subdirectory(ir) -if (NOT WIN32) add_subdirectory(details) -endif (NOT WIN32) # ddim lib proto_library(framework_proto SRCS framework.proto) @@ -68,11 +66,7 @@ if(WITH_GPU) else() cc_test(mixed_vector_test SRCS mixed_vector_test.cc DEPS place memory device_context tensor) endif() -if (NOT WIN32) - cc_library(lod_tensor SRCS lod_tensor.cc DEPS ddim place tensor framework_proto recordio version) -else() - cc_library(lod_tensor SRCS lod_tensor.cc DEPS ddim place tensor framework_proto version) -endif (NOT WIN32) +cc_library(lod_tensor SRCS lod_tensor.cc DEPS ddim place tensor framework_proto recordio version) cc_test(lod_tensor_test SRCS lod_tensor_test.cc DEPS lod_tensor memory) nv_test(lod_tensor_gpu_test SRCS lod_tensor_test.cu DEPS lod_tensor) @@ -122,13 +116,8 @@ cc_test(op_proto_maker_test SRCS op_proto_maker_test.cc DEPS op_proto_maker) cc_library(op_info SRCS op_info.cc DEPS attribute framework_proto) cc_library(shape_inference SRCS shape_inference.cc DEPS ddim attribute device_context) -if (NOT WIN32) cc_library(operator SRCS operator.cc DEPS op_info device_context tensor scope glog shape_inference data_transform lod_tensor profiler) -else() -cc_library(operator SRCS operator.cc DEPS op_info device_context tensor scope glog - shape_inference data_transform lod_tensor) -endif(NOT WIN32) cc_test(operator_test SRCS operator_test.cc DEPS operator op_registry device_context) @@ -183,12 +172,10 @@ else() cc_test(test_naive_executor SRCS naive_executor_test.cc DEPS naive_executor elementwise_add_op) endif() -if (NOT WIN32) cc_library(parallel_executor SRCS parallel_executor.cc DEPS threaded_ssa_graph_executor scope_buffered_ssa_graph_executor graph build_strategy fast_threaded_ssa_graph_executor) -endif() # NOT WIN32 cc_library(prune SRCS prune.cc DEPS framework_proto) cc_test(prune_test SRCS prune_test.cc DEPS op_info prune recurrent_op device_context) diff --git a/paddle/fluid/framework/details/fast_threaded_ssa_graph_executor.h b/paddle/fluid/framework/details/fast_threaded_ssa_graph_executor.h index 949616f02d5168e6abab932d608e4b20ee64304a..c3a8b85423403992e3a12ceb0a1acbae82d25dfa 100644 --- a/paddle/fluid/framework/details/fast_threaded_ssa_graph_executor.h +++ b/paddle/fluid/framework/details/fast_threaded_ssa_graph_executor.h @@ -13,9 +13,9 @@ // limitations under the License. #pragma once +#include #include #include -#include "ThreadPool.h" #include "paddle/fluid/framework/blocking_queue.h" #include "paddle/fluid/framework/details/exception_holder.h" #include "paddle/fluid/framework/details/execution_strategy.h" diff --git a/paddle/fluid/framework/eigen.h b/paddle/fluid/framework/eigen.h index 2b265a773fe967f5b2ab38ce795b0f599d859c2a..5bafa4345f42a1f6209b5ee31ae6ba2ded6a899c 100644 --- a/paddle/fluid/framework/eigen.h +++ b/paddle/fluid/framework/eigen.h @@ -13,11 +13,6 @@ See the License for the specific language governing permissions and limitations under the License. */ #pragma once -// logging.h and windows.h conflict -#define GLOG_NO_ABBREVIATED_SEVERITIES -// solve static linking error in windows -// https://github.com/google/glog/issues/301 -#define GOOGLE_GLOG_DLL_DECL #include "paddle/fluid/framework/tensor.h" #include "unsupported/Eigen/CXX11/Tensor" diff --git a/paddle/fluid/framework/op_registry.h b/paddle/fluid/framework/op_registry.h index ef2eb334a4e7f3f482ba6d62d3f325f109c69302..0e6e74293c30d5f8caa58fe6bfa63657d2669b46 100644 --- a/paddle/fluid/framework/op_registry.h +++ b/paddle/fluid/framework/op_registry.h @@ -23,11 +23,6 @@ limitations under the License. */ #include #include -#if defined(_WIN32) -#define GLOG_NO_ABBREVIATED_SEVERITIES // msvc conflict logging with windows.h -#define GOOGLE_GLOG_DLL_DECL -#endif - #include "glog/logging.h" // For VLOG() #include "paddle/fluid/framework/attribute.h" #include "paddle/fluid/framework/details/op_registry.h" diff --git a/paddle/fluid/framework/operator.cc b/paddle/fluid/framework/operator.cc index 2b35943d092518c7f45a8ed3b708532666a23353..1ec170b6f65f9c3ee0f80fb8904026b5438c94b2 100644 --- a/paddle/fluid/framework/operator.cc +++ b/paddle/fluid/framework/operator.cc @@ -11,8 +11,6 @@ distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License. */ -#define GLOG_NO_ABBREVIATED_SEVERITIES -#define GOOGLE_GLOG_DLL_DECL #include #include diff --git a/paddle/fluid/framework/operator.h b/paddle/fluid/framework/operator.h index 6918e030bf859bc8a55baed9d944e16217b0efb6..ef838332177c018865a922d570c697b4a94969b6 100644 --- a/paddle/fluid/framework/operator.h +++ b/paddle/fluid/framework/operator.h @@ -20,8 +20,6 @@ limitations under the License. */ #include #include #include -#define GLOG_NO_ABBREVIATED_SEVERITIES -#define GOOGLE_GLOG_DLL_DECL #include "glog/logging.h" // For VLOG #include "paddle/fluid/framework/attribute.h" diff --git a/paddle/fluid/inference/api/api_impl.h b/paddle/fluid/inference/api/api_impl.h index 4e4ab47ca9c5e37f2714ebd48d250c23c7e9b117..9dfa48d501f17fa654ec50049608b1a87c586cb6 100644 --- a/paddle/fluid/inference/api/api_impl.h +++ b/paddle/fluid/inference/api/api_impl.h @@ -14,12 +14,6 @@ limitations under the License. */ #pragma once -// logging.h and windows.h conflict -#define GLOG_NO_ABBREVIATED_SEVERITIES -// solve static linking error in windows -// https://github.com/google/glog/issues/301 -#define GOOGLE_GLOG_DLL_DECL - #include #include #include diff --git a/paddle/fluid/memory/allocation/cpu_allocator.h b/paddle/fluid/memory/allocation/cpu_allocator.h index 9e0044c47ae4ebde9c828e14d3d0e6c0cb1dc8dc..26d3643f4edff1f2d71b1c761e915a6dacb485ad 100644 --- a/paddle/fluid/memory/allocation/cpu_allocator.h +++ b/paddle/fluid/memory/allocation/cpu_allocator.h @@ -15,6 +15,12 @@ #pragma once #include "paddle/fluid/memory/allocation/allocator.h" +#ifdef _WIN32 +#define posix_memalign_free _aligned_free +#define posix_memalign(p, a, s) \ + (((*(p)) = _aligned_malloc((s), (a))), *(p) ? 0 : errno) +#endif + namespace paddle { namespace memory { namespace allocation { diff --git a/paddle/fluid/operators/CMakeLists.txt b/paddle/fluid/operators/CMakeLists.txt index 81c9239486ad00fe702af32c2171906baeeeae28..de4f23515d8591f28b80ad00322365f8cdce768b 100644 --- a/paddle/fluid/operators/CMakeLists.txt +++ b/paddle/fluid/operators/CMakeLists.txt @@ -22,9 +22,7 @@ if(WITH_DISTRIBUTE) add_subdirectory(distributed_ops) endif() -if (NOT WIN32) - add_subdirectory(reader) -endif() +add_subdirectory(reader) if (NOT WIN32) add_subdirectory(nccl) @@ -42,7 +40,7 @@ endif() register_operators(EXCLUDES warpctc_op conv_fusion_op DEPS ${OP_HEADER_DEPS}) # warpctc_op needs cudnn 7 above -if (WITH_GPU) +if (WITH_GPU AND NOT WIN32) if (${CUDNN_MAJOR_VERSION} VERSION_LESS 7) op_library(warpctc_op DEPS dynload_warpctc sequence_padding sequence_scale SRCS warpctc_op.cc warpctc_op.cu.cc) else() @@ -59,10 +57,12 @@ endif() set(COMMON_OP_DEPS ${OP_HEADER_DEPS}) -set(COMMON_OP_DEPS ${COMMON_OP_DEPS} selected_rows_functor selected_rows lod_tensor maxouting unpooling pooling lod_rank_table context_project sequence_pooling executor dynload_warpctc sequence_padding sequence_scale cos_sim_functor memory jit_kernel concat_and_split cross_entropy softmax vol2col im2col sampler) +set(COMMON_OP_DEPS ${COMMON_OP_DEPS} selected_rows_functor selected_rows lod_tensor maxouting unpooling pooling lod_rank_table context_project sequence_pooling executor) if (NOT WIN32) - set(COMMON_OP_DEPS ${COMMON_OP_DEPS} sequence2batch lstm_compute matrix_bit_code gru_compute activation_functions) + set(COMMON_OP_DEPS ${COMMON_OP_DEPS} dynload_warpctc) endif() +set(COMMON_OP_DEPS ${COMMON_OP_DEPS} sequence_padding sequence_scale cos_sim_functor memory jit_kernel concat_and_split cross_entropy softmax vol2col im2col sampler) +set(COMMON_OP_DEPS ${COMMON_OP_DEPS} sequence2batch lstm_compute matrix_bit_code gru_compute activation_functions) if (WITH_GPU) set(COMMON_OP_DEPS ${COMMON_OP_DEPS} depthwise_conv) endif() diff --git a/paddle/fluid/operators/hierarchical_sigmoid_op.h b/paddle/fluid/operators/hierarchical_sigmoid_op.h index 64096a717b12ed231344649f5eb76b7e4b9af4a6..79980cda53befc2bce3cbd79a15da58b39c922ad 100644 --- a/paddle/fluid/operators/hierarchical_sigmoid_op.h +++ b/paddle/fluid/operators/hierarchical_sigmoid_op.h @@ -111,7 +111,7 @@ class HierarchicalSigmoidGradOpKernel : public framework::OpKernel { auto pre_out_mat = EigenMatrix::From(*pre_out); auto pre_out_grad_mat = EigenMatrix::From(pre_out_grad); auto out_grad_mat = EigenMatrix::From(*out_grad); - Eigen::array bcast({{1, static_cast(pre_out_grad.dims()[1])}}); + Eigen::array bcast{1, static_cast(pre_out_grad.dims()[1])}; // softrelu derivative pre_out_grad_mat.device(place) = diff --git a/paddle/fluid/operators/math/CMakeLists.txt b/paddle/fluid/operators/math/CMakeLists.txt index 83ee9f6c51c64c6b000b20d73d41036b8590da5c..63363086adbf12c38ac09949ac20483116ccf4ee 100644 --- a/paddle/fluid/operators/math/CMakeLists.txt +++ b/paddle/fluid/operators/math/CMakeLists.txt @@ -1,6 +1,4 @@ -if (NOT WIN32) - add_subdirectory(detail) -endif(NOT WIN32) +add_subdirectory(detail) function(math_library TARGET) # math_library is a function to create math library. @@ -43,10 +41,8 @@ math_library(depthwise_conv) math_library(im2col) math_library(sampler) -if (NOT WIN32) # windows do not support avx functions yet. - math_library(gru_compute DEPS activation_functions math_function) - math_library(lstm_compute DEPS activation_functions) -endif (NOT WIN32) +math_library(gru_compute DEPS activation_functions math_function) +math_library(lstm_compute DEPS activation_functions) cc_library(blas SRCS blas.cc DEPS cblas framework_proto device_context) math_library(math_function DEPS blas) @@ -58,9 +54,9 @@ math_library(sequence_padding) math_library(sequence_pooling DEPS math_function) math_library(sequence_scale) math_library(softmax DEPS math_function) -if (NOT WIN32) - math_library(matrix_bit_code) -endif (NOT WIN32) + +math_library(matrix_bit_code) + math_library(unpooling) math_library(vol2col) @@ -76,13 +72,12 @@ if(WITH_GPU) endif() cc_test(concat_test SRCS concat_test.cc DEPS concat_and_split) cc_test(cpu_vec_test SRCS cpu_vec_test.cc DEPS blas cpu_info) -if (NOT WIN32) - set(JIT_KERNEL_SRCS jit_kernel.cc jit_kernel_blas.cc jit_kernel_exp.cc jit_kernel_rnn.cc jit_kernel_crf_decode.cc jit_kernel_layer_norm.cc) - set(JIT_KERNEL_DEPS cpu_info cblas gflags enforce) - if(WITH_XBYAK) - list(APPEND JIT_KERNEL_SRCS jit_gen.cc jit_code.cc) - list(APPEND JIT_KERNEL_DEPS xbyak) - endif() - cc_library(jit_kernel SRCS ${JIT_KERNEL_SRCS} DEPS ${JIT_KERNEL_DEPS}) - cc_test(jit_kernel_test SRCS jit_kernel_test.cc DEPS jit_kernel) -endif (NOT WIN32) + +set(JIT_KERNEL_SRCS jit_kernel.cc jit_kernel_blas.cc jit_kernel_exp.cc jit_kernel_rnn.cc jit_kernel_crf_decode.cc jit_kernel_layer_norm.cc) +set(JIT_KERNEL_DEPS cpu_info cblas gflags enforce) +if(WITH_XBYAK) + list(APPEND JIT_KERNEL_SRCS jit_gen.cc jit_code.cc) + list(APPEND JIT_KERNEL_DEPS xbyak) +endif() +cc_library(jit_kernel SRCS ${JIT_KERNEL_SRCS} DEPS ${JIT_KERNEL_DEPS}) +cc_test(jit_kernel_test SRCS jit_kernel_test.cc DEPS jit_kernel) diff --git a/paddle/fluid/operators/math/detail/activation_functions.h b/paddle/fluid/operators/math/detail/activation_functions.h index b127fbe8c8515e7fe57b07ea1d4291675ec4efca..2b3d38d95a18fad9b76e616cdf2cb6c3eb07da3a 100644 --- a/paddle/fluid/operators/math/detail/activation_functions.h +++ b/paddle/fluid/operators/math/detail/activation_functions.h @@ -15,6 +15,7 @@ limitations under the License. */ #pragma once #include #include + #include "paddle/fluid/platform/enforce.h" #include "paddle/fluid/platform/hostdevice.h" diff --git a/paddle/fluid/operators/math/matrix_bit_code.h b/paddle/fluid/operators/math/matrix_bit_code.h index 07854c83584f90db02b416b85a4aa61f5cdc0685..c329b8b6113e847ec1c57e63258a18b6f65d9396 100644 --- a/paddle/fluid/operators/math/matrix_bit_code.h +++ b/paddle/fluid/operators/math/matrix_bit_code.h @@ -67,7 +67,7 @@ inline constexpr size_t FindLastSet(size_t x) { : (std::is_same::value // NOLINT ? (x ? 8 * sizeof(x) - __builtin_clzl(x) : 0) : (x ? 8 * sizeof(x) - __builtin_clzll(x) : 0)); - +} #else // windows don't have built-in clz, ctz function template @@ -92,7 +92,6 @@ inline int clz(const T& value) { inline size_t FindLastSet(size_t x) { return sizeof(size_t) * 8 - clz(x); } #endif // !_WIN32 -} struct SimpleCode { SimpleCode(size_t code, size_t num_classes) : c_(code + num_classes) {} diff --git a/paddle/fluid/operators/reader/create_py_reader_op.cc b/paddle/fluid/operators/reader/create_py_reader_op.cc index 0f31ca1a94326956ae5e6dffd582daedeb55a9e3..901a92ab5b5c74b071be8b57a7653d90e2a4fb29 100644 --- a/paddle/fluid/operators/reader/create_py_reader_op.cc +++ b/paddle/fluid/operators/reader/create_py_reader_op.cc @@ -74,7 +74,7 @@ class CreatePyReaderOpMaker : public FileReaderMakerBase { "Name of the `LoDTensorBlockingQueueHolder` variable"); AddComment(R"DOC( - Create PyReader to support LoDTensor data feeding in Python side. + Create PyReader to support LoDTensor data feeding in Python side. )DOC"); } }; diff --git a/paddle/fluid/operators/roi_align_op.cc b/paddle/fluid/operators/roi_align_op.cc index c57a34c3a745e8fc03ca57dce478ecf60058a9a9..79f189222ef375a1e3f7b8c3e18619a1c4f2a829 100644 --- a/paddle/fluid/operators/roi_align_op.cc +++ b/paddle/fluid/operators/roi_align_op.cc @@ -35,10 +35,10 @@ class ROIAlignOp : public framework::OperatorWithKernel { "The format of input tensor is NCHW."); PADDLE_ENFORCE(rois_dims.size() == 2, "ROIs should be a 2-D LoDTensor of shape (num_rois, 4)" - "given as [[x1, y1, x2, y2], …]."); + "given as [[x1, y1, x2, y2], ...]."); PADDLE_ENFORCE(rois_dims[1] == 4, "ROIs should be a 2-D LoDTensor of shape (num_rois, 4)" - "given as [[x1, y1, x2, y2], …]."); + "given as [[x1, y1, x2, y2], ...]."); int pooled_height = ctx->Attrs().Get("pooled_height"); int pooled_width = ctx->Attrs().Get("pooled_width"); float spatial_scale = ctx->Attrs().Get("spatial_scale"); @@ -103,7 +103,7 @@ class ROIAlignOpMaker : public framework::OpProtoAndCheckerMaker { "(LoDTensor), " "ROIs (Regions of Interest) to pool over. " "should be a 2-D LoDTensor of shape (num_rois, 4)" - "given as [[x1, y1, x2, y2], …]. " + "given as [[x1, y1, x2, y2], ...]. " "(x1, y1) is the top left coordinates, and " "(x2, y2) is the bottom right coordinates."); AddOutput("Out", diff --git a/paddle/fluid/operators/roi_pool_op.cc b/paddle/fluid/operators/roi_pool_op.cc index 043ea680d1506e7b7e33ba5537a71f37feaf81be..3f6b2e46c7014a8c57701099fcc44c8d9e4f08e0 100644 --- a/paddle/fluid/operators/roi_pool_op.cc +++ b/paddle/fluid/operators/roi_pool_op.cc @@ -40,10 +40,10 @@ class ROIPoolOp : public framework::OperatorWithKernel { "The format of input tensor is NCHW."); PADDLE_ENFORCE(rois_dims.size() == 2, "ROIs should be a 2-D LoDTensor of shape (num_rois, 4)" - "given as [[x1, y1, x2, y2], …]."); + "given as [[x1, y1, x2, y2], ...]."); PADDLE_ENFORCE(rois_dims[1] == kROISize, "ROIs should be a 2-D LoDTensor of shape (num_rois, 4)" - "given as [[x1, y1, x2, y2], …]."); + "given as [[x1, y1, x2, y2], ...]."); int pooled_height = ctx->Attrs().Get("pooled_height"); int pooled_width = ctx->Attrs().Get("pooled_width"); @@ -110,7 +110,7 @@ class ROIPoolOpMaker : public framework::OpProtoAndCheckerMaker { "(LoDTensor), " "ROIs (Regions of Interest) to pool over. " "should be a 2-D LoDTensor of shape (num_rois, 4)" - "given as [[x1, y1, x2, y2], …]. " + "given as [[x1, y1, x2, y2], ...]. " "Where batch_id is the id of the data, " "(x1, y1) is the top left coordinates, and " "(x2, y2) is the bottom right coordinates."); diff --git a/paddle/fluid/operators/space_to_depth_op.cc b/paddle/fluid/operators/space_to_depth_op.cc index c047bc78ee315201d25a7294b7dae7d766a6c968..b579244673fa1618c282c4d4fedf2ba6d1726a82 100644 --- a/paddle/fluid/operators/space_to_depth_op.cc +++ b/paddle/fluid/operators/space_to_depth_op.cc @@ -86,7 +86,7 @@ class SpaceToDepthOpMaker : public framework::OpProtoAndCheckerMaker { .GreaterThan(1); AddComment(R"DOC( reorg operator used in Yolo v2. - The equation is: C2 = C1/blocksize * blocksize, W2 = W1 * blocksize + offset % blocksize, H2 = H1 * blocksize + offset / blocksize, + The equation is: C2 = C1/blocksize * blocksize, W2 = W1 * blocksize + offset % blocksize, H2 = H1 * blocksize + offset / blocksize, Reshape Input(X) into the shape according to Attr(blocksize). The data in Input(X) are unchanged. diff --git a/paddle/fluid/platform/CMakeLists.txt b/paddle/fluid/platform/CMakeLists.txt index 0d0613e1a4364e300640b62687c8a045e40b9ca9..93cb5eb2dc0b3480ebd05dcc6b36d8915d057bab 100644 --- a/paddle/fluid/platform/CMakeLists.txt +++ b/paddle/fluid/platform/CMakeLists.txt @@ -1,4 +1,3 @@ -if (NOT WIN32) proto_library(profiler_proto SRCS profiler.proto DEPS framework_proto) py_proto_compile(profiler_py_proto SRCS profiler.proto) @@ -6,11 +5,19 @@ add_custom_target(profiler_py_proto_init ALL COMMAND ${CMAKE_COMMAND} -E touch _ add_dependencies(profiler_py_proto profiler_py_proto_init) +if (NOT WIN32) add_custom_command(TARGET profiler_py_proto POST_BUILD COMMAND ${CMAKE_COMMAND} -E make_directory ${PADDLE_BINARY_DIR}/python/paddle/fluid/proto/profiler COMMAND cp *.py ${PADDLE_BINARY_DIR}/python/paddle/fluid/proto/profiler COMMENT "Copy generated python proto into directory paddle/fluid/proto/profiler." WORKING_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}) +else(NOT WIN32) +string(REPLACE "/" "\\" proto_dstpath "${PADDLE_BINARY_DIR}/python/paddle/fluid/proto/profiler/") +add_custom_command(TARGET profiler_py_proto POST_BUILD + COMMAND ${CMAKE_COMMAND} -E make_directory ${PADDLE_BINARY_DIR}/python/paddle/fluid/proto/profiler + COMMAND copy /Y *.py ${proto_dstpath} + COMMENT "Copy generated python proto into directory paddle/fluid/proto/profiler." + WORKING_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}) endif(NOT WIN32) if(WITH_GPU) @@ -60,12 +67,9 @@ cc_test(init_test SRCS init_test.cc DEPS device_context) nv_test(cudnn_helper_test SRCS cudnn_helper_test.cc DEPS dynload_cuda) nv_test(transform_test SRCS transform_test.cu DEPS memory place device_context) - -if (NOT WIN32) cc_library(device_tracer SRCS device_tracer.cc DEPS boost profiler_proto framework_proto ${GPU_CTX_DEPS}) cc_library(profiler SRCS profiler.cc DEPS device_context device_tracer) cc_test(profiler_test SRCS profiler_test.cc DEPS profiler) -endif(NOT WIN32) nv_test(float16_gpu_test SRCS float16_test.cu DEPS lod_tensor) cc_test(float16_test SRCS float16_test.cc DEPS lod_tensor) diff --git a/paddle/fluid/platform/cpu_helper.cc b/paddle/fluid/platform/cpu_helper.cc index 234a04b5c2eb5ee643e8a4e723b28331cd8e6ee0..f2d691b2931f5a57e70fd4762e9dea5665ed75c2 100644 --- a/paddle/fluid/platform/cpu_helper.cc +++ b/paddle/fluid/platform/cpu_helper.cc @@ -29,6 +29,13 @@ namespace platform { void SetNumThreads(int num_threads) { #ifdef PADDLE_USE_OPENBLAS +// windows has no support for openblas multi-thread +// please refer to: https://github.com/PaddlePaddle/Paddle/issues/7234 +#ifdef _WIN32 + if (num_threads > 1) { + num_threads = 1; + } +#endif int real_num_threads = num_threads > 1 ? num_threads : 1; openblas_set_num_threads(real_num_threads); #elif defined(PADDLE_WITH_MKLML) diff --git a/paddle/fluid/platform/device_tracer.h b/paddle/fluid/platform/device_tracer.h index f59fc40b71699a790978e22fd7e26da8d4d94c5f..eaf047d4744762f69d50bff8d467da8e3b8317cc 100644 --- a/paddle/fluid/platform/device_tracer.h +++ b/paddle/fluid/platform/device_tracer.h @@ -13,17 +13,11 @@ See the License for the specific language governing permissions and limitations under the License. */ #pragma once -#if !defined(_WIN32) -#include -#else -#include -#endif // !_WIN32 - -#include #include // NOLINT #include #include "paddle/fluid/platform/dynload/cupti.h" +#include "paddle/fluid/platform/port.h" #include "paddle/fluid/platform/profiler.pb.h" namespace paddle { @@ -32,15 +26,11 @@ namespace platform { /////////////////////// // WARN: Under Development. Don't depend on it yet. ////////////////////// -#if !defined(_WIN32) inline uint64_t PosixInNsec() { struct timeval tv; gettimeofday(&tv, nullptr); return 1000 * (static_cast(tv.tv_sec) * 1000000 + tv.tv_usec); } -#else -inline uint64_t PosixInNsec() { return static_cast(0); } -#endif // !_WIN32 // DeviceTracer performs the following tasks: // 1. Register cuda callbacks for various events: kernel, memcpy, etc. diff --git a/paddle/fluid/platform/dynload/cudnn.h b/paddle/fluid/platform/dynload/cudnn.h index 065b940b9ca6fb7522790d2145d1a93469169461..1a83ac7780a01fd3c20bc85baaf14e6ca3f8eb8c 100644 --- a/paddle/fluid/platform/dynload/cudnn.h +++ b/paddle/fluid/platform/dynload/cudnn.h @@ -13,8 +13,6 @@ See the License for the specific language governing permissions and limitations under the License. */ #pragma once -#define GLOG_NO_ABBREVIATED_SEVERITIES -#define GOOGLE_GLOG_DLL_DECL #include #include diff --git a/paddle/fluid/platform/enforce.h b/paddle/fluid/platform/enforce.h index a251bfcd9914422cb6300adbbcdef3dfa79f441c..a85972bdb72ca3119cc14f9e2b810c3875443538 100644 --- a/paddle/fluid/platform/enforce.h +++ b/paddle/fluid/platform/enforce.h @@ -18,12 +18,6 @@ limitations under the License. */ #include // for __cxa_demangle #endif // __GNUC__ -#if defined(_WIN32) -#define NOMINMAX // msvc max/min macro conflict with std::min/max -#define GLOG_NO_ABBREVIATED_SEVERITIES // msvc conflict logging with windows.h -#define GOOGLE_GLOG_DLL_DECL -#endif - #ifdef PADDLE_WITH_CUDA #include #include @@ -127,14 +121,14 @@ struct EOFException : public std::exception { #define UNLIKELY(condition) __builtin_expect(static_cast(condition), 0) #else // there is no equivalent intrinsics in msvc. -#define UNLIKELY(condition) (condition == 0) +#define UNLIKELY(condition) (condition) #endif #if !defined(_WIN32) #define LIKELY(condition) __builtin_expect(static_cast(condition), 1) #else // there is no equivalent intrinsics in msvc. -#define LIKELY(condition) (condition != 0) +#define LIKELY(condition) (condition) #endif template @@ -248,7 +242,6 @@ inline void throw_on_error(T e) { throw_on_error(e, ""); } -#if !defined(_WIN32) #define PADDLE_THROW(...) \ do { \ throw ::paddle::platform::EnforceNotMet( \ @@ -272,17 +265,6 @@ inline void throw_on_error(T e) { #define PADDLE_ENFORCE(...) ::paddle::platform::throw_on_error(__VA_ARGS__); #endif // REPLACE_ENFORCE_GLOG -#else // !_WIN32 -// disable enforce, caused by the varardic macro exception error -#define PADDLE_THROW(x) \ - do { \ - throw std::make_exception_ptr( \ - std::runtime_error("Windows disable the enforce.")); \ - } while (false) - -#define PADDLE_ENFORCE(x, ...) x -#endif // !_WIN32 - #define PADDLE_THROW_EOF() \ do { \ throw ::paddle::platform::EOFException("There is no next data.", __FILE__, \ @@ -302,20 +284,6 @@ inline void throw_on_error(T e) { * extra messages is also supported, for example: * PADDLE_ENFORCE(a, b, "some simple enforce failed between %d numbers", 2) */ -#if !defined(_WIN32) -#define PADDLE_ENFORCE_EQ(__VAL0, __VAL1, ...) \ - __PADDLE_BINARY_COMPARE(__VAL0, __VAL1, ==, !=, __VA_ARGS__) -#define PADDLE_ENFORCE_NE(__VAL0, __VAL1, ...) \ - __PADDLE_BINARY_COMPARE(__VAL0, __VAL1, !=, ==, __VA_ARGS__) -#define PADDLE_ENFORCE_GT(__VAL0, __VAL1, ...) \ - __PADDLE_BINARY_COMPARE(__VAL0, __VAL1, >, <=, __VA_ARGS__) -#define PADDLE_ENFORCE_GE(__VAL0, __VAL1, ...) \ - __PADDLE_BINARY_COMPARE(__VAL0, __VAL1, >=, <, __VA_ARGS__) -#define PADDLE_ENFORCE_LT(__VAL0, __VAL1, ...) \ - __PADDLE_BINARY_COMPARE(__VAL0, __VAL1, <, >=, __VA_ARGS__) -#define PADDLE_ENFORCE_LE(__VAL0, __VAL1, ...) \ - __PADDLE_BINARY_COMPARE(__VAL0, __VAL1, <=, >, __VA_ARGS__) - #define PADDLE_ENFORCE_NOT_NULL(__VAL, ...) \ do { \ if (UNLIKELY(nullptr == (__VAL))) { \ @@ -335,27 +303,19 @@ inline void throw_on_error(T e) { paddle::string::Sprintf("" __VA_ARGS__)); \ } \ } while (0) -#else -#define PADDLE_ENFORCE_EQ(__VAL0, __VAL1, ...) ((__VAL0) == (__VAL1)) -#define PADDLE_ENFORCE_NE(__VAL0, __VAL1, ...) ((__VAL0) != (__VAL1)) -#define PADDLE_ENFORCE_GT(__VAL0, __VAL1, ...) ((__VAL0) > (__VAL1)) -#define PADDLE_ENFORCE_GE(__VAL0, __VAL1, ...) ((__VAL0) >= (__VAL1)) -#define PADDLE_ENFORCE_LT(__VAL0, __VAL1, ...) ((__VAL0) < (__VAL1)) -#define PADDLE_ENFORCE_LE(__VAL0, __VAL1, ...) ((__VAL0) <= (__VAL1)) - -#define __PADDLE_BINARY_COMPARE(__VAL0, __VAL1, __CMP, __INV_CMP, ...) \ - do { \ - if (!((__VAL0)__CMP(__VAL1))) { \ - PADDLE_THROW("Windows disable the enforce. Enforce failed."); \ - } \ - } while (0) -#define PADDLE_ENFORCE_NOT_NULL(__VAL1, ...) \ - do { \ - if (nullptr == (__VAL1)) { \ - PADDLE_THROW("Windows disable the enforce. Enforce failed"); \ - } \ - } while (0) -#endif // !_WIN32 + +#define PADDLE_ENFORCE_EQ(__VAL0, __VAL1, ...) \ + __PADDLE_BINARY_COMPARE(__VAL0, __VAL1, ==, !=, __VA_ARGS__) +#define PADDLE_ENFORCE_NE(__VAL0, __VAL1, ...) \ + __PADDLE_BINARY_COMPARE(__VAL0, __VAL1, !=, ==, __VA_ARGS__) +#define PADDLE_ENFORCE_GT(__VAL0, __VAL1, ...) \ + __PADDLE_BINARY_COMPARE(__VAL0, __VAL1, >, <=, __VA_ARGS__) +#define PADDLE_ENFORCE_GE(__VAL0, __VAL1, ...) \ + __PADDLE_BINARY_COMPARE(__VAL0, __VAL1, >=, <, __VA_ARGS__) +#define PADDLE_ENFORCE_LT(__VAL0, __VAL1, ...) \ + __PADDLE_BINARY_COMPARE(__VAL0, __VAL1, <, >=, __VA_ARGS__) +#define PADDLE_ENFORCE_LE(__VAL0, __VAL1, ...) \ + __PADDLE_BINARY_COMPARE(__VAL0, __VAL1, <=, >, __VA_ARGS__) } // namespace platform } // namespace paddle diff --git a/paddle/fluid/platform/init.cc b/paddle/fluid/platform/init.cc index e07e9d3825243017159698c1959e626ef3e66dd7..0ccef6c6a8345e31cee3ef2422fe3f56c059c231 100644 --- a/paddle/fluid/platform/init.cc +++ b/paddle/fluid/platform/init.cc @@ -117,13 +117,6 @@ void InitDevices(bool init_p2p, const std::vector devices) { places.emplace_back(platform::CPUPlace()); platform::DeviceContextPool::Init(places); -// windows has no support for openblas multi-thread -#ifdef _WIN32 - if (FLAGS_paddle_num_threads > 1) { - FLAGS_paddle_num_threads = 1; - } -#endif - #ifndef PADDLE_WITH_MKLDNN platform::SetNumThreads(FLAGS_paddle_num_threads); #endif diff --git a/paddle/fluid/platform/init.h b/paddle/fluid/platform/init.h index 992ca5e6f6a966a331616a698e3bebd2eee129d5..0e30594672927253cc8083dcb88bb867d63ec729 100644 --- a/paddle/fluid/platform/init.h +++ b/paddle/fluid/platform/init.h @@ -16,9 +16,6 @@ limitations under the License. */ #include #include -#define GLOG_NO_ABBREVIATED_SEVERITIES -#define GOOGLE_GLOG_DLL_DECL - #include "gflags/gflags.h" #include "glog/logging.h" diff --git a/paddle/fluid/platform/port.h b/paddle/fluid/platform/port.h index 8823e97b0b696556b32724acd096e8fc79a49f53..ad070171df32fd436f24613561d9bc384f79195a 100644 --- a/paddle/fluid/platform/port.h +++ b/paddle/fluid/platform/port.h @@ -17,6 +17,7 @@ #include #include +#include #include #include @@ -27,8 +28,13 @@ #include // dladdr #include // backtrace #include +#include #include // std::accumulate #else +#define NOMINMAX // msvc max/min macro conflict with std::min/max +// solve static linking error in windows +// https://github.com/google/glog/issues/301 +#define GOOGLE_GLOG_DLL_DECL #include // _popen, _pclose #include #include @@ -57,6 +63,25 @@ static void *dlopen(const char *filename, int flag) { return reinterpret_cast(hModule); } +static int gettimeofday(struct timeval *tp, void *tzp) { + time_t clock; + struct tm tm; + SYSTEMTIME wtm; + + GetLocalTime(&wtm); + tm.tm_year = wtm.wYear - 1900; + tm.tm_mon = wtm.wMonth - 1; + tm.tm_mday = wtm.wDay; + tm.tm_hour = wtm.wHour; + tm.tm_min = wtm.wMinute; + tm.tm_sec = wtm.wSecond; + tm.tm_isdst = -1; + clock = mktime(&tm); + tp->tv_sec = clock; + tp->tv_usec = wtm.wMilliseconds * 1000; + + return (0); +} #endif // !_WIN32 static void ExecShellCommand(const std::string &cmd, std::string *message) { @@ -132,10 +157,12 @@ static void MkDir(const char *path) { } } #else - CreateDirectory(path, NULL); - auto errorno = GetLastError(); - if (errorno != ERROR_ALREADY_EXISTS) { - throw std::runtime_error(path_error); + BOOL return_value = CreateDirectory(path, NULL); + if (!return_value) { + auto errorno = GetLastError(); + if (errorno != ERROR_ALREADY_EXISTS) { + throw std::runtime_error(path_error); + } } #endif // !_WIN32 } diff --git a/paddle/fluid/platform/profiler.cc b/paddle/fluid/platform/profiler.cc index 56bf9e31a35fdec5b7f04849068ff96ac9776c0e..998242fb4a09138db24aa75759f4990ffdc4d4e2 100644 --- a/paddle/fluid/platform/profiler.cc +++ b/paddle/fluid/platform/profiler.cc @@ -13,8 +13,8 @@ See the License for the specific language governing permissions and limitations under the License. */ #include "paddle/fluid/platform/profiler.h" +#include "paddle/fluid/platform/port.h" -#include #include #include #include diff --git a/paddle/fluid/platform/profiler.h b/paddle/fluid/platform/profiler.h index e8eae874afa3d17f0d3374eef457cdbacb3f8424..f5d3490634f3199a23986ec3ae13d9fe3577ac35 100644 --- a/paddle/fluid/platform/profiler.h +++ b/paddle/fluid/platform/profiler.h @@ -69,7 +69,6 @@ void PushEvent(const std::string& name, const DeviceContext* dev_ctx); void PopEvent(const std::string& name, const DeviceContext* dev_ctx); -#if !defined(_WIN32) struct RecordEvent { // dev_ctx can be set to nullptr if device is cpu. RecordEvent(const std::string& name, const DeviceContext* dev_ctx); @@ -106,15 +105,6 @@ struct RecordBlock { std::string name_; uint64_t start_ns_; }; -#else -// windows do not support profiler temporarily. -struct RecordEvent { - RecordEvent(const std::string& name, const DeviceContext* dev_ctx) {} -}; -struct RecordBlock { - explicit RecordBlock(int block_id) {} -}; -#endif // Return the event list of all threads. Assumed the returned value calls // event_lists, event_lists[i][j] represents the j-th Event of i-th thread. diff --git a/paddle/fluid/platform/stream_callback_manager.h b/paddle/fluid/platform/stream_callback_manager.h index 0e88a439cf6ca83e3d98725f58875adeeea86be0..11c68f3449ee26b64c121acd081479b37c94fac4 100644 --- a/paddle/fluid/platform/stream_callback_manager.h +++ b/paddle/fluid/platform/stream_callback_manager.h @@ -45,16 +45,15 @@ class StreamCallbackManager { inline void AddCallback(Callback &&callback) const { auto *stream_callback_context = new StreamCallbackContext(this, std::forward(callback)); - PADDLE_ENFORCE( #if CUDA_VERSION >= 10000 - cudaLaunchHostFunc(stream_, StreamCallbackManager::StreamCallbackFunc, - stream_callback_context) + PADDLE_ENFORCE(cudaLaunchHostFunc(stream_, + StreamCallbackManager::StreamCallbackFunc, + stream_callback_context)); // NOLINT #else - cudaStreamAddCallback(stream_, - StreamCallbackManager::StreamCallbackFunc, - stream_callback_context, 0) + PADDLE_ENFORCE(cudaStreamAddCallback( + stream_, StreamCallbackManager::StreamCallbackFunc, + stream_callback_context, 0)); // NOLINT #endif - ); // NOLINT } void Wait() const { thread_pool_.reset(new ThreadPool(1)); } diff --git a/paddle/fluid/pybind/CMakeLists.txt b/paddle/fluid/pybind/CMakeLists.txt index 6417da077e63dd78857d29ddd3484c646849daf4..fb6ee2f4a53925f64b61b9fe03f5a4bd7203ed53 100644 --- a/paddle/fluid/pybind/CMakeLists.txt +++ b/paddle/fluid/pybind/CMakeLists.txt @@ -1,10 +1,6 @@ -set(PYBIND_DEPS pybind python proto_desc memory executor prune feed_fetch_method pass_builder) -set(PYBIND_SRCS pybind.cc exception.cc protobuf.cc const_value.cc) -if(NOT WIN32) - list(APPEND PYBIND_DEPS parallel_executor profiler) - list(APPEND PYBIND_SRCS recordio.cc) -endif(NOT WIN32) +set(PYBIND_DEPS pybind python proto_desc memory executor prune feed_fetch_method pass_builder parallel_executor profiler) +set(PYBIND_SRCS pybind.cc exception.cc protobuf.cc const_value.cc recordio.cc) if(WITH_PYTHON) if(WITH_AMD_GPU) hip_library(paddle_pybind SHARED diff --git a/paddle/fluid/pybind/pybind.cc b/paddle/fluid/pybind/pybind.cc index 5ef5bf4d6c99d8aa0ebc9bc26bbb93d8f3f369fc..6cc3a1739a5cc4d4e87679ec4e345edefba47a82 100644 --- a/paddle/fluid/pybind/pybind.cc +++ b/paddle/fluid/pybind/pybind.cc @@ -21,13 +21,6 @@ limitations under the License. */ #include #include -#if defined(_WIN32) -#define NOMINMAX -#define GLOG_NO_ABBREVIATED_SEVERITIES // msvc conflict logging with windows.h -#define GOOGLE_GLOG_DLL_DECL -#include -#endif - #include "paddle/fluid/framework/executor.h" #include "paddle/fluid/framework/feed_fetch_method.h" #include "paddle/fluid/framework/framework.pb.h" @@ -36,9 +29,7 @@ limitations under the License. */ #include "paddle/fluid/framework/lod_tensor.h" #include "paddle/fluid/framework/lod_tensor_array.h" #include "paddle/fluid/framework/op_registry.h" -#ifndef _WIN32 #include "paddle/fluid/framework/parallel_executor.h" -#endif #include "paddle/fluid/framework/prune.h" #include "paddle/fluid/framework/reader.h" #include "paddle/fluid/framework/selected_rows.h" @@ -359,22 +350,16 @@ All parameter, weight, gradient are variables in Paddle. return self.GetMutable(); }, py::return_value_policy::reference) - #endif -#ifndef _WIN32 .def("get_reader", [](Variable &self) -> framework::ReaderHolder * { PADDLE_ENFORCE(self.IsType()); return self.GetMutable(); }, - py::return_value_policy::reference) -#endif - ; // NOLINT + py::return_value_policy::reference); -#if !defined(_WIN32) py::class_(m, "Reader", "") .def("reset", &framework::ReaderHolder::ResetAll); -#endif using LoDTensorBlockingQueue = ::paddle::operators::reader::LoDTensorBlockingQueue; @@ -643,7 +628,6 @@ All parameter, weight, gradient are variables in Paddle. #endif #endif -#ifndef _WIN32 py::enum_(m, "ProfilerState", py::arithmetic()) .value("kDisabled", platform::ProfilerState::kDisabled) .value("kCPU", platform::ProfilerState::kCPU) @@ -664,7 +648,6 @@ All parameter, weight, gradient are variables in Paddle. m.def("disable_profiler", platform::DisableProfiler); m.def("is_profiler_enabled", platform::IsProfileEnabled); m.def("reset_profiler", platform::ResetProfiler); -#endif py::class_> pass(m, "Pass"); pass.def(py::init()) @@ -693,7 +676,6 @@ All parameter, weight, gradient are variables in Paddle. .def("remove_pass", [](ir::PassBuilder &self, size_t idx) { self.RemovePass(idx); }); -#ifndef _WIN32 // -- python binds for parallel executor. py::class_ pe(m, "ParallelExecutor"); py::class_ exec_strategy(pe, "ExecutionStrategy", R"DOC( @@ -921,7 +903,6 @@ All parameter, weight, gradient are variables in Paddle. }); BindRecordIOWriter(&m); -#endif return m.ptr(); } } // namespace pybind diff --git a/python/paddle/fluid/__init__.py b/python/paddle/fluid/__init__.py index f2f49f813a1840897024d851f2810786a153fb49..543acf2d349c7d02480204699f497536c7a4ca60 100644 --- a/python/paddle/fluid/__init__.py +++ b/python/paddle/fluid/__init__.py @@ -115,9 +115,8 @@ def __bootstrap__(): 'use_pinned_memory', 'check_nan_inf', 'benchmark', 'eager_delete_scope', 'use_mkldnn', 'use_ngraph', 'initial_cpu_memory_in_mb', 'init_allocated_mem', 'free_idle_memory', 'paddle_num_threads', - "dist_threadpool_size", 'cpu_deterministic', 'eager_delete_tensor_gb', - 'allocator_strategy', 'reader_queue_speed_test_mode', - 'print_sub_graph_dir' + "dist_threadpool_size", 'eager_delete_tensor_gb', 'allocator_strategy', + 'reader_queue_speed_test_mode', 'print_sub_graph_dir' ] if os.name != 'nt': read_env_flags.append('warpctc_dir') diff --git a/python/paddle/fluid/contrib/inferencer.py b/python/paddle/fluid/contrib/inferencer.py index b966ae01d039d7e9510dae73ecadb97b494f68c2..b8d5f4ffeadca0a7b103682f175d50dc46fa258a 100644 --- a/python/paddle/fluid/contrib/inferencer.py +++ b/python/paddle/fluid/contrib/inferencer.py @@ -15,15 +15,13 @@ from __future__ import print_function import contextlib -import os from .. import core from .. import executor from .. import framework from .. import io -if os.name != 'nt': - from .. import parallel_executor +from .. import parallel_executor from .. import unique_name from .trainer import check_and_get_place diff --git a/python/paddle/fluid/contrib/trainer.py b/python/paddle/fluid/contrib/trainer.py index 096821a5ba690074ecbf023cf87fed7e206d023f..8569e486f91786b5562e84dcdccf6d91da0612cc 100644 --- a/python/paddle/fluid/contrib/trainer.py +++ b/python/paddle/fluid/contrib/trainer.py @@ -28,8 +28,7 @@ from .. import framework from .. import io # optimizer is same as the parameter of Trainer.__init__. Rename it to opt_module from .. import optimizer as opt_module -if os.name != 'nt': - from .. import parallel_executor +from .. import parallel_executor from ..transpiler import distribute_transpiler __all__ = [ diff --git a/python/paddle/fluid/layers/io.py b/python/paddle/fluid/layers/io.py index a9075045a2d5282ecded1681bc9835feb15298ea..3f47053961bcc41b82f1b6776e9365166e78ddbf 100644 --- a/python/paddle/fluid/layers/io.py +++ b/python/paddle/fluid/layers/io.py @@ -347,72 +347,70 @@ def _copy_reader_create_op_(block, op): return new_op -if os.name != 'nt': - - @templatedoc(op_type='create_recordio_file_reader') - def open_recordio_file(filename, - shapes, - lod_levels, - dtypes, - pass_num=1, - for_parallel=True): - """ - ${comment} - - Args: - filename(${filename_type}): ${filename_comment}. - shapes(list): List of tuples which declaring data shapes. - lod_levels(${lod_levels_type}): ${lod_levels_comment}. - dtypes(list): List of strs which declaring data type. - pass_num(int): Number of passes to run. - for_parallel(Bool): Set it as True if you are going to run - subsequent operators in parallel. - - Returns: - ${out_comment}. - - Examples: - - >>> import paddle.fluid as fluid - >>> reader = fluid.layers.io.open_recordio_file( - >>> filename='./data.recordio', - >>> shapes=[(3,224,224), (1)], - >>> lod_levels=[0, 0], - >>> dtypes=['float32', 'int64']) - >>> # Via the reader, we can use 'read_file' layer to get data: - >>> image, label = fluid.layers.io.read_file(reader) - """ - dtypes = [convert_np_dtype_to_dtype_(dt) for dt in dtypes] - shape_concat = [] - ranks = [] +@templatedoc(op_type='create_recordio_file_reader') +def open_recordio_file(filename, + shapes, + lod_levels, + dtypes, + pass_num=1, + for_parallel=True): + """ + ${comment} - for shape in shapes: - shape_concat.extend(shape) - ranks.append(len(shape)) + Args: + filename(${filename_type}): ${filename_comment}. + shapes(list): List of tuples which declaring data shapes. + lod_levels(${lod_levels_type}): ${lod_levels_comment}. + dtypes(list): List of strs which declaring data type. + pass_num(int): Number of passes to run. + for_parallel(Bool): Set it as True if you are going to run + subsequent operators in parallel. - var_name = unique_name('open_recordio_file') + Returns: + ${out_comment}. - startup_blk = default_startup_program().current_block() - startup_var = startup_blk.create_var(name=var_name) - startup_blk.append_op( - type='create_recordio_file_reader', - outputs={'Out': [startup_var]}, - attrs={ - 'shape_concat': shape_concat, - 'lod_levels': lod_levels, - 'filename': filename, - 'ranks': ranks - }) + Examples: - startup_var.desc.set_dtypes(dtypes) - startup_var.persistable = True - main_prog_var = _copy_reader_var_( - default_main_program().current_block(), startup_var) + >>> import paddle.fluid as fluid + >>> reader = fluid.layers.io.open_recordio_file( + >>> filename='./data.recordio', + >>> shapes=[(3,224,224), (1)], + >>> lod_levels=[0, 0], + >>> dtypes=['float32', 'int64']) + >>> # Via the reader, we can use 'read_file' layer to get data: + >>> image, label = fluid.layers.io.read_file(reader) + """ + dtypes = [convert_np_dtype_to_dtype_(dt) for dt in dtypes] + shape_concat = [] + ranks = [] - if pass_num > 1: - main_prog_var = multi_pass(reader=main_prog_var, pass_num=pass_num) + for shape in shapes: + shape_concat.extend(shape) + ranks.append(len(shape)) + + var_name = unique_name('open_recordio_file') + + startup_blk = default_startup_program().current_block() + startup_var = startup_blk.create_var(name=var_name) + startup_blk.append_op( + type='create_recordio_file_reader', + outputs={'Out': [startup_var]}, + attrs={ + 'shape_concat': shape_concat, + 'lod_levels': lod_levels, + 'filename': filename, + 'ranks': ranks + }) - return monkey_patch_reader_methods(main_prog_var) + startup_var.desc.set_dtypes(dtypes) + startup_var.persistable = True + main_prog_var = _copy_reader_var_(default_main_program().current_block(), + startup_var) + + if pass_num > 1: + main_prog_var = multi_pass(reader=main_prog_var, pass_num=pass_num) + + return monkey_patch_reader_methods(main_prog_var) def random_data_generator(low, high, shapes, lod_levels, for_parallel=True): diff --git a/python/paddle/fluid/layers/nn.py b/python/paddle/fluid/layers/nn.py index 7b0a3e2c82b55f7fc646f970c2df6f66b696a865..e0cc09a4c76caee7496f935682ea0d1788bc4bcb 100644 --- a/python/paddle/fluid/layers/nn.py +++ b/python/paddle/fluid/layers/nn.py @@ -343,128 +343,126 @@ def embedding(input, return tmp -if os.name != 'nt': +@templatedoc(op_type="lstm") +def dynamic_lstm(input, + size, + h_0=None, + c_0=None, + param_attr=None, + bias_attr=None, + use_peepholes=True, + is_reverse=False, + gate_activation='sigmoid', + cell_activation='tanh', + candidate_activation='tanh', + dtype='float32', + name=None): + """ + ${comment} - @templatedoc(op_type="lstm") - def dynamic_lstm(input, - size, - h_0=None, - c_0=None, - param_attr=None, - bias_attr=None, - use_peepholes=True, - is_reverse=False, - gate_activation='sigmoid', - cell_activation='tanh', - candidate_activation='tanh', - dtype='float32', - name=None): - """ - ${comment} - - Args: - input (Variable): ${input_comment} - size (int): 4 * hidden size. - h_0(Variable): The initial hidden state is an optional input, default is zero. - This is a tensor with shape (N x D), where N is the - batch size and D is the hidden size. - c_0(Variable): The initial cell state is an optional input, default is zero. - This is a tensor with shape (N x D), where N is the - batch size. `h_0` and `c_0` can be NULL but only at the same time. - param_attr(ParamAttr|None): The parameter attribute for the learnable - hidden-hidden weights. - - - Weights = {:math:`W_{ch}, W_{ih}, \ - W_{fh}, W_{oh}`} - - The shape is (D x 4D), where D is the hidden - size. - - If it is set to None or one attribute of ParamAttr, - dynamic_lstm will create ParamAttr as param_attr. - If the Initializer of the param_attr is not set, the - parameter is initialized with Xavier. Default: None. - bias_attr (ParamAttr|None): The bias attribute for the learnable bias - weights, which contains two parts, input-hidden - bias weights and peephole connections weights if - setting `use_peepholes` to `True`. - - 1. `use_peepholes = False` - - Biases = {:math:`b_c, b_i, b_f, b_o`}. - - The shape is (1 x 4D). - 2. `use_peepholes = True` - - Biases = { :math:`b_c, b_i, b_f, b_o, W_{ic}, \ - W_{fc}, W_{oc}`}. - - The shape is (1 x 7D). - - If it is set to None or one attribute of ParamAttr, - dynamic_lstm will create ParamAttr as bias_attr. - If the Initializer of the bias_attr is not set, - the bias is initialized zero. Default: None. - use_peepholes (bool): ${use_peepholes_comment} - is_reverse (bool): ${is_reverse_comment} - gate_activation (str): ${gate_activation_comment} - cell_activation (str): ${cell_activation_comment} - candidate_activation (str): ${candidate_activation_comment} - dtype (str): Data type. Choices = ["float32", "float64"], default "float32". - name (str|None): A name for this layer(optional). If set None, the layer - will be named automatically. - - Returns: - tuple: The hidden state, and cell state of LSTM. The shape of both \ - is (T x D), and lod is the same with the `input`. - - Examples: - .. code-block:: python - - hidden_dim = 512 - forward_proj = fluid.layers.fc(input=input_seq, size=hidden_dim * 4, - bias_attr=False) - forward, _ = fluid.layers.dynamic_lstm( - input=forward_proj, size=hidden_dim * 4, use_peepholes=False) - """ - assert bias_attr is not False, "bias_attr should not be False in dynamic_lstmp." - helper = LayerHelper('lstm', **locals()) - size = size // 4 - weight = helper.create_parameter( - attr=helper.param_attr, shape=[size, 4 * size], dtype=dtype) - bias_size = [1, 7 * size] - if not use_peepholes: - bias_size[1] = 4 * size - bias = helper.create_parameter( - attr=helper.bias_attr, shape=bias_size, dtype=dtype, is_bias=True) + Args: + input (Variable): ${input_comment} + size (int): 4 * hidden size. + h_0(Variable): The initial hidden state is an optional input, default is zero. + This is a tensor with shape (N x D), where N is the + batch size and D is the hidden size. + c_0(Variable): The initial cell state is an optional input, default is zero. + This is a tensor with shape (N x D), where N is the + batch size. `h_0` and `c_0` can be NULL but only at the same time. + param_attr(ParamAttr|None): The parameter attribute for the learnable + hidden-hidden weights. - hidden = helper.create_variable_for_type_inference(dtype) - cell = helper.create_variable_for_type_inference(dtype) - batch_gate = helper.create_variable_for_type_inference(dtype) - batch_cell_pre_act = helper.create_variable_for_type_inference(dtype) - inputs = {'Input': input, 'Weight': weight, 'Bias': bias} - batch_size = input.shape[0] - if h_0: - assert h_0.shape == (batch_size, size), \ - 'The shape of h0 should be (batch_size, %d)' % size - inputs['H0'] = h_0 - if c_0: - assert c_0.shape == (batch_size, size), \ - 'The shape of c0 should be (batch_size, %d)' % size - inputs['C0'] = c_0 + - Weights = {:math:`W_{ch}, W_{ih}, \ + W_{fh}, W_{oh}`} + - The shape is (D x 4D), where D is the hidden + size. - helper.append_op( - type='lstm', - inputs=inputs, - outputs={ - 'Hidden': hidden, - 'Cell': cell, - 'BatchGate': batch_gate, - 'BatchCellPreAct': batch_cell_pre_act - }, - attrs={ - 'use_peepholes': use_peepholes, - 'is_reverse': is_reverse, - 'gate_activation': gate_activation, - 'cell_activation': cell_activation, - 'candidate_activation': candidate_activation - }) - return hidden, cell + If it is set to None or one attribute of ParamAttr, + dynamic_lstm will create ParamAttr as param_attr. + If the Initializer of the param_attr is not set, the + parameter is initialized with Xavier. Default: None. + bias_attr (ParamAttr|None): The bias attribute for the learnable bias + weights, which contains two parts, input-hidden + bias weights and peephole connections weights if + setting `use_peepholes` to `True`. + + 1. `use_peepholes = False` + - Biases = {:math:`b_c, b_i, b_f, b_o`}. + - The shape is (1 x 4D). + 2. `use_peepholes = True` + - Biases = { :math:`b_c, b_i, b_f, b_o, W_{ic}, \ + W_{fc}, W_{oc}`}. + - The shape is (1 x 7D). + + If it is set to None or one attribute of ParamAttr, + dynamic_lstm will create ParamAttr as bias_attr. + If the Initializer of the bias_attr is not set, + the bias is initialized zero. Default: None. + use_peepholes (bool): ${use_peepholes_comment} + is_reverse (bool): ${is_reverse_comment} + gate_activation (str): ${gate_activation_comment} + cell_activation (str): ${cell_activation_comment} + candidate_activation (str): ${candidate_activation_comment} + dtype (str): Data type. Choices = ["float32", "float64"], default "float32". + name (str|None): A name for this layer(optional). If set None, the layer + will be named automatically. + + Returns: + tuple: The hidden state, and cell state of LSTM. The shape of both \ + is (T x D), and lod is the same with the `input`. + + Examples: + .. code-block:: python + + hidden_dim = 512 + forward_proj = fluid.layers.fc(input=input_seq, size=hidden_dim * 4, + bias_attr=False) + forward, _ = fluid.layers.dynamic_lstm( + input=forward_proj, size=hidden_dim * 4, use_peepholes=False) + """ + assert bias_attr is not False, "bias_attr should not be False in dynamic_lstmp." + helper = LayerHelper('lstm', **locals()) + size = size // 4 + weight = helper.create_parameter( + attr=helper.param_attr, shape=[size, 4 * size], dtype=dtype) + bias_size = [1, 7 * size] + if not use_peepholes: + bias_size[1] = 4 * size + bias = helper.create_parameter( + attr=helper.bias_attr, shape=bias_size, dtype=dtype, is_bias=True) + + hidden = helper.create_variable_for_type_inference(dtype) + cell = helper.create_variable_for_type_inference(dtype) + batch_gate = helper.create_variable_for_type_inference(dtype) + batch_cell_pre_act = helper.create_variable_for_type_inference(dtype) + inputs = {'Input': input, 'Weight': weight, 'Bias': bias} + batch_size = input.shape[0] + if h_0: + assert h_0.shape == (batch_size, size), \ + 'The shape of h0 should be (batch_size, %d)' % size + inputs['H0'] = h_0 + if c_0: + assert c_0.shape == (batch_size, size), \ + 'The shape of c0 should be (batch_size, %d)' % size + inputs['C0'] = c_0 + + helper.append_op( + type='lstm', + inputs=inputs, + outputs={ + 'Hidden': hidden, + 'Cell': cell, + 'BatchGate': batch_gate, + 'BatchCellPreAct': batch_cell_pre_act + }, + attrs={ + 'use_peepholes': use_peepholes, + 'is_reverse': is_reverse, + 'gate_activation': gate_activation, + 'cell_activation': cell_activation, + 'candidate_activation': candidate_activation + }) + return hidden, cell def dynamic_lstmp(input, @@ -963,43 +961,39 @@ def linear_chain_crf(input, label, param_attr=None): return log_likelihood -if os.name != 'nt': - - @templatedoc() - def crf_decoding(input, param_attr, label=None): - """ - ${comment} +@templatedoc() +def crf_decoding(input, param_attr, label=None): + """ + ${comment} - Args: - input(${emission_type}): ${emission_comment} + Args: + input(${emission_type}): ${emission_comment} - param_attr(ParamAttr): The parameter attribute for training. + param_attr(ParamAttr): The parameter attribute for training. - label(${label_type}): ${label_comment} + label(${label_type}): ${label_comment} - Returns: - Variable: ${viterbi_path_comment} + Returns: + Variable: ${viterbi_path_comment} - Examples: - .. code-block:: python + Examples: + .. code-block:: python - crf_decode = layers.crf_decoding( - input=hidden, param_attr=ParamAttr(name="crfw")) - """ - helper = LayerHelper('crf_decoding', **locals()) - transition = helper.get_parameter(param_attr.name) - viterbi_path = helper.create_variable_for_type_inference( - dtype=helper.input_dtype()) - helper.append_op( - type='crf_decoding', - inputs={ - "Emission": [input], + crf_decode = layers.crf_decoding( + input=hidden, param_attr=ParamAttr(name="crfw")) + """ + helper = LayerHelper('crf_decoding', **locals()) + transition = helper.get_parameter(param_attr.name) + viterbi_path = helper.create_variable_for_type_inference( + dtype=helper.input_dtype()) + helper.append_op( + type='crf_decoding', + inputs={"Emission": [input], "Transition": transition, - "Label": label - }, - outputs={"ViterbiPath": [viterbi_path]}) + "Label": label}, + outputs={"ViterbiPath": [viterbi_path]}) - return viterbi_path + return viterbi_path @templatedoc() @@ -5593,48 +5587,42 @@ def label_smooth(label, return smooth_label -if os.name != 'nt': - - @templatedoc() - def roi_pool(input, - rois, - pooled_height=1, - pooled_width=1, - spatial_scale=1.0): - """ - ${comment} - - Args: - input (Variable): ${x_comment} - rois (Variable): ROIs (Regions of Interest) to pool over. - pooled_height (integer): ${pooled_height_comment} Default: 1 - pooled_width (integer): ${pooled_width_comment} Default: 1 - spatial_scale (float): ${spatial_scale_comment} Default: 1.0 - - Returns: - Variable: ${out_comment}. - - Examples: - .. code-block:: python - - pool_out = fluid.layers.roi_pool(input=x, rois=rois, 7, 7, 1.0) - """ - helper = LayerHelper('roi_pool', **locals()) - dtype = helper.input_dtype() - pool_out = helper.create_variable_for_type_inference(dtype) - argmaxes = helper.create_variable_for_type_inference(dtype='int32') - helper.append_op( - type="roi_pool", - inputs={"X": input, - "ROIs": rois}, - outputs={"Out": pool_out, - "Argmax": argmaxes}, - attrs={ - "pooled_height": pooled_height, - "pooled_width": pooled_width, - "spatial_scale": spatial_scale - }) - return pool_out +@templatedoc() +def roi_pool(input, rois, pooled_height=1, pooled_width=1, spatial_scale=1.0): + """ + ${comment} + + Args: + input (Variable): ${x_comment} + rois (Variable): ROIs (Regions of Interest) to pool over. + pooled_height (integer): ${pooled_height_comment} Default: 1 + pooled_width (integer): ${pooled_width_comment} Default: 1 + spatial_scale (float): ${spatial_scale_comment} Default: 1.0 + + Returns: + Variable: ${out_comment}. + + Examples: + .. code-block:: python + + pool_out = fluid.layers.roi_pool(input=x, rois=rois, 7, 7, 1.0) + """ + helper = LayerHelper('roi_pool', **locals()) + dtype = helper.input_dtype() + pool_out = helper.create_variable_for_type_inference(dtype) + argmaxes = helper.create_variable_for_type_inference(dtype='int32') + helper.append_op( + type="roi_pool", + inputs={"X": input, + "ROIs": rois}, + outputs={"Out": pool_out, + "Argmax": argmaxes}, + attrs={ + "pooled_height": pooled_height, + "pooled_width": pooled_width, + "spatial_scale": spatial_scale + }) + return pool_out @templatedoc() diff --git a/python/paddle/fluid/layers/ops.py b/python/paddle/fluid/layers/ops.py index 66eb1229aa3ec7a956146f12da2889d59b88671a..6c18af7283e19bd431c8d543255d900dc89cba09 100644 --- a/python/paddle/fluid/layers/ops.py +++ b/python/paddle/fluid/layers/ops.py @@ -100,26 +100,27 @@ Examples: >>> result = fluid.layers.hard_shrink(x=data, threshold=0.3) """ -if os.name != 'nt': - __all__ += ['cumsum'] - - _cum_sum_ = generate_layer_fn('cumsum') - - def cumsum(x, axis=None, exclusive=None, reverse=None): - locals_var = locals().keys() - kwargs = dict() - for name in locals_var: - val = locals()[name] - if val is not None: - kwargs[name] = val - return _cum_sum_(**kwargs) - - cumsum.__doc__ = _cum_sum_.__doc__ + """ - Examples: - - >>> data = fluid.layers.data(name="input", shape=[32, 784]) - >>> result = fluid.layers.cumsum(data, axis=0) - """ +__all__ += ['cumsum'] + +_cum_sum_ = generate_layer_fn('cumsum') + + +def cumsum(x, axis=None, exclusive=None, reverse=None): + locals_var = locals().keys() + kwargs = dict() + for name in locals_var: + val = locals()[name] + if val is not None: + kwargs[name] = val + return _cum_sum_(**kwargs) + + +cumsum.__doc__ = _cum_sum_.__doc__ + """ +Examples: + + >>> data = fluid.layers.data(name="input", shape=[32, 784]) + >>> result = fluid.layers.cumsum(data, axis=0) +""" __all__ += ['thresholded_relu']