diff --git a/doc/getstarted/quickstart_cn.rst b/doc/getstarted/quickstart_cn.rst index 51dd00f1e806e6423afe3ce53d80d53a187d2ca0..d511cead262dabafd095f68adb5ffc596a7fe596 100644 --- a/doc/getstarted/quickstart_cn.rst +++ b/doc/getstarted/quickstart_cn.rst @@ -1,6 +1,9 @@ 快速开始 ======== +快速安装 +-------- + PaddlePaddle支持使用pip快速安装，目前支持CentOS 6以上, Ubuntu 14.04以及MacOS 10.12，并安装有Python2.7。 执行下面的命令完成快速安装，版本为cpu_avx_openblas： @@ -16,6 +19,9 @@ PaddlePaddle支持使用pip快速安装，目前支持CentOS 6以上, Ubuntu 14. 更详细的安装和编译方法参考：:ref:`install_steps` 。 +快速使用 +-------- + 创建一个 housing.py 并粘贴此Python代码： .. code-block:: python diff --git a/doc/getstarted/quickstart_en.rst b/doc/getstarted/quickstart_en.rst index d1bcf82ea071e2c53760a5ccf6a5074a3ac0abd5..70f7fe0646068aa79cd72955c6848ac0250c2300 100644 --- a/doc/getstarted/quickstart_en.rst +++ b/doc/getstarted/quickstart_en.rst @@ -1,6 +1,9 @@ Quick Start ============ +Quick Install +------------- + You can use pip to install PaddlePaddle with a single command, supports CentOS 6 above, Ubuntu 14.04 above or MacOS 10.12, with Python 2.7 installed. Simply run the following command to install, the version is cpu_avx_openblas: @@ -17,6 +20,9 @@ If you need to install GPU version (cuda7.5_cudnn5_avx_openblas), run: For more details about installation and build: :ref:`install_steps` . +Quick Use +--------- + Create a new file called housing.py, and paste this Python code: diff --git a/doc/howto/cluster/index_cn.rst b/doc/howto/cluster/index_cn.rst index c68b2655b65b192814b94f0013fa92b0733b9afa..a60521b4a9646bdc6d9f1bf6da482acc989d8bf3 100644 --- a/doc/howto/cluster/index_cn.rst +++ b/doc/howto/cluster/index_cn.rst @@ -1,10 +1,22 @@ 分布式训练 ========== +本节将介绍如何使用PaddlePaddle在不同的集群框架下完成分布式训练。分布式训练架构如下图所示： + +.. image:: src/ps_cn.png + :width: 500 + +- 数据分片（Data shard): 用于训练神经网络的数据，被切分成多个部分，每个部分分别给每个trainer使用。 +- 计算节点（Trainer）: 每个trainer启动后读取切分好的一部分数据，开始神经网络的“前馈”和“后馈”计算，并和参数服务器通信。在完成一定量数据的训练后，上传计算得出的梯度（gradients），然后下载优化更新后的神经网络参数（parameters）。 +- 参数服务器（Parameter server）:每个参数服务器只保存整个神经网络所有参数的一部分。参数服务器接收从计算节点上传的梯度，并完成参数优化更新，再将更新后的参数下发到每个计算节点。 + +这样，通过计算节点和参数服务器的分布式协作，可以完成神经网络的SGD方法的训练。PaddlePaddle可以同时支持同步随机梯度下降（SGD）和异步随机梯度下降。 + +在使用同步SGD训练神经网络时，PaddlePaddle使用同步屏障（barrier），使梯度的提交和参数的更新按照顺序方式执行。在异步SGD中，则并不会等待所有trainer提交梯度才更新参数，这样极大地提高了计算的并行性：参数服务器之间不相互依赖，并行地接收梯度和更新参数，参数服务器也不会等待计算节点全部都提交梯度之后才开始下一步，计算节点之间也不会相互依赖，并行地执行模型的训练。可以看出，虽然异步SGD方式会提高参数更新并行度, 但是并不能保证参数同步更新，在任意时间某一台参数服务器上保存的参数可能比另一台要更新，与同步SGD相比，梯度会有噪声。 + .. toctree:: :maxdepth: 1 - introduction_cn.md preparations_cn.md cmd_argument_cn.md multi_cluster/index_cn.rst diff --git a/doc/howto/cluster/index_en.rst b/doc/howto/cluster/index_en.rst index af957e06cd7930ce63569a1bafdde47a1d34eb69..2640a09dcc904619bc97c9bd3f3d81a9dc307663 100644 --- a/doc/howto/cluster/index_en.rst +++ b/doc/howto/cluster/index_en.rst @@ -1,10 +1,22 @@ Distributed Training ==================== +In this section, we'll explain how to run distributed training jobs with PaddlePaddle on different types of clusters. The diagram below shows the main architecture of a distributed trainning job: + +.. image:: src/ps_en.png + :width: 500 + +- Data shard: training data will be split into multiple partitions, trainers use the partitions of the whole dataset to do the training job. +- Trainer: each trainer reads the data shard, and train the neural network. Then the trainer will upload calculated "gradients" to parameter servers, and wait for parameters to be optimized on the parameter server side. When that finishes, the trainer download optimized parameters and continues its training. +- Parameter server: every parameter server stores part of the whole neural network model data. They will do optimization calculations when gradients are uploaded from trainers, and then send updated parameters to trainers. + +PaddlePaddle can support both synchronize stochastic gradient descent (SGD) and asynchronous SGD. + +When training with synchronize SGD, PaddlePaddle uses an internal "synchronize barrier" which makes gradients update and parameter download in strict order. On the other hand, asynchronous SGD won't wait for all trainers to finish upload at a single step, this will increase the parallelism of distributed training: parameter servers do not depend on each other, they'll do parameter optimization concurrently. Parameter servers will not wait for trainers, so trainers will also do their work concurrently. But asynchronous SGD will introduce more randomness and noises in the gradient. + .. toctree:: :maxdepth: 1 - introduction_en.md preparations_en.md cmd_argument_en.md multi_cluster/index_en.rst diff --git a/doc/howto/cluster/introduction_cn.md b/doc/howto/cluster/introduction_cn.md deleted file mode 100644 index 562008a898414a6566d74d08cfeb18fb9d57582a..0000000000000000000000000000000000000000 --- a/doc/howto/cluster/introduction_cn.md +++ /dev/null @@ -1,13 +0,0 @@ -## 概述 - -本节将介绍如何使用PaddlePaddle在不同的集群框架下完成分布式训练。分布式训练架构如下图所示： - - - -- 数据分片（Data shard): 用于训练神经网络的数据，被切分成多个部分，每个部分分别给每个trainer使用。 -- 计算节点（Trainer）: 每个trainer启动后读取切分好的一部分数据，开始神经网络的“前馈”和“后馈”计算，并和参数服务器通信。在完成一定量数据的训练后，上传计算得出的梯度（gradients），然后下载优化更新后的神经网络参数（parameters）。 -- 参数服务器（Parameter server）:每个参数服务器只保存整个神经网络所有参数的一部分。参数服务器接收从计算节点上传的梯度，并完成参数优化更新，再将更新后的参数下发到每个计算节点。 - -这样，通过计算节点和参数服务器的分布式协作，可以完成神经网络的SGD方法的训练。PaddlePaddle可以同时支持同步随机梯度下降（SGD）和异步随机梯度下降。 - -在使用同步SGD训练神经网络时，PaddlePaddle使用同步屏障（barrier），使梯度的提交和参数的更新按照顺序方式执行。在异步SGD中，则并不会等待所有trainer提交梯度才更新参数，这样极大地提高了计算的并行性：参数服务器之间不相互依赖，并行地接收梯度和更新参数，参数服务器也不会等待计算节点全部都提交梯度之后才开始下一步，计算节点之间也不会相互依赖，并行地执行模型的训练。可以看出，虽然异步SGD方式会提高参数更新并行度, 但是并不能保证参数同步更新，在任意时间某一台参数服务器上保存的参数可能比另一台要更新，与同步SGD相比，梯度会有噪声。 diff --git a/doc/howto/cluster/introduction_en.md b/doc/howto/cluster/introduction_en.md deleted file mode 100644 index eb70d7cf35ab729e0da4c6a3a2e732c26905f584..0000000000000000000000000000000000000000 --- a/doc/howto/cluster/introduction_en.md +++ /dev/null @@ -1,13 +0,0 @@ -## Introduction - -In this section, we'll explain how to run distributed training jobs with PaddlePaddle on different types of clusters. The diagram below shows the main architecture of a distributed trainning job: - - - -- Data shard: training data will be split into multiple partitions, trainers use the partitions of the whole dataset to do the training job. -- Trainer: each trainer reads the data shard, and train the neural network. Then the trainer will upload calculated "gradients" to parameter servers, and wait for parameters to be optimized on the parameter server side. When that finishes, the trainer download optimized parameters and continues its training. -- Parameter server: every parameter server stores part of the whole neural network model data. They will do optimization calculations when gradients are uploaded from trainers, and then send updated parameters to trainers. - -PaddlePaddle can support both synchronize stochastic gradient descent (SGD) and asynchronous SGD. - -When training with synchronize SGD, PaddlePaddle uses an internal "synchronize barrier" which makes gradients update and parameter download in strict order. On the other hand, asynchronous SGD won't wait for all trainers to finish upload at a single step, this will increase the parallelism of distributed training: parameter servers do not depend on each other, they'll do parameter optimization concurrently. Parameter servers will not wait for trainers, so trainers will also do their work concurrently. But asynchronous SGD will introduce more randomness and noises in the gradient. diff --git a/doc/howto/cluster/src/ps_cn.png b/doc/howto/cluster/src/ps_cn.png new file mode 100644 index 0000000000000000000000000000000000000000..f9525739cc8bc6506adde642aafa0a85ae3ebebc Binary files /dev/null and b/doc/howto/cluster/src/ps_cn.png differ diff --git a/doc/howto/cluster/src/ps_en.png b/doc/howto/cluster/src/ps_en.png new file mode 100644 index 0000000000000000000000000000000000000000..6537d3d56589ca9f19a77a50a970e4b5275e6ce0 Binary files /dev/null and b/doc/howto/cluster/src/ps_en.png differ