remove conflict

benchmark/paddle/image/run_openblas_infer.sh

@@ -8,15 +8,19 @@ function clock_to_seconds() {
 }
 
 function infer() {
-  unset OMP_NUM_THREADS MKL_NUM_THREADS OMP_DYNAMIC KMP_AFFINITY
   topology=$1
   layer_num=$2
   bs=$3
-  thread=`nproc`
-  if [ $thread -gt $bs ]; then
-    thread=$bs
+  trainers=`nproc`
+  if [ $trainers -gt $bs ]; then
+    trainers=$bs
   fi
-  log="logs/infer-${topology}-${layer_num}-${thread}openblas-${bs}.log"
+  log="logs/infer-${topology}-${layer_num}-${trainers}openblas-${bs}.log"
+  threads=$((`nproc` / trainers))
+  if [ $threads -eq 0 ]; then
+    threads=1
+  fi
+  export OPENBLAS_NUM_THREADS=$threads
 
   models_in="models/${topology}-${layer_num}/pass-00000/"
   if [ ! -d $models_in ]; then
@@ -28,7 +32,7 @@ function infer() {
     --config="${topology}.py" \
     --use_mkldnn=False \
     --use_gpu=False \
-    --trainer_count=$thread \
+    --trainer_count=$trainers \
     --log_period=$log_period \
     --config_args="batch_size=${bs},layer_num=${layer_num},is_infer=True,num_samples=256" \
     --init_model_path=$models_in \

benchmark/paddle/image/run_openblas_train.sh

@@ -1,7 +1,7 @@
 set -e
 
 function train() {
-  unset OMP_NUM_THREADS MKL_NUM_THREADS OMP_DYNAMIC KMP_AFFINITY
+  export OPENBLAS_NUM_THREADS=1
   topology=$1
   layer_num=$2
   bs=$3

doc/api/v2/config/layer.rst

@@ -252,6 +252,11 @@ first_seq
 ..  autoclass:: paddle.v2.layer.first_seq
     :noindex:
 
+sub_seq
+---------
+..  autoclass:: paddle.v2.layer.sub_seq
+    :noindex:
+
 concat
 ------
 ..  autoclass:: paddle.v2.layer.concat

doc/api/v2/fluid/layers.rst

@@ -68,12 +68,6 @@ scale
     :noindex:
 
 
-reshape
----------
-..  autofunction:: paddle.v2.fluid.layers.reshape
-    :noindex:
-
-
 transpose
 ---------
 ..  autofunction:: paddle.v2.fluid.layers.transpose

doc/design/optimizer.md

@@ -79,7 +79,7 @@ class Optimizer(object):
     def minimize(self, loss, parameter_list):
         """Add operations to minimize `loss` by updating `parameter_list`.
 
-        This method combines interface `append_backward_ops()` and
+        This method combines interface `append_backward()` and
         `create_optimization_pass()` into one.
         """
         params_grads = self.create_backward_pass(loss, parameter_list)

doc/getstarted/build_and_install/docker_install_cn.rst

@@ -15,7 +15,7 @@
 获取PaddlePaddle的Docker镜像
 ------------------------------
 
-执行下面的命令获取最新的PaddlePaddle Docker镜像
+执行下面的命令获取最新的PaddlePaddle Docker镜像，版本为cpu_avx_mkl：
 
   .. code-block:: bash
 
@@ -27,7 +27,7 @@
 
      docker pull docker.paddlepaddle.org/paddle
 
-下载GPU版本的Docker镜像：
+下载GPU版本（cuda8.0_cudnn5_avx_mkl）的Docker镜像：
 
   .. code-block:: bash
 
@@ -54,7 +54,7 @@
 .. _docker_run:
 
 在Docker中执行PaddlePaddle训练程序
-------------------------------
+----------------------------------
 
 假设您已经在当前目录（比如在/home/work）编写了一个PaddlePaddle的程序 :code:`train.py` （可以参考
 `PaddlePaddleBook <http://www.paddlepaddle.org/docs/develop/book/01.fit_a_line/index.cn.html>`_ 
@@ -82,7 +82,7 @@
 .. _docker_run_book:
 
 使用Docker启动PaddlePaddle Book教程
-------------------------------
+-----------------------------------
 
 使用Docker可以快速在本地启动一个包含了PaddlePaddle官方Book教程的Jupyter Notebook，可以通过网页浏览。
 PaddlePaddle Book是为用户和开发者制作的一个交互式的Jupyter Notebook。

doc/getstarted/build_and_install/docker_install_en.rst

@@ -16,7 +16,7 @@ After you've read above tutorials you may proceed the following steps.
 Pull PaddlePaddle Docker Image
 ------------------------------
 
-Run the following command to download the latest Docker images:
+Run the following command to download the latest Docker images, the version is cpu_avx_mkl:
 
   .. code-block:: bash
 
@@ -28,7 +28,7 @@ For users in China, we provide a faster mirror:
 
      docker pull docker.paddlepaddle.org/paddle
 
-Download GPU version images:
+Download GPU version (cuda8.0_cudnn5_avx_mkl) images:
 
   .. code-block:: bash
 
@@ -58,7 +58,7 @@ and run:
 .. _docker_run:
 
 Launch your training program in Docker
-------------------------------
+--------------------------------------
 
 Assume that you have already written a PaddlePaddle program
 named :code:`train.py` under directory :code:`/home/work` (refer to 

doc/getstarted/build_and_install/pip_install_cn.rst

@@ -11,14 +11,14 @@ PaddlePaddle可以使用常用的Python包管理工具
 ------------------------------
 
 
-执行下面的命令即可在当前机器上安装PaddlePaddle的运行时环境，并自动下载安装依赖软件。
+执行下面的命令即可在当前机器上安装PaddlePaddle的运行时环境，并自动下载安装依赖软件，版本为cpu_avx_openblas。
 
   .. code-block:: bash
 
      pip install paddlepaddle
 
 
-如果需要安装支持GPU的版本，需要执行：
+如果需要安装支持GPU的版本（cuda7.5_cudnn5_avx_openblas），需要执行：
 
   .. code-block:: bash
 

doc/getstarted/build_and_install/pip_install_en.rst

@@ -12,14 +12,14 @@ Install Using pip
 ------------------------------
 
 Run the following command to install PaddlePaddle on the current
-machine, it will also download requirements.
+machine, it will also download requirements, the version is cpu_avx_openblas.
 
   .. code-block:: bash
 
      pip install paddlepaddle
 
 
-If you wish to install GPU version, just run:
+If you wish to install GPU version (cuda7.5_cudnn5_avx_openblas), just run:
 
   .. code-block:: bash
 

doc/getstarted/index_cn.rst

@@ -7,13 +7,13 @@
 ++++++++
 
 PaddlePaddle支持使用pip快速安装，目前支持CentOS 6以上, Ubuntu 14.04以及MacOS 10.12，并安装有Python2.7。
-执行下面的命令完成快速安装：
+执行下面的命令完成快速安装，版本为cpu_avx_openblas：
 
   .. code-block:: bash
 
      pip install paddlepaddle
 
-如果需要安装支持GPU的版本，需要执行：
+如果需要安装支持GPU的版本（cuda7.5_cudnn5_avx_openblas），需要执行：
 
   .. code-block:: bash
 

doc/getstarted/index_en.rst

@@ -8,13 +8,13 @@ Quick Install
 
 You can use pip to install PaddlePaddle with a single command, supports
 CentOS 6 above, Ubuntu 14.04 above or MacOS 10.12, with Python 2.7 installed.
-Simply run the following command to install:
+Simply run the following command to install, the version is cpu_avx_openblas:
 
   .. code-block:: bash
 
      pip install paddlepaddle
 
-If you need to install GPU version, run:
+If you need to install GPU version (cuda7.5_cudnn5_avx_openblas), run:
 
   .. code-block:: bash
 

paddle/framework/CMakeLists.txt

@@ -5,10 +5,18 @@ cc_library(ddim SRCS ddim.cc DEPS eigen3)
 cc_test(ddim_test SRCS ddim_test.cc DEPS ddim)
 nv_test(dim_test SRCS dim_test.cu DEPS ddim)
 
-cc_library(tensor SRCS tensor.cc DEPS ddim place paddle_memory device_context)
+if (WITH_GPU)
+  nv_library(tensor SRCS tensor.cc tensor_util.cu DEPS ddim place paddle_memory device_context framework_proto)
+else()
+  cc_library(tensor SRCS tensor.cc tensor_util.cc DEPS ddim place paddle_memory device_context framework_proto)
+endif ()
 
 cc_test(tensor_test SRCS tensor_test.cc DEPS tensor)
-cc_test(tensor_util_test SRCS tensor_util_test.cc DEPS tensor)
+if (WITH_GPU)
+  nv_test(tensor_util_test SRCS tensor_util_test.cc tensor_util_test.cu DEPS tensor)
+else()
+  cc_test(tensor_util_test SRCS tensor_util_test.cc DEPS tensor)
+endif()
 
 cc_test(eigen_test SRCS eigen_test.cc DEPS tensor)
 

paddle/framework/data_transform.cc

@@ -13,6 +13,7 @@ See the License for the specific language governing permissions and
 limitations under the License. */
 
 #include "paddle/framework/data_transform.h"
+#include "paddle/framework/lod_tensor.h"
 
 namespace paddle {
 namespace framework {

paddle/framework/data_transform.h

@@ -27,9 +27,8 @@ limitations under the License. */
 namespace paddle {
 namespace framework {
 
-using DataTransformFN =
-    std::function<void(const std::vector<platform::DeviceContext*> ctx,
-                       const Variable& in, Variable* out)>;
+using DataTransformFn = std::function<void(const platform::DeviceContext* ctx,
+                                           const Variable& in, Variable* out)>;
 using KernelTypePair = std::pair<OpKernelType, OpKernelType>;
 
 struct KernelTypePairHash {
@@ -47,7 +46,7 @@ struct KernelTypePairHash {
 };
 
 using DataTransformMap =
-    std::unordered_map<KernelTypePair, DataTransformFN, KernelTypePairHash>;
+    std::unordered_map<KernelTypePair, DataTransformFn, KernelTypePairHash>;
 
 class DataTransformFnMap {
  public:
@@ -58,25 +57,25 @@ class DataTransformFnMap {
   }
 
   void Insert(const OpKernelType& left, const OpKernelType& right,
-              const DataTransformFN& data_tranform_fn) {
+              const DataTransformFn& data_tranform_fn) {
     Insert(std::make_pair(left, right), data_tranform_fn);
   }
 
   void Insert(const KernelTypePair& kernel_type_pair,
-              const DataTransformFN& data_tranform_fn) {
+              const DataTransformFn& data_tranform_fn) {
     PADDLE_ENFORCE(!Has(kernel_type_pair),
                    "KernelTypePair %s has been registered", "");
     map_.insert({kernel_type_pair, data_tranform_fn});
   }
 
-  const DataTransformFN& Get(const KernelTypePair& key_pair) const {
+  const DataTransformFn& Get(const KernelTypePair& key_pair) const {
     auto data_transformer = GetNullable(key_pair);
     PADDLE_ENFORCE_NOT_NULL(data_transformer,
-                            "DataTransformFN should not be NULL");
+                            "DataTransformFn should not be NULL");
     return *data_transformer;
   }
 
-  const DataTransformFN* GetNullable(const KernelTypePair& key_pair) const {
+  const DataTransformFn* GetNullable(const KernelTypePair& key_pair) const {
     auto it = map_.find(key_pair);
     if (it == map_.end()) {
       return nullptr;