论文要点：GoogLeNet（模型M+公式F+代码C）

# GoogLeNet

1. 模型

图片来源：动手学深度学习 by 李沐

2. 代码

import torch
from torch import nn
from torch.nn import functional as F
from d2l import torch as d2l

class Inception_my(nn.Module):
    def __init__(self, in_channels, c1, c2, c3, c4, **kwargs):
        super(Inception_my, self).__init__(**kwargs)
        self.p1_1 = nn.Conv2d(in_channels, c1, kernel_size=1)
        self.p2_1 = nn.Conv2d(in_channels, c2[0], kernel_size=1)
        self.p2_2 = nn.Conv2d(c2[0], c2[1], kernel_size=3, padding=1)
        self.p3_1 = nn.Conv2d(in_channels, c3[0], kernel_size=1)
        self.p3_2 = nn.Conv2d(c3[0], c3[1], kernel_size=5, padding=2)
        self.p4_1 = nn.MaxPool2d(3, stride=1, padding=1)
        self.p4_2 = nn.Conv2d(in_channels, c4, kernel_size=1)
    
    def forward(self, X):
        p1 = F.relu(self.p1_1(X))
        p2 = F.relu(self.p2_2(F.relu(self.p2_1(X))))
        p3 = F.relu(self.p3_2(F.relu(self.p3_1(X))))
        p4 = F.relu(self.p4_2(self.p4_1(X)))
        return torch.concat((p1, p2, p3, p4), dim=1)  #(0, 1, 2, 3), dim=1表channels

# 实现5个stage
b1 = nn.Sequential(nn.Conv2d(1, 64, kernel_size=7, stride=2, padding=3), nn.ReLU(),
                   nn.MaxPool2d(3, stride=2, padding=1))
b2 = nn.Sequential(nn.Conv2d(64, 64, kernel_size=1), nn.ReLU(),
                  nn.Conv2d(64, 192, kernel_size=3, padding=1), 
                  nn.MaxPool2d(3, stride=2, padding=1))
b3 = nn.Sequential(Inception_my(192, 64, (96, 128), (16, 32), 32),
                  Inception_my(256, 128, (128, 192), (32, 96), 64),
                  nn.MaxPool2d(3, stride=2, padding=1))
b4 = nn.Sequential(Inception_my(480, 192, (96, 208), (16, 48), 64),
                   Inception_my(512, 160, (112, 224), (24, 64), 64),
                   Inception_my(512, 128, (128, 256), (24, 64), 64),
                   Inception_my(512, 112, (144, 288), (32, 64), 64),
                   Inception_my(528, 256, (160, 320), (32, 128), 128),
                   nn.MaxPool2d(kernel_size=3, stride=2, padding=1))
b5 = nn.Sequential(Inception_my(832, 256, (160, 320), (32, 128), 128),
                   Inception_my(832, 384, (192, 384), (48, 128), 128),
                   nn.AdaptiveAvgPool2d((1,1)),
                   nn.Flatten())
net = nn.Sequential(b1, b2, b3, b4, b5, nn.Linear(1024, 10))

X = torch.rand((1, 1, 96, 96), dtype=torch.float32)
for layer in net:
    X = layer(X)
    print(layer.__class__.__name__, 'output shape:', X.shape)

lr, num_epochs, batch_size = 0.1, 10, 128
train_iter, test_iter = d2l.load_data_fashion_mnist(batch_size)
d2l.train_ch6(net, train_iter, test_iter, num_epochs, lr, torch.try_gpu())