diff --git a/CMakeLists.txt b/CMakeLists.txt index e5e54cc8cfcfd0a9f7085a7bbc93140bb7bb3824..f7bc09202fec79743db700b29f3bd49ba5195179 100644 --- a/CMakeLists.txt +++ b/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") diff --git a/RELEASE.md b/RELEASE.md new file mode 100644 index 0000000000000000000000000000000000000000..a8a245ab442ba0fc63d1f1fda932e7590a6fe4ca --- /dev/null +++ b/RELEASE.md @@ -0,0 +1,69 @@ +# 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. diff --git a/doc/about/index.rst b/doc/about/index.rst new file mode 100644 index 0000000000000000000000000000000000000000..8a372d2bc2b2c54b021ed63941482cbad8d8f719 --- /dev/null +++ b/doc/about/index.rst @@ -0,0 +1,14 @@ +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 `_ of PaddlePaddle! diff --git a/doc/algorithm/index.rst b/doc/algorithm/index.rst deleted file mode 100644 index 6073add3c0cbb12529eabb0f8d8a051bcb84e628..0000000000000000000000000000000000000000 --- a/doc/algorithm/index.rst +++ /dev/null @@ -1,7 +0,0 @@ -Algorithm Tutorial -================== - -.. toctree:: - :maxdepth: 1 - - rnn/rnn.rst diff --git a/doc/algorithm/rnn/bi_lstm.jpg b/doc/algorithm/rnn/bi_lstm.jpg deleted file mode 120000 index a53296cf806f97f7f2520e1700c4fb93f6bfc9d8..0000000000000000000000000000000000000000 --- a/doc/algorithm/rnn/bi_lstm.jpg +++ /dev/null @@ -1 +0,0 @@ -../../demo/sentiment_analysis/bi_lstm.jpg \ No newline at end of file diff --git a/doc/algorithm/rnn/encoder-decoder-attention-model.png b/doc/algorithm/rnn/encoder-decoder-attention-model.png deleted file mode 120000 index db71321a43a37b774e7de0af3765a60345033743..0000000000000000000000000000000000000000 --- a/doc/algorithm/rnn/encoder-decoder-attention-model.png +++ /dev/null @@ -1 +0,0 @@ -../../demo/text_generation/encoder-decoder-attention-model.png \ No newline at end of file diff --git a/doc/ui/data_provider/index.rst b/doc/api/data_provider/index.rst similarity index 97% rename from doc/ui/data_provider/index.rst rename to doc/api/data_provider/index.rst index 3db5b57376257b83fc2a27c518b0db663682136d..5e7a49d63236ffa854e64c53921441bacebc13ae 100644 --- a/doc/ui/data_provider/index.rst +++ b/doc/api/data_provider/index.rst @@ -1,5 +1,5 @@ -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. diff --git a/doc/ui/data_provider/pydataprovider2.rst b/doc/api/data_provider/pydataprovider2.rst similarity index 99% rename from doc/ui/data_provider/pydataprovider2.rst rename to doc/api/data_provider/pydataprovider2.rst index e105d3be308705d228c0b188e15742a0f7325ab6..b42cbca576e4b5d67d50d0156939a01faae4533d 100644 --- a/doc/ui/data_provider/pydataprovider2.rst +++ b/doc/api/data_provider/pydataprovider2.rst @@ -1,5 +1,5 @@ -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 diff --git a/doc/api/index.rst b/doc/api/index.rst new file mode 100644 index 0000000000000000000000000000000000000000..ccee7a0f1f3e5290891dfa963ea24c7fdbd6275e --- /dev/null +++ b/doc/api/index.rst @@ -0,0 +1,36 @@ +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 diff --git a/doc/ui/predict/predict_sample.py b/doc/api/predict/predict_sample.py similarity index 100% rename from doc/ui/predict/predict_sample.py rename to doc/api/predict/predict_sample.py diff --git a/doc/ui/predict/swig_py_paddle_en.rst b/doc/api/predict/swig_py_paddle_en.rst similarity index 98% rename from doc/ui/predict/swig_py_paddle_en.rst rename to doc/api/predict/swig_py_paddle_en.rst index b743fc456914664168e1be6c7f18a419c38afa62..9845cd1607b425dc0a4ddc665aab40d96fa2fbe4 100644 --- a/doc/ui/predict/swig_py_paddle_en.rst +++ b/doc/api/predict/swig_py_paddle_en.rst @@ -1,5 +1,5 @@ -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: diff --git a/doc/ui/api/trainer_config_helpers/activations.rst b/doc/api/trainer_config_helpers/activations.rst similarity index 100% rename from doc/ui/api/trainer_config_helpers/activations.rst rename to doc/api/trainer_config_helpers/activations.rst diff --git a/doc/api/trainer_config_helpers/attrs.rst b/doc/api/trainer_config_helpers/attrs.rst new file mode 100644 index 0000000000000000000000000000000000000000..ac63127bf7d9db6351365ab7b58f43db12347a8e --- /dev/null +++ b/doc/api/trainer_config_helpers/attrs.rst @@ -0,0 +1,5 @@ +Parameter Attributes +======================= + +.. automodule:: paddle.trainer_config_helpers.attrs + :members: diff --git a/doc/ui/api/trainer_config_helpers/data_sources.rst b/doc/api/trainer_config_helpers/data_sources.rst similarity index 100% rename from doc/ui/api/trainer_config_helpers/data_sources.rst rename to doc/api/trainer_config_helpers/data_sources.rst diff --git a/doc/ui/api/trainer_config_helpers/evaluators.rst b/doc/api/trainer_config_helpers/evaluators.rst similarity index 100% rename from doc/ui/api/trainer_config_helpers/evaluators.rst rename to doc/api/trainer_config_helpers/evaluators.rst diff --git a/doc/ui/api/trainer_config_helpers/layers.rst b/doc/api/trainer_config_helpers/layers.rst similarity index 100% rename from doc/ui/api/trainer_config_helpers/layers.rst rename to doc/api/trainer_config_helpers/layers.rst diff --git a/doc/ui/api/trainer_config_helpers/networks.rst b/doc/api/trainer_config_helpers/networks.rst similarity index 100% rename from doc/ui/api/trainer_config_helpers/networks.rst rename to doc/api/trainer_config_helpers/networks.rst diff --git a/doc/ui/api/trainer_config_helpers/optimizers.rst b/doc/api/trainer_config_helpers/optimizers.rst similarity index 100% rename from doc/ui/api/trainer_config_helpers/optimizers.rst rename to doc/api/trainer_config_helpers/optimizers.rst diff --git a/doc/ui/api/trainer_config_helpers/poolings.rst b/doc/api/trainer_config_helpers/poolings.rst similarity index 100% rename from doc/ui/api/trainer_config_helpers/poolings.rst rename to doc/api/trainer_config_helpers/poolings.rst diff --git a/doc/cluster/index.rst b/doc/cluster/index.rst deleted file mode 100644 index 9062f85f98d2981b5c8dcf8149e32c2ccdac77f4..0000000000000000000000000000000000000000 --- a/doc/cluster/index.rst +++ /dev/null @@ -1,8 +0,0 @@ -Cluster Train -==================== - -.. toctree:: - :glob: - - opensource/cluster_train.md - internal/index.md diff --git a/doc/dev/index.rst b/doc/dev/index.rst deleted file mode 100644 index 0468dd492b6246cfe0771a05c3597ddee95b3ddd..0000000000000000000000000000000000000000 --- a/doc/dev/index.rst +++ /dev/null @@ -1,9 +0,0 @@ -Development Guide -================= - -.. toctree:: - :maxdepth: 1 - - layer.md - new_layer/new_layer.rst - ../source/index.md diff --git a/doc/dev/layer.md b/doc/dev/layer.md deleted file mode 100644 index 930fb0de1ac074b15d06197ed0e732f92288b411..0000000000000000000000000000000000000000 --- a/doc/dev/layer.md +++ /dev/null @@ -1,4 +0,0 @@ -# Layer Documents - -* [Layer Source Code Document](../source/gserver/layers/index.rst) -* [Layer Python API Document](../ui/api/trainer_config_helpers/index.rst) diff --git a/doc/getstarted/basic_usage/basic_usage.rst b/doc/getstarted/basic_usage/basic_usage.rst new file mode 100644 index 0000000000000000000000000000000000000000..dca7a6b1f4f017b302148c611122806f112564a9 --- /dev/null +++ b/doc/getstarted/basic_usage/basic_usage.rst @@ -0,0 +1,109 @@ +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 +--------------------- + +Now, to give you a hint of what using PaddlePaddle looks like, let's start with a fundamental learning problem - `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 +-------------------- + +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. + + .. code-block:: python + + # dataprovider.py + from paddle.trainer.PyDataProvider2 import * + import random + + # 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] + +3. Train a NeuralNetwork +------------------------- + +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: + + .. code-block:: python + + # trainer_config.py + from paddle.trainer_config_helpers import * + + # 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: + +- The first part shows how to feed data into PaddlePaddle. In general cases, PaddlePaddle reads raw data from a list of files, and then do some user-defined process to get real input. In this case, we only need to create a placeholder file since we are generating synthetic data on the fly. + +- 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``. + - **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: + + .. 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 +----------------------- + +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. + + .. code-block:: python + + 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 + + .. 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. + +There, you have recovered the underlying pattern between ``X`` and ``Y`` only from observed data. + + +5. Where to Go from Here +------------------------- + +- `Install and Build <../build_and_install/index.html>`_ +- `Tutorials <../demo/quick_start/index_en.html>`_ +- `Example and Demo <../demo/index.html>`_ diff --git a/doc/getstarted/basic_usage/parameters.png b/doc/getstarted/basic_usage/parameters.png new file mode 100644 index 0000000000000000000000000000000000000000..2ec67480951e21f0400bce1c34b3108dcd65c18c Binary files /dev/null and b/doc/getstarted/basic_usage/parameters.png differ diff --git a/doc/build/build_from_source.md b/doc/getstarted/build_and_install/build_from_source.md similarity index 100% rename from doc/build/build_from_source.md rename to doc/getstarted/build_and_install/build_from_source.md diff --git a/doc/build/cmake.png b/doc/getstarted/build_and_install/cmake.png similarity index 100% rename from doc/build/cmake.png rename to doc/getstarted/build_and_install/cmake.png diff --git a/doc/build/docker_install.rst b/doc/getstarted/build_and_install/docker_install.rst similarity index 100% rename from doc/build/docker_install.rst rename to doc/getstarted/build_and_install/docker_install.rst diff --git a/doc/build/index.rst b/doc/getstarted/build_and_install/index.rst similarity index 80% rename from doc/build/index.rst rename to doc/getstarted/build_and_install/index.rst index b4fe4596047c7d201fdf36bc76c26d5134611560..6187be9d7257b1690d223770f0f6a5b466cf2898 100644 --- a/doc/build/index.rst +++ b/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 diff --git a/doc/build/ubuntu_install.rst b/doc/getstarted/build_and_install/ubuntu_install.rst similarity index 100% rename from doc/build/ubuntu_install.rst rename to doc/getstarted/build_and_install/ubuntu_install.rst diff --git a/doc/getstarted/index.rst b/doc/getstarted/index.rst new file mode 100644 index 0000000000000000000000000000000000000000..5f2787066ea5cdb32a1eff2939ffd0585d7fbc63 --- /dev/null +++ b/doc/getstarted/index.rst @@ -0,0 +1,8 @@ +GET STARTED +============ + +.. toctree:: + :maxdepth: 2 + + build_and_install/index.rst + basic_usage/basic_usage.rst diff --git a/doc/cluster/opensource/cluster_train.md b/doc/howto/cluster/cluster_train.md similarity index 99% rename from doc/cluster/opensource/cluster_train.md rename to doc/howto/cluster/cluster_train.md index cb493a88f031850cb6a5eeed0ebe9e41bb7e01c3..1de34a6a99440bf45af8b1fec2c7a2361865fed3 100644 --- a/doc/cluster/opensource/cluster_train.md +++ b/doc/howto/cluster/cluster_train.md @@ -1,4 +1,4 @@ -# 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. diff --git a/doc/ui/cmd_argument/argument_outline.md b/doc/howto/cmd_parameter/arguments.md similarity index 100% rename from doc/ui/cmd_argument/argument_outline.md rename to doc/howto/cmd_parameter/arguments.md diff --git a/doc/ui/cmd_argument/detail_introduction.md b/doc/howto/cmd_parameter/detail_introduction.md similarity index 100% rename from doc/ui/cmd_argument/detail_introduction.md rename to doc/howto/cmd_parameter/detail_introduction.md diff --git a/doc/howto/cmd_parameter/index.md b/doc/howto/cmd_parameter/index.md new file mode 100644 index 0000000000000000000000000000000000000000..48cf835de142f19f41a9d077786d312100859592 --- /dev/null +++ b/doc/howto/cmd_parameter/index.md @@ -0,0 +1,5 @@ +# How to Set Command-line Parameters + +* [Use Case](use_case.md) +* [Arguments](arguments.md) +* [Detailed Descriptions](detail_introduction.md) diff --git a/doc/ui/cmd_argument/use_case.md b/doc/howto/cmd_parameter/use_case.md similarity index 100% rename from doc/ui/cmd_argument/use_case.md rename to doc/howto/cmd_parameter/use_case.md diff --git a/doc/build/contribute_to_paddle.md b/doc/howto/contribute_to_paddle.md similarity index 99% rename from doc/build/contribute_to_paddle.md rename to doc/howto/contribute_to_paddle.md index 1d03eb7362b1b6f2fcdac7b53f8b7f93fb75e49c..d1f12c6ab2fb9ddeed40b53c1b2c68a9ccb19105 100644 --- a/doc/build/contribute_to_paddle.md +++ b/doc/howto/contribute_to_paddle.md @@ -1,4 +1,4 @@ -# Contribute Code +# How to Contribute Code We sincerely appreciate your contributions. You can use fork and pull request workflow to merge your code. diff --git a/doc/howto/deep_model/index.rst b/doc/howto/deep_model/index.rst new file mode 100644 index 0000000000000000000000000000000000000000..06ef443f62f63723704d4f06ba5d321636c0b72b --- /dev/null +++ b/doc/howto/deep_model/index.rst @@ -0,0 +1,7 @@ +How to Configure Deep Models +============================ + +.. toctree:: + :maxdepth: 1 + + rnn/rnn.rst diff --git a/doc/algorithm/rnn/rnn.rst b/doc/howto/deep_model/rnn/rnn.rst similarity index 99% rename from doc/algorithm/rnn/rnn.rst rename to doc/howto/deep_model/rnn/rnn.rst index 01d2caefb5cdf4e949511fd0f5bbafe0e604e881..da29b8efadd299fe4fc74a71392cbc9a56e32be3 100644 --- a/doc/algorithm/rnn/rnn.rst +++ b/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`. diff --git a/doc/howto/index.rst b/doc/howto/index.rst new file mode 100644 index 0000000000000000000000000000000000000000..41877a64a56e289c39ca7513eb05fa6a41129487 --- /dev/null +++ b/doc/howto/index.rst @@ -0,0 +1,29 @@ +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 diff --git a/doc/dev/new_layer/FullyConnected.jpg b/doc/howto/new_layer/FullyConnected.jpg similarity index 100% rename from doc/dev/new_layer/FullyConnected.jpg rename to doc/howto/new_layer/FullyConnected.jpg diff --git a/doc/dev/new_layer/new_layer.rst b/doc/howto/new_layer/index.rst similarity index 99% rename from doc/dev/new_layer/new_layer.rst rename to doc/howto/new_layer/index.rst index af8b76a3075194ead9be40d2c943238b2cfadecc..922bda5b0d879b9041e3c0ca5d2518363a7cfa05 100644 --- a/doc/dev/new_layer/new_layer.rst +++ b/doc/howto/new_layer/index.rst @@ -1,6 +1,6 @@ -================== -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. diff --git a/doc/optimization/gpu_profiling.rst b/doc/howto/optimization/gpu_profiling.rst similarity index 100% rename from doc/optimization/gpu_profiling.rst rename to doc/howto/optimization/gpu_profiling.rst diff --git a/doc/howto/optimization/index.rst b/doc/howto/optimization/index.rst new file mode 100644 index 0000000000000000000000000000000000000000..e2822a00982b386e8900420e8c200ed1f92a2d9b --- /dev/null +++ b/doc/howto/optimization/index.rst @@ -0,0 +1,7 @@ +How to Tune GPU Performance +=========================== + +.. toctree:: + :maxdepth: 3 + + gpu_profiling.rst diff --git a/doc/optimization/nvvp1.png b/doc/howto/optimization/nvvp1.png similarity index 100% rename from doc/optimization/nvvp1.png rename to doc/howto/optimization/nvvp1.png diff --git a/doc/optimization/nvvp2.png b/doc/howto/optimization/nvvp2.png similarity index 100% rename from doc/optimization/nvvp2.png rename to doc/howto/optimization/nvvp2.png diff --git a/doc/optimization/nvvp3.png b/doc/howto/optimization/nvvp3.png similarity index 100% rename from doc/optimization/nvvp3.png rename to doc/howto/optimization/nvvp3.png diff --git a/doc/optimization/nvvp4.png b/doc/howto/optimization/nvvp4.png similarity index 100% rename from doc/optimization/nvvp4.png rename to doc/howto/optimization/nvvp4.png diff --git a/doc/source/api.rst b/doc/howto/source/api.rst similarity index 100% rename from doc/source/api.rst rename to doc/howto/source/api.rst diff --git a/doc/source/cuda/index.rst b/doc/howto/source/cuda/index.rst similarity index 100% rename from doc/source/cuda/index.rst rename to doc/howto/source/cuda/index.rst diff --git a/doc/source/cuda/matrix.rst b/doc/howto/source/cuda/matrix.rst similarity index 100% rename from doc/source/cuda/matrix.rst rename to doc/howto/source/cuda/matrix.rst diff --git a/doc/source/cuda/nn.rst b/doc/howto/source/cuda/nn.rst similarity index 100% rename from doc/source/cuda/nn.rst rename to doc/howto/source/cuda/nn.rst diff --git a/doc/source/cuda/utils.rst b/doc/howto/source/cuda/utils.rst similarity index 100% rename from doc/source/cuda/utils.rst rename to doc/howto/source/cuda/utils.rst diff --git a/doc/source/gserver/activations.rst b/doc/howto/source/gserver/activations.rst similarity index 100% rename from doc/source/gserver/activations.rst rename to doc/howto/source/gserver/activations.rst diff --git a/doc/source/gserver/dataproviders.rst b/doc/howto/source/gserver/dataproviders.rst similarity index 100% rename from doc/source/gserver/dataproviders.rst rename to doc/howto/source/gserver/dataproviders.rst diff --git a/doc/source/gserver/evaluators.rst b/doc/howto/source/gserver/evaluators.rst similarity index 100% rename from doc/source/gserver/evaluators.rst rename to doc/howto/source/gserver/evaluators.rst diff --git a/doc/source/gserver/gradientmachines.rst b/doc/howto/source/gserver/gradientmachines.rst similarity index 100% rename from doc/source/gserver/gradientmachines.rst rename to doc/howto/source/gserver/gradientmachines.rst diff --git a/doc/source/gserver/index.rst b/doc/howto/source/gserver/index.rst similarity index 100% rename from doc/source/gserver/index.rst rename to doc/howto/source/gserver/index.rst diff --git a/doc/source/gserver/layers.rst b/doc/howto/source/gserver/layers.rst similarity index 100% rename from doc/source/gserver/layers.rst rename to doc/howto/source/gserver/layers.rst diff --git a/doc/source/gserver/neworks.rst b/doc/howto/source/gserver/neworks.rst similarity index 100% rename from doc/source/gserver/neworks.rst rename to doc/howto/source/gserver/neworks.rst diff --git a/doc/source/index.rst b/doc/howto/source/index.rst similarity index 100% rename from doc/source/index.rst rename to doc/howto/source/index.rst diff --git a/doc/source/math/functions.rst b/doc/howto/source/math/functions.rst similarity index 100% rename from doc/source/math/functions.rst rename to doc/howto/source/math/functions.rst diff --git a/doc/source/math/index.rst b/doc/howto/source/math/index.rst similarity index 100% rename from doc/source/math/index.rst rename to doc/howto/source/math/index.rst diff --git a/doc/source/math/matrix.rst b/doc/howto/source/math/matrix.rst similarity index 100% rename from doc/source/math/matrix.rst rename to doc/howto/source/math/matrix.rst diff --git a/doc/source/math/utils.rst b/doc/howto/source/math/utils.rst similarity index 100% rename from doc/source/math/utils.rst rename to doc/howto/source/math/utils.rst diff --git a/doc/source/math/vector.rst b/doc/howto/source/math/vector.rst similarity index 100% rename from doc/source/math/vector.rst rename to doc/howto/source/math/vector.rst diff --git a/doc/source/parameter/index.rst b/doc/howto/source/parameter/index.rst similarity index 100% rename from doc/source/parameter/index.rst rename to doc/howto/source/parameter/index.rst diff --git a/doc/source/parameter/optimizer.rst b/doc/howto/source/parameter/optimizer.rst similarity index 100% rename from doc/source/parameter/optimizer.rst rename to doc/howto/source/parameter/optimizer.rst diff --git a/doc/source/parameter/parameter.rst b/doc/howto/source/parameter/parameter.rst similarity index 100% rename from doc/source/parameter/parameter.rst rename to doc/howto/source/parameter/parameter.rst diff --git a/doc/source/parameter/updater.rst b/doc/howto/source/parameter/updater.rst similarity index 100% rename from doc/source/parameter/updater.rst rename to doc/howto/source/parameter/updater.rst diff --git a/doc/source/pserver/client.rst b/doc/howto/source/pserver/client.rst similarity index 100% rename from doc/source/pserver/client.rst rename to doc/howto/source/pserver/client.rst diff --git a/doc/source/pserver/index.rst b/doc/howto/source/pserver/index.rst similarity index 100% rename from doc/source/pserver/index.rst rename to doc/howto/source/pserver/index.rst diff --git a/doc/source/pserver/network.rst b/doc/howto/source/pserver/network.rst similarity index 100% rename from doc/source/pserver/network.rst rename to doc/howto/source/pserver/network.rst diff --git a/doc/source/pserver/server.rst b/doc/howto/source/pserver/server.rst similarity index 100% rename from doc/source/pserver/server.rst rename to doc/howto/source/pserver/server.rst diff --git a/doc/source/trainer.rst b/doc/howto/source/trainer.rst similarity index 100% rename from doc/source/trainer.rst rename to doc/howto/source/trainer.rst diff --git a/doc/source/utils/customStackTrace.rst b/doc/howto/source/utils/customStackTrace.rst similarity index 100% rename from doc/source/utils/customStackTrace.rst rename to doc/howto/source/utils/customStackTrace.rst diff --git a/doc/source/utils/enum.rst b/doc/howto/source/utils/enum.rst similarity index 100% rename from doc/source/utils/enum.rst rename to doc/howto/source/utils/enum.rst diff --git a/doc/source/utils/index.rst b/doc/howto/source/utils/index.rst similarity index 100% rename from doc/source/utils/index.rst rename to doc/howto/source/utils/index.rst diff --git a/doc/source/utils/lock.rst b/doc/howto/source/utils/lock.rst similarity index 100% rename from doc/source/utils/lock.rst rename to doc/howto/source/utils/lock.rst diff --git a/doc/source/utils/queue.rst b/doc/howto/source/utils/queue.rst similarity index 100% rename from doc/source/utils/queue.rst rename to doc/howto/source/utils/queue.rst diff --git a/doc/source/utils/thread.rst b/doc/howto/source/utils/thread.rst similarity index 100% rename from doc/source/utils/thread.rst rename to doc/howto/source/utils/thread.rst diff --git a/doc/index.rst b/doc/index.rst index 76fb7a3ace8057d9cd34e03134c63ef0cd298cae..3555da1dfc81b29a89c7dfa6087d5fbb734a727b 100644 --- a/doc/index.rst +++ b/doc/index.rst @@ -4,8 +4,9 @@ PaddlePaddle Documentation .. toctree:: :maxdepth: 1 - introduction/index.md - user_guide.rst - dev/index.rst - algorithm/index.rst - optimization/index.rst + getstarted/index.rst + tutorials/index.md + howto/index.rst + api/index.rst + about/index.rst + \ No newline at end of file diff --git a/doc/introduction/index.md b/doc/introduction/index.md deleted file mode 100644 index 01f52031a1d0247cd0b885218c17001f23685239..0000000000000000000000000000000000000000 --- a/doc/introduction/index.md +++ /dev/null @@ -1,100 +0,0 @@ -# Introduction - -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 - -Now, to give you a hint of what using PaddlePaddle looks like, let's start with a fundamental learning problem - **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 - -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 - -# 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] -``` - -## 3. Train a NeuralNetwork in PaddlePaddle - -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: - -```python -# trainer_config.py -from paddle.trainer_config_helpers import * - -# 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: - -- The first part shows how to feed data into PaddlePaddle. In general cases, PaddlePaddle reads raw data from a list of files, and then do some user-defined process to get real input. In this case, we only need to create a placeholder file since we are generating synthetic data on the fly. - -- 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`. - - **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. - - -## 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. - -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. - -```python -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 -``` - -
![](./parameters.png)
- -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. - - -## 5. Where to Go from Here - -- Build and Installation -- Quick Start -- Example and Demo diff --git a/doc/introduction/parameters.png b/doc/introduction/parameters.png deleted file mode 120000 index f47e74c94fffabbd32f055febbadb1b18aa0c429..0000000000000000000000000000000000000000 --- a/doc/introduction/parameters.png +++ /dev/null @@ -1 +0,0 @@ -../../doc_cn/introduction/parameters.png \ No newline at end of file diff --git a/doc/optimization/index.rst b/doc/optimization/index.rst deleted file mode 100644 index c9e87e0778dfe44fa3d1bb84d0ad340aa6f25d08..0000000000000000000000000000000000000000 --- a/doc/optimization/index.rst +++ /dev/null @@ -1,7 +0,0 @@ -Performance Tuning -================== - -.. toctree:: - :maxdepth: 3 - - gpu_profiling.rst diff --git a/doc/demo/embedding_model/index.md b/doc/tutorials/embedding_model/index.md similarity index 100% rename from doc/demo/embedding_model/index.md rename to doc/tutorials/embedding_model/index.md diff --git a/doc/demo/embedding_model/neural-n-gram-model.png b/doc/tutorials/embedding_model/neural-n-gram-model.png similarity index 100% rename from doc/demo/embedding_model/neural-n-gram-model.png rename to doc/tutorials/embedding_model/neural-n-gram-model.png diff --git a/doc/demo/image_classification/cifar.png b/doc/tutorials/image_classification/cifar.png similarity index 100% rename from doc/demo/image_classification/cifar.png rename to doc/tutorials/image_classification/cifar.png diff --git a/doc/demo/image_classification/image_classification.md b/doc/tutorials/image_classification/image_classification.md similarity index 100% rename from doc/demo/image_classification/image_classification.md rename to doc/tutorials/image_classification/image_classification.md diff --git a/doc/demo/image_classification/image_classification.png b/doc/tutorials/image_classification/image_classification.png similarity index 100% rename from doc/demo/image_classification/image_classification.png rename to doc/tutorials/image_classification/image_classification.png diff --git a/doc/demo/image_classification/index.rst b/doc/tutorials/image_classification/index.rst similarity index 100% rename from doc/demo/image_classification/index.rst rename to doc/tutorials/image_classification/index.rst diff --git a/doc/demo/image_classification/lenet.png b/doc/tutorials/image_classification/lenet.png similarity index 100% rename from doc/demo/image_classification/lenet.png rename to doc/tutorials/image_classification/lenet.png diff --git a/doc/demo/image_classification/plot.png b/doc/tutorials/image_classification/plot.png similarity index 100% rename from doc/demo/image_classification/plot.png rename to doc/tutorials/image_classification/plot.png diff --git a/doc/demo/imagenet_model/resnet_block.jpg b/doc/tutorials/imagenet_model/resnet_block.jpg similarity index 100% rename from doc/demo/imagenet_model/resnet_block.jpg rename to doc/tutorials/imagenet_model/resnet_block.jpg diff --git a/doc/demo/imagenet_model/resnet_model.md b/doc/tutorials/imagenet_model/resnet_model.md similarity index 100% rename from doc/demo/imagenet_model/resnet_model.md rename to doc/tutorials/imagenet_model/resnet_model.md diff --git a/doc/demo/index.md b/doc/tutorials/index.md similarity index 96% rename from doc/demo/index.md rename to doc/tutorials/index.md index 289199d496eb3b527fa8c8261820bc8e4d301786..ebf5397391e65e096b265f44a0ad81942f0b9ec2 100644 --- a/doc/demo/index.md +++ b/doc/tutorials/index.md @@ -1,4 +1,4 @@ -# Examples and demos +# TUTORIALS There are serveral examples and demos here. ## Image diff --git a/doc/demo/quick_start/NetContinuous_en.png b/doc/tutorials/quick_start/NetContinuous_en.png similarity index 100% rename from doc/demo/quick_start/NetContinuous_en.png rename to doc/tutorials/quick_start/NetContinuous_en.png diff --git a/doc/demo/quick_start/NetConv_en.png b/doc/tutorials/quick_start/NetConv_en.png similarity index 100% rename from doc/demo/quick_start/NetConv_en.png rename to doc/tutorials/quick_start/NetConv_en.png diff --git a/doc/demo/quick_start/NetLR_en.png b/doc/tutorials/quick_start/NetLR_en.png similarity index 100% rename from doc/demo/quick_start/NetLR_en.png rename to doc/tutorials/quick_start/NetLR_en.png diff --git a/doc/demo/quick_start/NetRNN_en.png b/doc/tutorials/quick_start/NetRNN_en.png similarity index 100% rename from doc/demo/quick_start/NetRNN_en.png rename to doc/tutorials/quick_start/NetRNN_en.png diff --git a/doc/demo/quick_start/PipelineNetwork_en.jpg b/doc/tutorials/quick_start/PipelineNetwork_en.jpg similarity index 100% rename from doc/demo/quick_start/PipelineNetwork_en.jpg rename to doc/tutorials/quick_start/PipelineNetwork_en.jpg diff --git a/doc/demo/quick_start/PipelineTest_en.png b/doc/tutorials/quick_start/PipelineTest_en.png similarity index 100% rename from doc/demo/quick_start/PipelineTest_en.png rename to doc/tutorials/quick_start/PipelineTest_en.png diff --git a/doc/demo/quick_start/PipelineTrain_en.png b/doc/tutorials/quick_start/PipelineTrain_en.png similarity index 100% rename from doc/demo/quick_start/PipelineTrain_en.png rename to doc/tutorials/quick_start/PipelineTrain_en.png diff --git a/doc/demo/quick_start/Pipeline_en.jpg b/doc/tutorials/quick_start/Pipeline_en.jpg similarity index 100% rename from doc/demo/quick_start/Pipeline_en.jpg rename to doc/tutorials/quick_start/Pipeline_en.jpg diff --git a/doc/demo/quick_start/index_en.md b/doc/tutorials/quick_start/index_en.md similarity index 100% rename from doc/demo/quick_start/index_en.md rename to doc/tutorials/quick_start/index_en.md diff --git a/doc/demo/rec/ml_dataset.md b/doc/tutorials/rec/ml_dataset.md similarity index 100% rename from doc/demo/rec/ml_dataset.md rename to doc/tutorials/rec/ml_dataset.md diff --git a/doc/demo/rec/ml_regression.rst b/doc/tutorials/rec/ml_regression.rst similarity index 100% rename from doc/demo/rec/ml_regression.rst rename to doc/tutorials/rec/ml_regression.rst diff --git a/doc/demo/rec/rec_regression_network.png b/doc/tutorials/rec/rec_regression_network.png similarity index 100% rename from doc/demo/rec/rec_regression_network.png rename to doc/tutorials/rec/rec_regression_network.png diff --git a/doc/demo/semantic_role_labeling/curve.jpg b/doc/tutorials/semantic_role_labeling/curve.jpg similarity index 100% rename from doc/demo/semantic_role_labeling/curve.jpg rename to doc/tutorials/semantic_role_labeling/curve.jpg diff --git a/doc/demo/semantic_role_labeling/feature.jpg b/doc/tutorials/semantic_role_labeling/feature.jpg similarity index 100% rename from doc/demo/semantic_role_labeling/feature.jpg rename to doc/tutorials/semantic_role_labeling/feature.jpg diff --git a/doc/demo/semantic_role_labeling/index.rst b/doc/tutorials/semantic_role_labeling/index.rst similarity index 100% rename from doc/demo/semantic_role_labeling/index.rst rename to doc/tutorials/semantic_role_labeling/index.rst diff --git a/doc/demo/semantic_role_labeling/network_arch.png b/doc/tutorials/semantic_role_labeling/network_arch.png similarity index 100% rename from doc/demo/semantic_role_labeling/network_arch.png rename to doc/tutorials/semantic_role_labeling/network_arch.png diff --git a/doc/demo/semantic_role_labeling/semantic_role_labeling.md b/doc/tutorials/semantic_role_labeling/semantic_role_labeling.md similarity index 97% rename from doc/demo/semantic_role_labeling/semantic_role_labeling.md rename to doc/tutorials/semantic_role_labeling/semantic_role_labeling.md index e2793b2b3494160a7a80f07ec2127bd1f1a4f2e4..f5bdf64487aa189cefcd55d633cc6638912b9e31 100644 --- a/doc/demo/semantic_role_labeling/semantic_role_labeling.md +++ b/doc/tutorials/semantic_role_labeling/semantic_role_labeling.md @@ -1,200 +1,200 @@ -# Semantic Role labeling Tutorial # - -Semantic role labeling (SRL) is a form of shallow semantic parsing whose goal is to discover the predicate-argument structure of each predicate in a given input sentence. SRL is useful as an intermediate step in a wide range of natural language processing tasks, such as information extraction. automatic document categorization and question answering. An instance is as following [1]: - - [ A0 He ] [ AM-MOD would ][ AM-NEG n’t ] [ V accept] [ A1 anything of value ] from [A2 those he was writing about ]. - -- V: verb -- A0: acceptor -- A1: thing accepted -- A2: accepted-from -- A3: Attribute -- AM-MOD: modal -- AM-NEG: negation - -Given the verb "accept", the chunks in sentence would play certain semantic roles. Here, the label scheme is from Penn Proposition Bank. - -To this date, most of the successful SRL systems are built on top of some form of parsing results where pre-defined feature templates over the syntactic structure are used. This tutorial will present an end-to-end system using deep bidirectional long short-term memory (DB-LSTM)[2] for solving the SRL task, which largely outperforms the previous state-of-the-art systems. The system regards SRL task as the sequence labelling problem. - -## Data Description -The relevant paper[2] takes the data set in CoNLL-2005&2012 Shared Task for training and testing. Accordingto data license, the demo adopts the test data set of CoNLL-2005, which can be reached on website. - -To download and process the original data, user just need to execute the following command: - -```bash -cd data -./get_data.sh -``` -Several new files appear in the `data `directory as follows. -```bash -conll05st-release：the test data set of CoNll-2005 shared task -test.wsj.words：the Wall Street Journal data sentences -test.wsj.props: the propositional arguments -feature: the extracted features from data set -``` - -## Training -### DB-LSTM -Please refer to the Sentiment Analysis demo to learn more about the long short-term memory unit. - -Unlike Bidirectional-LSTM that used in Sentiment Analysis demo, the DB-LSTM adopts another way to stack LSTM layer. First a standard LSTM processes the sequence in forward direction. The input and output of this LSTM layer are taken by the next LSTM layer as input, processed in reversed direction. These two standard LSTM layers compose a pair of LSTM. Then we stack LSTM layers pair after pair to obtain the deep LSTM model. - -The following figure shows a temporal expanded 2-layer DB-LSTM network. -
-![pic](./network_arch.png) -
- -### Features -Two input features play an essential role in this pipeline: predicate (pred) and argument (argu). Two other features: predicate context (ctx-p) and region mark (mr) are also adopted. Because a single predicate word can not exactly describe the predicate information, especially when the same words appear more than one times in a sentence. With the predicate context, the ambiguity can be largely eliminated. Similarly, we use region mark mr = 1 to denote the argument position if it locates in the predicate context region, or mr = 0 if does not. These four simple features are all we need for our SRL system. Features of one sample with context size set to 1 is showed as following[2]: -
-![pic](./feature.jpg) -
- -In this sample, the coresponding labelled sentence is: - -[ A1 A record date ] has [ AM-NEG n't ] been [ V set ] . - -In the demo, we adopt the feature template as above, consists of : `argument`, `predicate`, `ctx-p (p=-1,0,1)`, `mark` and use `B/I/O` scheme to label each argument. These features and labels are stored in `feature` file, and separated by `\t`. - -### Data Provider - -`dataprovider.py` is the python file to wrap data. `hook()` function is to define the data slots for network. The Six features and label are all IndexSlots. -``` -def hook(settings, word_dict, label_dict, **kwargs): - settings.word_dict = word_dict - settings.label_dict = label_dict - #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(label_dict))] -``` -The corresponding data iterator is as following: -``` -@provider(init_hook=hook, should_shuffle=True, calc_batch_size=get_batch_size, - can_over_batch_size=False, cache=CacheType.CACHE_PASS_IN_MEM) -def process(settings, file_name): - with open(file_name, 'r') as fdata: - for line in fdata: - sentence, predicate, ctx_n2, ctx_n1, ctx_0, ctx_p1, ctx_p2, mark, label = \ - line.strip().split('\t') - - words = sentence.split() - sen_len = len(words) - word_slot = [settings.word_dict.get(w, UNK_IDX) for w in words] - - predicate_slot = [settings.predicate_dict.get(predicate)] * sen_len - ctx_n2_slot = [settings.word_dict.get(ctx_n2, UNK_IDX)] * sen_len - ctx_n1_slot = [settings.word_dict.get(ctx_n1, UNK_IDX)] * sen_len - ctx_0_slot = [settings.word_dict.get(ctx_0, UNK_IDX)] * sen_len - ctx_p1_slot = [settings.word_dict.get(ctx_p1, UNK_IDX)] * sen_len - ctx_p2_slot = [settings.word_dict.get(ctx_p2, UNK_IDX)] * sen_len - - marks = mark.split() - mark_slot = [int(w) for w in marks] - - 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 -``` -The `process`function yield 9 lists which are 8 features and label. - -### Neural Network Config -`db_lstm.py` is the neural network config file to load the dictionaries and define the data provider module and network architecture during the training procedure. - -Nine `data_layer` load instances from data provider. Eight features are transformed into embedddings respectively, and mixed by `mixed_layer` . Deep bidirectional LSTM layers extract features for the softmax layer. The objective function is cross entropy of labels. - -### Run Training -The script for training is `train.sh`, user just need to execute: -```bash - ./train.sh -``` -The content in `train.sh`: -``` -paddle train \ - --config=./db_lstm.py \ - --use_gpu=0 \ - --log_period=5000 \ - --trainer_count=1 \ - --show_parameter_stats_period=5000 \ - --save_dir=./output \ - --num_passes=10000 \ - --average_test_period=10000000 \ - --init_model_path=./data \ - --load_missing_parameter_strategy=rand \ - --test_all_data_in_one_period=1 \ -2>&1 | tee 'train.log' -``` - -- \--config=./db_lstm.py : network config file. -- \--use_gpu=false: use CPU to train, set true, if you install GPU version of PaddlePaddle and want to use GPU to train, until now crf_layer do not support GPU -- \--log_period=500: print log every 20 batches. -- \--trainer_count=1: set thread number (or GPU count). -- \--show_parameter_stats_period=5000: show parameter statistic every 100 batches. -- \--save_dir=./output: output path to save models. -- \--num_passes=10000: set pass number, one pass in PaddlePaddle means training all samples in dataset one time. -- \--average_test_period=10000000: do test on average parameter every average_test_period batches -- \--init_model_path=./data: parameter initialization path -- \--load_missing_parameter_strategy=rand: random initialization unexisted parameters -- \--test_all_data_in_one_period=1: test all data in one period - - -After training, the models will be saved in directory `output`. Our training curve is as following: -
-![pic](./curve.jpg) -
- -### Run testing -The script for testing is `test.sh`, user just need to execute: -```bash - ./test.sh -``` -The main part in `tesh.sh` -``` -paddle train \ - --config=./db_lstm.py \ - --model_list=$model_list \ - --job=test \ - --config_args=is_test=1 \ -``` - - - \--config=./db_lstm.py: network config file - - \--model_list=$model_list.list: model list file - - \--job=test: indicate the test job - - \--config_args=is_test=1: flag to indicate test - - \--test_all_data_in_one_period=1: test all data in 1 period - - -### Run prediction -The script for prediction is `predict.sh`, user just need to execute: -```bash - ./predict.sh - -``` -In `predict.sh`, user should offer the network config file, model path, label file, word dictionary file, feature file -``` -python predict.py - -c $config_file \ - -w $best_model_path \ - -l $label_file \ - -p $predicate_dict_file \ - -d $dict_file \ - -i $input_file \ - -o $output_file -``` - -`predict.py` is the main executable python script, which includes functions: load model, load data, data prediction. The network model will output the probability distribution of labels. In the demo, we take the label with maximum probability as result. User can also implement the beam search or viterbi decoding upon the probability distribution matrix. - -After prediction, the result is saved in `predict.res`. - -## Reference -[1] Martha Palmer, Dan Gildea, and Paul Kingsbury. The Proposition Bank: An Annotated Corpus of Semantic Roles , Computational Linguistics, 31(1), 2005. - -[2] Zhou, Jie, and Wei Xu. "End-to-end learning of semantic role labeling using recurrent neural networks." Proceedings of the Annual Meeting of the Association for Computational Linguistics. 2015. +# Semantic Role labeling Tutorial # + +Semantic role labeling (SRL) is a form of shallow semantic parsing whose goal is to discover the predicate-argument structure of each predicate in a given input sentence. SRL is useful as an intermediate step in a wide range of natural language processing tasks, such as information extraction. automatic document categorization and question answering. An instance is as following [1]: + + [ A0 He ] [ AM-MOD would ][ AM-NEG n’t ] [ V accept] [ A1 anything of value ] from [A2 those he was writing about ]. + +- V: verb +- A0: acceptor +- A1: thing accepted +- A2: accepted-from +- A3: Attribute +- AM-MOD: modal +- AM-NEG: negation + +Given the verb "accept", the chunks in sentence would play certain semantic roles. Here, the label scheme is from Penn Proposition Bank. + +To this date, most of the successful SRL systems are built on top of some form of parsing results where pre-defined feature templates over the syntactic structure are used. This tutorial will present an end-to-end system using deep bidirectional long short-term memory (DB-LSTM)[2] for solving the SRL task, which largely outperforms the previous state-of-the-art systems. The system regards SRL task as the sequence labelling problem. + +## Data Description +The relevant paper[2] takes the data set in CoNLL-2005&2012 Shared Task for training and testing. Accordingto data license, the demo adopts the test data set of CoNLL-2005, which can be reached on website. + +To download and process the original data, user just need to execute the following command: + +```bash +cd data +./get_data.sh +``` +Several new files appear in the `data `directory as follows. +```bash +conll05st-release：the test data set of CoNll-2005 shared task +test.wsj.words：the Wall Street Journal data sentences +test.wsj.props: the propositional arguments +feature: the extracted features from data set +``` + +## Training +### DB-LSTM +Please refer to the Sentiment Analysis demo to learn more about the long short-term memory unit. + +Unlike Bidirectional-LSTM that used in Sentiment Analysis demo, the DB-LSTM adopts another way to stack LSTM layer. First a standard LSTM processes the sequence in forward direction. The input and output of this LSTM layer are taken by the next LSTM layer as input, processed in reversed direction. These two standard LSTM layers compose a pair of LSTM. Then we stack LSTM layers pair after pair to obtain the deep LSTM model. + +The following figure shows a temporal expanded 2-layer DB-LSTM network. +
+![pic](./network_arch.png) +
+ +### Features +Two input features play an essential role in this pipeline: predicate (pred) and argument (argu). Two other features: predicate context (ctx-p) and region mark (mr) are also adopted. Because a single predicate word can not exactly describe the predicate information, especially when the same words appear more than one times in a sentence. With the predicate context, the ambiguity can be largely eliminated. Similarly, we use region mark mr = 1 to denote the argument position if it locates in the predicate context region, or mr = 0 if does not. These four simple features are all we need for our SRL system. Features of one sample with context size set to 1 is showed as following[2]: +
+![pic](./feature.jpg) +
+ +In this sample, the coresponding labelled sentence is: + +[ A1 A record date ] has [ AM-NEG n't ] been [ V set ] . + +In the demo, we adopt the feature template as above, consists of : `argument`, `predicate`, `ctx-p (p=-1,0,1)`, `mark` and use `B/I/O` scheme to label each argument. These features and labels are stored in `feature` file, and separated by `\t`. + +### Data Provider + +`dataprovider.py` is the python file to wrap data. `hook()` function is to define the data slots for network. The Six features and label are all IndexSlots. +``` +def hook(settings, word_dict, label_dict, **kwargs): + settings.word_dict = word_dict + settings.label_dict = label_dict + #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(label_dict))] +``` +The corresponding data iterator is as following: +``` +@provider(init_hook=hook, should_shuffle=True, calc_batch_size=get_batch_size, + can_over_batch_size=False, cache=CacheType.CACHE_PASS_IN_MEM) +def process(settings, file_name): + with open(file_name, 'r') as fdata: + for line in fdata: + sentence, predicate, ctx_n2, ctx_n1, ctx_0, ctx_p1, ctx_p2, mark, label = \ + line.strip().split('\t') + + words = sentence.split() + sen_len = len(words) + word_slot = [settings.word_dict.get(w, UNK_IDX) for w in words] + + predicate_slot = [settings.predicate_dict.get(predicate)] * sen_len + ctx_n2_slot = [settings.word_dict.get(ctx_n2, UNK_IDX)] * sen_len + ctx_n1_slot = [settings.word_dict.get(ctx_n1, UNK_IDX)] * sen_len + ctx_0_slot = [settings.word_dict.get(ctx_0, UNK_IDX)] * sen_len + ctx_p1_slot = [settings.word_dict.get(ctx_p1, UNK_IDX)] * sen_len + ctx_p2_slot = [settings.word_dict.get(ctx_p2, UNK_IDX)] * sen_len + + marks = mark.split() + mark_slot = [int(w) for w in marks] + + 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 +``` +The `process`function yield 9 lists which are 8 features and label. + +### Neural Network Config +`db_lstm.py` is the neural network config file to load the dictionaries and define the data provider module and network architecture during the training procedure. + +Nine `data_layer` load instances from data provider. Eight features are transformed into embedddings respectively, and mixed by `mixed_layer` . Deep bidirectional LSTM layers extract features for the softmax layer. The objective function is cross entropy of labels. + +### Run Training +The script for training is `train.sh`, user just need to execute: +```bash + ./train.sh +``` +The content in `train.sh`: +``` +paddle train \ + --config=./db_lstm.py \ + --use_gpu=0 \ + --log_period=5000 \ + --trainer_count=1 \ + --show_parameter_stats_period=5000 \ + --save_dir=./output \ + --num_passes=10000 \ + --average_test_period=10000000 \ + --init_model_path=./data \ + --load_missing_parameter_strategy=rand \ + --test_all_data_in_one_period=1 \ +2>&1 | tee 'train.log' +``` + +- \--config=./db_lstm.py : network config file. +- \--use_gpu=false: use CPU to train, set true, if you install GPU version of PaddlePaddle and want to use GPU to train, until now crf_layer do not support GPU +- \--log_period=500: print log every 20 batches. +- \--trainer_count=1: set thread number (or GPU count). +- \--show_parameter_stats_period=5000: show parameter statistic every 100 batches. +- \--save_dir=./output: output path to save models. +- \--num_passes=10000: set pass number, one pass in PaddlePaddle means training all samples in dataset one time. +- \--average_test_period=10000000: do test on average parameter every average_test_period batches +- \--init_model_path=./data: parameter initialization path +- \--load_missing_parameter_strategy=rand: random initialization unexisted parameters +- \--test_all_data_in_one_period=1: test all data in one period + + +After training, the models will be saved in directory `output`. Our training curve is as following: +
+![pic](./curve.jpg) +
+ +### Run testing +The script for testing is `test.sh`, user just need to execute: +```bash + ./test.sh +``` +The main part in `tesh.sh` +``` +paddle train \ + --config=./db_lstm.py \ + --model_list=$model_list \ + --job=test \ + --config_args=is_test=1 \ +``` + + - \--config=./db_lstm.py: network config file + - \--model_list=$model_list.list: model list file + - \--job=test: indicate the test job + - \--config_args=is_test=1: flag to indicate test + - \--test_all_data_in_one_period=1: test all data in 1 period + + +### Run prediction +The script for prediction is `predict.sh`, user just need to execute: +```bash + ./predict.sh + +``` +In `predict.sh`, user should offer the network config file, model path, label file, word dictionary file, feature file +``` +python predict.py + -c $config_file \ + -w $best_model_path \ + -l $label_file \ + -p $predicate_dict_file \ + -d $dict_file \ + -i $input_file \ + -o $output_file +``` + +`predict.py` is the main executable python script, which includes functions: load model, load data, data prediction. The network model will output the probability distribution of labels. In the demo, we take the label with maximum probability as result. User can also implement the beam search or viterbi decoding upon the probability distribution matrix. + +After prediction, the result is saved in `predict.res`. + +## Reference +[1] Martha Palmer, Dan Gildea, and Paul Kingsbury. The Proposition Bank: An Annotated Corpus of Semantic Roles , Computational Linguistics, 31(1), 2005. + +[2] Zhou, Jie, and Wei Xu. "End-to-end learning of semantic role labeling using recurrent neural networks." Proceedings of the Annual Meeting of the Association for Computational Linguistics. 2015. diff --git a/doc/demo/sentiment_analysis/bi_lstm.jpg b/doc/tutorials/sentiment_analysis/bi_lstm.jpg similarity index 100% rename from doc/demo/sentiment_analysis/bi_lstm.jpg rename to doc/tutorials/sentiment_analysis/bi_lstm.jpg diff --git a/doc/demo/sentiment_analysis/index.rst b/doc/tutorials/sentiment_analysis/index.rst similarity index 100% rename from doc/demo/sentiment_analysis/index.rst rename to doc/tutorials/sentiment_analysis/index.rst diff --git a/doc/demo/sentiment_analysis/lstm.png b/doc/tutorials/sentiment_analysis/lstm.png similarity index 100% rename from doc/demo/sentiment_analysis/lstm.png rename to doc/tutorials/sentiment_analysis/lstm.png diff --git a/doc/demo/sentiment_analysis/sentiment_analysis.md b/doc/tutorials/sentiment_analysis/sentiment_analysis.md similarity index 100% rename from doc/demo/sentiment_analysis/sentiment_analysis.md rename to doc/tutorials/sentiment_analysis/sentiment_analysis.md diff --git a/doc/demo/sentiment_analysis/stacked_lstm.jpg b/doc/tutorials/sentiment_analysis/stacked_lstm.jpg similarity index 100% rename from doc/demo/sentiment_analysis/stacked_lstm.jpg rename to doc/tutorials/sentiment_analysis/stacked_lstm.jpg diff --git a/doc/demo/text_generation/encoder-decoder-attention-model.png b/doc/tutorials/text_generation/encoder-decoder-attention-model.png similarity index 100% rename from doc/demo/text_generation/encoder-decoder-attention-model.png rename to doc/tutorials/text_generation/encoder-decoder-attention-model.png diff --git a/doc/demo/text_generation/index.rst b/doc/tutorials/text_generation/index.rst similarity index 100% rename from doc/demo/text_generation/index.rst rename to doc/tutorials/text_generation/index.rst diff --git a/doc/demo/text_generation/text_generation.md b/doc/tutorials/text_generation/text_generation.md similarity index 100% rename from doc/demo/text_generation/text_generation.md rename to doc/tutorials/text_generation/text_generation.md diff --git a/doc/ui/api/trainer_config_helpers/attrs.rst b/doc/ui/api/trainer_config_helpers/attrs.rst deleted file mode 100644 index 44919aba90df0b9da7c311a62339052c16c44ad1..0000000000000000000000000000000000000000 --- a/doc/ui/api/trainer_config_helpers/attrs.rst +++ /dev/null @@ -1,5 +0,0 @@ -Parameter and Extra Layer Attribute -=================================== - -.. automodule:: paddle.trainer_config_helpers.attrs - :members: diff --git a/doc/ui/api/trainer_config_helpers/index.rst b/doc/ui/api/trainer_config_helpers/index.rst deleted file mode 100644 index 8395eb75710b3e67ec0c5442f79c999bdacdff42..0000000000000000000000000000000000000000 --- a/doc/ui/api/trainer_config_helpers/index.rst +++ /dev/null @@ -1,14 +0,0 @@ -Model Config Interface -====================== - -.. toctree:: - :maxdepth: 1 - - optimizers.rst - data_sources.rst - layers.rst - activations.rst - poolings.rst - networks.rst - evaluators.rst - attrs.rst diff --git a/doc/ui/index.md b/doc/ui/index.md deleted file mode 100644 index 9c1ba27bdc14fa9ab762ffb97424a8a5946808f9..0000000000000000000000000000000000000000 --- a/doc/ui/index.md +++ /dev/null @@ -1,20 +0,0 @@ -# User Interface - -## Data Provider - -* [Introduction](data_provider/index.rst) -* [PyDataProvider2](data_provider/pydataprovider2.rst) - -## API Reference - -* [Model Config Interface](api/trainer_config_helpers/index.md) - -## Command Line Argument - -* [Use Case](cmd_argument/use_case.md) -* [Argument Outline](cmd_argument/argument_outline.md) -* [Detailed Descriptions](cmd_argument/detail_introduction.md) - -## Predict - -* [Python Prediction API](predict/swig_py_paddle_en.rst) diff --git a/doc/user_guide.rst b/doc/user_guide.rst deleted file mode 100644 index d4deb3ca5a4523b509ea5082f32be8a315570dea..0000000000000000000000000000000000000000 --- a/doc/user_guide.rst +++ /dev/null @@ -1,13 +0,0 @@ -User Guide -========== - -.. toctree:: - :maxdepth: 1 - - demo/quick_start/index_en.md - build/index.rst - build/contribute_to_paddle.md - ui/index.md - ui/api/trainer_config_helpers/index.rst - demo/index.md - cluster/index.md diff --git a/doc_cn/build_and_install/index.rst b/doc_cn/build_and_install/index.rst index 2205e282248c4e7f6d1173be47aadf160554c6be..48163fb36e561fe5fd8f6907379687a8b5c97f68 100644 --- a/doc_cn/build_and_install/index.rst +++ b/doc_cn/build_and_install/index.rst @@ -8,9 +8,7 @@ PaddlePaddle提供数个预编译的二进制来进行安装，包括Docker镜 .. toctree:: :maxdepth: 1 - :glob: - 使用Jumbo安装(对内) <../build/internal/install_from_jumbo.rst> install/docker_install.rst install/ubuntu_install.rst @@ -25,8 +23,5 @@ PaddlePaddle提供数个预编译的二进制来进行安装，包括Docker镜 .. toctree:: :maxdepth: 1 - :glob: - 源码下载(对内) <../build/internal/download_paddle_source_zh_cn.rst> - 从源码编译安装(对内) <../build/internal/build_from_source_zh_cn.rst> cmake/index.rst diff --git a/doc_cn/introduction/index.rst b/doc_cn/introduction/index.rst index f6eb5456c007ca03ea6002109b1f27b8a99faa0f..c996f5f4acd07011c98c3e1086080e85ed7dd1b4 100644 --- a/doc_cn/introduction/index.rst +++ b/doc_cn/introduction/index.rst @@ -1,102 +1,114 @@ -# 简介 +简介 +==== PaddlePaddle是源于百度的一个深度学习平台。这份简短的介绍将向你展示如何利用PaddlePaddle来解决一个经典的线性回归问题。 -## 1. 一个经典的任务 +1. 一个经典的任务 +----------------- -我们展示如何用PaddlePaddle解决单变量的线性回归问题。线性回归的输入是一批点`(x, y) `，其中 `y = wx + b + ε`， 而 ε 是一个符合高斯分布的随机变量。线性回归的输出是从这批点估计出来的参数 w 和 b。 +我们展示如何用PaddlePaddle解决 `单变量的线性回归 `_ 问题。线性回归的输入是一批点 `(x, y)` ，其中 `y = wx + b + ε`， 而 ε 是一个符合高斯分布的随机变量。线性回归的输出是从这批点估计出来的参数 `w` 和 `b` 。 一个例子是房产估值。我们假设房产的价格（y）是其大小（x）的一个线性函数，那么我们可以通过收集市场上房子的大小和价格，用来估计线性函数的参数w 和 b。 -## 2. 准备数据 +2. 准备数据 +----------- 假设变量 `x` 和 `y` 的真实关系为： `y = 2x + 0.3 + ε`，这里展示如何使用观测数据来拟合这一线性关系。首先，Python代码将随机产生2000个观测点，作为线性回归的输入。下面脚本符合PaddlePaddle期待的读取数据的Python程序的模式。 -```python -# dataprovider.py -from paddle.trainer.PyDataProvider2 import * -import random +.. code-block:: python -# 定义输入数据的类型: 2个浮点数 -@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. 训练模型 + # 定义输入数据的类型: 2个浮点数 + @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] + +3. 训练模型 +----------- 为了还原 `y = 2x + 0.3`，我们先从一条随机的直线 `y' = wx + b` 开始，然后利用观测数据调整 `w` 和 `b` 使得 `y'` 和 `y` 的差距不断减小，最终趋于接近。这个过程就是模型的训练过程，而 `w` 和 `b` 就是模型的参数，即我们的训练目标。 在PaddlePaddle里，该模型的网络配置如下。 -```python -# trainer_config.py -from paddle.trainer_config_helpers import * - -# 1. 定义数据来源，调用上面的process函数获得观测数据 -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. 学习算法。控制如何改变模型参数 w 和 b -settings(batch_size=12, learning_rate=1e-3, learning_method=MomentumOptimizer()) - -# 3. 神经网络配置 -x = data_layer(name='x', size=1) -y = data_layer(name='y', size=1) -# 线性计算网络层: ȳ = wx + b -ȳ = fc_layer(input=x, param_attr=ParamAttr(name='w'), size=1, act=LinearActivation(), bias_attr=ParamAttr(name='b')) -# 计算误差函数，即 ȳ 和真实 y 之间的距离 -cost = regression_cost(input= ȳ, label=y) -outputs(cost) -``` +.. code-block:: python + + # trainer_config.py + from paddle.trainer_config_helpers import * + + # 1. 定义数据来源，调用上面的process函数获得观测数据 + 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. 学习算法。控制如何改变模型参数 w 和 b + settings(batch_size=12, learning_rate=1e-3, learning_method=MomentumOptimizer()) + + # 3. 神经网络配置 + x = data_layer(name='x', size=1) + y = data_layer(name='y', size=1) + # 线性计算网络层: ȳ = wx + b + ȳ = fc_layer(input=x, param_attr=ParamAttr(name='w'), size=1, act=LinearActivation(), bias_attr=ParamAttr(name='b')) + # 计算误差函数，即 ȳ 和真实 y 之间的距离 + cost = regression_cost(input= ȳ, label=y) + outputs(cost) + 这段简短的配置展示了PaddlePaddle的基本用法： -- 第一部分定义了数据输入。一般情况下，PaddlePaddle先从一个文件列表里获得数据文件地址，然后交给用户自定义的函数（例如上面的`process`函数）进行读入和预处理从而得到真实输入。本文中由于输入数据是随机生成的不需要读输入文件，所以放一个空列表（`empty.list`）即可。 +- 第一部分定义了数据输入。一般情况下，PaddlePaddle先从一个文件列表里获得数据文件地址，然后交给用户自定义的函数（例如上面的 `process`函数）进行读入和预处理从而得到真实输入。本文中由于输入数据是随机生成的不需要读输入文件，所以放一个空列表（`empty.list`）即可。 - 第二部分主要是选择学习算法，它定义了模型参数改变的规则。PaddlePaddle提供了很多优秀的学习算法，这里使用一个基于momentum的随机梯度下降(SGD)算法，该算法每批量(batch)读取12个采样数据进行随机梯度计算来更新更新。 - 最后一部分是神经网络的配置。由于PaddlePaddle已经实现了丰富的网络层，所以很多时候你需要做的只是定义正确的网络层并把它们连接起来。这里使用了三种网络单元： + - **数据层**：数据层 `data_layer` 是神经网络的入口，它读入数据并将它们传输到接下来的网络层。这里数据层有两个，分别对应于变量 `x` 和 `y`。 - **全连接层**：全连接层 `fc_layer` 是基础的计算单元，这里利用它建模变量之间的线性关系。计算单元是神经网络的核心，PaddlePaddle支持大量的计算单元和任意深度的网络连接，从而可以拟合任意的函数来学习复杂的数据关系。 - - **回归误差代价层**：回归误差代价层 `regression_cost`是众多误差代价函数层的一种，它们在训练过程作为网络的出口，用来计算模型的误差，是模型参数优化的目标函数。 + - **回归误差代价层**：回归误差代价层 `regression_cost` 是众多误差代价函数层的一种，它们在训练过程作为网络的出口，用来计算模型的误差，是模型参数优化的目标函数。 + +定义了网络结构并保存为 `trainer_config.py` 之后，运行以下训练命令： + +.. code-block:: bash -定义了网络结构并保存为`trainer_config.py`之后，运行以下训练命令： - ``` - paddle train --config=trainer_config.py --save_dir=./output --num_passes=30 - ``` + paddle train --config=trainer_config.py --save_dir=./output --num_passes=30 PaddlePaddle将在观测数据集上迭代训练30轮，并将每轮的模型结果存放在 `./output` 路径下。从输出日志可以看到，随着轮数增加误差代价函数的输出在不断的减小，这意味着模型在训练数据上不断的改进，直到逼近真实解：` y = 2x + 0.3 ` -## 4. 模型检验 +4. 模型检验 +----------- 训练完成后，我们希望能够检验模型的好坏。一种常用的做法是用学习的模型对另外一组测试数据进行预测，评价预测的效果。在这个例子中，由于已经知道了真实答案，我们可以直接观察模型的参数是否符合预期来进行检验。 PaddlePaddle将每个模型参数作为一个numpy数组单独存为一个文件，所以可以利用如下方法读取模型的参数。 -```python -import numpy as np -import os +.. code-block:: python -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) + 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 -``` -
![](./parameters.png)
+ print 'w=%.6f, b=%.6f' % (load('output/pass-00029/w'), load('output/pass-00029/b')) + # w=1.999743, b=0.300137 + +.. image:: ./parameters.png + :align: center + :scale: 80 % 从图中可以看到，虽然 `w` 和 `b` 都使用随机值初始化，但在起初的几轮训练中它们都在快速逼近真实值，并且后续仍在不断改进，使得最终得到的模型几乎与真实模型一致。 -这样，我们用PaddlePaddle解决了单变量线性回归问题， 包括数据输入，模型训练和最后的结果验证。 +这样，我们用PaddlePaddle解决了单变量线性回归问题， 包括数据输入、模型训练和最后的结果验证。 -## 5. 推荐后续阅读 +5. 推荐后续阅读 +--------------- -- 安装/编译：PaddlePaddle的安装与编译文档。 -- 快速入门 ：使用商品评论分类任务，系统性的介绍如何一步步改进，最终得到产品级的深度模型。 -- 示例：各种实用案例，涵盖图像、文本、推荐等多个领域。 +- `安装/编译 <../build_and_install/index.html>`_ ：PaddlePaddle的安装与编译文档。 +- `快速入门 <../demo/quick_start/index.html>`_ ：使用商品评论分类任务，系统性的介绍如何一步步改进，最终得到产品级的深度模型。 +- `示例 <../demo/index.html>`_ ：各种实用案例，涵盖图像、文本、推荐等多个领域。 \ No newline at end of file diff --git a/doc_cn/ui/predict/swig_py_paddle.rst b/doc_cn/ui/predict/swig_py_paddle.rst index 89031dd72f5065b6919d873f5611a5e94e8b62e3..05f25345c5246687363dee1931310120b5723d0b 100644 --- a/doc_cn/ui/predict/swig_py_paddle.rst +++ b/doc_cn/ui/predict/swig_py_paddle.rst @@ -9,15 +9,24 @@ PaddlePaddle使用swig对常用的预测接口进行了封装，通过编译会 基于Python的模型预测，主要包括以下五个步骤。 1. 初始化PaddlePaddle环境 - 在程序开始阶段，通过调用 ``swig_paddle.initPaddle()`` 并传入相应的命令行参数初始化PaddlePaddle。 + + 在程序开始阶段，通过调用 ``swig_paddle.initPaddle()`` 并传入相应的命令行参数初始化PaddlePaddle。 + 2. 解析模型配置文件 - 初始化之后，可以通过调用 ``parse_config()`` 解析训练模型时用的配置文件。注意预测数据通常不包含label, 同时预测网络通常直接输出最后一层的结果而不是像训练网络一样再接一层cost layer，所以一般需要对训练用的模型配置文件稍作相应修改才能在预测时使用。 + + 初始化之后，可以通过调用 ``parse_config()`` 解析训练模型时用的配置文件。注意预测数据通常不包含label, 同时预测网络通常直接输出最后一层的结果而不是像训练网络一样再接一层cost layer，所以一般需要对训练用的模型配置文件稍作相应修改才能在预测时使用。 + 3. 构造paddle.GradientMachine - 通过调用 ``swig_paddle.GradientMachine.createFromConfigproto()`` 传入上一步解析出来的模型配置就可以创建一个 ``GradientMachine``。 + + 通过调用 ``swig_paddle.GradientMachine.createFromConfigproto()`` 传入上一步解析出来的模型配置就可以创建一个 ``GradientMachine``。 + 4. 准备预测数据 - swig_paddle中的预测接口的参数是自定义的C++数据类型，py_paddle里面提供了一个工具类 ``DataProviderConverter`` 可以用于接收和PyDataProvider2一样的输入数据并转换成预测接口所需的数据类型。 + + swig_paddle中的预测接口的参数是自定义的C++数据类型，py_paddle里面提供了一个工具类 ``DataProviderConverter`` 可以用于接收和PyDataProvider2一样的输入数据并转换成预测接口所需的数据类型。 + 5. 模型预测 - 通过调用 ``forwardTest()`` 传入预测数据，直接返回计算结果。 + + 通过调用 ``forwardTest()`` 传入预测数据，直接返回计算结果。 预测Demo @@ -34,7 +43,8 @@ Demo预测输出如下，其中value即为softmax层的输出。由于TEST_DATA .. code-block:: text - [{'id': None, 'value': array([[ 5.53018653e-09, 1.12194102e-05, 1.96644767e-09, + [{'id': None, 'value': array( + [[ 5.53018653e-09, 1.12194102e-05, 1.96644767e-09, 1.43630644e-02, 1.51111044e-13, 9.85625684e-01, 2.08823112e-10, 2.32777140e-08, 2.00186201e-09, 1.15501715e-08], diff --git a/paddle/scripts/docker/Dockerfile.m4 b/paddle/scripts/docker/Dockerfile.m4 index e14493ed9e842351125ab458db53fcc3f38233f6..761aa975d693631556c162dc29ae288ad6bd980b 100644 --- a/paddle/scripts/docker/Dockerfile.m4 +++ b/paddle/scripts/docker/Dockerfile.m4 @@ -1,7 +1,7 @@ FROM PADDLE_BASE_IMAGE MAINTAINER PaddlePaddle Dev Team COPY build.sh /root/ -ENV GIT_CHECKOUT=v0.9.0a0 +ENV GIT_CHECKOUT=v0.9.0 ENV WITH_GPU=PADDLE_WITH_GPU ENV IS_DEVEL=PADDLE_IS_DEVEL ENV WITH_DEMO=PADDLE_WITH_DEMO diff --git a/paddle/scripts/submit_local.sh.in b/paddle/scripts/submit_local.sh.in index 20ea2fedc4d464cdd5403af28bc917770c993b98..ace2c0dee972e338001a0e5a4045c32e64ff157e 100644 --- a/paddle/scripts/submit_local.sh.in +++ b/paddle/scripts/submit_local.sh.in @@ -29,6 +29,7 @@ function version(){ } function ver2num() { + set -e # convert version to number. if [ -z "$1" ]; then # empty argument printf "%03d%03d%03d%03d%03d" 0 @@ -41,6 +42,7 @@ function ver2num() { printf "%03d%03d%03d%03d%03d" $VERN fi fi + set +e } PADDLE_CONF_HOME="$HOME/.config/paddle"