Merge remote-tracking branch 'ups/develop' into mklml_funcs

Author: tensor-tang

CMakeLists.txt

@@ -57,7 +57,10 @@ option(GLIDE_INSTALL    "Download and install go dependencies "         ON)
 option(USE_NNPACK       "Compile PaddlePaddle with NNPACK library"      OFF)
 option(WITH_DISTRIBUTE  "Compile with grpc distributed support"         OFF)
 option(USE_EIGEN_FOR_BLAS   "Use matrix multiplication in Eigen"        OFF)
+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)
 
 # CMAKE_BUILD_TYPE
 if(NOT CMAKE_BUILD_TYPE)
@@ -202,7 +205,7 @@ endif(USE_NNPACK)
 
 add_subdirectory(proto)
 
-if(NOT MOBILE_INFERENCE)
+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)
@@ -230,3 +233,7 @@ if(WITH_DOC)
     find_python_module(recommonmark REQUIRED)
     add_subdirectory(doc)
 endif()
+
+if (WITH_CONTRIB)
+    add_subdirectory(paddle/contrib)
+endif()

Dockerfile

@@ -101,6 +101,3 @@ RUN echo 'root:root' | chpasswd
 RUN sed -ri 's/^PermitRootLogin\s+.*/PermitRootLogin yes/' /etc/ssh/sshd_config
 RUN sed -ri 's/UsePAM yes/#UsePAM yes/g' /etc/ssh/sshd_config
 EXPOSE 22
-
-# development image default do build work
-CMD ["bash", "/paddle/paddle/scripts/docker/build.sh"]

Dockerfile.android

@@ -40,5 +40,3 @@ RUN mkdir -p ${ANDROID_TOOLCHAINS_DIR} && \
     unzip -q android-ndk-r14b-linux-x86_64.zip && \
     mv android-ndk-r14b ${ANDROID_NDK_HOME} && \
     rm -rf /opt/android-ndk-tmp
-
-CMD ["bash", "/paddle/paddle/scripts/docker/build_android.sh"]

benchmark/fluid/README.md

@@ -24,22 +24,22 @@ Currently supported `--model` argument include:
 
 * Run the following command to start a benchmark job locally:
     ```bash
-      python fluid_benchmark.py --model mnist --parallel 1 --device GPU --with_test
+      python fluid_benchmark.py --model mnist  --device GPU
     ```
     You can choose to use GPU/CPU training. With GPU training, you can specify
-    `--parallel 1` to run multi GPU training.
+    `--gpus <gpu_num>` to run multi GPU training.
 * Run distributed training with parameter servers:
     * start parameter servers:
         ```bash
-        PADDLE_TRAINING_ROLE=PSERVER PADDLE_PSERVER_PORT=7164 PADDLE_PSERVER_IPS=127.0.0.1 PADDLE_TRAINERS=1 PADDLE_CURRENT_IP=127.0.0.1 PADDLE_TRAINER_ID=0 python fluid_benchmark.py --model mnist --parallel 0 --device GPU --update_method pserver
+        PADDLE_TRAINING_ROLE=PSERVER PADDLE_PSERVER_PORT=7164 PADDLE_PSERVER_IPS=127.0.0.1 PADDLE_TRAINERS=1 PADDLE_CURRENT_IP=127.0.0.1 PADDLE_TRAINER_ID=0 python fluid_benchmark.py --model mnist  --device GPU --update_method pserver
         ```
     * start trainers:
         ```bash
-        PADDLE_TRAINING_ROLE=PSERVER PADDLE_PSERVER_PORT=7164 PADDLE_PSERVER_IPS=127.0.0.1 PADDLE_TRAINERS=1 PADDLE_CURRENT_IP=127.0.0.1 PADDLE_TRAINER_ID=0 python fluid_benchmark.py --model mnist --parallel 0 --device GPU --update_method pserver
+        PADDLE_TRAINING_ROLE=TRAINER PADDLE_PSERVER_PORT=7164 PADDLE_PSERVER_IPS=127.0.0.1 PADDLE_TRAINERS=1 PADDLE_CURRENT_IP=127.0.0.1 PADDLE_TRAINER_ID=0 python fluid_benchmark.py --model mnist  --device GPU --update_method pserver
         ```
 * Run distributed training using NCCL2
     ```bash
-    PADDLE_PSERVER_PORT=7164 PADDLE_TRAINER_IPS=192.168.0.2,192.168.0.3  PADDLE_CURRENT_IP=127.0.0.1 PADDLE_TRAINER_ID=0 python fluid_benchmark.py --model mnist --parallel 0 --device GPU --update_method nccl2
+    PADDLE_PSERVER_PORT=7164 PADDLE_TRAINER_IPS=192.168.0.2,192.168.0.3  PADDLE_CURRENT_IP=127.0.0.1 PADDLE_TRAINER_ID=0 python fluid_benchmark.py --model mnist --device GPU --update_method nccl2
     ```
 
 ## Run Distributed Benchmark on Kubernetes Cluster
@@ -48,7 +48,7 @@ We provide a script `kube_gen_job.py` to generate Kubernetes yaml files to submi
 distributed benchmark jobs to your cluster. To generate a job yaml, just run:
 
 ```bash
-python kube_gen_job.py --jobname myjob --pscpu 4 --cpu 8 --gpu 8 --psmemory 20 --memory 40 --pservers 4 --trainers 4 --entry "python fluid_benchmark.py --model mnist --parallel 1 --device GPU --update_method pserver --with_test" --disttype pserver
+python kube_gen_job.py --jobname myjob --pscpu 4 --cpu 8 --gpu 8 --psmemory 20 --memory 40 --pservers 4 --trainers 4 --entry "python fluid_benchmark.py --model mnist --parallel 1 --device GPU --update_method pserver " --disttype pserver
 ```
 
 Then the yaml files are generated under directory `myjob`, you can run:

cmake/configure.cmake

@@ -41,6 +41,10 @@ if(USE_EIGEN_FOR_BLAS)
     add_definitions(-DPADDLE_USE_EIGEN_FOR_BLAS)
 endif(USE_EIGEN_FOR_BLAS)
 
+if(EIGEN_USE_THREADS)
+    add_definitions(-DEIGEN_USE_THREADS)
+endif(EIGEN_USE_THREADS)
+
 if(NOT WITH_PROFILER)
     add_definitions(-DPADDLE_DISABLE_PROFILER)
 endif(NOT WITH_PROFILER)

cmake/external/protobuf.cmake

@@ -212,6 +212,7 @@ FUNCTION(build_protobuf TARGET_NAME BUILD_FOR_HOST)
         ${CMAKE_COMMAND} ${PROTOBUF_SOURCES_DIR}/src/${TARGET_NAME}/cmake
             ${OPTIONAL_ARGS}
             -Dprotobuf_BUILD_TESTS=OFF
+            -DCMAKE_SKIP_RPATH=ON
             -DCMAKE_POSITION_INDEPENDENT_CODE=ON
             -DCMAKE_BUILD_TYPE=${THIRD_PARTY_BUILD_TYPE}
             -DCMAKE_INSTALL_PREFIX=${PROTOBUF_INSTALL_DIR}

doc/fluid/api/layers.rst

@@ -1003,9 +1003,9 @@ dice_loss
 ..  autofunction:: paddle.fluid.layers.dice_loss
     :noindex:
 
-bilinear_interp
+upsampling_bilinear2d
 ____
 
-..  autofunction:: paddle.fluid.layers.bilinear_interp
+..  autofunction:: paddle.fluid.layers.upsampling_bilinear2d
     :noindex: