{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 3.9 多层感知机的从零开始实现"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.4.1\n"
     ]
    }
   ],
   "source": [
    "import torch\n",
    "import numpy as np\n",
    "import sys\n",
    "sys.path.append(\"..\") # 为了导入上层目录的d2lzh_pytorch\n",
    "import d2lzh_pytorch as d2l\n",
    "\n",
    "print(torch.__version__)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3.9.1 获取和读取数据"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "batch_size = 256\n",
    "train_iter, test_iter = d2l.load_data_fashion_mnist(batch_size)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3.9.2 定义模型参数"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "num_inputs, num_outputs, num_hiddens = 784, 10, 256\n",
    "\n",
    "W1 = torch.tensor(np.random.normal(0, 0.01, (num_inputs, num_hiddens)), dtype=torch.float)\n",
    "b1 = torch.zeros(num_hiddens, dtype=torch.float)\n",
    "W2 = torch.tensor(np.random.normal(0, 0.01, (num_hiddens, num_outputs)), dtype=torch.float)\n",
    "b2 = torch.zeros(num_outputs, dtype=torch.float)\n",
    "\n",
    "params = [W1, b1, W2, b2]\n",
    "for param in params:\n",
    "    param.requires_grad_(requires_grad=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3.9.3 定义激活函数"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def relu(X):\n",
    "    return torch.max(input=X, other=torch.tensor(0.0))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3.9.4 定义模型"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def net(X):\n",
    "    X = X.view((-1, num_inputs))\n",
    "    H = relu(torch.matmul(X, W1) + b1)\n",
    "    return torch.matmul(H, W2) + b2"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3.9.5 定义损失函数"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "loss = torch.nn.CrossEntropyLoss()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3.9.6 训练模型"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 1, loss 0.0030, train acc 0.714, test acc 0.753\n",
      "epoch 2, loss 0.0019, train acc 0.821, test acc 0.777\n",
      "epoch 3, loss 0.0017, train acc 0.842, test acc 0.834\n",
      "epoch 4, loss 0.0015, train acc 0.857, test acc 0.839\n",
      "epoch 5, loss 0.0014, train acc 0.865, test acc 0.845\n"
     ]
    }
   ],
   "source": [
    "num_epochs, lr = 5, 100.0\n",
    "d2l.train_ch3(net, train_iter, test_iter, loss, num_epochs, batch_size, params, lr)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python [default]",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}