CNN/LeNet.ipynb

{
 "cells": [
  {
   "attachments": {
    "lenet-vert.png": {
     "image/png": "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"
    }
   },
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "LeNet（LeNet-5）由两个部分组成：\n",
    "\n",
    "1、卷积编码器：由两个卷积层组成;\n",
    "2、全连接层密集块：由三个全连接层组成。\n",
    "\n",
    "每个卷积块中的基本单元是一个卷积层、一个sigmoid激活函数和平均汇聚层。请注意，虽然ReLU和最大汇聚层更有效，但它们在20世纪90年代还没有出现。每个卷积层使用\n",
    "卷积核和一个sigmoid激活函数。这些层将输入映射到多个二维特征输出，通常同时增加通道的数量。第一卷积层有6个输出通道，而第二个卷积层有16个输出通道。每个\n",
    "池操作（步幅2）通过空间下采样将维数减少4倍。卷积的输出形状由批量大小、通道数、高度、宽度决定。\n",
    "\n",
    "为了将卷积块的输出传递给稠密块，我们必须在小批量中展平每个样本。换言之，我们将这个四维输入转换成全连接层所期望的二维输入。这里的二维表示的第一个维度索引小批量中的样本，第二个维度给出每个样本的平面向量表示。LeNet的稠密块有三个全连接层，分别有120、84和10个输出。因为我们在执行分类任务，所以输出层的10维对应于最后输出结果的数量。\n",
    "\n",
    "![lenet-vert.png](attachment:lenet-vert.png)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "from torch import nn\n",
    "import sys\n",
    "sys.path.append('..')\n",
    "from tools import tool_pytorch_017 as ty\n",
    "\n",
    "# 构建LeNet网络结构\n",
    "net = nn.Sequential(\n",
    "    nn.Conv2d(1, 6, kernel_size=5, padding=2), nn.Sigmoid(),\n",
    "    nn.AvgPool2d(kernel_size=2, stride=2),\n",
    "    nn.Conv2d(6, 16, kernel_size=5), nn.Sigmoid(),\n",
    "    nn.AvgPool2d(kernel_size=2, stride=2),\n",
    "    nn.Flatten(),\n",
    "    nn.Linear(16 * 5 * 5, 120), nn.Sigmoid(),\n",
    "    nn.Linear(120, 84), nn.Sigmoid(),\n",
    "    nn.Linear(84, 10))\n",
    "\n",
    "# 查看LeNet的分层结构输出\n",
    "X = torch.rand(size=(1, 1, 28, 28), dtype=torch.float32)\n",
    "for layer in net:\n",
    "    X = layer(X)\n",
    "    print(layer.__class__.__name__,'output shape: \\t',X.shape)\n",
    "    \n",
    "batch_size = 256\n",
    "# 加载数据集\n",
    "train_iter, test_iter = ty.load_data_fashion_mnist(batch_size=batch_size)\n",
    "\n",
    "def evaluate_accuracy_gpu(net, data_iter, device=None): #@save\n",
    "    \"\"\"使用GPU计算模型在数据集上的精度\"\"\"\n",
    "    if isinstance(net, nn.Module):\n",
    "        net.eval()  # 设置为评估模式\n",
    "        if not device:\n",
    "            device = next(iter(net.parameters())).device\n",
    "    # 正确预测的数量，总预测的数量\n",
    "    metric = ty.Accumulator(2)\n",
    "    with torch.no_grad():\n",
    "        for X, y in data_iter:\n",
    "            if isinstance(X, list):\n",
    "                # BERT微调所需的（之后将介绍）\n",
    "                X = [x.to(device) for x in X]\n",
    "            else:\n",
    "                X = X.to(device)\n",
    "            y = y.to(device)\n",
    "            metric.add(ty.accuracy(net(X), y), y.numel())\n",
    "    return metric[0] / metric[1]\n",
    "\n",
    "\n",
    "#@save\n",
    "def train_LeNet(net, train_iter, test_iter, num_epochs, lr, device):\n",
    "    \"\"\"训练LeNet模型\"\"\"\n",
    "    def init_weights(m):\n",
    "        if type(m) == nn.Linear or type(m) == nn.Conv2d:\n",
    "            nn.init.xavier_uniform_(m.weight)\n",
    "    net.apply(init_weights)\n",
    "    print('training on', device)\n",
    "    net.to(device)\n",
    "    optimizer = torch.optim.SGD(net.parameters(), lr=lr)\n",
    "    loss = nn.CrossEntropyLoss()\n",
    "    animator = ty.Animator(xlabel='epoch', xlim=[1, num_epochs],\n",
    "                            legend=['train loss', 'train acc', 'test acc'])\n",
    "    timer, num_batches = ty.Timer(), len(train_iter)\n",
    "    for epoch in range(num_epochs):\n",
    "        # 训练损失之和，训练准确率之和，样本数\n",
    "        metric = ty.Accumulator(3)\n",
    "        net.train()\n",
    "        for i, (X, y) in enumerate(train_iter):\n",
    "            timer.start()\n",
    "            optimizer.zero_grad()\n",
    "            X, y = X.to(device), y.to(device)\n",
    "            y_hat = net(X)\n",
    "            l = loss(y_hat, y)\n",
    "            l.backward()\n",
    "            optimizer.step()\n",
    "            with torch.no_grad():\n",
    "                metric.add(l * X.shape[0], ty.accuracy(y_hat, y), X.shape[0])\n",
    "            timer.stop()\n",
    "            train_l = metric[0] / metric[2]\n",
    "            train_acc = metric[1] / metric[2]\n",
    "            if (i + 1) % (num_batches // 5) == 0 or i == num_batches - 1:\n",
    "                animator.add(epoch + (i + 1) / num_batches,\n",
    "                             (train_l, train_acc, None))\n",
    "        test_acc = evaluate_accuracy_gpu(net, test_iter)\n",
    "        animator.add(epoch + 1, (None, None, test_acc))\n",
    "    print(f'loss {train_l:.3f}, train acc {train_acc:.3f}, '\n",
    "          f'test acc {test_acc:.3f}')\n",
    "    print(f'{metric[2] * num_epochs / timer.sum():.1f} examples/sec '\n",
    "          f'on {str(device)}')\n",
    "    \n",
    "lr, num_epochs = 0.9, 10\n",
    "train_LeNet(net, train_iter, test_iter, num_epochs, lr, ty.try_gpu())"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "python3.8",
   "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": "3.8.19"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}