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

… dockerreadme

.dockerignore

@@ -0,0 +1,15 @@
+*.DS_Store
+build/
+*.user
+.vscode
+.idea
+.project
+.cproject
+.pydevproject
+Makefile
+.test_env/
+third_party/
+*~
+bazel-*
+
+!build/*.deb

Dockerfile

@@ -3,20 +3,17 @@
 FROM nvidia/cuda:7.5-cudnn5-devel-ubuntu14.04
 MAINTAINER PaddlePaddle Authors <[email protected]>
 
-ARG DEBIAN_FRONTEND=noninteractive
 ARG UBUNTU_MIRROR
 RUN /bin/bash -c 'if [[ -n ${UBUNTU_MIRROR} ]]; then sed -i 's#http://archive.ubuntu.com/ubuntu#${UBUNTU_MIRROR}#g' /etc/apt/sources.list; fi'
 
 # ENV variables
 ARG BUILD_WOBOQ
-ARG BUILD_AND_INSTALL
 ARG WITH_GPU
 ARG WITH_AVX
 ARG WITH_DOC
 ARG WITH_STYLE_CHECK
 
 ENV BUILD_WOBOQ=${BUILD_WOBOQ:-OFF}
-ENV BUILD_AND_INSTALL=${BUILD_AND_INSTALL:-OFF}
 ENV WITH_GPU=${WITH_AVX:-OFF}
 ENV WITH_AVX=${WITH_AVX:-ON}
 ENV WITH_DOC=${WITH_DOC:-OFF}
@@ -31,7 +28,7 @@ RUN apt-get update && \
     apt-get install -y wget unzip tar xz-utils bzip2 gzip coreutils && \
     apt-get install -y curl sed grep graphviz libjpeg-dev zlib1g-dev && \
     apt-get install -y python-numpy python-matplotlib gcc g++ gfortran && \
-    apt-get install -y automake locales clang-format-3.8 && \
+    apt-get install -y automake locales clang-format-3.8 swig && \
     apt-get clean -y
 
 # git credential to skip password typing
@@ -51,8 +48,6 @@ RUN curl -sSL https://cmake.org/files/v3.4/cmake-3.4.1.tar.gz | tar -xz && \
     cd cmake-3.4.1 && ./bootstrap && make -j `nproc` && make install && \
     cd .. && rm -rf cmake-3.4.1
 
-RUN apt-get install -y swig
-
 VOLUME ["/usr/share/nginx/html/data", "/usr/share/nginx/html/paddle"]
 
 # Configure OpenSSH server. c.f. https://docs.docker.com/engine/examples/running_ssh_service

README.md

@@ -2,8 +2,8 @@
 
 
 [![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/)
-[![Documentation Status](https://img.shields.io/badge/中文文档-最新-brightgreen.svg)](http://www.paddlepaddle.org/cn/index.html)
+[![Documentation Status](https://img.shields.io/badge/docs-latest-brightgreen.svg?style=flat)](http://www.paddlepaddle.org/develop/doc/)
+[![Documentation Status](https://img.shields.io/badge/中文文档-最新-brightgreen.svg)](http://www.paddlepaddle.org/doc_cn/)
 [![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)
@@ -59,36 +59,36 @@ Please refer to our [release announcement](https://github.com/PaddlePaddle/Paddl
     the capability of PaddlePaddle to make a huge impact for your product.
 
 ## Installation
-Check out the [Install Guide](http://paddlepaddle.org/doc/build/) to install from
-pre-built packages (**docker image**, **deb package**) or
-directly build on **Linux** and **Mac OS X** from the source code.
+
+It is recommended to check out the
+[Docker installation guide](http://www.paddlepaddle.org/develop/doc/getstarted/build_and_install/docker_install_en.html)
+before looking into the
+[build from source guide](http://www.paddlepaddle.org/develop/doc/getstarted/build_and_install/build_from_source_en.html)
 
 ## Documentation
-Both [English Docs](http://paddlepaddle.org/doc/) and [Chinese Docs](http://paddlepaddle.org/doc_cn/) are provided for our users and developers.
 
-- [Quick Start](http://paddlepaddle.org/doc/demo/quick_start/index_en) <br>
-   You can follow the quick start tutorial to learn how use PaddlePaddle
-   step-by-step.
+We provide [English](http://www.paddlepaddle.org/develop/doc/) and
+[Chinese](http://www.paddlepaddle.org/doc_cn/) documentation.
+
+- [Deep Learning 101](http://book.paddlepaddle.org/index.en.html)
+
+  You might want to start from the this online interactive book that can run in Jupyter Notebook.
+
+- [Distributed Training](http://www.paddlepaddle.org/develop/doc/howto/usage/cluster/cluster_train_en.html)
+
+  You can run distributed training jobs on MPI clusters.
+
+- [Distributed Training on Kubernetes](http://www.paddlepaddle.org/develop/doc/howto/usage/k8s/k8s_en.html)
 
-- [Example and Demo](http://paddlepaddle.org/doc/demo/) <br>
-   We provide five demos, including: image classification, sentiment analysis,
-   sequence to sequence model, recommendation, semantic role labeling.
+   You can also run distributed training jobs on Kubernetes clusters.
 
-- [Distributed Training](http://paddlepaddle.org/doc/cluster) <br>
-  This system supports training deep learning models on multiple machines
-  with data parallelism.
+- [Python API](http://www.paddlepaddle.org/develop/doc/api/index_en.html)
 
-- [Python API](http://paddlepaddle.org/doc/ui/) <br>
-   PaddlePaddle supports using either Python interface or C++ to build your
-   system. We also use SWIG to wrap C++ source code to create a user friendly
-   interface for Python. You can also use SWIG to create interface for your
-   favorite programming language.
+   Our new API enables much shorter programs.
 
-- [How to Contribute](http://paddlepaddle.org/doc/build/contribute_to_paddle.html) <br>
-   We sincerely appreciate your interest and contributions. If you would like to
-   contribute, please read the contribution guide.
+- [How to Contribute](http://www.paddlepaddle.org/develop/doc/howto/dev/contribute_to_paddle_en.html)
 
-- [Source Code Documents](http://paddlepaddle.org/doc/source/) <br>
+   We appreciate your contributions!
 
 ## Ask Questions
 

cmake/FindSphinx.cmake

@@ -72,7 +72,7 @@ function( Sphinx_add_target target_name builder conf cache source destination )
     ${source}
     ${destination}
     COMMENT "Generating sphinx documentation: ${builder}"
-    COMMAND cd ${destination} && ln -s ./index_*.html index.html
+    COMMAND cd ${destination} && ln -sf ./index_*.html index.html
     )
 
   set_property(

paddle/gserver/dataproviders/DataProvider.h

@@ -164,15 +164,6 @@ class DataBatch {
     argu.value = value;
     data_.push_back(argu);
   }
-  /**
-   * @brief Append user defined data
-   * @param[in]  ptr     user defined data
-   */
-  void appendUserDefinedPtr(UserDefinedVectorPtr ptr) {
-    Argument argu;
-    argu.udp = ptr;
-    data_.push_back(argu);
-  }
 
   /*
    * @brief Append argument

paddle/gserver/layers/CostLayer.cpp

@@ -192,6 +192,59 @@ void SumOfSquaresCostLayer::backwardImp(Matrix& output,
   outputG.sumOfSquaresBp(output, *label.value);
 }
 
+//
+// class SmoothL1CostLayer
+//
+
+REGISTER_LAYER(smooth_l1, SmoothL1CostLayer);
+
+bool SmoothL1CostLayer::init(const LayerMap& layerMap,
+                             const ParameterMap& parameterMap) {
+  return CostLayer::init(layerMap, parameterMap);
+}
+
+void SmoothL1CostLayer::forwardImp(Matrix& output,
+                                   Argument& label,
+                                   Matrix& target) {
+  MatrixPtr targetCpu, outputCpu, labelCpu;
+  if (useGpu_) {
+    targetCpu =
+        Matrix::create(target.getHeight(), target.getWidth(), false, false);
+    outputCpu =
+        Matrix::create(output.getHeight(), output.getWidth(), false, false);
+    labelCpu = Matrix::create(
+        label.value->getHeight(), label.value->getWidth(), false, false);
+    targetCpu->copyFrom(target);
+    outputCpu->copyFrom(output);
+    labelCpu->copyFrom(*label.value);
+    targetCpu->smoothL1(*outputCpu, *(labelCpu));
+    target.copyFrom(*targetCpu);
+  } else {
+    target.smoothL1(output, *label.value);
+  }
+}
+
+void SmoothL1CostLayer::backwardImp(Matrix& output,
+                                    Argument& label,
+                                    Matrix& outputG) {
+  MatrixPtr outputGCpu, outputCpu, labelCpu;
+  if (useGpu_) {
+    outputGCpu =
+        Matrix::create(outputG.getHeight(), outputG.getWidth(), false, false);
+    outputCpu =
+        Matrix::create(output.getHeight(), output.getWidth(), false, false);
+    labelCpu = Matrix::create(
+        label.value->getHeight(), label.value->getWidth(), false, false);
+    outputGCpu->copyFrom(outputG);
+    outputCpu->copyFrom(output);
+    labelCpu->copyFrom(*label.value);
+    outputGCpu->smoothL1Bp(*outputCpu, *labelCpu);
+    outputG.copyFrom(*outputGCpu);
+  } else {
+    outputG.smoothL1Bp(output, *label.value);
+  }
+}
+
 //
 // class RankingCost
 //

paddle/gserver/layers/CostLayer.h

@@ -159,6 +159,29 @@ class SumOfSquaresCostLayer : public CostLayer {
                    Matrix& outputGrad) override;
 };
 
+/**
+ * This cost layer compute smooth L1 loss for real-valued regression
+ * tasks.
+ * \f[
+ * L =
+ *   (output - label)^2 * 0.5  / -1 < (output - label) < 1 /
+ *   (output - label) - 0.5    / otherwise  /
+ * \f]
+ */
+class SmoothL1CostLayer : public CostLayer {
+public:
+  explicit SmoothL1CostLayer(const LayerConfig& config) : CostLayer(config) {}
+
+  bool init(const LayerMap& layerMap,
+            const ParameterMap& parameterMap) override;
+
+  void forwardImp(Matrix& output, Argument& label, Matrix& cost) override;
+
+  void backwardImp(Matrix& outputValue,
+                   Argument& label,
+                   Matrix& outputGrad) override;
+};
+
 /**
  * A cost layer for learning to rank (LTR) task. This layer contains at leat
  * three inputs.

paddle/gserver/layers/SequencePoolLayer.cpp

@@ -56,17 +56,16 @@ void SequencePoolLayer::forward(PassType passType) {
   CHECK_EQ(newBatchSize_, starts->getSize() - 1);
 
   resetOutput(newBatchSize_, dim);
-  if (type_) {
-    CHECK(input.subSequenceStartPositions)
-        << "when trans_type = seq, input must hasSubseq";
-  }
+
   /* If type_ = kNonSeq, both seq has or not has sub-seq degrade to a non-seq,
    * thus, in this case, output_ has no sequenceStartPositions.
    * If type_ = kSeq, seq has sub-seq degrades to a seq, thus, only in this
    * case, we should compute the new sequenceStartPositions.
   */
   if (type_) {
-    output_.degradeSequence(input, useGpu_);
+    CHECK(input.subSequenceStartPositions)
+        << "when trans_type = seq, input must hasSubseq";
+    output_.degradeSequence(input);
   }
 }
 

paddle/gserver/tests/test_LayerGrad.cpp

@@ -1602,6 +1602,20 @@ TEST(Layer, PadLayer) {
   }
 }
 
+TEST(Layer, smooth_l1) {
+  TestConfig config;
+  config.layerConfig.set_type("smooth_l1");
+
+  config.inputDefs.push_back({INPUT_DATA, "layer_0", 1, 0});
+  config.inputDefs.push_back({INPUT_DATA_TARGET, "layer_1", 1, 0});
+  config.layerConfig.add_inputs();
+  config.layerConfig.add_inputs();
+
+  for (auto useGpu : {false, true}) {
+    testLayerGrad(config, "smooth_l1", 100, false, useGpu, false, 2.0);
+  }
+}
+
 int main(int argc, char** argv) {
   testing::InitGoogleTest(&argc, argv);
   initMain(argc, argv);

paddle/math/Matrix.cpp

@@ -3590,6 +3590,55 @@ void CpuMatrix::sumOfSquaresBp(Matrix& output, Matrix& label) {
   }
 }
 
+void CpuMatrix::smoothL1(Matrix& output, Matrix& label) {
+  CHECK(output.useGpu_ == false && label.useGpu_ == false)
+      << "Matrix type are not equal";
+
+  size_t numSamples = getHeight();
+  size_t dim = output.getWidth();
+  CHECK_EQ(label.getHeight(), numSamples);
+  CHECK_EQ(output.getHeight(), numSamples);
+  CHECK_EQ(label.getWidth(), dim);
+  CHECK_EQ(getWidth(), (size_t)1);
+  real* out = output.getData();
+  real* cost = getData();
+  real* lbl = label.getData();
+
+  for (size_t i = 0; i < numSamples; ++i, out += dim, cost += dim, lbl += dim) {
+    for (size_t j = 0; j < dim; ++j) {
+      cost[j] = std::fabs(out[j] - lbl[j]);
+      if (cost[j] < 1.0)
+        cost[j] = 0.5 * cost[j] * cost[j];
+      else
+        cost[j] = cost[j] - 0.5;
+    }
+  }
+}
+
+void CpuMatrix::smoothL1Bp(Matrix& output, Matrix& label) {
+  CHECK(output.useGpu_ == false && label.useGpu_ == false)
+      << "Matrix type are not equal";
+
+  size_t numSamples = getHeight();
+  size_t dim = output.getWidth();
+  CHECK_EQ(label.getHeight(), numSamples);
+  CHECK_EQ(output.getHeight(), numSamples);
+  CHECK_EQ(label.getWidth(), dim);
+  CHECK_EQ(getWidth(), (size_t)1);
+  real* out = output.getData();
+  real* cost = getData();
+  real* lbl = label.getData();
+
+  // f'(x) = x         if |x| < 1
+  //       = sign(x)   otherwise
+  for (size_t i = 0; i < numSamples; ++i, out += dim, cost += dim, lbl += dim) {
+    for (size_t j = 0; j < dim; ++j) {
+      cost[j] = out[j] - lbl[j];
+      if (std::fabs(cost[j]) >= 1) cost[j] = (0 < cost[j]) - (cost[j] < 0);
+    }
+  }
+}
+
 void CpuMatrix::tanh(Matrix& output) {
   CHECK(isContiguous());
   CHECK(output.isContiguous());

paddle/math/Matrix.h

@@ -783,6 +783,14 @@ class Matrix : public BaseMatrix {
     LOG(FATAL) << "Not implemented";
   }
 
+  virtual void smoothL1(Matrix& output, Matrix& label) {
+    LOG(FATAL) << "Not implemented";
+  }
+
+  virtual void smoothL1Bp(Matrix& outputV, Matrix& label) {
+    LOG(FATAL) << "Not implemented";
+  }
+
   virtual void tanh(Matrix& output) { LOG(FATAL) << "Not implemented"; }
 
   virtual void tanhDerivative(Matrix& output) {
@@ -1720,6 +1728,9 @@ class CpuMatrix : public Matrix {
   /// gradient of sumOfSquares.
   void sumOfSquaresBp(Matrix& outputV, Matrix& label);
 
+  void smoothL1(Matrix& output, Matrix& label);
+  void smoothL1Bp(Matrix& output, Matrix& label);
+
   void tanh(Matrix& output);
   void tanhDerivative(Matrix& output);