VGG的开销真大啊…显卡在咆哮

import torch
from torch import nn
from d2l import torch as d2l


# VGG块
def vgg_block(num_convs, in_channels, out_channels):
    layers = []
    for _ in range(num_convs):
        layers.append(nn.Conv2d(in_channels, out_channels, kernel_size=3, padding=1))
        layers.append(nn.ReLU())
        in_channels = out_channels
    layers.append(nn.MaxPool2d(kernel_size=2, stride=2))
    return nn.Sequential(*layers)


# VGG网络
conv_arch = ((1, 64), (1, 128), (2, 256), (2, 512), (2, 512))


def vgg(conv_arch):
    conv_blks = []  # VGG块
    in_channels = 1
    for (num_convs, out_channels) in conv_arch:
        conv_blks.append(vgg_block(num_convs, in_channels, out_channels))
        in_channels = out_channels

    return nn.Sequential(
        *conv_blks, nn.Flatten(),
        # 全连接层部分
        nn.Linear(out_channels * 7 * 7, 4096), nn.ReLU(), nn.Dropout(0.5),
        nn.Linear(4096, 4096), nn.ReLU(), nn.Dropout(0.5),
        nn.Linear(4096, 10))

net = vgg(conv_arch)

# 训练模型
ratio = 4
small_conv_arch = [(pair[0], pair[1] //ratio) for pair in conv_arch] # 通道数除以4
net = vgg(small_conv_arch)

lr, num_epochs, batch_size = 0.05, 10, 128
train_iter, test_iter = d2l.load_data_fashion_mnist(batch_size, resize=224)
d2l.train_ch6(net, train_iter, test_iter, num_epochs, lr, d2l.try_gpu())
d2l.plt.show()

Output

loss 0.177, train acc 0.935, test acc 0.918
1482.9 examples/sec on cuda:0

\
发现效果居然出奇的好