{
"cells": [
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "-"
}
},
"source": [
"# 含并行连结的网络（GoogLeNet）\n",
"\n",
"Inception块"
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {
"origin_pos": 2,
"tab": [
"pytorch"
]
},
"outputs": [],
"source": [
"import torch\n",
"from torch import nn\n",
"from torch.nn import functional as F\n",
"from d2l import torch as d2l\n",
"\n",
"\n",
"class Inception(nn.Module):\n",
" def __init__(self, in_channels, c1, c2, c3, c4, **kwargs):\n",
" super(Inception, self).__init__(**kwargs)\n",
" self.p1_1 = nn.Conv2d(in_channels, c1, kernel_size=1)\n",
" #1X1卷积\n",
" self.p2_1 = nn.Conv2d(in_channels, c2[0], kernel_size=1)\n",
" self.p2_2 = nn.Conv2d(c2[0], c2[1], kernel_size=3, padding=1)\n",
" #1X1卷积，3X3卷积\n",
" self.p3_1 = nn.Conv2d(in_channels, c3[0], kernel_size=1)\n",
" self.p3_2 = nn.Conv2d(c3[0], c3[1], kernel_size=5, padding=2)\n",
" #1X1卷积，5X5卷积\n",
" self.p4_1 = nn.MaxPool2d(kernel_size=3, stride=1, padding=1)\n",
" self.p4_2 = nn.Conv2d(in_channels, c4, kernel_size=1)\n",
" #maxpool,1x1卷积\n",
" def forward(self, x):\n",
" p1 = F.relu(self.p1_1(x))\n",
" p2 = F.relu(self.p2_2(F.relu(self.p2_1(x))))\n",
" p3 = F.relu(self.p3_2(F.relu(self.p3_1(x))))\n",
" p4 = F.relu(self.p4_2(self.p4_1(x)))\n",
" #定义四个模块，然后将输出在通道维度上拼接在一起\n",
" return torch.cat((p1, p2, p3, p4), dim=1)"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "slide"
}
},
"source": [
"GoogLeNet模型"
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {
"origin_pos": 22,
"tab": [
"pytorch"
]
},
"outputs": [],
"source": [
"b1 = nn.Sequential(nn.Conv2d(1, 64, kernel_size=7, stride=2, padding=3),\n",
" nn.ReLU(),\n",
" nn.MaxPool2d(kernel_size=3, stride=2, padding=1))\n",
"\n",
"b2 = nn.Sequential(nn.Conv2d(64, 64, kernel_size=1),\n",
" nn.ReLU(),\n",
" nn.Conv2d(64, 192, kernel_size=3, padding=1),\n",
" nn.ReLU(),\n",
" nn.MaxPool2d(kernel_size=3, stride=2, padding=1))\n",
"\n",
"b3 = nn.Sequential(Inception(192, 64, (96, 128), (16, 32), 32),\n",
" Inception(256, 128, (128, 192), (32, 96), 64),\n",
" nn.MaxPool2d(kernel_size=3, stride=2, padding=1))\n",
"\n",
"b4 = nn.Sequential(Inception(480, 192, (96, 208), (16, 48), 64),\n",
" Inception(512, 160, (112, 224), (24, 64), 64),\n",
" Inception(512, 128, (128, 256), (24, 64), 64),\n",
" Inception(512, 112, (144, 288), (32, 64), 64),\n",
" Inception(528, 256, (160, 320), (32, 128), 128),\n",
" nn.MaxPool2d(kernel_size=3, stride=2, padding=1))\n",
"\n",
"b5 = nn.Sequential(Inception(832, 256, (160, 320), (32, 128), 128),\n",
" Inception(832, 384, (192, 384), (48, 128), 128),\n",
" nn.AdaptiveAvgPool2d((1,1)),\n",
" nn.Flatten())\n",
"\n",
"net = nn.Sequential(b1, b2, b3, b4, b5, nn.Linear(1024, 10))"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "slide"
}
},
"source": [
"为了使Fashion-MNIST上的训练短小精悍，我们将输入的高和宽从224降到96"
]
},
{
"cell_type": "code",
"execution_count": 7,
"metadata": {
"origin_pos": 26,
"tab": [
"pytorch"
]
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Sequential output shape:\t torch.Size([1, 64, 24, 24])\n",
"Sequential output shape:\t torch.Size([1, 192, 12, 12])\n",
"Sequential output shape:\t torch.Size([1, 480, 6, 6])\n",
"Sequential output shape:\t torch.Size([1, 832, 3, 3])\n",
"Sequential output shape:\t torch.Size([1, 1024])\n",
"Linear output shape:\t torch.Size([1, 10])\n"
]
}
],
"source": [
"X = torch.rand(size=(1, 1, 96, 96))\n",
"for layer in net:\n",
" X = layer(X)\n",
" print(layer.__class__.__name__,'output shape:\\t', X.shape)"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "slide"
}
},
"source": [
"训练模型"
]
},
{
"cell_type": "code",
"execution_count": 8,
"metadata": {
"origin_pos": 29,
"tab": [
"pytorch"
]
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"loss 0.246, train acc 0.906, test acc 0.884\n",
"3448.9 examples/sec on cuda:0\n"
]
},
{
"data": {
"image/svg+xml": [
"\n",
"\n",
"\n",
"\n"
],
"text/plain": [
""
]
},
"metadata": {
"needs_background": "light"
},
"output_type": "display_data"
}
],
"source": [
"lr, num_epochs, batch_size = 0.1, 10, 128\n",
"train_iter, test_iter = d2l.load_data_fashion_mnist(batch_size, resize=96)\n",
"d2l.train_ch6(net, train_iter, test_iter, num_epochs, lr, d2l.try_gpu())"
]
}
],
"metadata": {
"celltoolbar": "Slideshow",
"kernelspec": {
"display_name": "Python 3 (ipykernel)",
"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": "3.8.12"
},
"rise": {
"autolaunch": true,
"enable_chalkboard": true,
"overlay": "",
"scroll": true
}
},
"nbformat": 4,
"nbformat_minor": 4
}
