Merge branch 'develop' of https://github.com/baidu/Paddle into benchmark_cfg_doc

.travis.yml

@@ -42,7 +42,7 @@ addons:
 before_install:
   - |
     if [ ${JOB} == "BUILD_AND_TEST" ]; then
-      if ! git diff --name-only $TRAVIS_COMMIT_RANGE | grep -qvE '(\.md$)'
+      if ! git diff --name-only $TRAVIS_COMMIT_RANGE | grep -qvE '(\.md$)|(\.rst$)|(\.jpg$)|(\.png$)'
       then
         echo "Only markdown docs were updated, stopping build process."
         exit

CMakeLists.txt

@@ -3,7 +3,7 @@ cmake_minimum_required(VERSION 2.8)
 project(paddle CXX C)
 set(PADDLE_MAJOR_VERSION 0)
 set(PADDLE_MINOR_VERSION 9)
-set(PADDLE_PATCH_VERSION 0a0)
+set(PADDLE_PATCH_VERSION 0)
 set(PADDLE_VERSION ${PADDLE_MAJOR_VERSION}.${PADDLE_MINOR_VERSION}.${PADDLE_PATCH_VERSION})
 
 set(CMAKE_MODULE_PATH ${CMAKE_MODULE_PATH} "${CMAKE_SOURCE_DIR}/cmake")
@@ -36,6 +36,7 @@ option(WITH_RDMA "Compile PaddlePaddle with rdma support" OFF)
 option(WITH_GLOG "Compile PaddlePaddle use glog, otherwise use a log implement internally" ${LIBGLOG_FOUND})
 option(WITH_GFLAGS "Compile PaddlePaddle use gflags, otherwise use a flag implement internally" ${GFLAGS_FOUND})
 option(WITH_TIMER "Compile PaddlePaddle use timer" OFF)
+option(WITH_PROFILER "Compile PaddlePaddle use gpu profiler" OFF)
 option(WITH_TESTING "Compile and run unittest for PaddlePaddle" ${GTEST_FOUND})
 option(WITH_DOC "Compile PaddlePaddle with documentation" OFF)
 option(WITH_SWIG_PY "Compile PaddlePaddle with py PaddlePaddle prediction api" ${SWIG_FOUND})
@@ -115,7 +116,6 @@ else()
     endif(WITH_AVX)
 
     if(WITH_DSO)
-        set(CUDA_LIBRARIES "")
         add_definitions(-DPADDLE_USE_DSO)
     endif(WITH_DSO)
 
@@ -135,6 +135,10 @@ if(NOT WITH_TIMER)
     add_definitions(-DPADDLE_DISABLE_TIMER)
 endif(NOT WITH_TIMER)
 
+if(NOT WITH_PROFILER)
+    add_definitions(-DPADDLE_DISABLE_PROFILER)
+endif(NOT WITH_PROFILER)
+
 if(WITH_AVX)
     set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${AVX_FLAG}")
     set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${AVX_FLAG}")

RELEASE.md

+# Release v0.9.0
+
+## New Features:
+
+* New Layers
+  * bilinear interpolation layer.
+  * spatial pyramid-pool layer.
+  * de-convolution layer.
+  * maxout layer.
+* Support rectangle padding, stride, window and input for Pooling Operation.
+* Add —job=time in trainer, which can be used to print time info without compiler option -WITH_TIMER=ON.
+* Expose cost_weight/nce_layer in `trainer_config_helpers`
+* Add FAQ, concepts, h-rnn docs.
+* Add Bidi-LSTM and DB-LSTM to quick start demo @alvations
+* Add usage track scripts.
+
+## Improvements
+
+* Add Travis-CI for Mac OS X. Enable swig unittest in Travis-CI. Skip Travis-CI when only docs are changed.
+* Add code coverage tools.
+* Refine convolution layer to speedup and reduce GPU memory.
+* Speed up PyDataProvider2
+* Add ubuntu deb package build scripts.
+* Make Paddle use git-flow branching model.
+* PServer support no parameter blocks.
+
+## Bug Fixes
+
+* add zlib link to py_paddle
+* add input sparse data check for sparse layer at runtime
+* Bug fix for sparse matrix multiplication
+* Fix floating-point overflow problem of tanh
+* Fix some nvcc compile options
+* Fix a bug in yield dictionary in DataProvider
+* Fix SRL hang when exit.
+
+# Release v0.8.0beta.1
+New features:
+
+* Mac OSX is supported by source code. #138
+   * Both GPU and CPU versions of PaddlePaddle are supported.
+
+* Support CUDA 8.0
+
+* Enhance `PyDataProvider2`
+   * Add dictionary yield format. `PyDataProvider2` can yield a dictionary with key is data_layer's name, value is features.
+   * Add `min_pool_size` to control memory pool in provider.
+
+* Add `deb` install package & docker image for no_avx machines.
+   * Especially for cloud computing and virtual machines
+
+* Automatically disable `avx` instructions in cmake when machine's CPU don't support `avx` instructions.
+
+* Add Parallel NN api in trainer_config_helpers.
+
+* Add `travis ci` for Github
+
+Bug fixes:
+
+* Several bugs in trainer_config_helpers. Also complete the unittest for trainer_config_helpers
+* Check if PaddlePaddle is installed when unittest.
+* Fix bugs in GTX series GPU
+* Fix bug in MultinomialSampler
+
+Also more documentation was written since last release.
+
+# Release v0.8.0beta.0
+
+PaddlePaddle v0.8.0beta.0 release. The install package is not stable yet and it's a pre-release version.

demo/image_classification/train.sh

@@ -24,7 +24,7 @@ paddle train \
 --test_all_data_in_one_period=1 \
 --use_gpu=1 \
 --trainer_count=1 \
---num_passes=200 \
+--num_passes=300 \
 --save_dir=$output \
 2>&1 | tee $log
 

demo/model_zoo/embedding/pre_DictAndModel.sh

@@ -18,7 +18,5 @@ set -x
 # download the dictionary and pretrained model 
 for file in baidu.dict model_32.emb model_64.emb model_128.emb model_256.emb
 do 
-  # following is the google drive address
-  # you can also directly download from https://pan.baidu.com/s/1o8q577s
-  wget https://www.googledrive.com/host/0B7Q8d52jqeI9ejh6Q1RpMTFQT1k/embedding/$file --no-check-certificate
+  wget http://paddlepaddle.bj.bcebos.com/model_zoo/embedding/$file
 done

demo/model_zoo/resnet/get_model.sh

@@ -24,9 +24,7 @@ echo "Downloading ResNet models..."
 
 for file in resnet_50.tar.gz resnet_101.tar.gz resnet_152.tar.gz mean_meta_224.tar.gz 
 do 
-  # following is the google drive address
-  # you can also directly download from https://pan.baidu.com/s/1o8q577s
-  wget https://www.googledrive.com/host/0B7Q8d52jqeI9ejh6Q1RpMTFQT1k/imagenet/$file --no-check-certificate
+  wget http://paddlepaddle.bj.bcebos.com/model_zoo/imagenet/$file
   tar -xvf $file 
   rm $file
 done

demo/quick_start/data/README.md

+This dataset consists of electronics product reviews associated with
+binary labels (positive/negative) for sentiment classification.
+
+The preprocessed data can be downloaded by script `get_data.sh`.
+The data was derived from reviews_Electronics_5.json.gz at
+
+http://snap.stanford.edu/data/amazon/productGraph/categoryFiles/reviews_Electronics_5.json.gz
+
+If you want to process the raw data, you can use the script `proc_from_raw_data/get_data.sh`.

demo/quick_start/data/get_data.sh

@@ -17,14 +17,11 @@ set -e
 DIR="$( cd "$(dirname "$0")" ; pwd -P )"
 cd $DIR
 
-echo "Downloading Amazon Electronics reviews data..."
-# http://jmcauley.ucsd.edu/data/amazon/
-wget http://snap.stanford.edu/data/amazon/productGraph/categoryFiles/reviews_Electronics_5.json.gz
+# Download the preprocessed data
+wget http://paddlepaddle.bj.bcebos.com/demo/quick_start_preprocessed_data/preprocessed_data.tar.gz
 
-echo "Downloading mosesdecoder..."
-#https://github.com/moses-smt/mosesdecoder
-wget https://github.com/moses-smt/mosesdecoder/archive/master.zip
+# Extract package
+tar zxvf preprocessed_data.tar.gz
 
-unzip master.zip
-rm master.zip
-echo "Done."
+# Remove compressed package
+rm preprocessed_data.tar.gz

demo/quick_start/data/pred.list

-./data/pred.txt

demo/quick_start/data/pred.txt

-the device is cute , but that 's just about all that 's good. the specs are what you 'd expect : it 's a wifi mic , with some noise filter options. the app has the option to upload your baby 's name and photo , which is a cutesy touch. but the app is otherwise unstable and useless unless you upgrade for $ 60 / year.set up involves downloading the app , turning on the mic , switching your phone to the wifi network of the mic , telling the app your wifi settings , switching your wifi back to your home router. the app is then directly connected to your mic.the app is adware ! the main screen says " cry notifications on / off : upgrade to evoz premium and receive a text message of email when your baby is crying " .but the adware points out an important limitation , this monitor is only intended to be used from your home network. if you want to access it remotely , get a webcam. this app would make a lot more sense of the premium features were included with the hardware .
-don 't be fooled by my one star rating. if there was a zero , i would have selected it. this product was a waste of my money.it has never worked like the company said it supposed to. i only have one device , an iphone 4gs. after charging the the iphone mid way , the i.sound portable power max 16,000 mah is completely drained. the led light no longer lit up. when plugging the isound portable power max into a wall outlet to charge , it would charge for about 20-30 minutes and then all four battery led indicator lit up showing a full charge. i would leave it on to charge for the full 8 hours or more but each time with the same result upon using. don 't buy this thing. put your money to good use elsewhere .

demo/quick_start/preprocess.sh → demo/quick_start/data/proc_from_raw_data/get_data.sh

@@ -16,10 +16,26 @@
 # 1. size of pos : neg = 1:1.
 # 2. size of testing set = min(25k, len(all_data) * 0.1), others is traning set.
 # 3. distinct train set and test set.
-# 4. build dict
 
 set -e
 
+DIR="$( cd "$(dirname "$0")" ; pwd -P )"
+cd $DIR
+
+# Download data
+echo "Downloading Amazon Electronics reviews data..."
+# http://jmcauley.ucsd.edu/data/amazon/
+wget http://snap.stanford.edu/data/amazon/productGraph/categoryFiles/reviews_Electronics_5.json.gz
+echo "Downloading mosesdecoder..."
+# https://github.com/moses-smt/mosesdecoder
+wget https://github.com/moses-smt/mosesdecoder/archive/master.zip
+
+unzip master.zip
+rm master.zip
+
+##################
+# Preprocess data 
+echo "Preprocess data..."
 export LC_ALL=C
 UNAME_STR=`uname`
 
@@ -29,11 +45,11 @@ else
   SHUF_PROG='gshuf'
 fi
 
-mkdir -p data/tmp
-python preprocess.py -i data/reviews_Electronics_5.json.gz
+mkdir -p tmp
+python preprocess.py -i reviews_Electronics_5.json.gz
 # uniq and shuffle
-cd data/tmp
-echo 'uniq and shuffle...'
+cd tmp
+echo 'Uniq and shuffle...'
 cat pos_*|sort|uniq|${SHUF_PROG}> pos.shuffed
 cat neg_*|sort|uniq|${SHUF_PROG}> neg.shuffed
 
@@ -53,11 +69,11 @@ cat train.pos train.neg | ${SHUF_PROG} >../train.txt
 cat test.pos test.neg | ${SHUF_PROG} >../test.txt
 
 cd -
-echo 'data/train.txt' > data/train.list
-echo 'data/test.txt' > data/test.list
+echo 'train.txt' > train.list
+echo 'test.txt' > test.list
 
 # use 30k dict
-rm -rf data/tmp
-mv data/dict.txt data/dict_all.txt
-cat data/dict_all.txt | head -n 30001 > data/dict.txt
-echo 'preprocess finished'
+rm -rf tmp
+mv dict.txt dict_all.txt
+cat dict_all.txt | head -n 30001 > dict.txt
+echo 'Done.'

demo/quick_start/preprocess.py → demo/quick_start/data/proc_from_raw_data/preprocess.py

@@ -14,7 +14,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 """
-1. (remove HTML before or not)tokensizing
+1. Tokenize the words and punctuation 
 2. pos sample : rating score 5; neg sample: rating score 1-2.
 
 Usage:
@@ -76,7 +76,11 @@ def tokenize(sentences):
     sentences : a list of input sentences.
     return: a list of processed text.
     """
-    dir = './data/mosesdecoder-master/scripts/tokenizer/tokenizer.perl'
+    dir = './mosesdecoder-master/scripts/tokenizer/tokenizer.perl'
+    if not os.path.exists(dir):
+        sys.exit(
+            "The ./mosesdecoder-master/scripts/tokenizer/tokenizer.perl does not exists."
+        )
     tokenizer_cmd = [dir, '-l', 'en', '-q', '-']
     assert isinstance(sentences, list)
     text = "\n".join(sentences)
@@ -104,7 +108,7 @@ def tokenize_batch(id):
         num_batch, instance, pre_fix = parse_queue.get()
         if num_batch == -1:  ### parse_queue finished
             tokenize_queue.put((-1, None, None))
-            sys.stderr.write("tokenize theread %s finish\n" % (id))
+            sys.stderr.write("Thread %s finish\n" % (id))
             break
         tokenize_instance = tokenize(instance)
         tokenize_queue.put((num_batch, tokenize_instance, pre_fix))

demo/semantic_role_labeling/data/get_data.sh

@@ -14,10 +14,10 @@
 # limitations under the License.
 set -e
 wget http://www.cs.upc.edu/~srlconll/conll05st-tests.tar.gz
-wget https://www.googledrive.com/host/0B7Q8d52jqeI9ejh6Q1RpMTFQT1k/semantic_role_labeling/verbDict.txt --no-check-certificate
-wget https://www.googledrive.com/host/0B7Q8d52jqeI9ejh6Q1RpMTFQT1k/semantic_role_labeling/targetDict.txt --no-check-certificate
-wget https://www.googledrive.com/host/0B7Q8d52jqeI9ejh6Q1RpMTFQT1k/semantic_role_labeling/wordDict.txt --no-check-certificate
-wget https://www.googledrive.com/host/0B7Q8d52jqeI9ejh6Q1RpMTFQT1k/semantic_role_labeling/emb --no-check-certificate
+wget http://paddlepaddle.bj.bcebos.com/demo/srl_dict_and_embedding/verbDict.txt
+wget http://paddlepaddle.bj.bcebos.com/demo/srl_dict_and_embedding/targetDict.txt 
+wget http://paddlepaddle.bj.bcebos.com/demo/srl_dict_and_embedding/wordDict.txt 
+wget http://paddlepaddle.bj.bcebos.com/demo/srl_dict_and_embedding/emb
 tar -xzvf conll05st-tests.tar.gz
 rm conll05st-tests.tar.gz
 cp ./conll05st-release/test.wsj/words/test.wsj.words.gz  .

demo/semantic_role_labeling/dataprovider.py

@@ -25,12 +25,13 @@ def hook(settings, word_dict, label_dict, predicate_dict, **kwargs):
     #all inputs are integral and sequential type
     settings.slots = [
         integer_value_sequence(len(word_dict)),
-        integer_value_sequence(len(predicate_dict)),
         integer_value_sequence(len(word_dict)),
         integer_value_sequence(len(word_dict)),
         integer_value_sequence(len(word_dict)),
         integer_value_sequence(len(word_dict)),
-        integer_value_sequence(len(word_dict)), integer_value_sequence(2),
+        integer_value_sequence(len(word_dict)), 
+        integer_value_sequence(len(predicate_dict)),
+        integer_value_sequence(2),
         integer_value_sequence(len(label_dict))
     ]
 
@@ -63,5 +64,5 @@ def process(settings, file_name):
 
             label_list = label.split()
             label_slot = [settings.label_dict.get(w) for w in label_list]
-            yield word_slot, predicate_slot, ctx_n2_slot, ctx_n1_slot, \
-                  ctx_0_slot, ctx_p1_slot, ctx_p2_slot, mark_slot, label_slot
+            yield word_slot, ctx_n2_slot, ctx_n1_slot, \
+                  ctx_0_slot, ctx_p1_slot, ctx_p2_slot, predicate_slot, mark_slot, label_slot

demo/semantic_role_labeling/predict.py

@@ -55,18 +55,14 @@ class Prediction():
 
         slots = [
             integer_value_sequence(len_dict),
-            integer_value_sequence(len_pred),
             integer_value_sequence(len_dict),
             integer_value_sequence(len_dict),
             integer_value_sequence(len_dict),
             integer_value_sequence(len_dict),
             integer_value_sequence(len_dict), 
+            integer_value_sequence(len_pred),
             integer_value_sequence(2)
             ]
-            integer_value_sequence(len_dict), integer_value_sequence(len_dict),
-            integer_value_sequence(len_dict), integer_value_sequence(len_dict),
-            integer_value_sequence(len_dict), integer_value_sequence(2)
-        ]
         self.converter = DataProviderConverter(slots)
 
     def load_dict_label(self, dict_file, label_file, predicate_dict_file):
@@ -104,8 +100,8 @@ class Prediction():
                 marks = mark.split()
                 mark_slot = [int(w) for w in marks]
                 
-                yield word_slot, predicate_slot, ctx_n2_slot, ctx_n1_slot, \
-                      ctx_0_slot, ctx_p1_slot, ctx_p2_slot,  mark_slot
+                yield word_slot, ctx_n2_slot, ctx_n1_slot, \
+                      ctx_0_slot, ctx_p1_slot, ctx_p2_slot, predicate_slot, mark_slot
 
     def predict(self, data_file, output_file):
         """

demo/semantic_role_labeling/predict.sh

@@ -18,7 +18,7 @@ set -e
 function get_best_pass() {
   cat $1  | grep -Pzo 'Test .*\n.*pass-.*' | \
   sed  -r 'N;s/Test.* cost=([0-9]+\.[0-9]+).*\n.*pass-([0-9]+)/\1 \2/g' | \
-  sort | head -n 1
+  sort -n | head -n 1
 }   
 
 log=train.log

demo/semantic_role_labeling/test.sh

@@ -18,7 +18,7 @@ set -e
 function get_best_pass() {
   cat $1  | grep -Pzo 'Test .*\n.*pass-.*' | \
   sed  -r 'N;s/Test.* cost=([0-9]+\.[0-9]+).*\n.*pass-([0-9]+)/\1 \2/g' |\
-  sort | head -n 1
+  sort -n | head -n 1
 }
 
 log=train.log

demo/sentiment/test.sh

@@ -17,7 +17,7 @@ set -e
 function get_best_pass() {
   cat $1  | grep -Pzo 'Test .*\n.*pass-.*' | \
   sed  -r 'N;s/Test.* classification_error_evaluator=([0-9]+\.[0-9]+).*\n.*pass-([0-9]+)/\1 \2/g' |\
-  sort | head -n 1
+  sort -n | head -n 1
 }
 
 log=train.log

demo/seqToseq/data/paraphrase_data.sh

@@ -16,9 +16,7 @@ set -e
 set -x
 
 # download the in-house paraphrase dataset
-# following is the google drive address
-# you can also directly download from https://pan.baidu.com/s/1o8q577s
-wget https://www.googledrive.com/host/0B7Q8d52jqeI9ejh6Q1RpMTFQT1k/embedding/paraphrase.tar.gz --no-check-certificate
+wget http://paddlepaddle.bj.bcebos.com/model_zoo/embedding/paraphrase.tar.gz
 
 # untar the dataset
 tar -zxvf paraphrase.tar.gz

demo/seqToseq/data/wmt14_model.sh

@@ -16,9 +16,7 @@ set -e
 set -x
 
 # download the pretrained model
-# following is the google drive address
-# you can also directly download from https://pan.baidu.com/s/1o8q577s
-wget https://www.googledrive.com/host/0B7Q8d52jqeI9ejh6Q1RpMTFQT1k/wmt14_model.tar.gz --no-check-certificate
+wget http://paddlepaddle.bj.bcebos.com/model_zoo/wmt14_model.tar.gz
 
 # untar the model
 tar -zxvf wmt14_model.tar.gz

doc/about/index.rst

+ABOUT
+=======
+
+PaddlPaddle is an easy-to-use, efficient, flexible and scalable deep learning platform,
+which is originally developed by Baidu scientists and engineers for the purpose of applying deep learning to many products at Baidu.
+
+PaddlePaddle is now open source but far from complete, which is intended to be built upon, improved, scaled, and extended.
+We hope to build an active open source community both by providing feedback and by actively contributing to the source code.
+
+
+Credits
+--------
+
+We owe many thanks to `all contributors and developers <https://github.com/PaddlePaddle/Paddle/blob/develop/authors>`_ of PaddlePaddle!

doc/algorithm/rnn/bi_lstm.jpg

-../../demo/sentiment_analysis/bi_lstm.jpg
\ No newline at end of file

doc/algorithm/rnn/encoder-decoder-attention-model.png

-../../demo/text_generation/encoder-decoder-attention-model.png
\ No newline at end of file

doc/ui/data_provider/index.rst → doc/api/data_provider/index.rst

-DataProvider Introduction
-=========================
+Introduction
+==============
 DataProvider is a module that loads training or testing data into cpu or gpu
 memory for the following triaining or testing process.
 

doc/ui/data_provider/pydataprovider2.rst → doc/api/data_provider/pydataprovider2.rst

-How to use PyDataProvider2
-==========================
+PyDataProvider2
+=================
 
 We highly recommand users to use PyDataProvider2 to provide training or testing
 data to PaddlePaddle. The user only needs to focus on how to read a single

doc/api/index.rst

+API
+====
+
+DataProvider API
+----------------
+
+..  toctree::
+  :maxdepth: 1
+
+  data_provider/index.rst
+  data_provider/pydataprovider2.rst
+
+Model Config API
+----------------
+
+..  toctree::
+  :maxdepth: 1
+
+  trainer_config_helpers/index.rst
+  trainer_config_helpers/optimizers.rst
+  trainer_config_helpers/data_sources.rst
+  trainer_config_helpers/layers.rst
+  trainer_config_helpers/activations.rst 
+  trainer_config_helpers/poolings.rst
+  trainer_config_helpers/networks.rst
+  trainer_config_helpers/evaluators.rst
+  trainer_config_helpers/attrs.rst
+
+
+Applications API
+----------------
+
+..  toctree::
+  :maxdepth: 1
+
+  predict/swig_py_paddle_en.rst
\ No newline at end of file

doc/ui/predict/swig_py_paddle_en.rst → doc/api/predict/swig_py_paddle_en.rst

-Python Prediction API
-=====================
+Python Prediction
+==================
 
 PaddlePaddle offers a set of clean prediction interfaces for python with the help of
 SWIG. The main steps of predict values in python are:

doc/ui/api/trainer_config_helpers/attrs.rst → doc/api/trainer_config_helpers/attrs.rst

-Parameter and Extra Layer Attribute
-===================================
+Parameter Attributes
+=======================
 
 ..  automodule:: paddle.trainer_config_helpers.attrs
     :members:

doc/cluster/index.rst

-Cluster Train
-====================
-
-.. toctree::
-  :glob:
-
-  opensource/cluster_train.md
-  internal/index.md

doc/dev/index.rst

-Development Guide
-=================
-
-..  toctree::
-  :maxdepth: 1
-
-  layer.md
-  new_layer/new_layer.rst
-  ../source/index.md

doc/dev/layer.md

-# Layer Documents
-
-* [Layer Source Code Document](../source/gserver/layers/index.rst)
-* [Layer Python API Document](../ui/api/trainer_config_helpers/index.rst)

doc/introduction/index.md → doc/getstarted/basic_usage/basic_usage.rst

-# Introduction
+Basic Usage
+=============
 
 PaddlePaddle is a deep learning platform open-sourced by Baidu. With PaddlePaddle, you can easily train a classic neural network within a couple lines of configuration, or you can build sophisticated models that provide state-of-the-art performance on difficult learning tasks like sentiment analysis, machine translation, image caption and so on.
 
-## 1. A Classic Problem
+1. A Classic Problem
+---------------------
 
-Now, to give you a hint of what using PaddlePaddle looks like, let's start with a fundamental learning problem - <a href="https://en.wikipedia.org/wiki/Simple_linear_regression">**simple linear regression**</a> : you have observed a set of two-dimensional data points of `X` and `Y`, where `X` is an explanatory variable and `Y` is corresponding dependent variable, and you want to recover the underlying correlation between `X` and `Y`. Linear regression can be used in many practical scenarios. For example, `X` can be a variable about house size, and `Y` a variable about house price. You can build a model that captures relationship between them by observing real estate markets.
+Now, to give you a hint of what using PaddlePaddle looks like, let's start with a fundamental learning problem - `simple linear regression <https://en.wikipedia.org/wiki/Simple_linear_regression>`_: you have observed a set of two-dimensional data points of ``X`` and ``Y``, where ``X`` is an explanatory variable and ``Y`` is corresponding dependent variable, and you want to recover the underlying correlation between ``X`` and ``Y``. Linear regression can be used in many practical scenarios. For example, ``X`` can be a variable about house size, and ``Y`` a variable about house price. You can build a model that captures relationship between them by observing real estate markets.
 
-## 2. Prepare the Data
+2. Prepare the Data
+--------------------
 
-Suppose the true relationship can be characterized as `Y = 2X + 0.3`, let's see how to recover this pattern only from observed data. Here is a piece of python code that feeds synthetic data to PaddlePaddle. The code is pretty self-explanatory, the only extra thing you need to add for PaddlePaddle is a definition of input data types.
+Suppose the true relationship can be characterized as ``Y = 2X + 0.3``, let's see how to recover this pattern only from observed data. Here is a piece of python code that feeds synthetic data to PaddlePaddle. The code is pretty self-explanatory, the only extra thing you need to add for PaddlePaddle is a definition of input data types.
 
-```python
-# dataprovider.py
-from paddle.trainer.PyDataProvider2 import *
-import random
+    .. code-block:: python
 
-# define data types of input: 2 real numbers
-@provider(input_types=[dense_vector(1), dense_vector(1)],use_seq=False)
-def process(settings, input_file):
-    for i in xrange(2000):
-        x = random.random()
-        yield [x], [2*x+0.3]
-```
+        # dataprovider.py
+        from paddle.trainer.PyDataProvider2 import *
+        import random
 
-## 3. Train a NeuralNetwork in PaddlePaddle
+        # define data types of input: 2 real numbers
+        @provider(input_types=[dense_vector(1), dense_vector(1)],use_seq=False)
+        def process(settings, input_file):
+            for i in xrange(2000):
+                x = random.random()
+                yield [x], [2*x+0.3]
 
-To recover this relationship between `X` and `Y`, we use a neural network with one layer of linear activation units and a square error cost layer. Don't worry if you are not familiar with these terminologies, it's just saying that we are starting from a random line `Y' = wX + b` , then we gradually adapt `w` and `b` to minimize the difference between `Y'` and `Y`. Here is what it looks like in PaddlePaddle:
+3. Train a NeuralNetwork
+-------------------------
 
-```python
-# trainer_config.py
-from paddle.trainer_config_helpers import *
+To recover this relationship between ``X`` and ``Y``, we use a neural network with one layer of linear activation units and a square error cost layer. Don't worry if you are not familiar with these terminologies, it's just saying that we are starting from a random line ``Y' = wX + b`` , then we gradually adapt ``w`` and ``b`` to minimize the difference between ``Y'`` and ``Y``. Here is what it looks like in PaddlePaddle:
 
-# 1. read data. Suppose you saved above python code as dataprovider.py
-data_file = 'empty.list'
-with open(data_file, 'w') as f: f.writelines(' ')
-define_py_data_sources2(train_list=data_file, test_list=None, 
-        module='dataprovider', obj='process',args={})
+    .. code-block:: python
 
-# 2. learning algorithm
-settings(batch_size=12, learning_rate=1e-3, learning_method=MomentumOptimizer())
+        # trainer_config.py
+        from paddle.trainer_config_helpers import *
 
-# 3. Network configuration
-x = data_layer(name='x', size=1)
-y = data_layer(name='y', size=1)
-y_predict = fc_layer(input=x, param_attr=ParamAttr(name='w'), size=1, act=LinearActivation(), bias_attr=ParamAttr(name='b'))
-cost = regression_cost(input=y_predict, label=y)
-outputs(cost)
-```
+        # 1. read data. Suppose you saved above python code as dataprovider.py
+        data_file = 'empty.list'
+        with open(data_file, 'w') as f: f.writelines(' ')
+        define_py_data_sources2(train_list=data_file, test_list=None, 
+                module='dataprovider', obj='process',args={})
+
+        # 2. learning algorithm
+        settings(batch_size=12, learning_rate=1e-3, learning_method=MomentumOptimizer())
+
+        # 3. Network configuration
+        x = data_layer(name='x', size=1)
+        y = data_layer(name='y', size=1)
+        y_predict = fc_layer(input=x, param_attr=ParamAttr(name='w'), size=1, act=LinearActivation(), bias_attr=ParamAttr(name='b'))
+        cost = regression_cost(input=y_predict, label=y)
+        outputs(cost)
 
 Some of the most fundamental usages of PaddlePaddle are demonstrated:
 
@@ -55,46 +59,51 @@ Some of the most fundamental usages of PaddlePaddle are demonstrated:
 -  The second part describes learning algorithm. It defines in what ways adjustments are made to model parameters. PaddlePaddle provides a rich set of optimizers, but a simple momentum based optimizer will suffice here, and it processes 12 data points each time.
 
 -  Finally, the network configuration. It usually is as simple as "stacking" layers. Three kinds of layers are used in this configuration:
-	-  **Data Layer**: a network always starts with one or more data layers. They provide input data to the rest of the network. In this problem, two data layers are used respectively for `X` and `Y`.
+	-  **Data Layer**: a network always starts with one or more data layers. They provide input data to the rest of the network. In this problem, two data layers are used respectively for ``X`` and ``Y``.
 	-  **FC Layer**: FC layer is short for Fully Connected Layer, which connects all the input units to current layer and does the actual computation specified as activation function. Computation layers like this are the fundamental building blocks of a deeper model.
 	-  **Cost Layer**: in training phase, cost layers are usually the last layers of the network. They measure the performance of current model, and provide guidence to adjust parameters.
 
 Now that everything is ready, you can train the network with a simple command line call:
- ```
- paddle train --config=trainer_config.py --save_dir=./output --num_passes=30
- ```
 
-This means that PaddlePaddle will train this network on the synthectic dataset for 30 passes, and save all the models under path `./output`. You will see from the messages printed out during training phase that the model cost is decreasing as time goes by, which indicates we are getting a closer guess.
+    .. code-block:: bash
+ 
+        paddle train --config=trainer_config.py --save_dir=./output --num_passes=30
+ 
+
+This means that PaddlePaddle will train this network on the synthectic dataset for 30 passes, and save all the models under path ``./output``. You will see from the messages printed out during training phase that the model cost is decreasing as time goes by, which indicates we are getting a closer guess.
+
 
+4. Evaluate the Model
+-----------------------
 
-## 4. Evaluate the Model
+Usually, a different dataset that left out during training phase should be used to evalute the models. However, we are lucky enough to know the real answer: ``w=2, b=0.3``, thus a better option is to check out model parameters directly.
 
-Usually, a different dataset that left out during training phase should be used to evalute the models. However, we are lucky enough to know the real answer: `w=2, b=0.3`, thus a better option is to check out model parameters directly.
+In PaddlePaddle, training is just to get a collection of model parameters, which are ``w`` and ``b`` in this case. Each parameter is saved in an individual file in the popular ``numpy`` array format. Here is the code that reads parameters from last pass.
 
-In PaddlePaddle, training is just to get a collection of model parameters, which are `w` and `b` in this case. Each parameter is saved in an individual file in the popular `numpy` array format. Here is the code that reads parameters from last pass.
+    .. code-block:: python
 
-```python
-import numpy as np
-import os
+        import numpy as np
+        import os
 
-def load(file_name):
-    with open(file_name, 'rb') as f:
-        f.read(16) # skip header for float type.
-        return np.fromfile(f, dtype=np.float32)
-        
-print 'w=%.6f, b=%.6f' % (load('output/pass-00029/w'), load('output/pass-00029/b'))
-# w=1.999743, b=0.300137
-```
+        def load(file_name):
+            with open(file_name, 'rb') as f:
+                f.read(16) # skip header for float type.
+                return np.fromfile(f, dtype=np.float32)
+                
+        print 'w=%.6f, b=%.6f' % (load('output/pass-00029/w'), load('output/pass-00029/b'))
+        # w=1.999743, b=0.300137
 
-<center> ![](./parameters.png) </center>
+    .. image:: parameters.png
+        :align: center
 
-Although starts from a random guess, you can see that value of `w` changes quickly towards 2 and `b` changes quickly towards 0.3. In the end, the predicted line is almost identical with real answer.
+Although starts from a random guess, you can see that value of ``w`` changes quickly towards 2 and ``b`` changes quickly towards 0.3. In the end, the predicted line is almost identical with real answer.
 
-There, you have recovered the underlying pattern between `X` and `Y` only from observed data.
+There, you have recovered the underlying pattern between ``X`` and ``Y`` only from observed data.
 
 
-## 5. Where to Go from Here
+5. Where to Go from Here
+-------------------------
 
-- <a href="../build/index.html"> Build and Installation </a>
-- <a href="../demo/quick_start/index_en.html">Quick Start</a>
-- <a href="../demo/index.html">Example and Demo</a>
+- `Install and Build <../build_and_install/index.html>`_
+- `Tutorials <../demo/quick_start/index_en.html>`_
+- `Example and Demo <../demo/index.html>`_

doc/build/build_from_source.md → doc/getstarted/build_and_install/build_from_source.md

@@ -95,7 +95,7 @@ As a simple example, consider the following:
     ```bash
     # necessary
     sudo apt-get update
-    sudo apt-get install -y g++ make cmake build-essential libatlas-base-dev python python-pip libpython-dev m4 libprotobuf-dev protobuf-compiler python-protobuf python-numpy git
+    sudo apt-get install -y g++ make cmake swig build-essential libatlas-base-dev python python-pip libpython-dev m4 libprotobuf-dev protobuf-compiler python-protobuf python-numpy git
     # optional
     sudo apt-get install libgoogle-glog-dev
     sudo apt-get install libgflags-dev
@@ -149,15 +149,15 @@ If still not found, you can manually set it based on CMake error information fro
 
 As a simple example, consider the following:
 
-- **Only CPU**
+- **Only CPU with swig**
 
   ```bash
-  cmake  .. -DWITH_GPU=OFF
+  cmake  .. -DWITH_GPU=OFF -DWITH_SWIG_PY=ON
   ```
-- **GPU**
+- **GPU with swig**
 
   ```bash
-  cmake .. -DWITH_GPU=ON
+  cmake .. -DWITH_GPU=ON -DWITH_SWIG_PY=ON
   ```
 
 - **GPU with doc and swig**
@@ -170,15 +170,13 @@ Finally, you can build PaddlePaddle:
 
 ```bash
 # you can add build option here, such as:    
-cmake .. -DWITH_GPU=ON -DCMAKE_INSTALL_PREFIX=<path to install>
+cmake .. -DWITH_GPU=ON -DCMAKE_INSTALL_PREFIX=<path to install> -DWITH_SWIG_PY=ON
 # please use sudo make install, if you want to install PaddlePaddle into the system
 make -j `nproc` && make install
 # set PaddlePaddle installation path in ~/.bashrc
 export PATH=<path to install>/bin:$PATH
 ```
 
-**Note:**
-
 If you set `WITH_SWIG_PY=ON`, related python dependencies also need to be installed.
 Otherwise, PaddlePaddle will automatically install python dependencies
 at first time when user run paddle commands, such as `paddle version`, `paddle train`.

doc/build/index.rst → doc/getstarted/build_and_install/index.rst

@@ -8,8 +8,6 @@ Install PaddlePaddle
     :maxdepth: 1
     :glob:
 
-    install_*
-    internal/install_from_jumbo.md
     docker_install.rst
     ubuntu_install.rst
 
@@ -24,5 +22,4 @@ Build from Source
     :maxdepth: 1
     :glob:
 
-    build_from_source.md
-    contribute_to_paddle.md
+    build_from_source.md
\ No newline at end of file

doc/getstarted/index.rst

+GET STARTED
+============
+
+..  toctree::
+  :maxdepth: 2
+
+  build_and_install/index.rst
+  basic_usage/basic_usage.rst

doc/cluster/opensource/cluster_train.md → doc/howto/cluster/cluster_train.md

-# Distributed Training
+# How to Run Distributed Training
 
 In this article, we explain how to run distributed Paddle training jobs on clusters.  We will create the distributed version of the single-process training example, [recommendation](https://github.com/baidu/Paddle/tree/develop/demo/recommendation).
 
@@ -9,7 +9,7 @@ In this article, we explain how to run distributed Paddle training jobs on clust
 1. Aforementioned scripts use a Python library [fabric](http://www.fabfile.org/) to run SSH commands.  We can use `pip` to install fabric:
 
    ```bash
-pip install fabric
+   pip install fabric
    ```
 
 1. We need to install PaddlePaddle on all nodes in the cluster.  To enable GPUs, we need to install CUDA in `/usr/local/cuda`; otherwise Paddle would report errors at runtime.

doc/howto/cmd_parameter/index.md

+# How to Set Command-line Parameters
+
+* [Use Case](use_case.md)
+* [Arguments](arguments.md)
+* [Detailed Descriptions](detail_introduction.md)

doc/build/contribute_to_paddle.md → doc/howto/contribute_to_paddle.md

-# Contribute Code
+# How to Contribute Code
 
 We sincerely appreciate your contributions. You can use fork and pull request
 workflow to merge your code. 

doc/algorithm/index.rst → doc/howto/deep_model/index.rst

-Algorithm Tutorial
-==================
+How to Configure Deep Models
+============================
 
 ..  toctree::
   :maxdepth: 1

doc/algorithm/rnn/rnn.rst → doc/howto/deep_model/rnn/rnn.rst

@@ -42,7 +42,7 @@ Simple Gated Recurrent Neural Network
 
 Recurrent neural network process a sequence at each time step sequentially. An example of the architecture of LSTM is listed below.
 
-.. image:: ./bi_lstm.jpg
+.. image:: ../../../tutorials/sentiment_analysis/bi_lstm.jpg
 	 :align: center
 
 Generally speaking, a recurrent network perform the following operations from :math:`t=1` to :math:`t=T`, or reversely from :math:`t=T` to :math:`t=1`.
@@ -101,7 +101,7 @@ Sequence to Sequence Model with Attention
 -----------------------------------------
 We will use the sequence to sequence model with attention as an example to demonstrate how you can configure complex recurrent neural network models. An illustration of the sequence to sequence model with attention is shown in the following figure.
 
-.. image:: ./encoder-decoder-attention-model.png
+.. image:: ../../../tutorials/text_generation/encoder-decoder-attention-model.png
  	 :align: center
 
 In this model, the source sequence :math:`S = \{s_1, \dots, s_T\}` is encoded with a bidirectional gated recurrent neural networks. The hidden states of the bidirectional gated recurrent neural network :math:`H_S = \{H_1, \dots, H_T\}` is called *encoder vector* The decoder is a gated recurrent neural network. When decoding each token :math:`y_t`, the gated recurrent neural network generates a set of weights :math:`W_S^t = \{W_1^t, \dots, W_T^t\}`, which are used to compute a weighted sum of the encoder vector. The weighted sum of the encoder vector is utilized to condition the generation of the token :math:`y_t`.

doc/howto/index.rst

+HOW TO
+=======
+
+Usage
+-------
+
+..  toctree::
+  :maxdepth: 1
+
+  cmd_parameter/index.md
+  deep_model/index.rst
+  cluster/cluster_train.md
+
+Development
+------------
+
+..  toctree::
+  :maxdepth: 1
+
+  new_layer/index.rst
+  contribute_to_paddle.md
+
+Optimization
+-------------
+
+..  toctree::
+  :maxdepth: 1
+
+  optimization/index.rst

doc/dev/new_layer/new_layer.rst → doc/howto/new_layer/index.rst

-==================
-Writing New Layers
-==================
+=======================
+How to Write New Layers
+=======================
 
 This tutorial will guide you to write customized layers in PaddlePaddle. We will utilize fully connected layer as an example to guide you through the following steps for writing a new layer.
 

doc/howto/optimization/gpu_profiling.rst

+Profiling on PaddlePaddle
+=========================
+
+This tutorial will guide you step-by-step through how to conduct profiling and performance tuning using built-in timer, **nvprof** and **nvvp**.
+
+- What is profiling?
+- Why we need profiling?
+- How to do profiling?
+- Profile tools
+- Hands-on Tutorial
+- Profiling tips
+
+What's profiling?
+=================
+In software engineering, profiling is a form of dynamic program analysis that measures the space (memory) or time
+complexity of a program, the usage of particular instructions, or the frequency and duration of function calls.
+Most commonly, profiling information serves to aid program optimization.
+
+Briefly, profiler is used to measure application performance. Program analysis tools are extremely important for
+understanding program behavior. Simple profiling can tell you that how long does an operation take? For advanced
+profiling, it can interpret why does an operation take a long time?
+
+Why we need profiling?
+======================
+Since training deep neural network typically take a very long time to get over, performance is gradually becoming
+the most important thing in deep learning field. The first step to improve performance is to understand what parts
+are slow.  There is no point in improving performance of a region which doesn’t take much time!
+
+
+How to do profiling?
+====================
+To achieve maximum performance, there are five steps you can take to reach your goals.
+
+- Profile the code
+- Find the slow parts
+- Work out why they’re slow
+- Make them fast
+- Profile the code again
+
+Usually, processor has two key performance limits include float point throughput and
+memory throughput. For GPU,  it also need more parallelism to fulfill its potential.
+This is why they can be so fast.
+
+Profiler Tools
+==============
+For general GPU profiling, a bunch of tools are provided from both NVIDIA and third party.
+
+**nvprof** is Nvidia profiler and **nvvp** is (GUI based) Nvidia visual profiler.
+In this tutorial, we will focus on nvprof and nvvp.
+
+:code:`test_GpuProfiler` from :code:`paddle/math/tests` directory will be used to evaluate
+above profilers. 
+
+.. literalinclude:: ../../paddle/math/tests/test_GpuProfiler.cpp
+   :language: c++
+   :lines: 111-124
+   :linenos:
+
+The above code snippet includes two methods, you can use any of them to profile the regions of interest.
+
+1. :code:`REGISTER_TIMER_INFO` is a built-in timer wrapper which can calculate the time overhead of both cpu functions and cuda kernels.
+
+2. :code:`REGISTER_GPU_PROFILER` is a general purpose wrapper object of :code:`cudaProfilerStart` and :code:`cudaProfilerStop` to avoid
+program crashes when CPU version of PaddlePaddle invokes them.
+
+You can find more details about how to use both of them in the next session.
+
+Hands-on Approach
+=================
+
+Built-in Timer
+--------------
+
+To enable built-in timer in PaddlePaddle, first you have to add :code:`REGISTER_TIMER_INFO` into the regions of you interest.
+Then, all information could be stamped in the console via :code:`printStatus` or :code:`printAllStatus` function.
+As a simple example, consider the following:
+
+1. Add :code:`REGISTER_TIMER_INFO` and :code:`printAllStatus` functions (see the emphasize-lines).
+
+    .. literalinclude:: ../../paddle/math/tests/test_GpuProfiler.cpp
+        :language: c++
+        :lines: 111-124
+        :emphasize-lines: 8-10,13
+        :linenos:
+
+2. Configure cmake with **WITH_TIMER** and recompile PaddlePaddle.
+
+    .. code-block:: bash
+
+        cmake .. -DWITH_TIMER=ON
+        make
+
+3. Execute your code and observe the results (see the emphasize-lines). 
+
+    .. code-block:: bash
+        :emphasize-lines: 1,12-15
+
+        > ./paddle/math/tests/test_GpuProfiler                                                                             
+        I1117 11:13:42.313065 2522362816 Util.cpp:155] commandline: ./paddle/math/tests/test_GpuProfiler                                             
+        I1117 11:13:42.845065 2522362816 Util.cpp:130] Calling runInitFunctions                                                                      
+        I1117 11:13:42.845208 2522362816 Util.cpp:143] Call runInitFunctions done.                                                                   
+        [==========] Running 1 test from 1 test case.                                                                                                
+        [----------] Global test environment set-up.                                                                                                 
+        [----------] 1 test from Profiler                                                                                                            
+        [ RUN      ] Profiler.BilinearFwdBwd                                                                                                         
+        I1117 11:13:42.845310 2522362816 test_GpuProfiler.cpp:114] Enable GPU Profiler Stat: [testBilinearFwdBwd] "numSamples = 10, channels = 16, im
+        gSizeX = 64, imgSizeY = 64"                                                                                                                  
+        I1117 11:13:42.850154 2522362816 ThreadLocal.cpp:37] thread use undeterministic rand seed:20659751                                           
+        I1117 11:13:42.981501 2522362816 Stat.cpp:130] ======= StatSet: [GlobalStatInfo] status ======                                               
+        I1117 11:13:42.981539 2522362816 Stat.cpp:133] Stat=testBilinearFwdBwd     total=136.141    avg=136.141    max=136.141    min=136.141   count=1                                                                                                                                  
+        I1117 11:13:42.981572 2522362816 Stat.cpp:141] ======= BarrierStatSet status ======                                                          
+        I1117 11:13:42.981575 2522362816 Stat.cpp:154] --------------------------------------------------                                            
+        [       OK ] Profiler.BilinearFwdBwd (136 ms)                                                                                                
+        [----------] 1 test from Profiler (136 ms total)                                                                                             
+                                                                                                                                                    
+        [----------] Global test environment tear-down                                                                                               
+        [==========] 1 test from 1 test case ran. (136 ms total)                                                                                     
+        [  PASSED  ] 1 test.
+
+nvprof profiler
+---------------
+
+To use this command line profiler **nvprof**, you can simply issue the following command:
+
+1. Add :code:`REGISTER_GPU_PROFILER` function (see the emphasize-lines).
+
+    .. literalinclude:: ../../paddle/math/tests/test_GpuProfiler.cpp
+        :language: c++
+        :lines: 111-124
+        :emphasize-lines: 6-7
+        :linenos:
+
+2. Configure cmake with **WITH_PROFILER** and recompile PaddlePaddle.
+
+    .. code-block:: bash
+
+        cmake .. -DWITH_PROFILER=ON
+        make
+
+3. Use Nvidia profiler **nvprof** to profile the binary.
+
+    .. code-block:: bash
+
+        nvprof  ./paddle/math/tests/test_GpuProfiler
+
+Then, you can get the following profiling result:
+
+.. code-block:: bash
+
+    ==78544== Profiling application: ./paddle/math/tests/test_GpuProfiler                                                                                                      
+    ==78544== Profiling result:                                                                                                                                                
+    Time(%)     Time     Calls       Avg       Min       Max  Name                                                                                                            
+    27.60%  9.6305ms         5  1.9261ms  3.4560us  6.4035ms  [CUDA memcpy HtoD]                                                                                              
+    26.07%  9.0957ms         1  9.0957ms  9.0957ms  9.0957ms  KeBilinearInterpBw                                                                                            
+    23.78%  8.2977ms         1  8.2977ms  8.2977ms  8.2977ms  KeBilinearInterpFw                                                                                        
+    22.55%  7.8661ms         2  3.9330ms  1.5798ms  6.2863ms  [CUDA memcpy DtoH]                                                                                              
+                                                                                                                                                                            
+    ==78544== API calls:                                                                                                                                                       
+    Time(%)     Time     Calls       Avg       Min       Max  Name                                                                                                            
+    46.85%  682.28ms         8  85.285ms  12.639us  682.03ms  cudaStreamCreateWithFlags                                                                                       
+    39.83%  580.00ms         4  145.00ms     302ns  550.27ms  cudaFree                                                                                                        
+    9.82%   143.03ms         9  15.892ms  8.7090us  142.78ms  cudaStreamCreate                                                                                                
+    1.23%   17.983ms         7  2.5690ms  23.210us  6.4563ms  cudaMemcpy                                                                                                      
+    1.23%   17.849ms         2  8.9247ms  8.4726ms  9.3768ms  cudaStreamSynchronize                                                                                           
+    0.66%   9.5969ms         7  1.3710ms  288.43us  2.4279ms  cudaHostAlloc                                                                                                   
+    0.13%   1.9530ms        11  177.54us  7.6810us  591.06us  cudaMalloc                                                                                                      
+    0.07%   1.0424ms         8  130.30us  1.6970us  453.72us  cudaGetDevice                                                                                                   
+    0.04%   527.90us        40  13.197us     525ns  253.99us  cudaEventCreateWithFlags                                                                                        
+    0.03%   435.73us       348  1.2520us     124ns  42.704us  cuDeviceGetAttribute                                                                                            
+    0.03%   419.36us         1  419.36us  419.36us  419.36us  cudaGetDeviceCount                                                                                              
+    0.02%   260.75us         2  130.38us  129.32us  131.43us  cudaGetDeviceProperties                                                                                         
+    0.02%   222.32us         2  111.16us  106.94us  115.39us  cudaLaunch                                                                                                      
+    0.01%   214.06us         4  53.514us  28.586us  77.655us  cuDeviceGetName                                                                                                 
+    0.01%   115.45us         4  28.861us  9.8250us  44.526us  cuDeviceTotalMem                                                                                                
+    0.01%   83.988us         4  20.997us     578ns  77.760us  cudaSetDevice                                                                                                   
+    0.00%   38.918us         1  38.918us  38.918us  38.918us  cudaEventCreate                                                                                                 
+    0.00%   34.573us        31  1.1150us     279ns  12.784us  cudaDeviceGetAttribute                                                                                          
+    0.00%   17.767us         1  17.767us  17.767us  17.767us  cudaProfilerStart                                                                                               
+    0.00%   15.228us         2  7.6140us  3.5460us  11.682us  cudaConfigureCall                                                                                               
+    0.00%   14.536us         2  7.2680us  1.1490us  13.387us  cudaGetLastError                                                                                                
+    0.00%   8.6080us        26     331ns     173ns     783ns  cudaSetupArgument                                                                                               
+    0.00%   5.5470us         6     924ns     215ns  2.6780us  cuDeviceGet                                                                                                     
+    0.00%   5.4090us         6     901ns     328ns  3.3320us  cuDeviceGetCount                                                                                                
+    0.00%   4.1770us         3  1.3920us  1.0630us  1.8300us  cuDriverGetVersion                                                                                              
+    0.00%   3.4650us         3  1.1550us  1.0810us  1.2680us  cuInit                                                                                                          
+    0.00%      830ns         1     830ns     830ns     830ns  cudaRuntimeGetVersion
+
+
+nvvp profiler
+-------------
+
+For visual profiler **nvvp**, you can either import the output of :code:`nvprof –o ...` or
+run application through GUI.
+
+**Note: nvvp also support CPU profiling** (Click the box in nvvp to enable profile execution on CPU).
+
+..  image:: nvvp1.png
+    :align: center
+    :scale: 33%
+
+From the perspective of kernel functions, **nvvp** can even illustrate why does an operation take a long time?
+As shown in the following figure, kernel's block usage, register usage and shared memory usage from :code:`nvvp`
+allow us to fully utilize all warps on the GPU.
+
+..  image:: nvvp2.png
+    :align: center
+    :scale: 33%
+
+From the perspective of application, **nvvp** can give you some suggestions to address performance bottleneck.
+For instance, some advice in data movement and compute utilization from the below figure can guide you to tune performance.
+
+..  image:: nvvp3.png
+    :align: center
+    :scale: 33%
+
+..  image:: nvvp4.png
+    :align: center
+    :scale: 33%
+
+Profiling tips
+==============
+
+- The **nvprof** and **nvvp** output is a very good place to start.
+- The timeline is a good place to go next.
+- Only dig deep into a kernel if it’s taking a significant amount of your time.
+- Where possible, try to match profiler output with theory.
+    1) For example, if I know I’m moving 1GB, and my kernel takes 10ms, I expect the profiler to report 100GB/s.
+    2) Discrepancies are likely to mean your application isn’t doing what you thought it was.
+- Know your hardware: If your GPU can do 6 TFLOPs, and you’re already doing 5.5 TFLOPs, you won’t go much faster!
+
+
+Profiling is a key step in optimization. Sometimes quite simple changes can lead to big improvements in performance.
+Your mileage may vary!
+
+Reference
+=========
+Jeremy Appleyard, `GPU Profiling for Deep Learning <http://www.robots.ox.ac.uk/~seminars/seminars/Extra/2015_10_08_JeremyAppleyard.pdf>`_, 2015

doc/howto/optimization/index.rst

+How to Tune GPU Performance
+===========================
+
+.. toctree::
+  :maxdepth: 3
+
+  gpu_profiling.rst

doc/index.rst

@@ -4,7 +4,9 @@ PaddlePaddle Documentation
 ..  toctree::
   :maxdepth: 1
 
-  introduction/index.md
-  user_guide.rst
-  dev/index.rst
-  algorithm/index.rst
+  getstarted/index.rst
+  tutorials/index.md
+  howto/index.rst
+  api/index.rst
+  about/index.rst 
+ 
\ No newline at end of file