Merge remote-tracking branch 'ups/develop' into refine/set_num_threads

fix conflicts

Author: tensor-tang

CMakeLists.txt

@@ -61,8 +61,11 @@ option(EIGEN_USE_THREADS "Compile with multi-threaded Eigen"            OFF)
 option(WITH_ARM_FP16    "Use half precision support on armv8.2-a cpu"   OFF)
 option(WITH_FAST_BUNDLE_TEST    "Bundle tests that can be run in a single process together to reduce launch overhead"   OFF)
 option(WITH_CONTRIB     "Compile the third-party contributation"        OFF)
+option(REPLACE_ENFORCE_GLOG "Replace PADDLE_ENFORCE with glog/CHECK for better debug." OFF)
 option(WITH_ANAKIN      "Compile with Anakin library"                   OFF)
 option(WITH_GRPC     "Use grpc as the default rpc framework"            ${WITH_DISTRIBUTE})
+option(WITH_BRPC_RDMA     "Use brpc rdma as the rpc protocal"           OFF)
+option(WITH_SYSTEM_BLAS   "Use system blas library"           OFF)
 
 # CMAKE_BUILD_TYPE
 if(NOT CMAKE_BUILD_TYPE)
@@ -131,6 +134,10 @@ if (NOT DEFINED WITH_MKLDNN)
         set(WITH_MKLDNN OFF)
     endif()
 endif()
+
+if (REPLACE_ENFORCE_GLOG)
+  add_definitions("-DREPLACE_ENFORCE_GLOG")
+endif()
 ########################################################################################
 
 include(external/mklml)     # download mklml package
@@ -153,12 +160,24 @@ include(external/cares)
 if(WITH_DISTRIBUTE)
     if(WITH_GRPC)
         include(external/grpc)
+        message(STATUS "Use grpc framework.")
     else()
+        message(STATUS "Use brpc framework.")
         include(external/leveldb)
         include(external/brpc)
     endif()
 endif()
 
+if(WITH_BRPC_RDMA)
+    message(STATUS "Use brpc with rdma.")
+    if(WITH_GRPC)
+        message(FATAL_ERROR "Can't use grpc with brpc rdma.")
+    endif()
+    if(NOT WITH_DISTRIBUTE)
+        message(FATAL_ERROR "Can't use brpc rdma in no distribute env.")
+    endif()
+endif()
+
 include(external/snappy)    # download snappy
 include(external/snappystream)
 include(external/threadpool)
@@ -178,7 +197,7 @@ include(inference_lib)      # add paddle fluid inference libraries
 
 
 include_directories("${PADDLE_SOURCE_DIR}")
-include_directories("${PADDLE_SOURCE_DIR}/paddle/cuda/include")
+include_directories("${PADDLE_SOURCE_DIR}/paddle/legacy/cuda/include")
 include_directories("${CMAKE_CURRENT_BINARY_DIR}/proto")
 include_directories("${CMAKE_CURRENT_BINARY_DIR}/go/pserver/client/c")
 
@@ -222,7 +241,7 @@ add_subdirectory(proto)
 if(NOT MOBILE_INFERENCE AND NOT WITH_FLUID_ONLY)
     # "add_subdirectory(go)" should be placed after the following loine,
     # because it depends on paddle/optimizer.
-    add_subdirectory(paddle/optimizer)
+    add_subdirectory(paddle/legacy/optimizer)
 endif()
 
 # "add_subdirectory(paddle)" and "add_subdirectory(python)" should be

CONTRIBUTING.md

@@ -159,4 +159,4 @@ This will enable VLOG messages generated by `buddy_allocator.{h,cc}` and in the
 - verbose level 1: [framework](https://github.com/PaddlePaddle/Paddle/tree/develop/paddle/framework)
 - verbose level 3: [operators](https://github.com/PaddlePaddle/Paddle/tree/develop/paddle/operators)
 - verbose level 5: [memory](https://github.com/PaddlePaddle/Paddle/tree/develop/paddle/memory), [platform](https://github.com/PaddlePaddle/Paddle/tree/develop/paddle/platform)
-- verbose level 7: [math](https://github.com/PaddlePaddle/Paddle/tree/develop/paddle/math)
+- verbose level 7: [math](https://github.com/PaddlePaddle/Paddle/tree/develop/paddle/legacy/math)

README.md

@@ -4,7 +4,6 @@
 [![Build Status](https://travis-ci.org/PaddlePaddle/Paddle.svg?branch=develop)](https://travis-ci.org/PaddlePaddle/Paddle)
 [![Documentation Status](https://img.shields.io/badge/docs-latest-brightgreen.svg?style=flat)](http://www.paddlepaddle.org/docs/develop/documentation/en/getstarted/index_en.html)
 [![Documentation Status](https://img.shields.io/badge/中文文档-最新-brightgreen.svg)](http://www.paddlepaddle.org/docs/develop/documentation/zh/getstarted/index_cn.html)
-[![Coverage Status](https://coveralls.io/repos/github/PaddlePaddle/Paddle/badge.svg?branch=develop)](https://coveralls.io/github/PaddlePaddle/Paddle?branch=develop)
 [![Release](https://img.shields.io/github/release/PaddlePaddle/Paddle.svg)](https://github.com/PaddlePaddle/Paddle/releases)
 [![License](https://img.shields.io/badge/license-Apache%202-blue.svg)](LICENSE)
 
@@ -19,6 +18,8 @@ learning to many products at Baidu.
 Our vision is to enable deep learning for everyone via PaddlePaddle.
 Please refer to our [release announcement](https://github.com/PaddlePaddle/Paddle/releases) to track the latest feature of PaddlePaddle.
 
+### Lastest PaddlePaddle Version: [Fluid](https://github.com/PaddlePaddle/Paddle/tree/develop/paddle/fluid)
+
 ## Features
 
 - **Flexibility**

benchmark/fluid/args.py

@@ -122,5 +122,13 @@ def parse_args():
         type=str,
         default="",
         help='Directory that contains all the training recordio files.')
+    parser.add_argument(
+        '--use_inference_transpiler',
+        action='store_true',
+        help='If set, use inference transpiler to optimize the program.')
+    parser.add_argument(
+        '--no_random',
+        action='store_true',
+        help='If set, keep the random seed and do not shuffle the data.')
     args = parser.parse_args()
     return args

benchmark/fluid/fluid_benchmark.py

@@ -131,6 +131,7 @@ def train(avg_loss, infer_prog, optimizer, train_reader, test_reader, batch_acc,
     exe = fluid.Executor(place)
     exe.run(startup_prog)
 
+    # Use inference_transpiler to speedup
     if not args.use_reader_op:
         feed_var_list = [
             var for var in train_prog.global_block().vars.itervalues()
@@ -181,6 +182,10 @@ def train(avg_loss, infer_prog, optimizer, train_reader, test_reader, batch_acc,
         print("Pass: %d, Loss: %f" % (pass_id, np.mean(train_losses))),
         # evaluation
         if not args.no_test and batch_acc and not args.use_reader_op:
+            if args.use_inference_transpiler:
+                t = fluid.InferenceTranspiler()
+                t.transpile(infer_prog, place)
+
             pass_test_acc = test(exe, infer_prog, test_reader, feeder,
                                  batch_acc)
             print(", Test Accuracy: %f" % pass_test_acc)
@@ -311,6 +316,8 @@ def main():
     args = parse_args()
     print_arguments(args)
     print_paddle_envs()
+    if args.no_random:
+        fluid.default_startup_program().random_seed = 1
 
     # the unique trainer id, starting from 0, needed by trainer
     # only

benchmark/fluid/models/resnet.py

@@ -197,12 +197,12 @@ def get_model(args):
     optimizer = fluid.optimizer.Momentum(learning_rate=0.01, momentum=0.9)
 
     batched_train_reader = paddle.batch(
-        paddle.reader.shuffle(
+        train_reader if args.no_random else paddle.reader.shuffle(
             train_reader, buf_size=5120),
         batch_size=args.batch_size * args.gpus,
         drop_last=True)
     batched_test_reader = paddle.batch(
-        train_reader, batch_size=args.batch_size, drop_last=True)
+        test_reader, batch_size=args.batch_size, drop_last=True)
 
     return avg_cost, inference_program, optimizer, batched_train_reader,\
                    batched_test_reader, batch_acc

cmake/cblas.cmake

@@ -83,18 +83,20 @@ else()
   set(REFERENCE_CBLAS_LIB_SEARCH_PATHS ${REFERENCE_CBLAS_ROOT}/lib)
 endif()
 
-find_path(REFERENCE_CBLAS_INCLUDE_DIR NAMES cblas.h PATHS
+if(WITH_SYSTEM_BLAS)
+  find_path(REFERENCE_CBLAS_INCLUDE_DIR NAMES cblas.h PATHS
         ${REFERENCE_CBLAS_INCLUDE_SEARCH_PATHS})
-find_library(REFERENCE_CBLAS_LIBRARY NAMES cblas PATHS
+  find_library(REFERENCE_CBLAS_LIBRARY NAMES cblas PATHS
         ${REFERENCE_CBLAS_LIB_SEARCH_PATHS})
 
-if(REFERENCE_CBLAS_INCLUDE_DIR AND REFERENCE_CBLAS_LIBRARY)
-  set(CBLAS_FOUND ON)
-  set(CBLAS_PROVIDER REFERENCE)
-  set(CBLAS_INC_DIR ${REFERENCE_CBLAS_INCLUDE_DIR})
-  set(CBLAS_LIBRARIES ${REFERENCE_CBLAS_LIBRARY})
-  add_definitions(-DPADDLE_USE_REFERENCE_CBLAS)
-  message(STATUS "Found reference-cblas (include: ${CBLAS_INC_DIR}, library: ${CBLAS_LIBRARIES})")
+  if(REFERENCE_CBLAS_INCLUDE_DIR AND REFERENCE_CBLAS_LIBRARY)
+    set(CBLAS_FOUND ON)
+    set(CBLAS_PROVIDER REFERENCE)
+    set(CBLAS_INC_DIR ${REFERENCE_CBLAS_INCLUDE_DIR})
+    set(CBLAS_LIBRARIES ${REFERENCE_CBLAS_LIBRARY})
+    add_definitions(-DPADDLE_USE_REFERENCE_CBLAS)
+    message(STATUS "Found reference-cblas (include: ${CBLAS_INC_DIR}, library: ${CBLAS_LIBRARIES})")
+  endif()
 endif()
 
 if(IOS_USE_VECLIB_FOR_BLAS AND VECLIB_FOUND)

cmake/configure.cmake

@@ -174,3 +174,7 @@ endif(WITH_GOLANG)
 if(WITH_GRPC)
     add_definitions(-DPADDLE_WITH_GRPC)
 endif(WITH_GRPC)
+
+if(WITH_BRPC_RDMA)
+    add_definitions(-DPADDLE_WITH_BRPC_RDMA)
+endif(WITH_BRPC_RDMA)

cmake/external/anakin.cmake

@@ -26,13 +26,15 @@ function(fetch_include_recursively root_dir)
     endforeach()
 endfunction()
 
-# download library
-message(STATUS "Download Anakin library from ${ANAKIN_LIBRARY_URL}")
-execute_process(COMMAND bash -c "mkdir -p ${ANAKIN_INSTALL_DIR}")
-execute_process(COMMAND bash -c "rm -rf ${ANAKIN_INSTALL_DIR}/*")
-execute_process(COMMAND bash -c "cd ${ANAKIN_INSTALL_DIR}; wget -q ${ANAKIN_LIBRARY_URL}")
-execute_process(COMMAND bash -c "mkdir -p ${ANAKIN_INSTALL_DIR}")
-execute_process(COMMAND bash -c "cd ${ANAKIN_INSTALL_DIR}; tar xzf anakin_release_simple.tar.gz")
+if (NOT EXISTS "${ANAKIN_INSTALL_DIR}")
+    # download library
+    message(STATUS "Download Anakin library from ${ANAKIN_LIBRARY_URL}")
+    execute_process(COMMAND bash -c "mkdir -p ${ANAKIN_INSTALL_DIR}")
+    execute_process(COMMAND bash -c "rm -rf ${ANAKIN_INSTALL_DIR}/*")
+    execute_process(COMMAND bash -c "cd ${ANAKIN_INSTALL_DIR}; wget -q ${ANAKIN_LIBRARY_URL}")
+    execute_process(COMMAND bash -c "mkdir -p ${ANAKIN_INSTALL_DIR}")
+    execute_process(COMMAND bash -c "cd ${ANAKIN_INSTALL_DIR}; tar xzf anakin_release_simple.tar.gz")
+endif()
 
 if (WITH_ANAKIN)
     message(STATUS "Anakin for inference is enabled")

cmake/external/brpc.cmake

@@ -14,6 +14,15 @@
 
 INCLUDE(ExternalProject)
 
+find_library(SSL_LIBRARY NAMES ssl)
+ADD_LIBRARY(ssl SHARED IMPORTED GLOBAL)
+SET_PROPERTY(TARGET ssl PROPERTY IMPORTED_LOCATION ${SSL_LIBRARY})
+
+find_library(CRYPTO_LIBRARY NAMES crypto)
+ADD_LIBRARY(crypto SHARED IMPORTED GLOBAL)
+SET_PROPERTY(TARGET crypto PROPERTY IMPORTED_LOCATION ${CRYPTO_LIBRARY})
+
+
 SET(BRPC_SOURCES_DIR ${THIRD_PARTY_PATH}/brpc)
 SET(BRPC_INSTALL_DIR ${THIRD_PARTY_PATH}/install/brpc)
 SET(BRPC_INCLUDE_DIR "${BRPC_INSTALL_DIR}/include" CACHE PATH "brpc include directory." FORCE)
@@ -22,14 +31,14 @@ SET(BRPC_LIBRARIES "${BRPC_INSTALL_DIR}/lib/libbrpc.a" CACHE FILEPATH "brpc libr
 INCLUDE_DIRECTORIES(${BRPC_INCLUDE_DIR})
 
 # Reference https://stackoverflow.com/questions/45414507/pass-a-list-of-prefix-paths-to-externalproject-add-in-cmake-args
-set(prefix_path "${THIRD_PARTY_PATH}/install/gflags|${THIRD_PARTY_PATH}/install/leveldb|${THIRD_PARTY_PATH}/install/snappy|${THIRD_PARTY_PATH}/install/gtest|${THIRD_PARTY_PATH}/install/protobuf")
+set(prefix_path "${THIRD_PARTY_PATH}/install/gflags|${THIRD_PARTY_PATH}/install/leveldb|${THIRD_PARTY_PATH}/install/snappy|${THIRD_PARTY_PATH}/install/gtest|${THIRD_PARTY_PATH}/install/protobuf|${THIRD_PARTY_PATH}/install/zlib")
 
 # If minimal .a is need, you can set  WITH_DEBUG_SYMBOLS=OFF
 ExternalProject_Add(
     extern_brpc
     ${EXTERNAL_PROJECT_LOG_ARGS}
-    GIT_REPOSITORY  "https://github.com/brpc/brpc"
-    GIT_TAG         "6d153dd7ff00f960ae6895c9c5fff0ce9f07aff2"
+    GIT_REPOSITORY  "https://github.com/gongweibao/brpc"
+    GIT_TAG         "7dc04defad1fd4173aae170c3fcbde131b65155a"
     PREFIX          ${BRPC_SOURCES_DIR}
     UPDATE_COMMAND  ""
     CMAKE_ARGS      -DCMAKE_CXX_COMPILER=${CMAKE_CXX_COMPILER}
@@ -42,14 +51,16 @@ ExternalProject_Add(
                     -DCMAKE_BUILD_TYPE=${THIRD_PARTY_BUILD_TYPE}
                     -DCMAKE_PREFIX_PATH=${prefix_path}
                     -DBRPC_WITH_GLOG=ON
+                    -DIOBUF_WITH_HUGE_BLOCK=ON
+                    -DBRPC_WITH_RDMA=${WITH_BRPC_RDMA}
                     ${EXTERNAL_OPTIONAL_ARGS}
     LIST_SEPARATOR |
     CMAKE_CACHE_ARGS -DCMAKE_INSTALL_PREFIX:PATH=${BRPC_INSTALL_DIR}
                      -DCMAKE_INSTALL_LIBDIR:PATH=${BRPC_INSTALL_DIR}/lib
                      -DCMAKE_POSITION_INDEPENDENT_CODE:BOOL=ON
                      -DCMAKE_BUILD_TYPE:STRING=${THIRD_PARTY_BUILD_TYPE}
 )
-ADD_DEPENDENCIES(extern_brpc protobuf leveldb gflags glog gtest snappy)
+ADD_DEPENDENCIES(extern_brpc protobuf ssl crypto leveldb gflags glog gtest snappy)
 ADD_LIBRARY(brpc STATIC IMPORTED GLOBAL)
 SET_PROPERTY(TARGET brpc PROPERTY IMPORTED_LOCATION ${BRPC_LIBRARIES})
 ADD_DEPENDENCIES(brpc extern_brpc)