ppt_notes/00_2_PaddlePaddle快速入门.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "# PaddlePaddle快速入门\n",
    "![](https://ai-studio-static-online.cdn.bcebos.com/466b717826ff438a917074f4e62ca844e5f291ca1f8b48d7b78838d513ad7536)\n",
    "\n",
    "学会使用深度学习框架是解决深度学习任务的基本能力，这里我们推荐成熟易用的国产开源框架PaddlePaddle。以下内容将带你你快速入门PaddlePaddle，你可以尝试跑通一个小demo来熟悉PaddlePaddle的基本命令。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "# PaddlePaddle基础命令\n",
    "PaddlePaddle是百度开源的深度学习框架，类似的深度学习框架还有谷歌的Tensorflow、Facebook的Pytorch等，在入门深度学习时，学会并使用一门常见的框架，可以让学习效率大大提升。在PaddlePaddle中，计算的对象是张量，我们可以先使用PaddlePaddle来计算一个[[1, 1], [1, 1]] * [[1, 1], [1, 1]]。\n",
    "\n",
    "## 计算常量的加法：1+1\n",
    "首先导入PaddlePaddle库\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'2.1.0'"
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import paddle\n",
    "\n",
    "paddle.__version__"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "定义两个张量的常量x1和x2，并指定它们的形状是[2, 2]，并赋值为1铺满整个张量，类型为int64."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# 定义两个张量\n",
    "x1 = paddle.ones([2,2], dtype='int64')\n",
    "x2 = paddle.ones([2,2], dtype='int64')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "接着定义一个操作，该计算是将上面两个张量进行加法计算，并返回一个求和的算子。PaddlePaddle提供了大量的操作，比如加减乘除、三角函数等。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Tensor(shape=[2, 2], dtype=int64, place=CPUPlace, stop_gradient=True,\n",
      "       [[2, 2],\n",
      "        [2, 2]])\n"
     ]
    }
   ],
   "source": [
    "# 将两个张量求和\n",
    "y1 = paddle.add(x1, x2)\n",
    "\n",
    "# 查看结果\n",
    "print(y1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "# 使用PaddlePaddle做线性回归\n",
    "\n",
    "在上面的教学中，教大家学会用PaddlePaddle做基本的算子运算，下面来教大家如何用PaddlePaddle来做简单的线性回归，包括从定义网络到使用自定义的数据进行训练，最后验证我们网络的预测能力。\n",
    "\n",
    "首先导入PaddlePaddle库和一些工具类库。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'2.1.0'"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import paddle\n",
    "import numpy as np\n",
    "\n",
    "paddle.__version__"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "我们使用numpy定义一组数据，这组数据的每一条数据有13个，为了做示例，其中除了第一个数外都填充了0。这组数据是符合y = 2 * x + 1，但是程序是不知道的，我们之后使用这组数据进行训练，看看强大的神经网络是否能够训练出一个拟合这个函数的模型。最后定义了一个预测数据，是在训练完成，使用这个数据作为x输入，看是否能够预测于正确值相近结果。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# 定义训练和测试数据\n",
    "x_data = np.array([[1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], \n",
    "                   [2.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], \n",
    "                   [3.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], \n",
    "                   [4.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], \n",
    "                   [5.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]]).astype('float32')\n",
    "y_data = np.array([[3.0], [5.0], [7.0], [9.0], [11.0]]).astype('float32')\n",
    "test_data = np.array([[6.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]]).astype('float32')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "定义一个简单的线性网络，这个网络非常简单，结构是：\n",
    "```text\n",
    "[输入层] --> [隐层] --> [激活函数] --> [输出层]\n",
    "```\n",
    "\n",
    "更具体的就是一个输出大小为100的全连接层、之后接激活函数ReLU和一个输出大小为1的全连接层，就这样构建了一个非常简单的网络。\n",
    "\n",
    "这里定义输入层的形状为13，这是因为波士顿房价数据集的每条数据有13个属性，我们之后自定义的数据集也是为了符合这一个维度。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# 定义一个简单的线性网络\n",
    "net = paddle.nn.Sequential(\n",
    "    paddle.nn.Linear(13, 100),\n",
    "    paddle.nn.ReLU(),\n",
    "    paddle.nn.Linear(100, 1)\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "接着是定义训练使用的优化方法，这里使用的是随机梯度下降优化方法。PaddlePaddle提供了大量的优化函数接口，除了本项目使用的随机梯度下降法（SGD），还有Momentum、Adagrad、Adagrad等等，读者可以更加自己项目的需求使用不同的优化方法。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# 定义优化方法\n",
    "optimizer = paddle.optimizer.SGD(learning_rate=0.01, parameters=net.parameters())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "我们可以使用数据进行训练了，我们这次训练了10个pass，读者可根据情况设置更多的训练轮数，通常来说训练的次数和模型收敛有一定的关系。\n",
    "\n",
    "因为本项目是一个线性回归任务，所以我们在训练的时候使用的是平方差损失函数。因为`paddle.nn.functional.square_error_cost`求的是一个Batch的损失值，所以我们还要对他求一个平均值。PaddlePaddle提供了很多的损失函数的接口，比如交叉熵损失函数`paddle.nn.CrossEntropyLoss`。\n",
    "\n",
    "在训练过程中，我们可以看到输出的损失值在不断减小，证明我们的模型在不断收敛。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Pass:0, Cost:56.83829\n",
      "Pass:1, Cost:16.90196\n",
      "Pass:2, Cost:2.45377\n",
      "Pass:3, Cost:0.06338\n",
      "Pass:4, Cost:0.03584\n",
      "Pass:5, Cost:0.03505\n",
      "Pass:6, Cost:0.03432\n",
      "Pass:7, Cost:0.03361\n",
      "Pass:8, Cost:0.03291\n",
      "Pass:9, Cost:0.03223\n"
     ]
    }
   ],
   "source": [
    "# 将numpy类型数据转换成tensor之后才能用于模型训练\n",
    "inputs = paddle.to_tensor(x_data)\n",
    "labels = paddle.to_tensor(y_data)\n",
    "\n",
    "# 开始训练100个pass\n",
    "for pass_id in range(10):\n",
    "    out = net(inputs)\n",
    "    loss = paddle.mean(paddle.nn.functional.square_error_cost(out, labels))\n",
    "\n",
    "    loss.backward()\n",
    "    optimizer.step()\n",
    "    optimizer.clear_grad()\n",
    "\n",
    "    print(\"Pass:%d, Cost:%0.5f\" % (pass_id, loss))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "训练完成之后，我们使用上面克隆主程序得到的预测程序了预测我们刚才定义的预测数据。根据我们上面定义数据时，满足规律y = 2 * x + 1，所以当x为6时，y应该时13，最后输出的结果也是应该接近13的。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "当x为6.0时，y为：13.27442\n"
     ]
    }
   ],
   "source": [
    "# 开始预测\n",
    "predict_inputs = paddle.to_tensor(test_data)\n",
    "result = net(predict_inputs)\n",
    "\n",
    "print(\"当x为6.0时，y为：%0.5f\" % result)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "# 快来上手第一个Demo吧！——用PaddlePaddle做房价预测"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "# **Step1：准备数据。**\n",
    "\n",
    "- 1. uci-housing数据集介绍\n",
    "\n",
    "数据集共506行,每行14列。前13列用来描述房屋的各种信息，最后一列为该类房屋价格中位数。\n",
    "\n",
    "PaddlePaddle提供了读取uci_housing数据集的接口，`paddle.text.datasets.UCIHousing`。\n",
    "\n",
    "- 2. 数据集加载\n",
    "\n",
    "PaddlePaddle中使用`paddle.io.DataLoader`来进行数据的加载操作，通过参数batch_size控制批次大小，shuffle控制是否打乱顺序。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Cache file /home/aistudio/.cache/paddle/dataset/uci_housing/housing.data not found, downloading http://paddlemodels.bj.bcebos.com/uci_housing/housing.data \n",
      "Begin to download\n",
      "............\n",
      "Download finished\n"
     ]
    }
   ],
   "source": [
    "# 导入基本的库\n",
    "import os\n",
    "import paddle\n",
    "import numpy as np\n",
    "\n",
    "\n",
    "BATCH_SIZE=20\n",
    "\n",
    "train_dataset = paddle.text.datasets.UCIHousing(mode='train')\n",
    "valid_dataset = paddle.text.datasets.UCIHousing(mode='test')\n",
    "\n",
    "#用于训练的数据加载器，每次随机读取批次大小的数据，剩余不满足批大小的数据丢弃\n",
    "train_loader = paddle.io.DataLoader(train_dataset, batch_size=BATCH_SIZE, shuffle=True, drop_last=True)\n",
    "\n",
    "#用于测试的数据加载器，每次随机读取批次大小的数据\n",
    "valid_loader = paddle.io.DataLoader(valid_dataset, batch_size=BATCH_SIZE, shuffle=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "打印查看uci_housing数据"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(array([-0.0405441 ,  0.06636363, -0.32356226, -0.06916996, -0.03435197,\n",
      "        0.05563625, -0.03475696,  0.02682186, -0.37171334, -0.21419305,\n",
      "       -0.33569506,  0.10143217, -0.21172912], dtype=float32), array([24.], dtype=float32))\n"
     ]
    }
   ],
   "source": [
    "#用于打印，查看uci_housing数据\r\n",
    "print(train_dataset[0])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "# **Step2:网络配置**\n",
    "\n",
    "**（1）网络搭建**：对于线性回归来讲，它就是一个从输入到输出的简单的全连接层。\n",
    "\n",
    "对于波士顿房价数据集，假设属性和房价之间的关系可以被属性间的线性组合描述。\n",
    "\n",
    "![](https://ai-studio-static-online.cdn.bcebos.com/f3f567e7aa7a499fb3abb767aaaa1b943eed373cf4694d87beb382e161ea8edc)\n",
    "\n",
    "![](https://ai-studio-static-online.cdn.bcebos.com/f6c44a96e1624828829bbd438c29c17ead9ecc45c68b4310bfbbb0a4dc96c3fe)\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# 输入数据形状为[13]，输出形状[1]\r\n",
    "net = paddle.nn.Linear(13, 1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "**(2)定义损失函数**\n",
    "\n",
    "此处使用均方差损失函数。\n",
    "\n",
    "square_error_cost(input,lable):接受输入预测值和目标值，并返回方差估计,即为（y-y_predict）的平方"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "**(3)定义优化函数**\n",
    "\n",
    "此处使用的是随机梯度下降。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "optimizer = paddle.optimizer.SGD(learning_rate=0.001, parameters=net.parameters())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "# **Step3.模型训练** and **Step4.模型评估**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "**（1）定义绘制训练过程的损失值变化趋势的方法draw_train_process**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/matplotlib/__init__.py:107: DeprecationWarning: Using or importing the ABCs from 'collections' instead of from 'collections.abc' is deprecated, and in 3.8 it will stop working\n",
      "  from collections import MutableMapping\n",
      "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/matplotlib/rcsetup.py:20: DeprecationWarning: Using or importing the ABCs from 'collections' instead of from 'collections.abc' is deprecated, and in 3.8 it will stop working\n",
      "  from collections import Iterable, Mapping\n",
      "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/matplotlib/colors.py:53: DeprecationWarning: Using or importing the ABCs from 'collections' instead of from 'collections.abc' is deprecated, and in 3.8 it will stop working\n",
      "  from collections import Sized\n"
     ]
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\r\n",
    "\r\n",
    "\r\n",
    "iter = 0\r\n",
    "iters = []\r\n",
    "train_costs = []\r\n",
    "\r\n",
    "def draw_train_process(iters, train_costs):\r\n",
    "    title=\"training cost\"\r\n",
    "    plt.title(title, fontsize=24)\r\n",
    "    plt.xlabel(\"iter\", fontsize=14)\r\n",
    "    plt.ylabel(\"cost\", fontsize=14)\r\n",
    "    plt.plot(iters, train_costs, color='red', label='training cost') \r\n",
    "    plt.grid()\r\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "**（2）训练并保存模型**\n",
    "\n",
    "遍历轮次和数据集loader，将批次数据送入net里面进行计算，最终经过loss计算，在进行反向传播和参数优化。\n",
    "\n",
    "注：enumerate() 函数用于将一个可遍历的数据对象(如列表、元组或字符串)组合为一个索引序列，同时列出数据和数据下标，"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Pass:0, Cost:694.04108\n",
      "Test:0, Cost:211.20151\n",
      "Pass:1, Cost:718.70618\n",
      "Test:1, Cost:81.96242\n",
      "Pass:2, Cost:475.78058\n",
      "Test:2, Cost:176.78989\n",
      "Pass:3, Cost:402.72916\n",
      "Test:3, Cost:251.75000\n",
      "Pass:4, Cost:502.68036\n",
      "Test:4, Cost:113.36610\n",
      "Pass:5, Cost:568.16602\n",
      "Test:5, Cost:177.52846\n",
      "Pass:6, Cost:476.13885\n",
      "Test:6, Cost:99.38324\n",
      "Pass:7, Cost:263.62946\n",
      "Test:7, Cost:131.53128\n",
      "Pass:8, Cost:408.37491\n",
      "Test:8, Cost:165.41617\n",
      "Pass:9, Cost:348.94781\n",
      "Test:9, Cost:125.17019\n",
      "Pass:10, Cost:235.03036\n",
      "Test:10, Cost:128.47586\n",
      "Pass:11, Cost:373.77365\n",
      "Test:11, Cost:109.12988\n",
      "Pass:12, Cost:382.86288\n",
      "Test:12, Cost:52.62062\n",
      "Pass:13, Cost:228.76428\n",
      "Test:13, Cost:135.51625\n",
      "Pass:14, Cost:147.32423\n",
      "Test:14, Cost:23.88525\n",
      "Pass:15, Cost:266.08676\n",
      "Test:15, Cost:51.34695\n",
      "Pass:16, Cost:187.61273\n",
      "Test:16, Cost:218.02568\n",
      "Pass:17, Cost:201.46187\n",
      "Test:17, Cost:36.58718\n",
      "Pass:18, Cost:164.64169\n",
      "Test:18, Cost:136.90654\n",
      "Pass:19, Cost:109.09910\n",
      "Test:19, Cost:43.69257\n",
      "Pass:20, Cost:165.95181\n",
      "Test:20, Cost:67.69698\n",
      "Pass:21, Cost:67.52733\n",
      "Test:21, Cost:100.40565\n",
      "Pass:22, Cost:148.87378\n",
      "Test:22, Cost:44.28013\n",
      "Pass:23, Cost:272.91302\n",
      "Test:23, Cost:64.55545\n",
      "Pass:24, Cost:320.36285\n",
      "Test:24, Cost:8.77742\n",
      "Pass:25, Cost:181.69868\n",
      "Test:25, Cost:36.35912\n",
      "Pass:26, Cost:68.41258\n",
      "Test:26, Cost:39.68889\n",
      "Pass:27, Cost:119.07393\n",
      "Test:27, Cost:21.35777\n",
      "Pass:28, Cost:61.71670\n",
      "Test:28, Cost:5.72559\n",
      "Pass:29, Cost:53.73602\n",
      "Test:29, Cost:25.21531\n",
      "Pass:30, Cost:81.89694\n",
      "Test:30, Cost:40.26062\n",
      "Pass:31, Cost:51.65599\n",
      "Test:31, Cost:1.77657\n",
      "Pass:32, Cost:18.46457\n",
      "Test:32, Cost:18.31264\n",
      "Pass:33, Cost:58.57187\n",
      "Test:33, Cost:15.08143\n",
      "Pass:34, Cost:102.43970\n",
      "Test:34, Cost:13.90950\n",
      "Pass:35, Cost:81.94363\n",
      "Test:35, Cost:16.54850\n",
      "Pass:36, Cost:43.64971\n",
      "Test:36, Cost:30.42578\n",
      "Pass:37, Cost:59.23506\n",
      "Test:37, Cost:22.46311\n",
      "Pass:38, Cost:91.55733\n",
      "Test:38, Cost:33.10298\n",
      "Pass:39, Cost:50.88599\n",
      "Test:39, Cost:5.41668\n",
      "Pass:40, Cost:17.12974\n",
      "Test:40, Cost:29.49467\n",
      "Pass:41, Cost:34.43634\n",
      "Test:41, Cost:9.65246\n",
      "Pass:42, Cost:31.48676\n",
      "Test:42, Cost:1.60214\n",
      "Pass:43, Cost:91.27415\n",
      "Test:43, Cost:6.00659\n",
      "Pass:44, Cost:49.95349\n",
      "Test:44, Cost:8.52874\n",
      "Pass:45, Cost:19.81659\n",
      "Test:45, Cost:100.12550\n",
      "Pass:46, Cost:62.98671\n",
      "Test:46, Cost:97.64473\n",
      "Pass:47, Cost:65.81707\n",
      "Test:47, Cost:5.76155\n",
      "Pass:48, Cost:76.49935\n",
      "Test:48, Cost:125.73111\n",
      "Pass:49, Cost:20.85202\n",
      "Test:49, Cost:1.26879\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/matplotlib/cbook/__init__.py:2349: DeprecationWarning: Using or importing the ABCs from 'collections' instead of from 'collections.abc' is deprecated, and in 3.8 it will stop working\n",
      "  if isinstance(obj, collections.Iterator):\n",
      "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/matplotlib/cbook/__init__.py:2366: DeprecationWarning: Using or importing the ABCs from 'collections' instead of from 'collections.abc' is deprecated, and in 3.8 it will stop working\n",
      "  return list(data) if isinstance(data, collections.MappingView) else data\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "EPOCH_NUM=50\r\n",
    "\r\n",
    "#训练EPOCH_NUM轮\r\n",
    "for pass_id in range(EPOCH_NUM):                                  \r\n",
    "    # 开始训练并输出最后一个batch的损失值\r\n",
    "    train_cost = 0\r\n",
    "\r\n",
    "    #遍历train_reader迭代器\r\n",
    "    for batch_id, data in enumerate(train_loader()):\r\n",
    "        inputs = paddle.to_tensor(data[0])\r\n",
    "        labels = paddle.to_tensor(data[1])\r\n",
    "        out = net(inputs)\r\n",
    "        train_loss = paddle.mean(paddle.nn.functional.square_error_cost(out, labels))\r\n",
    "        train_loss.backward()\r\n",
    "        optimizer.step()\r\n",
    "        optimizer.clear_grad()\r\n",
    "\r\n",
    "        #每运行40步，输出一次信息,\r\n",
    "        #注意batch_id=0时也输出, 即 0, 40, 80, ...\r\n",
    "        if batch_id % 40 == 0:\r\n",
    "            print(\"Pass:%d, Cost:%0.5f\" % (pass_id, train_loss))    \r\n",
    "        \r\n",
    "        iter = iter + BATCH_SIZE\r\n",
    "        iters.append(iter)\r\n",
    "        train_costs.append(train_loss.numpy()[0])\r\n",
    "       \r\n",
    "    # 开始测试并输出最后一个batch的损失值\r\n",
    "    test_loss = 0\r\n",
    "\r\n",
    "    #遍历test_reader迭代器\r\n",
    "    for batch_id, data in enumerate(valid_loader()):               \r\n",
    "        inputs = paddle.to_tensor(data[0])\r\n",
    "        labels = paddle.to_tensor(data[1])\r\n",
    "        out = net(inputs)\r\n",
    "        test_loss = paddle.mean(paddle.nn.functional.square_error_cost(out, labels))\r\n",
    "        \r\n",
    "    #打印最后一个batch的损失值\r\n",
    "    print('Test:%d, Cost:%0.5f' % (pass_id, test_loss))     \r\n",
    "    \r\n",
    "#保存模型\r\n",
    "paddle.save(net.state_dict(), 'fit_a_line.pdparams')\r\n",
    "\r\n",
    "draw_train_process(iters,train_costs)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "# **Step5.模型预测**\n",
    "\n",
    "**(1)可视化真实值与预测值方法定义**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "infer_results = []\r\n",
    "groud_truths = []\r\n",
    "\r\n",
    "#绘制真实值和预测值对比图\r\n",
    "def draw_infer_result(groud_truths, infer_results):\r\n",
    "    title='Boston'\r\n",
    "    plt.title(title, fontsize=24)\r\n",
    "    x = np.arange(1,20) \r\n",
    "    y = x\r\n",
    "    plt.plot(x, y)\r\n",
    "    plt.xlabel('ground truth', fontsize=14)\r\n",
    "    plt.ylabel('infer result', fontsize=14)\r\n",
    "    plt.scatter(groud_truths, infer_results, color='green',label='training cost') \r\n",
    "    plt.grid()\r\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "**(2)开始预测**\n",
    "\n",
    "通过paddle.load加载已经训练好的模型，来对从未遇见过的数据进行预测。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Index:0, Infer Result: 13.28, Ground Truth: 8.50\n",
      "Index:1, Infer Result: 13.60, Ground Truth: 5.00\n",
      "Index:2, Infer Result: 12.96, Ground Truth: 11.90\n",
      "Index:3, Infer Result: 14.35, Ground Truth: 27.90\n",
      "Index:4, Infer Result: 13.48, Ground Truth: 17.20\n",
      "Index:5, Infer Result: 14.10, Ground Truth: 27.50\n",
      "Index:6, Infer Result: 13.78, Ground Truth: 15.00\n",
      "Index:7, Infer Result: 13.48, Ground Truth: 17.20\n",
      "Index:8, Infer Result: 11.57, Ground Truth: 17.90\n",
      "Index:9, Infer Result: 13.39, Ground Truth: 16.30\n",
      "Index:10, Infer Result: 11.18, Ground Truth: 7.00\n",
      "Index:11, Infer Result: 12.51, Ground Truth: 7.20\n",
      "Index:12, Infer Result: 13.04, Ground Truth: 7.50\n",
      "Index:13, Infer Result: 12.64, Ground Truth: 10.40\n",
      "Index:14, Infer Result: 12.86, Ground Truth: 8.80\n",
      "Index:15, Infer Result: 13.46, Ground Truth: 8.40\n",
      "Index:16, Infer Result: 14.27, Ground Truth: 16.70\n",
      "Index:17, Infer Result: 14.11, Ground Truth: 14.20\n",
      "Index:18, Infer Result: 14.44, Ground Truth: 20.80\n",
      "Index:19, Infer Result: 13.16, Ground Truth: 13.40\n",
      "Index:20, Infer Result: 13.69, Ground Truth: 11.70\n",
      "Index:21, Infer Result: 12.63, Ground Truth: 8.30\n",
      "Index:22, Infer Result: 14.15, Ground Truth: 10.20\n",
      "Index:23, Infer Result: 13.77, Ground Truth: 10.90\n",
      "Index:24, Infer Result: 13.51, Ground Truth: 11.00\n",
      "Index:25, Infer Result: 13.03, Ground Truth: 9.50\n",
      "Index:26, Infer Result: 14.01, Ground Truth: 14.50\n",
      "Index:27, Infer Result: 13.94, Ground Truth: 14.10\n",
      "Index:28, Infer Result: 14.67, Ground Truth: 16.10\n",
      "Index:29, Infer Result: 13.76, Ground Truth: 14.30\n",
      "Index:30, Infer Result: 13.62, Ground Truth: 11.70\n",
      "Index:31, Infer Result: 13.16, Ground Truth: 13.40\n",
      "Index:32, Infer Result: 13.25, Ground Truth: 9.60\n",
      "Index:33, Infer Result: 12.33, Ground Truth: 8.70\n",
      "Index:34, Infer Result: 12.04, Ground Truth: 8.40\n",
      "Index:35, Infer Result: 13.59, Ground Truth: 12.80\n",
      "Index:36, Infer Result: 13.73, Ground Truth: 10.50\n",
      "Index:37, Infer Result: 14.10, Ground Truth: 17.10\n",
      "Index:38, Infer Result: 14.22, Ground Truth: 18.40\n",
      "Index:39, Infer Result: 14.15, Ground Truth: 15.40\n",
      "Index:40, Infer Result: 13.12, Ground Truth: 10.80\n",
      "Index:41, Infer Result: 12.83, Ground Truth: 11.80\n",
      "Index:42, Infer Result: 14.02, Ground Truth: 14.90\n",
      "Index:43, Infer Result: 14.28, Ground Truth: 12.60\n",
      "Index:44, Infer Result: 14.20, Ground Truth: 14.10\n",
      "Index:45, Infer Result: 13.96, Ground Truth: 13.00\n",
      "Index:46, Infer Result: 13.44, Ground Truth: 13.40\n",
      "Index:47, Infer Result: 14.33, Ground Truth: 15.20\n",
      "Index:48, Infer Result: 14.39, Ground Truth: 16.10\n",
      "Index:49, Infer Result: 14.78, Ground Truth: 17.80\n",
      "Index:50, Infer Result: 13.33, Ground Truth: 14.90\n",
      "Index:51, Infer Result: 13.53, Ground Truth: 14.10\n",
      "Index:52, Infer Result: 13.08, Ground Truth: 12.70\n",
      "Index:53, Infer Result: 13.31, Ground Truth: 13.50\n",
      "Index:54, Infer Result: 14.26, Ground Truth: 14.90\n",
      "Index:55, Infer Result: 14.63, Ground Truth: 20.00\n",
      "Index:56, Infer Result: 14.28, Ground Truth: 16.40\n",
      "Index:57, Infer Result: 14.69, Ground Truth: 17.70\n",
      "Index:58, Infer Result: 14.81, Ground Truth: 19.50\n",
      "Index:59, Infer Result: 15.05, Ground Truth: 20.20\n",
      "Index:60, Infer Result: 15.28, Ground Truth: 21.40\n",
      "Index:61, Infer Result: 15.09, Ground Truth: 19.90\n",
      "Index:62, Infer Result: 13.48, Ground Truth: 19.00\n",
      "Index:63, Infer Result: 14.17, Ground Truth: 19.10\n",
      "Index:64, Infer Result: 14.85, Ground Truth: 19.10\n",
      "Index:65, Infer Result: 15.31, Ground Truth: 20.10\n",
      "Index:66, Infer Result: 14.98, Ground Truth: 19.90\n",
      "Index:67, Infer Result: 15.31, Ground Truth: 19.60\n",
      "Index:68, Infer Result: 15.38, Ground Truth: 23.20\n",
      "Index:69, Infer Result: 15.78, Ground Truth: 29.80\n",
      "Index:70, Infer Result: 14.19, Ground Truth: 13.80\n",
      "Index:71, Infer Result: 13.91, Ground Truth: 13.30\n",
      "Index:72, Infer Result: 14.69, Ground Truth: 16.70\n",
      "Index:73, Infer Result: 13.50, Ground Truth: 12.00\n",
      "Index:74, Infer Result: 14.51, Ground Truth: 14.60\n",
      "Index:75, Infer Result: 15.03, Ground Truth: 21.40\n",
      "Index:76, Infer Result: 15.88, Ground Truth: 23.00\n",
      "Index:77, Infer Result: 16.16, Ground Truth: 23.70\n",
      "Index:78, Infer Result: 16.33, Ground Truth: 25.00\n",
      "Index:79, Infer Result: 16.05, Ground Truth: 21.80\n",
      "Index:80, Infer Result: 15.63, Ground Truth: 20.60\n",
      "Index:81, Infer Result: 15.93, Ground Truth: 21.20\n",
      "Index:82, Infer Result: 15.10, Ground Truth: 19.10\n",
      "Index:83, Infer Result: 15.66, Ground Truth: 20.60\n",
      "Index:84, Infer Result: 16.09, Ground Truth: 15.20\n",
      "Index:85, Infer Result: 15.40, Ground Truth: 7.00\n",
      "Index:86, Infer Result: 14.76, Ground Truth: 8.10\n",
      "Index:87, Infer Result: 16.22, Ground Truth: 13.60\n",
      "Index:88, Infer Result: 16.82, Ground Truth: 20.10\n",
      "Index:89, Infer Result: 20.30, Ground Truth: 21.80\n",
      "Index:90, Infer Result: 20.46, Ground Truth: 24.50\n",
      "Index:91, Infer Result: 20.25, Ground Truth: 23.10\n",
      "Index:92, Infer Result: 19.19, Ground Truth: 19.70\n",
      "Index:93, Infer Result: 19.93, Ground Truth: 18.30\n",
      "Index:94, Infer Result: 20.20, Ground Truth: 21.20\n",
      "Index:95, Infer Result: 19.71, Ground Truth: 17.50\n",
      "Index:96, Infer Result: 19.88, Ground Truth: 16.80\n",
      "Index:97, Infer Result: 21.37, Ground Truth: 22.40\n",
      "Index:98, Infer Result: 21.10, Ground Truth: 20.60\n",
      "Index:99, Infer Result: 21.52, Ground Truth: 23.90\n",
      "Index:100, Infer Result: 21.39, Ground Truth: 22.00\n",
      "Index:101, Infer Result: 21.08, Ground Truth: 11.90\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import paddle\r\n",
    "import numpy as np\r\n",
    "import matplotlib.pyplot as plt\r\n",
    "\r\n",
    "\r\n",
    "valid_dataset = paddle.text.UCIHousing(mode='test')\r\n",
    "infer_loader = paddle.io.DataLoader(valid_dataset, batch_size=200)\r\n",
    "\r\n",
    "infer_net = paddle.nn.Linear(13, 1)\r\n",
    "param = paddle.load('fit_a_line.pdparams')\r\n",
    "infer_net.set_dict(param)\r\n",
    "\r\n",
    "\r\n",
    "data = next(infer_loader())\r\n",
    "inputs = paddle.to_tensor(data[0])\r\n",
    "results = infer_net(inputs)\r\n",
    "\r\n",
    "for idx, item in enumerate(zip(results, data[1])):\r\n",
    "    print(\"Index:%d, Infer Result: %.2f, Ground Truth: %.2f\" % (idx, item[0], item[1]))\r\n",
    "    infer_results.append(item[0].numpy()[0])\r\n",
    "    groud_truths.append(item[1].numpy()[0])\r\n",
    "    \r\n",
    "draw_infer_result(groud_truths, infer_results)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "到这里你已经学会了PaddlePaddle的基本命令和第一个小例子！恭喜你已经入门啦~ 如果想学习更多入门内容欢迎查看AI Studio 上的更多内容，有精选项目，也有优质课程，加油哦！\n",
    "\n",
    "NLP12日打卡营即将开始，你准备好了吗！"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "PaddlePaddle 2.0.0b0 (Python 3.5)",
   "language": "python",
   "name": "py35-paddle1.2.0"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}