深度学习（十四）卷积神经网络网络中的网络（NiN） pytorch新手入门及常见问题解决

最新推荐文章于 2024-10-09 00:00:00 发布

浮云白日303

最新推荐文章于 2024-10-09 00:00:00 发布

阅读量343

点赞数 6

文章标签：深度学习 cnn pytorch 人工智能神经网络 python 学习

本文链接：https://blog.csdn.net/m0_68737530/article/details/136002043

版权

本文介绍了如何在PyTorch中实现NIN(NetworkinNetwork)架构，用于处理Fashion-MNIST数据集的图像分类任务。代码展示了网络结构、层间操作以及训练过程。

摘要由CSDN通过智能技术生成

##NIN

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

def nn_block(in_channel,out_channel,kernal,strides,padding):
    return nn.Sequential(nn.Conv2d(in_channel,out_channel,kernal,strides,padding),nn.ReLU(),
                  nn.Conv2d(out_channel,out_channel,kernel_size=1),nn.ReLU(),
                  nn.Conv2d(out_channel,out_channel,kernel_size=1),nn.ReLU())

net = nn.Sequential(nn_block(1,96,kernal=11,strides=4,padding=0),
                    nn.MaxPool2d(3,stride=2),
                    nn_block(96,256,kernal=5,strides=1,padding=2),
                    nn.MaxPool2d(3,stride=2),
                    nn_block(256,384,kernal=3,strides=1,padding=1),
                    nn.MaxPool2d(3,stride=2),nn.Dropout(0.5),
                   nn_block(384,10,kernal=3,strides=1,padding=1),
                    nn.AdaptiveAvgPool2d((1,1)),nn.Flatten())

#查看每个块的形状
x = torch.rand(size=(1,1,224,224))
for layer in net:
    x=layer(x)
    print(layer.__class__.__name__,"out shape:\\t",x.shape)

xuexilv,xunlianxishu,btach = 0.1,10,128

train_iter,test_iter = d2l.load_data_fashion_mnist(btach,resize=224)
d2l.train_ch6(net,train_iter,test_iter,xunlianxishu,xuexilv,d2l.try_gpu())
d2l.plt.show()